Datasets:

mimiklee

/

masterthesis-longt5-sum-8k

like 0

METHODS --- Data: the American Time Use Survey The primary dataset is the American Time Use Survey (ATUS). The ATUS is an annual, nationally representative cross-sectional survey, administered by the Bureau of Labor Statistics since commencing in 2003. A monthly ATUS sample is randomly drawn from respondents recently completing the bureau's Current Population Survey. Respondents chronologically list what they consider to be their primary activities and the activities' durations beginning with 4 am on the previous day through 4 am on the day of the interview, a 24-h period referred to as the respondents' "diary day". A limitation of the data is that only the activity which respondents consider to be their primary engagement at a given time is recorded; other activities performed simultaneously and considered secondary are omitted. This information is categorized on a per-minute time scale into a three-tiered system of approximately 400 activities (e.g., an activity may be classified as "Work & Work-Related Activities" in the first tier, "Working" in the second tier, and "Work, main job" in the third). The ATUS also records where each activity took place, or for travel activities, the transit mode. Additionally, researchers have constructed MET intensity values for each ATUS activity category. 20 A "MET" (or "metabolic equivalent") is a unit commonly used to gauge the intensity of a physical activity and is roughly defined as the ratio of energy expenditure in an activity to energy expenditure while at rest. A MET value of 3.0 is considered the threshold for moderate intensity. --- Primary Measures (a) Health-Related Activity Outcomes Four health-related activities are modeled: (1) aggregate physical activity, the summation of time spent in 35 individual exercise and sports activities and time for any other activity for which the MET value is 3.0 or greater; (2) total time spent preparing food; (3) the total time spent eating as a primary activity; and (4) total time spent sleeping. These time usages are appropriate measures of healthy behaviors. Physical activity includes deliberate exercise as well as other sources daily activity, which may be an important source for many. Given the absence of precise nutritional intake, food preparation and the time spent eating with family proxy for diet quality. More time in these activities suggests that fewer pre-prepared or processed foods consumed. Additionally, to provide benchmark comparisons to the healthy-related behaviors, the time spent (5) socializing or communicating with others and (6) watching television, which together comprise approximately one third of the sample's waking leisure time, are also measured. --- (b) Commuting The principal explanatory factor is the total daily time spent commuting. Classifying a commute is not straight-forward because commuters often run other errands over the course of their journey. 21 To accommodate commuting trips with multiple destinations, a commute is defined as all travel time for any purpose from the time the respondent leaves home until arrival at work, and vice versa. A respondent's total daily commuting time is the summation of all qualifying travel time. --- (c) Covariates Several demographic control variables are included: age, gender, race and Hispanic status, education, marital status, number of children, an indicator for the presence of a child aged 1 year or less, hourly wage, household income, school enrollment status, and employment status by occupation and industry. Covariates also include fixed effects for the respondents' metropolitan area of residence and the diary day date (the year, month, day of the week, and whether it was a holiday). Additionally, two indicators are constructed to identify commuters using active travel modes (walking or bicycling) at least 30 min daily or using public transportation during any portion of their commute. Lastly, commuting is a workrelated activity. Because in addition to any trade-off with commuting labor time itself constrains healthy behaviors, 22 the total daily time spent working is also included. --- Sample Construction The sample is constructed as follows: when ATUS samples from survey years 2003 through 2010 are pooled, 112,038 observations comprise the full set. Given the focus on commuting, a work-related activity, the sample is limited to working-age (21-65) adults residing within identifiable urban labor markets, specifically the respondent's Core Based Statistical Area. These criteria result in the omission of 23,788 individuals living in rural areas and 20,343 individuals outside of the age range. Next, 3,957 identified as beginning or ending the day away from home or working night shifts were omitted. Commuters falling under loosely defined "traditional" commuting schedules were eligible for inclusion if they arrived at work between 4:30 am and 6 pm, and also arrived at home between 10 am and 11:30 pm. Furthermore, 28,019 respondents who were either not employed or not employed full-time and thus not subject to comparable time constraints were dropped. A further 3,932 employed respondents who did not report wages and 5,877 who did not report household income were not included. Finally, 1,261 individuals with time usages in any health-related activity, socializing, television, labor, or commuting time exceeding the 99th percentile (measured at the full pooled sample) were excluded as outliers. The final dataset includes 24,861 observations. --- Statistical Methods The empirical objective is to estimate how the participation in several health-related activities varies by the amount of time spent commuting. The analysis uses seemingly unrelated regression (SUR), which acknowledges the interrelatedness among activities within a given day. In particular, the SUR procedure allows for correlation among the error terms within an individual's daily activities. Commuting time is entered into the model in linear, quadratic, and cubic terms, and labor time is entered in linear and quadratic. Additionally, interaction terms are included between commuting time and labor time, and also between commuting time and the indicators for public transportation mode and active commuting. A quadratic term is included for age and respondents' hourly wage and household income are logged. The SUR results are then available to produce covariate-adjusted means for each health-related activity at each commuting length. The adjusted means are then used to accomplish two primary aims: (1) to calculate how much involvement in a specific activity decreases when commuting time increases, to gauge that activity's responsiveness to commuting increases, and (2) to calculate the percentages of commuting time that are derived from decreases in each health-related activity. These percentages indicate in which activities commuters would otherwise be involved rather than traveling to work. Commuting time increases are calculated using the mean commuting time as the reference comparison. --- RESULTS Table 1 presents time use activity and covariate descriptions and sample means and categorical percentages. Summary statistics are presented for the overall sample and additionally within several brackets of daily commuting time in 1-h intervals: 0, 1-59, 60-119, 120-179, and 180 min or more total commuting time on the diary day. Time usages are sample averages and are not conditional on having engaged in the specific activity. The mean time spent daily commuting is 40.0 min overall, whereas conditional on having commuted the mean daily commuting time is 62.2 min and the median is 55 min. The covariates' raw means and rates suggest that respondents with longer commutes are a higher proportion male (54.9% of the 1-59-min commuting bracket group compared to 60.3% of the 120-179-min commuting time bracket group), minority race, and ethnicity (whites are 68.6% of the 1-59 bracket and 61.7% of the 120-179 bracket), have higher household income ($60.7 thousand in the 1-59 bracket and $72.7 thousand in the 120-179 bracket), and of higher educational attainment (37.7% have a college degree or higher in the 1-59 bracket, and 46.6% of the 120-179 bracket). Table 2 provides additional descriptive statistics pertinent to commuting. Within the sample, 4.9% or those working on their diary day spend 0 min commuting, 54.2% spend less than 60 min, 89.9% spend less than 120 min, and 1.6% spend 3 h or more commuting to work. The vast majority of the sample commutes to work solely by automobile: 92.8% of the sample overall. Few commuters walk or bicycle to work (2.2%) or use public transportation (3.4%), but commuters using alternative modes are clustered among respondents reporting longer commuting times, representing almost a third of respondents spending 180 min or more daily commuting. 20. The sample is restricted to respondents employed full-time residing in urban counties Table 3 presents adjusted activity means evaluated at half-hour commuting intervals, from 0 to 180 min total daily commuting time. Means are first presented for the four health-related activities individually, and the fifth column displays the aggregate adjusted health-related activity time totals. Means for the time spent socializing are presented in the sixth and seventh columns, respectively. Individually, each health-related activity decreases with increased time spent commuting. Between commutes of 0 and 180 min, mean physical activity decreases 27.3 to 30.1 min (48.4%), food preparation decreases 6.3 to 13.0 min (32.6%), family eating decreases 10.7 to 24.2 min (30.6%), and sleeping decreases 64.5 to 441.0 min (12.8%). Alongside each adjusted mean, within parentheses, ratios of activity times at each commute time relative to activity times at the average commuting time (62.2 min daily) are presented to gauge each activity's responsiveness to commuting time changes. An average commuter whose total commuting time increased 1 h daily to 120 min would experience a 23% reduction in physical activity, a 17% reduction in food preparation, a 8% reduction in time eating with family, and a 3% reduction in sleep time. By comparison, time spent socializing would decrease 3% and time watching television would decrease 14%. In the extreme evaluation of 180 min spent commuting, physical activity time decreases 44% and food preparation time decreases 31% relative to at the mean commuting time, whereas comparatively television time decreases 20%, and an individual spending 3 h commuting daily still views on average 102.0 min of television. Table 4 presents the percentage of time reallocated in each activity as a percentage of the total daily commuting time's difference from the mean commuting time. This is equivalent to the percentage of commuting time attributable to the decrease in participation time of that particular activity. The fifth column displays the summation of the previous four percentages and indicates the total percentage of the daily commuting time which is drawn from the four health-related activities. The total-health-related column illustrates that the majority of commuting timeconsistently above 56%-is attributable to decreases in these four health-related activity categories. The health-related trade-off with commuting is largely comprised of the trade-off with sleep, which is 35.3% of 30 min spent commuting, 30.6% of 60 min spent commuting, and 32.6% of 180 min spent commuting. The trade-off with sleep is fairly constant over the range of commuting time, whereas an increasing proportion of time is drawn from physical activities and food --- DISCUSSION The results indicate that longer commutes are increasingly associated with behavioral patterns which over time may contribute to poor health outcomes. Average commuting times are associated with modest reductions in health-related activities. Decreases in the four health-related activities-physical activity, food preparation, time eating with family, and sleeping-constitute the majority of time from which is commuting reallocated from. The greatest percentage of these four categories is taken from sleeping time. Above the mean commuting time, a greater proportion of commuting time is derived from decreases in food preparation and physical activities of moderate or greater intensity than at the mean. Lastly, even among respondents that report spending 2 or 3 h each day commuting, the average time spent watching television exceeds 100 min the same day. This estimate suggests that sizeable discretionary time still remains that Americans could utilize to improve their health and other aspects of their lives. The presented calculations are for a single average day, only. There are also likely cumulative effects, the calculation of which unfortunately requires extrapolation from the data. Assuming 260 working days per year and linear cumulative tradeoffs, a 90-min-per-day commuter will trade-off 24.2 h of physical activity annually at a moderate or greater intensity relative to a worker with the average 60-min-perday commute time. Even at moderate intensities 24.2 h is the equivalent to thousands of unexpended calories. Slight behavioral changes cumulate to dramatic results: the growth in median body weight since the 1980s is equivalent to only 100-150 additional calories per day-"three Oreo cookies or one can of Pepsi". 18 The modest trade-off estimates support trade-offs with commuting as a small but meaningful factor among an array of complex causes of obesity. These physical activity trade-offs are in addition to any further dietary or sleep impacts. However, because the functional relationships of time inputs of these activities into health are not well understood, health consequences are more difficult to assess. For example, individuals with longer commutes spend less time preparing food and are likely consuming more processed and prepared foods. Yet, the lack of any detailed nutritional intake prohibits further estimating any health implications. Similarly, no research exists to guide quantifying the precise health impacts of 15min lost sleep each working day. Commuting is an activity with diverse practices. Certainly, active modes such as walking or bicycling provide some physical activity via transit and thus may influence health-related behaviors differently than commuting by automobile. Similarly, public transportation enables some commuters to multitask, such as working en route, potentially freeing up additional leisure time for healthy behaviors. However, these practices are predominantly clustered in major cities with the infrastructure to support non-car commuting modes. Active travel in sprawled communities is often inconvenient or infeasible. The vast majority of sample respondents (93%) commuted solely using automobiles, and a large number likely did not have short-term alternative options. Future researchers should evaluate coping strategies long car-dependent commuters could implement immediately, such as parking further from their worksite and walking the remaining distance. Anecdotally, workers in foreign cities experience commutes comparable to or exceeding Americans' travel, yet abroad, obesity and related diseases rates' are often lower. This suggests that beyond the direct time loss to commuting, a crucial element is also the context in which that time loss occurs. Responses to time constraints may differ by culture, attitudes towards health, or the environment. In the USA, leisure time lost to commuting occurs within a context of increasingly low-cost, energy dense foods, which are linked to obesity. 23 Workers with long commutes living in a different food environment may not respond the same. An extension to this study could investigate cross-cultural differences in commuting trade-offs. Lastly, this study's estimates likely only establish an upper-bound on commuting trade-offs. Commuting time is not randomly assigned and self-selection may yield biased estimates. Already, it is noted that the failure to acknowledge unobserved preferences for health or location might lead to misguided interpretation in the association between sprawl and obesity. 24,25 In the commuting context, individuals caring less about their health might purposefully select long commutes to enjoy cheaper housing costs. It is difficult to disentangle such factors from long-run conditions such as commuting time and these results are only intended as an initial attempt to quantify trade-offs. --- CONCLUSIONS This paper describes Americans' trade-offs between commuting time and healthrelated activities. Individuals with longer commutes are increasingly less engaged in health-related activities. These results provide a first look at trade-offs between health-related activities and commuting. Continuing research in this area will inform understanding of work-life balance and how environmental context interacts with lifestyle choices that are risk factors for obesity.

To further understand documented associations between obesity and urban sprawl, this research describes individuals' trade-offs between health-related activities and commuting time. A cross-section of 24,861 working-age individuals employed full-time and residing in urban counties is constructed from the American Time Use Survey (2003-2010). Data are analyzed using seemingly unrelated regressions to quantify health-related activity decreases in response to additional time spent commuting. Outcomes are total daily minutes spent in physical activity at a moderate or greater intensity, preparing food, eating meals with family, and sleeping. Commuting time is measured as all travel time between home and work and vice versa. The mean commuting time is 62 min daily, the median is 55 min, and 10.1% of workers commute 120 min or more. Spending an additional 60 min daily commuting above average is associated with a 6% decrease in aggregate healthrelated activities and spending an additional 120 min is associated with a 12% decrease. The greatest percentage of commuting time comes from sleeping time reductions (28-35%). Additionally, larger proportions of commuting time are taken from physical activity and food preparation relative to the mean commuting length: of 60 min spent commuting, 16.1% is taken from physical activity and 4.1% is taken from food preparation; of 120 min commuting, 20.3% is taken from physical activity and 5.6% is taken from food preparation. The results indicate that longer commutes are associated with behavioral patterns which over time may contribute to obesity and other poor health outcomes. These findings will assist both urban planners and researchers wishing to understand time constraints' impacts on health.

Introduction Measuring inter na tional migra tion is chal leng ing (Bilsborrow et al. 1997). The lack of timely and com pre hen sive data about migrants, com bined with the vary ing mea sures and def ni tions of migra tion used by dif fer ent countries, is a bar rier to accu rately esti mat ing inter na tional migra tion (Bijak 2010;Willekens 1994Willekens, 2019)). In recent years, schol ars have started using Bayes ian meth ods to com bine dif fer ent sources of migration data in order to pro vide bet ter esti ma tes of the migrant stock-the total num ber of migrants pres ent in a coun try at a cer tain date (Azose and Raftery 2019). In this arti cle, we aim to improve esti ma tes by complementing sur vey data with social media data. This is impor tant because, when design ing migra tion pol i cies, it is cru cial to have access to valid sources of data on inter na tional migra tion. We pro pose using a Bayes ian data assess ment model that com bines data from the Labour Force Survey (LFS) and the Facebook Advertising Platform to assess the num ber of Euro pean migrants in the United Kingdom (UK). The goal is to dem on strate how such a model can pro duce a more accu rate esti mate of Euro pean migra tion. We use the UK as an exam ple as it is a Western coun try for which the migra tion data are of poor qual ity. We employ the Integrated Model of Euro pean Migration (IMEM), which is a Bayes ian model for esti mat ing migra tion. This frame work was cre ated by Raymer et al. (2013) for com bin ing the flows reported by the send ing countries with the flows reported by the receiv ing countries to esti mate a num ber closer to the true value of the flows. The IMEM model with mod i f ca tions has been used by Disney (2015) to com bine mul ti ple migra tion sur vey data sets in the UK, and by Wi<unk>niowski (2017) to com bine the LFS data in the case of Pol ish migra tion to the UK. More recently, Del Fava et al. (2019) expanded the model by draw ing on admin is tra tive and house hold sur vey data for 31 Euro pean countries. The main fea ture of the IMEM approach is that it pro vi des a framework that assesses the lim i ta tions of the avail able data sets in terms of the def ni tion of migrants used. Assessments of the bias and the accu racy of these data sets are used to cre ate appro pri ate prior dis tri bu tions to adjust for the iden ti fed data issues. At the same time, a new strand of research has emerged that has been repurposing dig i tal data to com ple ment tra di tional demo graphic data sources and to improve their cov er age and time li ness of pro duc tion. Since dig i tal trace data are often geolocated, migra tion has received par tic u lar atten tion in this lit er a ture. As Cesare et al. (2018) have suggested, using dig i tal trace data sources has advan tages, such as the speed and low cost of data col lec tion, but also lim i ta tions, with issues regard ing the lack of accessi bil ity, trans par ency, and rep re sen ta tive ness. Drawing on data from the Facebook Advertising Platform and the LFS, we inves ti gate whether the dig i tal traces that individ u als leave on Facebook can be used to esti mate stocks of migrants in the UK. This is by no means the frst study that has tried to com bine dig i tal traces with sur vey data (Alexander et al. 2019(Alexander et al., 2020;;Zagheni et al. 2018). However, in this arti cle, we pro pose for the frst time an over arch ing frame work that includes both a the o ret i cal model that con sid ers push and pull fac tors related to migra tion the o ries and a data assess ment model that aims to reduce the bias from the data that enters the model. This frame work pro vi des a more con textspe cifc model for exam in ing migra tion to the UK from sev eral send ing countries. Moreover, our study pro vi des impor tant insights into the com plex real ity of inter na tional migra tion to the UK by shed ding light on the demo graph ics of migrants by coun try of ori gin, which are hard to obtain using cur rently avail able off cial sta tis tics. The atten tion is lim ited to migrants from Euro pean countries because, in the UK con text, these migrant stocks are the hardest to esti mate owing to the "free dom of move ment" that char ac ter izes the Euro pean Union (EU). At least until Decem ber 2020, there has been no require ment for EU migrants in the UK to reg is ter their res i dence. Thus, up to now, sur vey data have been used to esti mate the stock of migrants from the EU. We want to com ple ment these existing, but incom plete, off cial esti ma tes of migrant stocks by ana lyz ing dig i tal trace data. As an illus tra tion, we pro duce an esti mate of the total num ber of EU migrants for 2018 and 2019. Estimating Migrant Stocks Using Digital Traces and Survey Data There are two addi tional rea sons why it is inter est ing to look at the migra tion system of the UK. First, the UK Offce of National Statistics (ONS) bases its esti ma tes of inter na tional migra tion on sur veys. In August 2019, the ONS reclassifed their esti ma tes as exper i men tal sta tis tics, empha siz ing that the esti ma tes might be inac cu rate (ONS 2019a). Furthermore, the sci en tifc lit er a ture has suggested that these sur veys are affected by dif fer ent sources of bias (Coleman 1983;Kupiszewska and Nowok 2008;Kupiszewska et al. 2010;Rendall et al. 2003). In Europe, the UK is an exam ple of a coun try in which there is only a "bronze stan dard," mean ing that the UK migration data sources are infe rior to the "gold stan dard" but are of "suf f cient qual ity for val i da tion" (Azose and Raftery 2019). Second, although the UK has expe ri enced a net pos i tive increase in migra tion from Euro pean countries over the past two decades (Champion and Falkingham 2016), the ONS reported an under count of 16% for the net migra tion esti ma tes for the EU8 countries (Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia, and Slovenia) in 2016, suggesting that the rel evant migra tion sta tis tics are of insuf f cient qual ity (ONS 2019b). Using dig i tal traces might pro vide insights into UK migra tion trends by sex and coun try of ori gin by enabling research ers to pro duce esti ma tes of stocks of Euro pean migrants in the UK. Moreover, Willekens (1994Willekens (, 2019) ) has called for the cre a tion of a syn thetic migration data base that com bines data from dif fer ent sources. The pur pose of this database would be to "cre ate the best pos si ble esti ma tes of the true num ber of migrants" (Willekens 2019). The pres ent arti cle seeks to con trib ute to this "learn ing pro cess" by answer ing the fol low ing research ques tion: What can Facebook adver tis ing data contrib ute to ONS migra tion esti ma tes, in a con text in which there is no "ground truth" data against which model esti ma tes can be val i dated? --- Data --- Traditional Data and Their Limitations A gold stan dard for migra tion esti ma tes does not yet exist. In fact, Swed ish reg is ter data, long con sid ered the gold stan dard among demo graphic data sets, have been proved to over count migrants (Monti et al. 2019). Using tra di tional data to esti mate the migrant stock, such as data from censuses, admin is tra tive sources, and sur veys, pres ents lim i ta tions related to the def ni tion of migrant, the cov er age of the migrant pop u la tion, and the accu racy of the esti ma tes (Willekens 2019). Moreover, tra di tional sources of migra tion data are not timely. The United Nations suggested using the fol low ing def ni tion of an inter na tional migrant in order to har mo nize data sources on migra tion world wide (United Nations 1998): "per son who moves from their coun try of usual res i dence for a period of at least 12 months." An indi vid ual who lives abroad for a period of 3-12 months is con sid ered a short-term migrant. While Europewide data sources fol low the stan dard def ni tion of an inter na tional migrant (Euro pean Parliament and Council of the Euro pean Union 2007), indi vidual Euro pean countries use a vari ety of sys tems to track the num ber of inter na tional migrants liv ing within their bor ders. While censuses are con sid ered the best source of data for esti mat ing migrant num bers, these data have at least three lim i ta tions (Willekens 1994(Willekens, 2019)). First is that cen sus data are col lected every 10 years, and so they do not pro vide a timely pic ture of migra tion. Second, the cen sus records immigrants liv ing in the coun try, but does not account for the emi grants that have left the coun try. And third, the cen sus does not ask for impor tant data such as the indi vid ual's age at time of migra tion or return migra tion. Administrative data sources, such as pop u la tion reg is ters, can also be used to estimate migrants. Only a hand ful of countries use sur vey data to esti mate inter na tional migra tion. The advan tage of sur vey data col lected from migrants is that they might pro vide addi tional infor ma tion that is not included in the cen sus or admin is tra tive data sources. However, sur vey data might fail to ade quately cover the migrant pop u la tion. In the absence of reg is ters, the UK largely relies on a sur vey-based sys tem to collect infor ma tion on its migrant pop u la tion. The two main sources used to esti mate inter na tional migra tion to the UK are the International Passenger Survey (IPS) and the Labour Force Survey. The IPS has been run ning since 1961, and it was orig i nally intro duced to esti mate lev els of over seas travel and tour ism. It is cur rently the off cial source of data for esti mat ing inflows and out flows of inter na tional migrants. The ONS itself admit ted that the IPS "has been stretched beyond its orig i nal pur pose" (ONS 2019c) and can not be used as the only source when seek ing to esti mate interna tional migra tion in the UK. The sec ond main data source is the LFS, a Europe-wide quar terly house hold sur vey that aims to esti mate labor mar ket con di tions such as employ ment lev els. Through a boost of this sur vey pro vided by the Annual Population Survey (APS), the ONS col lects data on the stocks of for eign-born and for eign cit i zens liv ing in the UK at the local author ity level. The APS records infor ma tion on the length of time migrants have already spent in the UK. The LFS inter views 41,000 UK house holds per quar ter (ONS 2018a) and com bines these data with data from two quar terly waves of the LFS to cre ate a sam ple cov er ing 360,000 indi vid u als and 170,000 house holds per year. The data are released three months after the end of the sur vey. The lim i ta tions of the sam pling frame work, the sys tem atic bias, and the cov erage of both the IPS and LFS have been described in sev eral stud ies (Coleman 1983;Kupiszewska et al. 2010;Kupiszewska and Nowok 2008;Rendall et al. 2003). In addi tion, the ONS has recently started a work pro gram that aims to com bine data from addi tional admin is tra tive sources with data from the IPS and LFS in order to obtain a com pre hen sive mea sure of migra tion (ONS 2018b). --- Digital Traces and Their Limitations New social media data sources might be used to improve off cial migra tion sta tis tics, as these sources can pro vide infor ma tion on the back grounds and other demographic char ac ter is tics of migrants. Digital traces can be col lected quickly using the Application Programming Interface (API), which links research ers, as the cli ent, to the server, where the data we are inter ested in are stored in the form of a data base (Cooksey 2014;Sloan and Quan-Haase 2017). The abil ity to know in real time how many of the users are in a spe cifc loca tion can help us to nowcast migra tion. In addi tion, social media data can be geolocated. For exam ple, email loca tion data have been used to esti mate inter na tional migra tion rates (Zagheni and Weber 2012). These data are cheap because they are col lected by repurposing data sets Estimating Migrant Stocks Using Digital Traces and Survey Data orig i nally intended for adver tis ing. Thus, by rely ing on these data, we no lon ger need to cre ate new data infra struc tures to col lect data. Moreover, these new data sources can pro vide us with insights that will enable us to expand the def ni tion of an inter na tional migrant. Different countries use dif fer ent def ni tions of a migrant that vary depending on the length of time an indi vid ual must spend out side of their usual coun try of res i dence. Thus, the def ni tion of migrant is still not har mo nized world wide (Kupiszewska and Nowok 2008;Willekens 1994). Fiorio et al. (2021) have high lighted the pos si bil ity of using geotagged Twitter data to inves ti gate short-term mobil ity and long-term migra tion. They suggested that draw ing on dig i tal trace data could help to refne migra tion the ory and mod el ing. In addi tion, these data can be aug mented through data from ded i cated sur veys of pop u la tions that are too hard or too expen sive to reach with a tra di tional sam pling frame work (Pötzschke and Braun 2017;Rosenzweig et al. 2020). Nevertheless, these sources also have impor tant lim i ta tions. In some cases, research ers do not have direct access to all these new data sets and need to cre ate partner ships with pri vate com pa nies to obtain the desired level of access (Blumenstock 2012). Digital trace data from LinkedIn might pro vide insights into trends in highly skilled migra tion to the United States (State et al. 2014), while data from the Web of Science have been used to fol low trends and pat terns of inter na tional migra tion among schol ars (Aref et al. 2019). However, these sources do not pro vide data that are rep re sen ta tive of the entire pop u la tion. Hargittai (2018) ana lyzed the poten tial bias of dif fer ent plat forms in the United States, includ ing Facebook, LinkedIn, Twitter, Tumblr, and Reddit. She found that Facebook is the most rep re sen ta tive social media plat form across edu ca tional and internet skill lev els, while the other social media plat forms are used by smaller and more spe cifc U.S. pop u la tion groups. The work of Hargittai builds on the cri tique by Lazer et al. (2014) of the assump tion that we can sub sti tute tra di tional data sources with dig i tal trace data by show ing that using these new data sources with out con sid er ing their bias is prob lem atic. These authors have also pointed out the algo rithm dynam ics and the unsta ble char ac ter is tics of dig i tal traces, as the com pa nies that gen er ate the data we are seek ing to use are con stantly mod i fy ing their algo rithms and are in full con trol of the infor ma tion the research ers ulti mately receive. In this arti cle, we focus on Facebook Advertising Platform data. Facebook provi des adver tis ers with infor ma tion on its users, includ ing each user's age, sex, level of edu ca tion, and lan guage. For this rea son, Facebook has been described as a biased dig i tal cen sus (Cesare et al. 2018;Zagheni et al. 2017). Facebook's main busi ness is adver tis ing, and the data pro vided on the Facebook Advertising Platform are made avail able to adver tis ers to help them plan their online cam paigns. Facebook has a strong incen tive to accu rately report the char ac ter is tics of its users, because its ability to do so has become the main focus of its busi ness, as the com pany is aware that adver tis ers might change plat forms if they can not tar get the right audi ences through their ser vices. We are repurposing data from this adver tis ing plat form for demographic research. The var i able that is cur rently used to esti mate inter na tional migrants is defned by Facebook as "peo ple that used to live in coun try x and now live in coun try y." This var i able was frst used in Zagheni et al. (2017), where it was com pared to data from the Amer i can Community Survey. Until Decem ber 2018, the var i able was referred to as "expat from coun try x," show ing that the word ing of Facebook's def ni tion of migrant has changed over time. However, their doc u men ta tion does not pro vide infor ma tion on which indi vid ual char ac ter is tics have been used to cre ate the var i able, or whether the algo rithm iden ti fy ing a user as a migrant was changed along with the change in the word ing of the def ni tion in 2018. Two stud ies have inves ti gated how Facebook pro cesses this cat e gory. In the frst, research ers at Facebook suggested that Facebook users are con sid ered "expats" on the basis of the loca tion of their hometown and the struc ture of their friend ship net works (Herda<unk>delen and Marelli 2017). In the sec ond study, Spyratos et al. (2019) ran a sur vey in which 114 Facebook users were asked whether Facebook's Advertising Platform identifes them as an "expat" on the targeting plat form. The authors con cluded that Facebook uses other types of infor ma tion that are not spec i fed in the users' pro fles, includ ing geolocation out puts. The fnal clue can be found in Facebook's form 10K, which is a U.S. Securities and Exchange Commission (U.S. SEC) doc u ment that pro vi des a sum mary of Facebook, Inc.'s fnan cial per for mance on the stock mar ket. In these doc u ments, Facebook wrote that "the geo graphic loca tion of our users is esti mated based on a num ber of fac tors, such as user's IP address and self-disclosed loca tion" (U.S. SEC 2019, 2020). In the cur rent arti cle, we addi tion ally lever age the var i able "lan guage" from the Facebook Advertising Platform. Facebook reported that it is pos si ble to "tar get peo ple with language other than com mon lan guage for a loca tion." 1The Facebook mar ket ing API pro vi des two met rics: Daily Active Users (DAUs) and Monthly Active Users (MAUs). On Facebook for devel op ers,2 DAUs are defned as the "esti mated num ber of peo ple that have been active on your selected plat forms and sat isfy your targeting spec in the past day," while MAUs are defned as the "esti mated num ber of peo ple that have been active on your selected plat forms and sat isfy your targeting spec in the past month." The same U.S. SEC doc u ment (U.S. SEC 2019, 2020) reported esti ma tes of the bias of MAUs in 2018 and 2019, esti mat ing that 11% of accounts were dupli cated and 5% of accounts were false. Most of these anom a lies were detected in Southeast Asia. We are using the MAUs esti ma tes, because the Facebook doc u ment makes clear that this mea sure is more sta ble than the DAUs met ric. The MAUs met ric does not report num bers under 1,000 to pre vent the targeting of small groups of indi vid u als. Through the Facebook mar ket ing API, we included in the cur rent study all Facebook users in an aggre gated and anonymized for mat. --- Comparison Between LFS Data and Facebook Data The two main data sources we used are the LFS and the Facebook Advertising Platform. We included 20 of the EU27 countries in our study: Austria, Belgium, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, the Netherlands, Poland, Portugal, Romania, Slovakia, Spain, and Sweden. Malta and Luxembourg were excluded because of their small size, Bulgaria and Croatia were excluded because Facebook does not pro vide esti ma tes of expat num bers for them, and Estonia and Slovenia were excluded for miss ing val ues in the data used as covariates. Estimating Migrant Stocks Using Digital Traces and Survey Data Moreover, Cyprus was excluded because the Facebook "expat" esti ma tes might include all the users liv ing there (Gendronneau et al. 2019). The aggre gated esti ma tes of Euro pean migrants from the Facebook Advertising Platform were col lected in the third week of July 2018 and July 2019. We used pySocialWatcher, which is a Python package, to down load the data (DAUs and MAUs) from the Facebook API (Araujo et al. 2017). The data from the LFS were pro vided by the ONS for the period of June-July 2018 and June-July 2019. For the pur pose of this anal y sis, we have assumed that the age struc ture of the LFS and Facebook migrant users did not change much between 2018 and 2019. Figure 1 shows a com par i son between these two data sources for the two years included in the anal y sis. Three var i ables from three data sources are shown: the migrant var i able and lan guage var i able from Facebook, and esti ma tes of migrant stocks by coun try of birth from the LFS. We can see a cor re la tion between the Facebook migrant and lan guage var i ables for many countries. The cor re la tion between these var i ables is 0.92 for both years, while the cor re la tion between the Facebook migrant var i able and the LFS esti ma tes is 0.91 in 2018 and 0.88 in 2019. However, there are excep tions: • for countries with a lan guage that is also spo ken in other countries (e.g., Ger man in Germany, Austria, Switzerland, and Belgium; or French in France, Switzerland, and Belgium); and • for Greece, where we notice that the "expat" var i able on Facebook does not cap ture the Greek migrants. (The Greek lan guage is spo ken in Greece and part of Cyprus.) Facebook variable "expat" Facebook variable language Labour Force Survey Fig. 1 Facebook's aggregated estimates (in 1,000s) for the "expat" and language variables and Labour Force Survey data of migrant stocks from 20 European countries of origin in 2018 and 2019 Figure 1 shows a vis i ble drop in the Facebook migrant var i able esti ma tes between 2018 and 2019. This is not due to out-migra tion from the UK, but rather due to an algo rithm change that affected the Facebook esti ma tes. In Figure B2 in the online appen dix, we high light the shift that hap pened in the mid dle of March 2019, which led to an aver age change in the esti ma tes of 48%. --- Additional Data Sources In this anal y sis, addi tional sources are used as covariates that can help us esti mate migrant stocks. We used data on inflows and out flows of migrants from and to the UK from the IPS for 2017 and 2018. We used infor ma tion on the pop u la tions of the countries of ori gin from the pro jec tions pro duced by Eurostat, together with the Eurostat esti ma tes of unem ploy ment and gross domes tic prod uct (GDP) per cap ita. The pop u la tion data are used for the anal y sis for the years 2018 and 2019, while the other two data sets are used for the anal y sis for the years 2017 and 2018. Data from the UK set tled and presettled sta tus scheme are added to make an addi tional com par i son. These data come from the UK Home Offce. This scheme allows Euro pean migrants already resid ing in the UK to apply for presettled sta tus if they have been liv ing in the UK for less than fve years, and for set tled sta tus if they have been liv ing there for fve years or more. The mea sure of appli ca tions to the scheme pro vi des an indi ca tion of the num ber of Euro pe ans who want to con tinue to have the right to remain in the UK after Brexit has been fnal ized. The data rep re sent an esti mate for the total num ber of appli ca tions and includes the period from August 28, 2018, to Decem ber 31, 2019. --- Methodology --- General Model Architecture The aim of the IMEM frame work is to esti mate the true or latent flow of inter na tional migrants across send ing and receiv ing countries by com bin ing biased data (Raymer et al. 2013). The orig i nal IMEM model com bines flows from send ing and receiv ing countries across the EU. Our aim is to pro vide an esti mate of the true stock of Euro pean migrants in the UK based on a com bi na tion of the LFS and Facebook Advertising Platform data. The esti mate of true stock is the num ber of migrants who would be counted if our col lec tion sys tem were able to per fectly mea sure all migrants (Disney 2015). While the true num ber of migrants is not known, through the use of Bayes ian meth ods we might esti mate a prob a bil ity dis tri bu tion for the true num ber of migrants that reflects our knowl edge about it. These true or latent esti ma tes from the model incor po rate all the infor ma tion col lected from the var i ous data sources, as well as our prior data about the migra tion pro cess. Thus, the point esti mate of the true num ber of migrants would be a sum mary of this dis tri bu tion (i.e., the median). The model is divided into two parts: the Measurement Error Model (MEM) and the Theory-Based Model (TBM). In the MEM, the Facebook Advertising Platform and LFS data are com bined; in the TBM, other var i ables are also con sid ered in the esti ma tion Estimating Migrant Stocks Using Digital Traces and Survey Data of the true stock. In this frame work, the IMEM quantifes the lim i ta tions of the data sources and pro vi des the appro pri ate prior dis tri bu tion in order to reduce the bias. The lim i ta tions of the data are assessed in terms of the fol low ing (Disney 2015;Raymer et al. 2013): • Definition: How closely does the inter na tional migrant mea sure match the United Nation's def ni tion of an inter na tional migrant? • Coverage: What pro por tion of the total immi gra tion stock does the data cover? • Bias: Is there any sys tem atic bias in the data? In Figure 2, the model is explained using a dia gram that is divided into four parts: input, data assess ment, model, and out put. In the input col umn, the data sources are sur vey data, dig i tal trace data, and migra tion the ory covariates for the TBM. The data assess ment is followed by a sum mary of the lim i ta tions of the data in terms of def ni tion, bias, and cov er age. In the model box, the true stock at the cen ter of the fg ure is esti mated by the TBM and the MEM, which com bine the stock esti ma tes from the LFS with those from the Facebook Advertising Platform, while incor po rat ing con sider ations related to def ni tion, bias, and accu racy. Finally, in the out put, the diag nos tics and results are shown. The model is constructed as fo

The online ver sion of this arti cle (https: / /doi .org /10 .1215 /00703370 -9578562) con tains sup ple men tary mate rial.

: Is there any sys tem atic bias in the data? In Figure 2, the model is explained using a dia gram that is divided into four parts: input, data assess ment, model, and out put. In the input col umn, the data sources are sur vey data, dig i tal trace data, and migra tion the ory covariates for the TBM. The data assess ment is followed by a sum mary of the lim i ta tions of the data in terms of def ni tion, bias, and cov er age. In the model box, the true stock at the cen ter of the fg ure is esti mated by the TBM and the MEM, which com bine the stock esti ma tes from the LFS with those from the Facebook Advertising Platform, while incor po rat ing con sider ations related to def ni tion, bias, and accu racy. Finally, in the out put, the diag nos tics and results are shown. The model is constructed as fol lows. The num ber of Euro pean migrants (stocks), z ijt k, from a cer tain coun try, i, in the UK with a cer tain char ac ter is tic, j, is observed. In this case the char ac ter is tic selected is sex. This is done using data from Facebook, F, and from the LFS, L, and the value k is then used to rep re sent either L or F depending on which data are used to mea sure the Euro pean migrants stock (z ijt k ). The year, t, in this case is 2018 and 2019. The data sets used can thus be described in the form of matri ces Z F (Eq. ( 1)) for Facebook and Z L (Eq. ( 2)) for the LFS. The model bor rows strength across the two years: Z F = z 11t F z 12t F... z 1Jt F z 21t F z 22t F... z 2 Jt F!! "! z I1t F z I 2t F... z I Jt F <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk>,(1) Z L = z 11t L z 12t L... z 1Jt L z 21t L z 22t L... z 2 Jt L!! "! z I1t L z I 2t L... z I Jt L <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk>. (2 ) For every time t, the value of Y ijt (Eq. ( 3)) is the ran dom var i able esti mate of the true stock. It is a matrix with dimen sion I <unk> J: The value of z ijt k is assumed to fol low a Poisson dis tri bu tion (Eq. ( 4)). The Poisson dis tri bu tion is a prob a bil ity dis tri bu tion of the num ber of times an event is expected to occur. Here, the dis tri bu tion of Euro pean migrants is based on expec ta tions from the Facebook and LFS data. The dis tri bu tion is Y = y 11t y 12t... y 1Jt y 21t y 22t... y 2 Jt!! "! y I1t y I 2t... y I Jt <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk> <unk>. (3 z ijt k <unk> Po(<unk> ijt k ). (4 ) Figure 3 illus trates the hier ar chi cal struc ture of the model, which is explained in detail in the fol low ing sec tion. The model is esti mated using JAGS in R (Plummer et al. 2016). In JAGS, the nor mal dis tri bu tions are defned in terms of the mean, <unk>, and pre ci sion (i.e., one over the var i ance), <unk>. The JAGS nota tion is used. --- Fig. 2 (continued) --- Measurement Error Models The Measurement Error Models describe how the observed val ues relate to the true count. The gen eral equa tion (5) of the Measurement Error Model is log <unk> ijt k = log y ijt + <unk> k + <unk> k + <unk> ijt k + <unk> ijt k + <unk> ijt k + <unk> ijt k. (5 ) The equa tion is com posed of fve terms, <unk> k, <unk> k, <unk> ijt k, <unk> ijt k, and <unk> ijt k, which are used to con vert the data from Facebook and the LFS to com ply with the United Nation's def ni tion of an inter na tional migrant, and to reduce the under es ti ma tion linked to the bias or cov er age of the data. The frst param e ter, <unk> k, cap tures the dif fer ences in rela tion to the def ni tion of migrants. The bias in the data is cap tured by <unk> k, while the cov er age of the Facebook data is con sid ered in <unk> ijt k. The param e ter <unk> ijt k deflates the Facebook esti ma tes of 2018 by the algo rithm change that hap pened in 2019. The param e ter <unk> ijt k inflates the Facebook esti ma tes with knowl edge pro vided by the Facebook esti ma tes of peo ple speak ing a cer tain lan guage. The term <unk> ijt k is the error term with nor mal distri bu tion N(0, <unk> ijt ), and the pre ci sion <unk> ijt has Gamma dis tri bu tion G(100, 1) (where 100 is the shape param e ter and 1 is the rate param e ter), which has a mean equal to 100 and pre ci sion equal to 1 (e.g., var i ance equal to 100). Table 1 sum ma rizes the param e triza tion of the model and the direc tion of the prior dis tri bu tions. --- Data Assessment of the Labour Force Survey The LFS def nes a longterm inter na tional migrant in the same way as the United Nations (ONS 2018a) and pro vi des data on each migrant's coun try of birth and citi zen ship. For our pur poses, the coun try of birth cri te rion is used because it cap tures indi vid u als with a migrant back ground, includ ing those who acquired cit i zen ship through nat u ral i za tion. Because the LFS is used to esti mate the stock of migrants in the UK, many research ers have inves ti gated the qual ity of the sur vey's esti mates and found that they under es ti mate migrants. Rendall et al. (2003), for exam ple, reported that the 2001 LFS underreported inter na tional migrants by 26% com pared with the 2001 cen sus. Other research has shown that the bias in the LFS might be as high as 30% for nation al i ties with smaller stocks, such as Greeks and Lith u a nians (Kupiszewska et al. 2010), and that the sur vey has a non re sponse rate of more than 15% (Mart<unk> and Ródenas 2007). Furthermore, the sam pling frame work of the LFS does not cover the entire tar get pop u la tion (Kupiszewska et al. 2010) as stu dents and more mobile migrants might not fully appear in the sam ple. Table 2 com pares data from the LFS col lected between Jan u ary and Decem ber 2011 with the Brit ish cen sus that occurred on March 27, 2011. The data are aggregated for England and Wales only. The rel a tive per cent age change between the LFS and the cen sus gives a sense of the bias between the two. It has to be stressed that the ONS has already attempted to recalibrate the LFS esti ma tes with the results of the cen sus. Despite this, there is still a prob lem with both undercounting and overcounting. The range of the bias is between -21% and 15%. This sug gests that the LFS Measurement Error Equation is log <unk> ijt L = log y ijt + <unk> g (i) L + <unk> ijt L. (6 ) Regarding this assess ment, the LFS data are deflated by only one param e ter, <unk> L, which con sid ers both the bias and the cov er age of the data. A sep a rate param e ter, such as <unk> L, is redun dant as the def ni tion of inter na tional migrant in the LFS fol lows the United Nations stan dard. The lit er a ture (Kupiszewska et al. 2010;Mart<unk> and Ródenas 2007;Rendall et al. 2003) sug gests that for countries with small migrant pop u la tions in the UK, LFS migrant esti ma tes may be around 30% lower than the true num bers. This per cent age is reduced, to around 15%, for those nation al i ties with large pop u la tions in the UK. Table 2 pro vi des a mea sure of the bias at a coun try level. The ONS reports that the qual ity of the LFS esti ma tes decreases over time when dis tanced from the cen sus year (ONS 2020). The clas si f ca tion relies on the lit er a ture, the data from Table 2, and the assess ment from the ONS, as well as our own exper tise. The LFS bias is anchored to the rel a tive per cent age change between the LFS and the cen sus, and an increase of bias over time is also con sid ered. As a mat ter of fact, the countries are divided into three groups: • Low-Bias at 4%: Austria, Belgium, Czech Republic, Latvia, Sweden; • Medium-Bias at 12%: France, Germany, Greece, Hungary, Lithuania; • High-Bias at 30%: Denmark, Finland, Ireland, Italy, Netherlands, Poland, Portugal, Romania, Slovakia, Spain. As a con se quence, the <unk> L param e ter is assigned according to a param e ter g(i), where g(i) = 1, if the undercount is assumed to be low; 2, if the undercount is assumed to be medium; 3, if the undercount is assumed to be high. <unk> <unk> <unk> <unk> <unk> (7) The prior dis tri bu tion is set to <unk> i L <unk> N -0.04, 100( ), if the undercount is assumed to be low; N -0.13, 100 ( ), if the undercount is assumed to be medium; N -0.35, 100( ), if the undercount is assumed to be high. The means on the prior <unk> L are assumed to be time-invari ant: they are con sid ered as an approx i ma tion of the bias and thus small time var i ances are not accounted for. The term <unk> ijt k is the error term with nor mal dis tri bu tion N(0, <unk> ijt ), and the pre ci sion <unk> ijt has Gamma dis tri bu tion G(100, 1), as pre vi ously described. <unk> <unk> <unk> <unk> <unk> <unk> <unk> (8) --- Data Assessment of the Facebook Advertising Platform Given our ear lier descrip tion of the Facebook data, a param e ter was cre ated for the data's def ni tion, bias, and cov er age. The Facebook <unk> F is a pri ori assumed to be normally dis trib uted with N(0, 100), while <unk> F has a nor mal dis tri bu tion N(0.04, 100). The mean of <unk> F is set at 4% to deflate the Facebook esti ma tes to account for fake and dupli cate accounts. This value is lower than the 11% suggested by Facebook itself, because we assume that the per cent age of fake and dupli cated accounts labeled as belong ing to migrants is lower in Europe. The mean of the cov er age param e ter <unk> ijt F (Eq. ( 9)) is the rate of non-Facebook users in the coun try of ori gin of the Euro pean migrants, since the aim is to cor rect by this adjust ment. It is com puted as <unk> ijt F = log 1- Number of Facebook users ijt Eurostat population size ijt <unk> <unk> <unk> <unk> <unk> <unk>.(9) Additionally, the dig i tal trace data are described as unsta ble. Indeed, it seems that Facebook reviewed its algo rithm on expats in the mid dle of March 2019, and there was a drop in the migrant esti ma tes after this time. The change is coun try-and sexspe cifc. For this rea son, a param e ter was intro duced for the rate algo rithm <unk> ijt F (Eq. ( 10)), which aims to adjust the Facebook data for this bias caused by the change in the algo rithm: <unk> ijt F = log Estimates before ij -Estimates after ij Estimates before ij <unk> <unk> <unk> <unk> <unk> <unk>.(10) A param e ter was used for Greece that inflates the esti ma tes of the Facebook expat var i able (Eq. ( 11)), which reports a low num ber of "peo ple that used to live in Greece and now live in the UK." However, the lan guage var i able, which Facebook uses to "tar get peo ple with lan guage other than com mon lan guage for a loca tion," pro vi des some infor ma tion that can be used to adjust the num ber of Greeks liv ing in the UK. As the Greek lan guage is also spo ken by Cypriot migrants, the esti ma tes are deflated by a ratio cal cu lated using LFS data of the num ber of Greek and Cypriot migrants. Unfortunately, this is another sign that dig i tal trace data are not per fect, as it seems that Facebook is not account ing for Greek migrants with the migrant var i able (see also Figure A1 in the online appen dix). <unk> ijt F = log FB Language ijt FB Migrant ijt <unk> LFS Greece migrant ijt LFS Greece migrant ijt + LFS Cyprus migrant ijt <unk> <unk> <unk> <unk> <unk> <unk>. (11 ) After this assess ment, the Facebook Measurement Error Equation is log <unk> ijt F = log y ijt + <unk> F + <unk> F + <unk> ijt F + <unk> ijt F + <unk> ijt F + <unk> ijt F. (12 ) --- Theory-Based Model In this part of the model, covariates that might help to explain the true stock of European migrants in the UK are intro duced: log y ijt = <unk> 0 + <unk> 1 P ijt + <unk> 2 I ijt + <unk> 3 O ijt + <unk> 4 log G ijt + <unk> 5 log U ijt + <unk> ijt,(13) where <unk> = (<unk> 0,..., <unk> 5 ) is a vec tor of param e ters; <unk> 0 is assumed to be normally distrib uted <unk> 0 <unk> N(0, 0.01), pro vid ing a weakly infor ma tive prior on the con stant term, while <unk> (1,..., 5) <unk> N(0, 1) is assumed to be more infor ma tive. The error term <unk> ijt has a nor mal dis tri bu tion N(0, <unk> ijt ), with pre ci sion <unk> ijt fol low ing a Gamma dis tri bu tion Gamma(100, 1). The covariates used in the mod els for 2018 and 2019 include: P: a nor mal ized mea sure of pop u la tion size in the coun try of ori gin, divided by the mean of the pop u la tion in the same countries con sid ered in the model (data are from the lat est esti ma tes by Eurostat in 2018 and in 2019); I: a nor mal ized mea sure of the inflows from Euro pean countries to the UK, divided by the mean of the inflows of migrants from the countries con sid ered in the model (data are from the IPS in 2017 and in 2018); O: a nor mal ized mea sure of the out flows to the Euro pean countries from the UK, divided by the mean of the out flows of emi grants from the countries con sid ered in the model (data are from the IPS in 2017 and in 2018); G: ratio of GDP growth rate in the Euro pean coun try of ori gin in 2017 and in 2018, divided by the GDP growth rate in the UK (data are from Eurostat); and U: ratio of the unem ploy ment rate in the Euro pean coun try of ori gin in 2017 and in 2018, divided by the unem ploy ment rate in the UK (data are from Eurostat). The nor mal ized mea sure of the pop u la tion size is a pre dic tor of the pos si ble number of migrants informed by a grav ity model; that is, the larger the pop u la tion, the larger the num ber of pos si ble migrants. The nor mal ized mea sures of inflows and out flows from the IPS pro vide an indi ca tion of the lev els of fluc tu a tion in terms of arriv als and depar tures for every nation al ity, and thus help to cap ture fluc tu a tions in the stocks. The ratio of the GDP growth rate to the unem ploy ment rate pro vi des infor ma tion on how the econ omy of the coun try of ori gin com pares to that of the UK, and there fore is a form of eco nomic grav ity indi ca tor. Study code and data are accessible at https://github.com/chiccorampazzo/fb-migration-uk. --- Results We pres ent two sets of mod els. The frst is for the total num ber of Euro pean migrants in the UK, and the sec ond dis ag gre gates the esti ma tes by sex. They are run simul ta neously by year (2018 and 2019) to bor row strength across the years. In the frst model, the aim is to explain the mag ni tude of the under count of the LFS data rel a tive to the estima tes pro duced by the model for the two years. All esti ma tes of the mod els con verge. Detailed results and diag nos tic sta tis tics are in the online appen dix. Estimating Migrant Stocks Using Digital Traces and Survey Data --- Model for Total Num bers Figure 4 shows data from three data sets and our esti ma tes: the Facebook adver tis ing data are in blue, the LFS data are in yel low, the set tled sta tus appli ca tion data are in red, and the model esti ma tes are in green. The set tled sta tus data are presented for com par i son only. LFS data are shown with a 95% con f dence inter val, while model esti ma tes are shown with the interquartile (IQR). There are three main mes sages that can be discerned from this fg ure. First, the dif fer ences between the Facebook data in 2018 and 2019 are read ily vis i ble, and are related to the algo rithm change car ried out by Facebook. However, the prior dis tri bution on the algo rithm param e ter seems to fx this bias, as the dif fer ences between the 2018 and the 2019 esti ma tes were rel a tively small. Second, while the LFS data are rel atively con sis tent across the two years, a decreas ing trend in the num ber of EU migrants in the UK is vis i ble. Third, the model esti ma tes are higher than the LFS esti ma tes. In some cases, the IQR range of the model esti ma tes includes the LFS esti ma tes. The param e ter on Greece seems to be effec tive in bring ing the esti ma tes closer to the LFS val ues. In the online appen dix, the pos te rior char ac ter is tics of the true stock esti ma tes for all of the mod els and the R are reported, the lat ter being a mea sure that helps deter mine whether chains have con verged depending on whether it is close to one (Gelman et al. 2013). All of the chains have con verged when R is strictly equal to one (except for Romania in 2018 and Poland in 2019, where R is 1.01 as shown in the online appen dix). The algo rithm for esti mat ing all of the other param e ters has con verged as well. Table 3 gives a com par i son of the undercounted LFS esti ma tes with the model esti ma tes. While the ONS has esti mated an under count of 16%, the model esti ma tes an under count of 25% for 2018 and 20% for 2019. The under count for 2018 has larger inter vals, likely owing to the prior on the algo rithm change. Additionally, the model for 2019 esti ma tes a higher num ber of migrants of cer tain nation al i ties (e.g., Pol ish, Ital ian, and Hun gar ian) and a lower num ber of migrants of other nation al i ties (e.g., Roma nian, Ger man, and Czech). The interquartile range of these dis tri bu tions is large, high light ing the uncer tainty in the esti ma tes. However, the mod els for the two years indi cate that the under count and the uncer tainty are in the same direc tion. --- Model Disaggregated by Sex In this part of the model, the esti ma tes are disaggregated by sex, because it is important to study the age and sex dif fer ences of migrants, as there might be large dif ferences across sexes. The model pro posed works for sex dis ag gre ga tion, and Figure 5 shows the esti ma tes. In this case, the com par i son with migrants who have applied for the set tled sta tus scheme is not avail able because the data from the Home Offce are not disaggregated by sex. --- Sensitivity Analysis Some sen si tiv ity checks of the model are pro vided. First, the model was run while including only the LFS data. For the model spec i fed in this arti cle, the under count is esti mated at 25% in 2018 and at 20% in 2019. In Table 4, the under count of this new spec i f ca tion of the model is reported, esti mated at a median level of 8% in 2018 and 22% in 2019. These two median lev els are not close to those pro duced by the model that com bines Facebook and LFS data, with a smaller under count in 2018 and a larger one in 2019. Overall, the uncer tainty of the under count esti mate is greater when using only LFS data. The sec ond sen si tiv ity check was to mod ify the param e ters from the Facebook and LFS Measurement Error Models. In the mod els used here, the param e ters are informed by pre vi ous research and cal cu la tions on the data, except the <unk> F, which is the bias param e ter for Facebook. It is assumed the value is lower than the per cent age of fake and dupli cate accounts world wide. In the sen si tiv ity anal y sis, the Facebook bias param e ter was frst mod i fed to 0%, indi cat ing no bias in the Facebook esti ma tes, and then to 11%. In this par a graph, we ana lyze the dif fer ent spec i f ca tion of the mod els looking at Table 4. The under count with no bias attrib uted to the Facebook esti ma tes is 22% for 2018 and 19% for 2019, which is slightly lower than that spec i fed in the suggested model. The under count with a higher <unk> F is 25% for 2018 and 20% for 2019. The under counts with a <unk> F at 4% and at 11% are very sim i lar. The model is sen si tive to the choice of the assumed bias of the LFS param e ter. We mod i fed the bias of the LFS to 4% (the min i mum level assumed) and to 30% (the max i mum level assumed) for all the countries. With the low min i mum bias level assumed, the under count reaches neg a tive median val ues, while it is larger when the max i mum bias level is assumed. We also tried dif fer ent spec i f ca tions of the pre ci sion dis tri bu tion term, which is assumed to fol low a Gamma(100, 1) in the presented model. The model was spec i fed with a Gamma(1, 1), which is less infor ma tive than Gamma(100, 1). The gra di ent of the median of the under count is sim i lar to the one in the presented model, though the uncer tainty is larger. There is some impact of the prior selec tion on the uncer tainty of the esti ma tes. Finally, in Figure 6, the esti ma tes from the model on the total esti ma tes (model 1) are com pared to the sum of the esti ma tes from the sex dis ag gre ga tion model. Though the esti ma tes are close to each other, there are cases in which the sum from the sex dis ag grega tion model is not com pletely aligned with the dis tri bu tion from model 1. This is due to inconsistencies in the Facebook and LFS data disaggregated by sex. The esti ma tes from our mod els seem to be sta ble to dif fer ent prior dis tri bu tions, yet the pre ci sion of those prior dis tri bu tions had to be care fully cho sen to ensure model con ver gence, while exploring rea son able areas of the param e ter space with respect to the pre ci sion param e ters. --- Discussion The model esti mated the migrant stocks for 2018 and 2019. In the 2018 model, a prior dis tri bu tion was used to account for an algo rithm change that Facebook implemented in March 2019, which led to a decrease in the esti mate of Euro pean migrant num bers. This algo rithm change was not uni form, how ever, as it var ied by coun try and sex of the migrants. This fnd ing high lights the impor tance of mon i tor ing dig i tal traces and cau tions that using dig i tal traces alone is not suf f cient to gen er ate bet ter esti ma tes of stocks of migrants. The param e ters asso ci ated with the algo rithm change and the Greek fac tor (i.e., the fac tor that Greeks are under rep re sented in the Facebook migrant var i able) were shown to be effec tive in bring ing the model esti ma tes in line with the LFS esti ma tes. Our inclu sion of the Home Offce's data related to set tle ment and pre set tle ment appli ca tions as an addi tional com par i son proved inter est ing. For Pol ish migrants, the num ber of appli cants to these schemes was lower than the LFS esti mate, while for Roma nian migrants, this num ber was the same as the LFS esti mate. The num ber of appli cants is expected to be lower than the LFS esti mate of migrants because apply ing for the scheme before the end of the tran si tion period is not man da tory. It was observed, how ever, that in some cases the set tled sta tus appli ca tion num ber was higher than the LFS esti mate but closer to the model esti ma tes, suggesting that the model might have been pro duc ing a more accu rate esti mate than the LFS. For Ital ian migrants, for exam ple, the num ber of set tled sta tus appli ca tions was close to the median esti mate from the pro posed model. Conversely, the model esti ma tes for Por tu guese migrants were closer to the LFS esti ma tes and lower than the esti ma tes of appli cants for set tled or presettled sta tus. Interestingly, the results for the model esti ma tes for Germany were also lower than the LFS esti ma tes, but were closer to the esti ma tes of those who fled a set tle ment or pre set tle ment appli ca tion. Almost no Irish nation als applied to the set tled or presettled scheme owing to the bilat eral agree ments between the Republic of Ireland and the UK. An esti mate of the total num ber of Euro pean migrants by sex is also pro vided. The sum of the esti ma tes from this sec ond model was equal to the total from the frst. There was uncer tainty in our esti ma tes, espe cially for the countries of ori gin with the highest num ber of migrants in the UK: Poland and Romania. This might sug gest that for nation al i ties where the level of uncer tainty is higher, the sam ple of house holds and migrants interviewed should be increased. A pos si ble solu tion to reduce the uncer tainty would be to include a prior dis tri bu tion in the model driven by expert opin ion, as well as more infor ma tive pri ors on the Facebook and LFS data once they become avail able. Moreover, the anal y sis showed one of the main lim i ta tions of dig i tal trace data: the lack of trans par ency on how pri vate dig i tal com pa nies pro duce their esti ma tes. Indeed, it is not clear how exactly Facebook labels users as "peo ple that used to live in coun try x and now live in coun try y," or how they deter mine which lan guages are spoken by the users on their plat form. Furthermore, there are no details avail able about the algo rithm change Facebook implemented in March 2019. --- Conclusions Our over arch ing research ques tion was: What can Facebook adver tis ing data contrib ute to ONS migra tion esti ma tes in a

The online ver sion of this arti cle (https: / /doi .org /10 .1215 /00703370 -9578562) con tains sup ple men tary mate rial.

the uncer tainty would be to include a prior dis tri bu tion in the model driven by expert opin ion, as well as more infor ma tive pri ors on the Facebook and LFS data once they become avail able. Moreover, the anal y sis showed one of the main lim i ta tions of dig i tal trace data: the lack of trans par ency on how pri vate dig i tal com pa nies pro duce their esti ma tes. Indeed, it is not clear how exactly Facebook labels users as "peo ple that used to live in coun try x and now live in coun try y," or how they deter mine which lan guages are spoken by the users on their plat form. Furthermore, there are no details avail able about the algo rithm change Facebook implemented in March 2019. --- Conclusions Our over arch ing research ques tion was: What can Facebook adver tis ing data contrib ute to ONS migra tion esti ma tes in a con text in which there are no "ground truth" data against which model esti ma tes can be val i dated? This ques tion was answered by explor ing the two data sources and pro duc ing a prob a bi lis tic mea sure of Euro pean migra tion. Although this study found greater uncer tainty in the esti ma tes that were already known to be biased, it con trib utes to the "learn ing pro cess" hoped for by Willekens (1994Willekens (, 2019)), which can lead to the exten sion of this frame work. The obvious next step would be to expand the model to dis ag gre gate the esti ma tes by age and sex. This anal y sis made three con tri bu tions to dig i tal and com pu ta tional demog ra phy. First, it pro posed to apply a frame work that is already in use in migra tion research to dig i tal traces. The pro posed model is a flex i ble frame work, in which it is pos si ble to include new infor ma tion as soon as it becomes avail able, includ ing additional dig i tal trace data from other adver tis ing plat forms such as Instagram, Snapchat, and LinkedIn, as well as from other admin is tra tive sources. Second, it addressed the biases of both tra di tional and dig i tal trace data. The use of a prior dis tri bu tion was shown to fx these issues in a prob a bi lis tic fash ion. Third, it pro duced an esti mate of the under count of migra tion lev els. Overall, the model esti mated an under count of 25% for 2018 and 20% for 2019 based on the LFS data. For migrants to the UK from the EU8 countries, the ONS had esti mated an under count of 16% for March 2016. It would be pos si ble to com pute this mea sure using data from both the LFS and Facebook at the time of the next cen sus (which in the UK is sched uled for 2021). In this way, the model could be used to help nowcast migra tion in a timely man ner, thus com par ing the esti ma tes to those of the cen sus. Facebook's cov er age of the gen eral pop u la tion varies by age and sex (self-reported by Facebook's users). A Pew Research Center report (Pew Research Center 2018) showed that while Facebook is used across all age-groups, the num bers of youn ger users have been declin ing. Facebook has, how ever, noted that some youn ger users reg is ter with an inac cu rate age (U.S. SEC 2019SEC, 2020)). In addi tion to the age compo si tion of Facebook users, we should con sider the cov er age dif fer ences between men and women. Fatehkia et al. (2018) and Garcia et al. (2018) explored pat terns in the use of Facebook to describe the dig i tal gen der gap that exists even in devel oped countries. While the gap is grow ing smaller, there are still more men than women on Estimating Migrant Stocks Using Digital Traces and Survey Data Facebook (Fatehkia et al. 2018). Including an age and sex dis ag gre ga tion is a fur ther step that we leave for future research. Traditionally, demo graphic meth ods have relied on approaches like the basic demographic bal anc ing equa tion, in which the terms have to add up. That may not be neces sary, how ever, when the under ly ing data have dif fer ent types of biases. At the same time, more and more data sources that con tain impor tant sig nals of change (as well as biases) are becom ing avail able. This study con trib utes to the demo graphic lit er a ture by pro pos ing an approach to study ing migra tion that is able to com bine and make sense of new and dif fer ent data sources in a way that builds on clas sic demo graphic approaches, while repurposing them within a Bayes ian sta tis ti cal frame work. <unk>

The online ver sion of this arti cle (https: / /doi .org /10 .1215 /00703370 -9578562) con tains sup ple men tary mate rial.

Introduction Bullying and fighting at school are very common concerns of peer violence among adolescents in many international studies [1][2][3]. Bullying is defined as a "specific type of aggression in which the behavior is intended to harm or disturb, the behavior occurs repeatedly over time, and there is an imbalance of power, with a more powerful person or group attacking a less powerful one" [4]. The World Health Organization states that youth violence is as a result of an interaction between individual and social factors, especially relationships with family, peers, and teachers [5]. The 2006 UNESCO study on violence against children reported that 20-65% of schoolchildren are affected by verbal bullying and it is the most prevalent form of violence at school [6]. The existence of bullying at school has a negative impact on the school climate and violates the rights of children to learn in a safe environment without fear. Among adolescents, the most common risk factors for bullying perpetration are younger age, male gender, substance use and other health-threatening behaviours [7]. Studies show multiple predictors of bullying (not necessarily considered as causes), and individual characteristics and social contexts are crucial [8][9][10][11][12]. Family and home environment, school climate, and community factors significantly predict involvement for bullies, indicating the important role social context plays in the development and maintenance of bullying [13]. Specifically, the perceived emotional support from parents, peers, and teachers is associated with teenagers' aggressive behaviours [7,14,15] and the primary source of social support for adolescents varies as a function of age-in early adolescence, parental support is important but in late adolescence, peer support becomes more salient [16,17]. Studies analyzing parents as a social influence show that greater parental warmth and support associates with less involvement in all forms of adolescent bullying [18,19]. Bullying perpetration and fighting are largely studied as a school-based phenomenon, so in this context the school climate and relationships with classmates and with teachers play significant roles in a child's social support network, providing social support in its multiple forms [20]. Bullying is a "group process" at the peer-group level, and adolescents belonging to bullying groups increase their aggressive behaviours overtime through reciprocal rewarding and reinforcing of each other's behaviours. Therefore, bullies who are dominant very often perceive themselves as popular among classmates and other peers [21]. At schools where bullying is a problem, children do not feel accepted, supported, respected, or fairly treated [2,22]. However, peer support is not the only relevant factor of support. Studies show that active teacher support is related to lower engagement in pupils' health risk and antisocial behaviours such as bullying and fighting [23,24]. Teachers and school staff are in a unique and influential position to promote healthy relationships and to intervene in violent situations [25]. Even though it has been showed that the main predictors of bullying are delinquent behaviours as a risk factor and good social climate at school as a protective factor, the strong social bonds in the community are also supportive and can reduce the level of bullying at schools [26]. Students feeling more connected to their schools demonstrate reductions in violent behaviour over time [10]. Therefore, it can be seen that social factors are relevant in the onset and prevention of aggressive behaviours, though other individual behavioural factors (such as alcohol consumption or smoking) can also be among indicators associated with bullying or fighting. There is a need to study the role of perceived social support, besides the influence of individual factors, in planning and implementing youth violence prevention programs, also at the country level [3,7,13]. Therefore, our study was aimed to investigate how different factors play together in the occurrence of aggressive behaviours at school. For this, we analysed individual and social factors as well as main health risk behaviours in nine selected countries. --- Methods --- Data Collection Data for this study were collected during the 2013/2014 Health Behaviour in School-aged Children (HBSC) survey, which is the collaborative project of World Health Organization Regional Office for Europe. The analysed data focus on social context (relationships with family, peers, and school teachers) as well as risk behaviours such as smoking, drunkenness, fighting and bullying in adolescents. Data were collected by school-based surveys using a standard international study protocol [27]. The samples at national levels were representative for the age groups of 11-, 13-and 15-year-olds attending school in each country and region. Cluster probability sampling (systematic or random) of school classes was carried out in each country and region. Sampling of schools was carried out where lists of classes were not available, followed by sampling of classes within schools. Each country data records were revised by the HBSC Data Management Centre at the Department of Health Promotion and Development, University of Bergen, Norway. Appropriate ethical approval was granted in all countries and regions (approval code). Standardised information about the study was provided to parents and school children with the invitation to participate. Where possible, special adjustments were made to accommodate pupils who could not complete the questionnaire under standard conditions (through provision of, for instance, large-print versions or a reader). --- Sample The analysed sample consisted of 9 countries (Armenia, Estonia, Israel, Latvia, Lithuania, Moldova, Poland, Russia, and Ukraine), involving 43,667 school children in total. The samples within the countries ranged from 3679 to 6193 participants. Detailed characteristics of study sample by age, gender, and country are presented in Table 1. --- Instruments School support was assessed using the amended Teacher and Classmate Support Scale [28]. Teachers support was evaluated using 3 question items: "I feel that my teachers accept me as I am", "I feel that my teachers care about me as a person", and "I feel a lot of trust in my teachers". Response options ranged from 1-"very strongly disagree" to 5-"strongly agree" for both teacher and classmate scales. Classmates support scale consisted of 3 items: "The students in my class(es) enjoy being together", "Most of the students in my class(es) are kind and helpful", and "Other students accept me as I am". For analyses, the mean scale score for both teacher and classmate support scales was dichotomised: values from 1 to 2 pts were considered as "Low support", and a score higher than 2 as "High support". Confirmatory factor analysis from a number of the European countries [28,29] supported a two factor structure for these support scales and confirmed test-retest reliability and measurement invariance across countries. Family and friends support was measured using the Multidimensional Scale of Perceived Social Support (MSPSS) [30]. The MSPSS has been well-validated and used in multiple studies and across different cultural contexts [31][32][33]. Friends support scale was evaluated using 4 items: "My friends really try to help me", "I can count on my friends when things go wrong", "I have friends with whom I can share my joys and sorrows", and "I can talk about my problems with my friends". Response options for both friend and family support ranged from 1-"very strongly disagree" to 7-"very strongly agree". Dichotomisation of this scale was defined using the scale response descriptors, where mean scale scores ranging from 5 to 7 pts were considered as "High support", and a score lower than 5 as "Low support". Family support was evaluated by 4 items from the family subscale: "My family really tries to help me", "I get the emotional help and support I need from my family", "I can talk about my problems with my family", and "My family is willing to help me make decisions". Dichotomisation of family support scale was defined using the scale response descriptors, where mean scale scores ranging from 1 to 2 were considered as "High support"; a score lower than 1-categorised as "Low support". Lithuania used different scale intervals (according to different response options), where the mean scale scores higher than 5 pts were considered as "High support", and scores lower than 5 as "Low support". The internal consistency of all support scales across countries was acceptable, being 0.72 at lowest. The Cronbach alphas were consistently higher for family support (0.86-0.95) and friends' support (0.80-0.93), indicating not only necessarily higher consistency, but higher number of items as well. Detailed data on internal consistency are presented in Table 2. For overall social support analysis, the four scales were composed to one composite indicator-Overall Social Support Score. This score ranged from 0 to 4, indicating how many support types were high for every particular participant under study. Sufficient social support was considered as 3 or 4 points on Overall Social Support Score. The Family Affluence Scale (FAS-III) [34] is an indicator of young people's socio-economic status comprising six items on material assets in the family: "Does your family own a car, van or truck?", "Do you have your own bedroom?", "During the past 12 months, how many times did you travel away on holiday with your family?", "How many computers does your family own?","How many baths/showers are there in your house", and "Does your family have a dishwasher at home?" Scale scores were calculated by summing up the scores of all six items. A measure of relative FAS (proportions of low-medium-high in each country) of 20-60-20 percentiles was used in the current study, according to analysis applied in the HBCS international report from the 2013/2014 survey [35]. Armenia and Lithuania did not include two and one FAS items, respectively, lowering the theoretically maximum scale score; nevertheless, the FAS subgroups for those countries were also based on 20-60-20 percentiles of their distributions (Table 3). Fighting was evaluated with the question: "During the past 12 months, how many times were you involved in a physical fight?" Responses of at least 3 times in the last 12 months have been classified as frequent physical fighting and a dichotomous variable was constructed in line with this [36]. To measure bullying perpetration, adolescents were asked "How often have you taken part in bullying another student(s) at school in the past couple of months?" These questions were preceded by a definition of bullying [37], which has been well-used and validated in empirical studies in multiple countries [38,39]. Reports of at least 2 or 3 times a month in the past couple of months have been considered as chronic bullying [40,41]. Smoking was evaluated through the question adapted from Monitoring the Future Study and the European School Survey Project on Alcohol and Other Drugs (ESPAD) [42]. Participants answered on a 4-point Likert scale from 1-"never" to 4-"every day". A dichotomous variable of smoking was created by recoding ("rarely or never" or "at least once a week"). Self-report smoking measures have been found to have good reliability and validity [43]. Drunkenness was measured by the following question: "Have you ever had so much alcohol that you were really drunk?" Based on Swiss ESPAD data, the question on subjective drunkenness better measures risky single occasion drinking among adolescents rather than by asking about the frequency [44]. This item has been well used and found to have a good predictive and criterion validity [44]. As in previous papers using HBSC data, a dichotomous variable was created in order to identify adolescents involved in problematic alcohol use (indicating "never or once in the past month" or "twice or more") and due to the skewed nature of the distribution [45]. The majority of the measures were dichotomised in order to have comparable odds ratios across different factors. In cases where different factors would have a different number of response categories, the differences will be more striking due to larger distances between extreme categories. --- Statistical Analysis The study data were analysed using the "IBM SPSS 20.0" statistical package. The descriptive analyses included means <unk> standard deviations (SD) as well as percentages. Inferential analysis was based on logistic regression. First, univariate regression was conducted to check the associations between analysed factors and outcomes (fighting or bullying). In multivariate regression, separate models were built to predict higher levels of proneness to fighting and bullying. The latter two were considered as binary outcomes. Putative risk factors included in multivariate models were perceived support from teachers, friends, family, and peers, sociodemographic indicators (gender, age group, and family affluence), and risk behaviours (smoking and drunkenness). The strength of associations was expressed in odds ratios (OR). For inferential analysis, separate support scales were used to calculate an Overall Social Support Score (OSSS), reflecting on how many of social supports scales a particular adolescent had high social support. This was conducted due to strong associations between different types of support (odds ratios ranging at levels of 2 to 3 and more) that later in multivariate analyses may give underestimated associations due to strong interactions. Since we analysed four types of support, the OSSS score ranged from 0 (none of the supports is high) to 4 (all analysed social supports are high). For countrywise comparisons this was further dichotomised to an indicator of Sufficient Social Support, referring to the proportion of the population with at least 3 types of social support rated as "high". The multiple regression models were run separately for all countries together and for every country under analysis to check for possible (in)consistencies of risk effects. The statistical significance level was set at 5%. --- Results --- Descriptive Analysis This study analysis included nine countries, mainly from Eastern Europe (Armenia, Estonia, Latvia, Lithuania, Moldova, Poland, Russia, and Ukraine) and included Israel as well. The bullying in the analysed countries was more prevalent than fighting, with the exception of Armenia, Israel, and Poland. The prevalence among countries differed several times, with fighting being most expressed in Armenia (19%) and bullying-in Latvia and Lithuania (23%). Current smoking was most prevalent in Russia (almost 9%), while drunkenness-in Lithuania (16%). The detailed information on the prevalence of abovementioned behaviours is presented in Table 4. The social support in this study was analysed assessing four types of support-classmates, family, friends, and teachers support (Table 5). In most countries, it was high family support that was most expressed among different support types. It can also be noted that high friend support was usually more prevalent compared to classmate support, while the levels of the latter were similar to the teachers support. For further analyses the social support was regarded as a total score composed of classmates, family, friends, and teachers support (as overall social support-OSSS). OSSS indicated the number of different support types that were rated as high. It was found that the highest prevalence of full social support (i.e., all four types of social support are high) was in Armenia (36%), Israel and Moldova (33% each). For country wise comparisons additionally, we used a dichotomous social support measure-Sufficient Social Support, indicating how many participants had at least three of four analysed types of social support rated as high. --- Overall Support as a Predictor of Fighting Further, the analysis of potential risk and protective factors for fighting was conducted. For this, the univariate model was calculated first, with later adjustments for sociodemographic factors and risk behaviours. The findings suggest (Table 6) that the strongest risk factor for fighting was male gender: boys are 4.9 times more likely to have been involved in fighting than girls. A younger age was also associated with a higher risk of fighting. Among other independent risk factors for fighting, it was found that smoking and the experience of being drunk were both associated with a double likelihood of being involved in fighting. Social support was also found as an independent factor associated with fighting (Table 6). In multivariate analysis, we used OSSS and it was found that the less social support, the higher the risk of fighting among adolescents: the respondents who had the lowest level of overall social support were 1.5 times more likely to have been involved into fighting compared to school children who had the highest perceived social support. --- Overall Social Support as a Predictor of Bullying The analysis of risk factors for bullying revealed (Table 7) that male gender and risk behaviours such as smoking or being drunk were significant predictors of bullying: these three factors were associated with an approximately double risk of bullying among school children. However, differently from fighting, gender had a less expressed effect. On the other hand, the social support's potential effect on bullying showed almost a linear pattern-the increase of one additional high support type was related with a quite similar increase of risk estimate (Table 7). In particular, a high risk of bullying was found for adolescents who reported the absence of high social support from all-family, friends, classmates, and teachers (OR = 3.1). This was the most vulnerable subgroup independent of gender, risk behaviours and other factors in the model. --- Effects of Specific Types of Social Support on Aggressive Behaviours In order to see specifically if different types of social support act similarly in predicting the likelihood of fighting and bullying, we analysed them in a multivariate model. The results show (Table 8) that in the case of fighting it was the high family support that acted most strongly (OR = 1.3), with the slightly lower relevance of teacher support (OR = 1.2). However, age and age-mates support (classmates and friends) do not seem to have a protective or risk effect against fighting. Regarding bullying, the high support of family was also dominant in predicting the risk of bullying, and more so than in case of fighting (OR = 1.5). The school-related support (by classmates and teachers) was also significant, though a slightly weaker predictor of bullying than family support. --- Protective Effects of Social Support by Country We established that there is an overall pattern that the social support seems to act as a protective factor against fighting and bullying almost in a linear manner (Tables 6 and7). Therefore, for further analyses on the countries comparison, we analysed the social support as a continuous variable with the score ranging from 0 (none of the four analysed social support scales indicating high support) to 4 (all four social support scales indicating high support). We found that social support as a potential protective factor acts specifically depending on the country. The results showed (Table 9) that the strongest protective effect of overall social support against fighting was found in Israel and Poland (OR = 0.8) and differed significantly from the overall estimate (OR = 0.88; 95% CI 0.86-0.91). These two countries had the prevalence of fighting as one of the lowest among analysed countries. Exceptionally, the data showed that the country with the highest prevalence of fighting, i.e., Armenia, does not face a protective effect from social support (OR = 1.02), unlike other analysed countries. The most expressed protective effect of social support against bullying was found in Estonia and Israel (OR = 0.7); however, the difference from the overall estimate was non-significant. Of note, these countries had higher than average prevalence of sufficient social support and lower prevalence of bullying. Almost the opposite was revealed to occur in countries with the highest prevalence of bullying-Latvia and Lithuania. --- Discussion The public health approach to the problem of violent behaviours among adolescents seeks to identify not only the risk but also protective factors in order to determine when in the lifespan they occur and how to develop the prevention programs [46,47]. It is a complex phenomenon that involves the social framework, relationships within the family, friends, and teachers [8,11,48]. There are complex interactions of individual and social determinants [12] predicting that the adolescent becomes violent. The underlying predictors of school violence and bullying include gender and social norms and wider contextual and structural factors [11,49]. For example, in some Eastern countries, boys are encouraged to be stronger, more aggressive and even fight to prove their masculinity; the social norms existing in many communities in those countries do not support males being too gentle and peaceful. It might be partially caused by difficult historical situation and many wars experienced in the past history and other factors. Current survey data, in general, proved the idea published in the report of UNESCO in 2017 that physical violence in industrialised countries is less common in schools than bullying [50]. Although the individual characteristics and dynamics that underlie social value and power may vary across cultures and countries, the abuse of power to distress or control another person is consistently observed [51]. School violence and bullying are often invisible or ignored by teachers and parents. In some contexts, adults view physical punishment, fighting, and bullying as a normal part of discipline or growing up and are not aware of the negative impact they have on the education, health and well-being of children and adolescents [50]. Comprehensive assessments of children's involvement in bullying and victimisation must take into account their relationships with parents since parents are the primary socialisation agents. Close relationships with adults, especially parental support and time spent with parents, are protective factors of bullying perpetration among adolescents [11,52], which is consistent with our study results. In our study, family support seems to act as the strongest potential protector against fighting and bullying. In the family context, some studies show that parental support is also negatively related to the perpetration of a physical fight and verbal, relational bullying [19]. In studies specifically focused on protective factors of bullying and victimisation [8,52], researchers have found that family support acts as a strong risk-based protective factor, especially when there is a strong attachment to parents in the family. Parents need support and encouragement to promote children's social competencies and healthy relationships. Though the family is usually considered as a primary factor for social support [49,52,53], in case of adolescent aggressive behaviours, it was found that teacher support may also play a significant role [52,53]. Levels of bullying are higher in schools where teachers use authoritarian and inflexible practices to cope with student misbehaviour [54]. Authoritarian practices of enforcing discipline distance the students from their teachers and decrease trust in them. However, both teachers and students report that teachers do not know how to effectively intervene, which prevents students from seeking help and contributes to teachers ignoring bullying [55]. Worsening relationships with teachers and disliking school increased the likelihood of violent behavior [18]. Therefore, it is essential to increase teachers' awareness on how to act in order to prevent violent behaviours of adolescents. Moreover, this prevention should not be based only on direct prevention of bullying and fighting, but rather on improvement of social relationships between school children and teachers. School bonding is related to both bullying perpetration and victimisation, with possible bi-directional influences [49]. Many studies show that peer support is positively related to substance use or risk behaviour [56,57] and is a protective factor for bullying behaviours [17,18]. As shown in this study data, the relationship between peers, teachers and bullying is complex, where received social support from each of these groups acts as a very strong and comprehensive protection factor. School staff and students require understanding about bullying and peers' roles in promoting or preventing bullying. Having friends is a protective factor; therefore, children who are victimised benefit from supportive peer relationships. Promoting social skills and attitudes that are supportive of victimised youth is critical to creating a climate in which peers will intervene. In line with previous studies [58] the present study also showed that risk behaviours such as smoking or being drunk were significant predictors of bullying. Risk behaviors include substance use, and aggression and violence have negative effects both on adolescents, their families and community [59]. In our study, the social support was found to be multidimensional-each form of social support was significantly related to lower levels of violent behaviour, but also the more different forms of social support were associated with lower involvement in aggressive behaviours. Dose-response relationships found in our study when talking about overall social support versus fighting and bullying indirectly may indicate ahigher likelihood of causality based on Hill's causality criterion of biological gradient [60,61]. Our results show that overall social support as a potential protective factor acts specifically depending on the country, and it was pattern that countries, which have lowest rates of bullying and fighting, differed significantly in overall social support (much higher than in other countries). The HBSC survey data analysis on the example of nine selected countries demonstrates implied but not always verifiable truth that what fits the overall region or dataset does not necessarily fit particular countries or subsamples. This strongly depends on the homogeneity of risk estimates. Therefore, the conclusions based on large-scale merged international data of different countries should be re-addressed scientifically or at least dealt with some caution when it comes to implication at national levels. Nevertheless, the heterogeneity of risk estimates across countries should not serve as a suggestion to abolish or raise doubt on international or merged analyses-rather, in cases where overall estimates on the burden of disease or risk factors are needed, the merged datasets that are based on unified study protocols should be of primary relevance and reliability. When talking about the limitations of this study, it should be noted that self-reported aggressive behaviours may not always reflect the true prevalence of such behaviours. On the other hand, the data are based on a rich dataset from different countries. Therefore, it provides an opportunity to examine possible protective effects of social support in a consistent manner across different social environments peculiar for specific countries. An integrated approach to school violence prevention programs in schools with teachers, peers and parents involvement is needed to provide counselling and skills to cope with and resolve violence among pupils and to reduce engagement in smoking and alcohol consumption. Policy regulation development as well as building a health and safety promoting environment should effectively prevent violent behaviours among adolescents. More research is needed to look at similarities and differences in the development of fighting and bullying, as well as potential social protective factors across cultures in order to identify how cultural factors may be related to bullying involvement, with a view to learning effective strategies to promote children's social skills and healthy relationships. Also, research is needed on protective psychosocial and behavioural factors as well as factors related to fighting and bullying in countries with various cultural and social norms with a view to learning effective strategies to promote children's social skills and healthy relationships. This suggests that in some cases the analysis of separate countries may be reasonable, together with overall international comparisons. In addition, our study provides an insight into the multicausality of aggressive behaviours among school children, where the social environment and support are likely to play a role in preventing or triggering these behaviours independently from such well-established risk behaviours like alcohol drinking or smoking. --- Conclusions With some exceptions, the majority of the study findings confirm the findings of other researchers. Bullying, across 9 countries, was more prevalent than fighting, with the exception of Armenia, Israel, and Poland. The prevalence among countries differed by several times, with fighting being most expressed in Armenia and bullying-in Latvia and Lithuania. The strongest risk factor for bullying and fighting was related with male gender (less expressed for bullying), smoking and alcohol consumption. High support by friends and teachers seems to be quite a strong protective factor for school childrens violent bahaviours. Thus, social support was found to be multidimensional and to act as a possible protective factor in an almost linear manner-the increase of one additional high support type (classmates, family, friends and teachers) was significantly related to lower levels of violent behaviour. Also, a number of various forms of social support was associated with lower involvement in aggressive behaviours. Regardless the socioeconomic, historical, and cultural differences among selected countries, the enhancement and reinforcement of the social support from possible many different resources should be taken into consideration in prevention programs against school violence behaviours. --- Author Contributions: K.<unk>., T.V., M.M.-C., M.M., E.M., A.Z. were involved in conducting the data collection. K.<unk>., T.V., M.M.-C., M.M., E.M., and J.L. proposed the study concept. The framework for analysis and description of results was mainly conducted by K.<unk>., J.L., T.V., and M.M.-C. Statistical analyses were conducted by K.<unk>. and J.L. The Introduction and Discussion were written by K.<unk>., T.V., J.L., M.M.-C., M.M., and E.M. All authors revised the manuscript critically and have read and approved the final manuscript. --- Conflicts of Interest: The authors declare no conflict of interest.

Background: This study aims to explore how sufficient social support can act as a possible preventive factor against fighting and bullying in school-aged children in 9 European countries. Methods: Data for this study were collected during the 2013/2014 Health Behaviour in School-aged Children (HBSC) survey. The sample consisted of 9 European countries, involving 43,667 school children in total, aged 11, 13 and 15 years. The analysed data focus on social context (relations with family, peers, and school) as well as risk behaviours such as smoking, drunkenness, fighting and bullying in adolescents. The relationships between social support and violent behaviour variables were estimated using multiple regression models and multivariate analyses. Results: Bullying, across 9 countries, was more prevalent than fighting, except for Armenia, Israel, and Poland. The prevalence among countries differed considerably, with fighting being most expressed in Armenia and bullying-in Latvia and Lithuania. The strongest risk factors for bullying and fighting were male gender (less expressed for bullying), smoking and alcohol consumption. In addition, for bullying the social support was similarly strong factor like above-mentioned factors, while for fighting-less significant, but still independent. All forms of social support were significantly relate with lower violent behaviour of school children, and family support was associated most strongly. Regardless the socioeconomic, historical, and cultural differences among selected countries, the enhancement and reinforcement of the social support from possible many different resources should be taken into consideration in prevention programs against school violence behaviours.

INTRODUCTION With the emergence and dynamic development of social networking platforms, the way of doing shopping has altered considerably. Sreejesh et al. (2020) highlighted the modification of their function from communicative to interactive one. Zafar et al. (2019) stated that social media serve as a trigger to do impulsive purchases. Moreover, Varkaris & Neuhofer (2017) noted that compared to other SNS-s Instagram has a huge impact on the decision-making process of consumers. The fashion retail market due to intense competition has forced companies to meet modern requirements. Jegham & Bouzaabia (2022) characterize this industry as ultra-modern so to succeed in this field, there is a need to comprehend individuals' consumer behavior who tend to do impulsive purchasing (Dover, 2019). Shopping is one of the areas where young consumers from generation Z start to act independently by making choices and expressing themselves (Djafarova & Bowes, 2021). The main distinguishing feature of Generation Z compared to other generations is that they were born in the digital age (Seemiller & Grace, 2016). For this reason, scholars introduced several variations such as "digital natives", "post-millennials" or "iGeneration" to name them with an aim to describe fully and better outline their inherent characteristics (Wiedmer, 2015). Compared to other generations iGens are tech-geeks, well-educated and more confident (Bhavya et al, 2022). The behavior of this generation is closely related to technology and social media. According to Kim et al. (2020) by 2025, the Asia-Pacific region's population will consist of 25% centennials. As they grow up and become active members of society, there is a need to have a close look at their preferences and interests. They cannot imagine their life without the Internet, so their shopping decisions and the search for needed items are made there (Singh, 2014). Dabija & Lung (2019) stated that since digital natives most of the time stay online, so they consider online shopping as an obvious choice. Some scientists reported that young adults incline toward impulsive purchases because fashion items are one of the tools for their communication in society. Stachowiak-Krzy<unk>an & Ankiel (2019) also paid attention to the remarkable role of youth in the fashion industry. In recent years, Instagram has hugely affected consumer behavior; however, far too little attention has been paid to this relation, and precisely the impact of stimuli on this social media has not been investigated in depth (Kim & Kim, 2019). By applying the S-O-R framework Leong et al. (2018) have examined the effects of stimuli on impulsive purchasing on Facebook. In 1974, Mehrabian and Russell introduced S-O-R model -stimulus, organism and response -that was mainly used in psychology (Parboteeah et al., 2009). Nevertheless, this framework plays a critical role in addressing the issue of an individual's impulsive action. Therefore, its advanced versions are widely used in marketing. Despite the importance of research on post-millennial consuming manners, there remains a paucity of evidence on emerging markets. Villa & Jason (2017) state that no detailed investigation from an angle of national research in this field. Thangavel et al. (2021) argue that the research to date on the Generation Cohort theory has tended to focus mostly on consumers from Western and European countries. Such approaches, however, fail to apply to developing nations. This indicates a need to understand this issue from the perspective of emerging markets. Moreover, a systematic consideration of how iGens' use of Instagram affects impulsive purchases in Kazakhstan is still lacking. There are two primary aims of this study: 1. An analysis of the increasing role of Instagram in the context of generation Z and impulsive purchases in Kazakhstan. 2. Whether the modified version of S-O-R model proposed by Djafarova & Bowes (2021) applies to the purchase habits of Kazakhstani post-millennials. --- LITERATURE REVIEW --- Instagram and generation Z in Kazakhstan Instagram is a type of social media application available free in application stores. It is mostly dedicated to visual content and allows individuals to share their photos and videos with friends and other Instagram users (Wally & Koshy, 2014;Tingetal, 2015). As World popul at ion review (2022) indicated, in 2022 Instagram had more than 2 billion users worldwide. According to Briskman (2022), this app was downloaded most in the 4th quarter of 2021. There were 10.9 million Instagram users in Kazakhstan in January 2023 and almost a third of them (31.6%) are youth 13-24 y.o. (Cat, 2023). Jegham & Bouzaabia (2022) highlight Instagram's significance in the sales of fashion items. It can be explained by the desire of brands to visually present their apparel. Due to this primarily visual function of Instagram, most fashion companies give priority to this app. Furthermore, its user-friendliness is another advantage for brands to consider it as a sales channel. Following the global tendency, Kazakhstan's market is also experiencing an escalating appearance of centennials. As a report from the Ministry of Labor of the Republic of Kazakhstan states, Millennials and Generation Z in 2022 make up 60% of the workforce. By 2030, this figure will reach 80% (Primeminister, 2022). From an economic point of view, this is a large category of consumers. The vast majority of iGens prefer everything at a rapid pace. Much of the current research on their shopping habits pays particular attention to impulsiveness. It is now well established from a variety of studies, there is a clear link between Instagram and consumers' spontaneous purchases because they tend not to evaluate while using it. Moreover, this app functions as a stimulus for impulsive buying (Xiang et al., 2016;Aprilia & Setiadi, 2017). Research findings revealed that about 60 % of Z members made spontaneous purchases (Brewis, 2020). --- Content on Instagram Scholars suggested several variations where Instagram may lead consumers to purchase spontaneously. Firstly, recommendations from influencers and friends (Zhu et al., 2020; Xiang et al., 2016). Secondly, promotional advertisements (Triwidsari et al., 2017). Lastly, a third option -is the appearance of videos and photos on the brands' Instagram (Handayani et al., 2018). However, to date, only a limited number of research was dedicated to its impact on impulsive buying decisions from the perspective of the fashion industry (Djafarova & Bowes, 2021). The content on Instagram mostly can be derived from two types -user-generated and brandgenerated content. The first type of content on Instagram is User-generated content (USG). Gao et al. (2021) defined it as "a form of content created by users of a system or a service and made available publicly on that system". So as Beveridge (2022) noted "any content -text, videos, images, reviews, etc. -created by people, rather than brands" can be categorized as a UGC. From the research evidence, it is clear that UGC influences the urge to buy impulsively more compared to BGC (Mayrhofer et al., 2019). 62% of centennials use Instagram on a daily basis (Wise, 2022). The posts with bought items and recommendations are recognized as extremely powerful in changing consumer behaviors (Aragincillo & Orus, 2019) Therefore, it is perceived that UGC is more credible than BGC (Nash, 2019). Brand-generated content (BGC) means various forms of content such as photos or videos that are created directly by companies for the launch and promotion of their goods with the main goal to engage and enlarge buyers' interest (Liang et al., 2020). Sometimes it may be designated as "marketer-generated content" or "firm-generated content". In the context of Instagram, the "publications" and "stories" reached recognition from both -companies and consumers. 500 million users view stories on Instagram, the most popular ones usually belong to the content from companies' accounts (Casalo et al., 2021). Pongpaew et al. (2017) explained that BGC improves not only consumer engagement on the accounts of a brand but also assists to develop brandconsumer and consumer-consumer contacts. The authors such as Fromm & Read (2018), Evans et al. ( 2019), Stachowiak-Krzyzan (2019) also discussed BGC from various perspectives. Additionally, influencers and opinion leaders can address messages on Instagram creating the next type of BGC. According to Bruhn et al. (2012), this form of communication is also powerful. In their studies Casalo et al. (2018) and Boerman (2020) have shown that Instagram is a popular and attractive app among influencers; moreover, they may engage a big number of online followers. According to Jegham & Bouzaabia (2022) such criteria as the trustworthiness of opinion leaders and their genuine relationship towards their followers help to reach a positive attitude. Fakhreddin & Foroudi (2022) also stated the characteristics such as quality and ingenuity that will lead influencers to prosper. In some cases, influencers' sponsored advertisements may cause negative viewpoints about the product, as users mostly are aware of micro-celebrities possible benefits from companies (Kim & Kim, 2020). Stimulus-Organism-Response model and impulsive purchase Mehrabian and Russel (1974) introduced this framework -S-O-R stands for Stimulus -Organism -Response. As shown in Figure 1 the S-O-R model identifies the interrelation between stimuli, organism and response, so external factors influence organisms triggering emotions in individuals resulting in behavioral responses from them. Some researcher categorized two forms of stimulus (S) -external such as situational stimuli and marketing stimuli or internalcharacteristics of consumers. (Chan et al., 2017). This includes individual's reactions like investigation or approach that may appear as the result of interaction with organisms and stimuli. Previously, this paradigm is often used by researchers. As they highlighted this model is considerably helpful to explain a connection between consumer behavior and environmental stimuli. With the recognition of motivation in impulsive purchases as the stimulus, not only online purchases but also offline ones were represented by applying This section has attempted to provide a summary of the literature relating to impulsive purchases within Instagram by centennials. There are relatively few studies that have assessed Instagram's impact on impulsive purchase intentions, therefore scholars state about research gaps. The features of earlier studies in this field were characterized by a lack of general analysis -for instance, some researchers investigated only women, whereas the opinions of men would be also vital or some of them mainly focused only on a particular aspect. Furthermore, to our knowledge, no study has examined the context of emerging markets like Kazakhstan although digitization in this country is recognized as competitive. Thus, this study aims to determine the most influential stimuli and to test the S-O-R model to the consumer behavior of Generation Z in the fashion industry within Instagram. --- METHODOLOGY Previously many researchers have utilized qualitative methods to achieve a more profound outline of centennials' motives and behaviors (Chen, 2018;Djafarova & Trofimenko, 2019). However, both qualitative and quantitative methods were used in this research. Denscombe (2010) defined mixed methods as the "use of qualitative and quantitative approaches within a single research project". Aiming the improvement of findings quality scholars may use a combination of different methods where each method has its specific features. This type of research allows us to look into issues more intensely and to understand the interdependences between methodology and findings. On the other hand, performing the research using several methods requires the proficiency of the researcher and takes time as well. Moreover, there is a possible problem of incompatibility of the results from various methods. The study uses qualitative analysis to gain insights into behavioral intentions of young iGen consumers. A focus group approach can be more useful for identifying and characterizing the motivations and factors that lead them to do impulsive purchases. Moreover, the main body of literature in this field applied qualitative research methods (Chen, 2018;Nash, 2019). In general, six focus groups were conducted. Total of 39 respondents -24 of them were female and 15 were male. All of the participants were aged between 18 and 22. The primary inclusion criteria for the focus group participants were being in the age frame of generation Z. The study was held in November-December of the year 2022 in Almaty City. The participants were recruited from Al-Farabi National University among the Higher school of economics and business students. Prior to commencing the study, the participants received an explanation of the research. In addition, ethical issues and anonymity were also clarified. Semi-structured interviews were conducted in six focus groups. The set of predetermined questions generally can be divided into three levels -1) about Instagram and social media usage, purpose and frequency; 2) about the activity of influencers and bloggers, brands and fashion companies' Instagram profiles; 3) about users' reactions to advertisements and recommendations, comparing influences of UGC and BGC, intentions to impulse purchases. The quantitative method was applied to obtain data from participants about Instagram's influence on impulse purchases. The main goal of holding online-based surveys was to prove numerically the results of focus groups. The survey was created and conducted on the Google platform and was distributed online; thus, participation is voluntary. Participants first provided informed consent about the general purpose of the study. The questionnaire was prepared in the Kazakh and Russian languages. Table 1 below illustrates the demographic data of the respondents. All the participants were in the age frame of generation Z between 18 and 22. The data were collected from 106 centennials -68,9% of the sample were female and 31,1% were male. In age criteria, most of the respondents are 18 years old (37,7%), and 20 years old (33%). --- FINDINGS AND DISCUSSION In this part, the results of focus groups and online survey will be discussed. The first set of questions during the focus groups aimed to shed light on the usage of Instagram, frequency of use and purpose. Regarding the question about the presence of an Instagram account, almost all participants excluding a male (who stated his indifference to it) answered positively. Another male (focus group 4) shared his unsuccessful experience of deleting an Instagram account, which was re-uploaded 5 hours later. As he explained, "I could not do anything and it felt like I lost something meaningful, something was wrong". To assess the frequency of using Instagram, the second question was asked. The vast majority replied minimum of three-four hours. Among the respondents, a female (focus group 1) disclosed spending 10-12 hours on Instagram but she then clarified that she does it simultaneously with other activities, for example, while cooking. For the third question about the purpose of Instagram usage among the frequent answers were -"to have fun in spare time", "to chat and to be in contact with friends", and "to be aware of the latest news". The responses of the online survey about daily usage of social media platforms are presented in Figure 1. These results are in agreement with those obtained by focus groups indicating that social media platforms are an integral part of iGens' lives. So as can be seen from Table 2. more than a third of respondents entertain on social media for 3-4 hours and a massive division of centennials (44,3%) have a preference for using social media platforms for 4 and more hours. The share of respondents who use social media less than 2 hours a day is 1/5, that means that for the vast majority (approximately 80%) of interviewees spending more than 3-4 hours on Instagram is considered as a normal activity (Table 2). Additionally, to analyse the popularity of social media platforms among the centennials the question was asked to indicate three of the most used type of social media platforms. The responses were distributed in the following way in Table 3. As it has become clear from the discussions of focus groups informants believe that social networking platforms such as Facebook and VKontakte are not of particular interest and they are intended for a more mature and older audience. The online survey demonstrate the similar results. For example, Facebook was chosen only by 4,7% of respondents. As can be seen in Table 3, among the leaders are three of them -Instagram (91,5%), Youtube (76,4%) and Tik-Tok (72,6%) -the most popular platforms from the perspective of youth. So, the answers demonstrate that Instagram has become the favoured app by the members of generation Z. The purpose of the second part of the questions was to analyze the informants' opinions about the activity of bloggers and influencers, brands and fashion companies' Instagram profiles. Respondents of focus groups were asked to indicate whether they follow bloggers -33 % answered positively, 20 % pointed out that they follow no more than five bloggers in total and 38 % replied negatively. An individual who answered negatively reasoned his response that "these days the bloggers tend to advertise not only their blog or product, but they advertise everything for money". The results indicate that the trustworthiness of bloggers and influencers is questionable so this reflects those of Djafarova & Bowes (2021) who found their activities less credible. The reason is nowadays users are aware of their remunerations. In contrast, the opinion of respondents about the Instagram accounts of fashion brands and fashion retail shops were much more positive. The vast majority (72 %) answered that "yes, follow them", adding positive comments such as "all the latest news about sales or arrivals I knew from the Instagram account of the retail shop", "If there was no retail shop's Instagram account, I would not know about it". The third set of questions is designed to define the users' reactions to advertisements and recommendations, comparing influences of UGC and BGC, and motivations to buy apparel items impulsively. There is a surprising result to the question "What will have more influence: 1) an advertisement of a blogger; 2) a post of a fashion retail shop; 3) a friend's recommendationalmost all participants noted that a friend's recommendation will have greater impact. They explained that friend is a reliable person who will not recommend something bad. This outcome correspondence with the viewpoint of Stachowiak-Krzy<unk>an & Ankiel (2019) who indicated that the opinions of friends and peers are extremely indispensable for young consumers. Mostly, they tend to make decisions by following their pieces of advice. However, some responses pointed out the importance of all mentioned types of advertisement. A combination of all three will influence to purchase impulsively in the way that each form has its features, so every form will boost another's influence. As one interviewee said: 'If I saw a post of a fashion retail shop, then a blogger proposed it, finally, if my friend will suggest me as a result I will buy this item'. The answers to the online survey about the influence of different types of stimuli are presented in Table 4. These results do not support the discussions of focus groups which indicated a friend's recommendation as an absolute leader among others. According to responses of online survey participants the stimuli like a post of the brand (37,1%) and friend's recommendations (46,7%) have a greater influence on them rather than a post of fashion blogger (16,2%). These results indicate that friend's recommendations are also important however, they are not as high as in the results of focus groups. Additionally, the respondents hold the opinion that post of brand or retail shop on Instagram (37,1%) also may have an influence on them in purchase of fashion apparel (Table 4). The present study was designed to determine whether Instagram plays a prominent role in the lives of centennials and whether it influences to apparel purchasing habits of them. The most obvious finding to emerge from the analysis of the results of both focus groups and online survey is that Instagram has become a powerful platform in various aspects. The participants consider it as a predominant tool in terms of not only socializing and entertaining but also as a key instrument that has an effect in shaping their impulsive shopping behaviors. Moreover, another piece of evidence for it is the amount of time they spend on Instagram. Surprisingly, this research found that the trustworthiness and popularity of bloggers and micro-influencers are not highly estimated; on the contrary, the accounts of fashion brands and retail shops gained more positive responses and credibility. Likewise, the informants highlighted that friends' recommendations will be a decisive factor in purchasing an apparel item. Nearly 80 % of participants agreed that Instagram hugely affects their purchasing behavior and dressing style. The majority stated about changing their styles, saving time and being aware of the latest trends as positive aspects of Instagram. This finding is consistent with that of Lidholm et al. (2017) who stated about considerable changings that social media platforms have made. The abovementioned findings provide support for the hypothesis that Instagram has a huge influence on the everyday activities of generation Z. Turning now to the next hypothesis about the applicability of the revised S-O-R model to purchasing habits of Kazakhstani post-millennials the findings illustrate a positive correlation. In Figure 2 presented a version of S-O-R model. (Mehrabian & Russel, 1974) The data from research suggest that stimuli such as a brand's account on Instagram, bloggers' activity and recommendations of friends trigger emotional responses (such as positive reactions, interest and willingness to buy) from centennials that in turn result in behavioral responses. The behavioral response can be seen as a spontaneous purchase of fashion cloth by them. Moreover, the results of the research indicate the interest to buy impulsively arise in both gender -female and male. Therefore, this is partly in contrast to the prevailing stereotype that fashion purchases are only interesting for women. In contrary the findings of researchers who demonstrated the micro-celebrities impact on the impulsive purchase of fashion among generation Z females, this paper reported a positive attitude of them towards brand activity on Instagram. However, there is a need to conduct further research to develop and maintain more detailed outcome. --- CONCLUSIONS Instagram is recognized to be an influential social media due to its distinctive feature such as it is a more concentrated on visual appearance that has a significant impact on advertisements in the fashion industry. Instagram's popularity in everyday life and its influence on impulse purchases among generation Z are indisputable. It plays a prominent role in presenting and triggering purchase intentions among young consumers in a virtual world of apparel items. However, despite the fact that an Instagram is a popular app in Kazakhstan there is a lack of research in terms of its use among centennial and particularly Instagram's impact on purchasing of fashion apparel. To fill this gap this paper aimed to justify this platform's increasing presence and describe its importance in impulsive purchases by centennials. The Methodology of this research was designed to understand Instagram's role in spontaneous purchases of fashion items by young members. So to obtain the results there were held both -qualitative and quantitative methods. The focus groups and online survey results allowed to make the following conclusions: -Instagram has a substantial impact on the daily lives of centennial; -this app plays a critical role in being aware of the latest fashion trends; -the research discovered Instagram's capacity to alter consumer habits and fashion perceptions of iGens; -the study also demonstrated an increasing interest in changing the tastes and consuming behaviour of males; -brands' advertisements and friends' recommendations on Instagram are perceived more positively than publications of bloggers, therefore their creativity and ability to engage consumers will be appropriate; -positive emotions derived from Instagram mostly can influence centennials' impulsive purchases of fashion items; -the results show that S-O-R framework can be applied to understand an impulsive purchase of centennials. To conclude, according to the findings it can be made assumptions that in order to win attention of centennials fashion brands and retail shops may broadly use Instagram. Because capacity of this app and its concentration on visual appearance of apparel allow to young people to assess the characteristics of fashion items and make decision about purchase. The future research may be addressed to identify whether the answers of a study with a bigger sample size of respondents will demonstrate similar results. Moreover, the answer to the question if the research will be held in other big cities or small towns will the results differ may cause an interest and bring valuable insights too. Additionally, a quantitative research with large sample size might be helpful to verify the applicability of findings to broader representatives of centennials. The results are helpful for managers to comprehend centennials' purchase behavior on Instagram and to develop satisfactory marketing strategies that will lead to fruitful results. --- LIMITATIONS AND RECOMMENDATIONS Although this research comprises sufficient results of focus groups, the authors note that because of the comparatively small number of respondents in online survey not all the conclusions might be a case to the broader audience of centennials in the country. Therefore, future research may be addressed to identify whether the answers of a study with a bigger sample size of respondents will demonstrate similar results or if the research in other big cities or small towns will be different may cause interest and bring valuable insights too. --- PRACTICAL IMPLICATIONS The findings are helpful for managers to comprehend centennials' purchase behaviors on Instagram and to develop satisfactory marketing strategies that will lead to fruitful results. --- AUTHOR BIOGRAPHIES Aigerim Kazhmuratova -Cand. Sc. (Econ.), Senior Lecturer, Al-Farabi Kazakh National University, Almaty, Kazakhstan. <unk>mail: [email protected] *Zhazira Kakitayeva -PhD candidate, Al-Farabi Kazakh National University, Almaty, Kazakhstan. Email: [email protected] Zhazira Tymbayeva -Cand. Sc. (Econ.), Associate Professor, Satbayev University, Almaty, Kazakhstan. Email: [email protected], ORCID ID: https://orcid.org/0000-0002-7705-9874 Dinara Satybaldiyeva -PhD, Associate Professor, Satbayev University, Almaty, Kazakhstan. Email: [email protected], ORCID ID: https://orcid.org/0000-0001-6494-0681 Leona Tam -PhD, Professor of Marketing, University of Technology Sydney, Australia. Email: [email protected], ORCID ID: https://orcid.org/0000-0003-1116-998X

The proliferation of digital technologies has made a huge impact on the development of consumers' behavior and purchasing habits. If to consider the rising presence and the role of centennials in the market, it is highly important to comprehend their preferences and consuming behavior. The primary purpose of this article is to examine the increasing role and impact of Instagram on the impulsive purchase of fashion items in the context of generation Z in Kazakhstan and to identify what kind of stimuli may affect the spontaneous decisions of centennials on buying apparel items. The methodology comprises six focus groups with 39 participants and an online survey which involved 106 centennials. The research was conducted in Almaty city among the population Z aged 18 and 22. The results indicate a prominent influence of Instagram in the everyday activities of respondents; moreover, it showed a positive correlation between this platform and the motivations of centennials to buy impulsively. The following results were obtained: 1) activities of fashion brands and friends' recommendations on Instagram act as stimuli on young consumers from generation Z; 2) activities of fashion brands and friends are likely to trigger positive reactions among centennials; 3) positive emotions derived from Instagram mainly can influence on centennials' impulsive purchases of fashion items. The outcome of this research can be helpful in the marketing departments of companies to understand the centennials' consuming habits and their use of social networks.

Introduction An important determinant of a student's behaviour and performance is the school's teaching and learning environment. 1 Evidence suggests that a positive learning environment as perceived by students impacts their academic performance and can lead to increased success in both the academic and professional domains. 2 Innovations in medical curricula (which include a blend of classroom, workplace, clinical and community-based learning) and increasing diversity of the student population in medical courses have led to increased recognition of a need to evaluate the educational environment of medical schools. Evaluation helps to assess if these curricula are beneficial to students and adding to their skills as compared to the traditional counterparts, and to draw lessons for continuous improvement. 3 The combination of workforce shortages, increasing burden of chronic disease, more treatable conditions, advances in medical technology and an ageing population have led to increasing demands on the healthcare system. 4 The inequitable distribution of healthcare workers with shortages in rural communities presents a major area of concern for addressing the global burden of diseases and quality healthcare delivery with universal coverage. 5 Mostly, urban areas show a heavy concentration of healthcare workers whilst the population in rural areas experiences a greater burden of diseases. As evident from the literature, there have been recommendations in the past that new models of healthcare delivery should be examined to address these issues. 4 Staffing health facilities with sufficient numbers of appropriately trained health professionals is a major challenge and a prerequisite to implement the National Health Insurance successfully in South Africa. 6 To fulfil the National Department of Health's vision of 'Health care for all' it is necessary to develop and employ new health professional cadres to meet the health needs of the population, ensure retention and improve workforce productivity. 7 Whilst mid-level health workers have been successfully addressing medical workforce shortage in high-income countries such as the United States of America, Europe and Australia, 4,8,9 they hold great potential for addressing human resources shortages in low-and middle-income countries, as suggested by evidence from mid-level practitioner employment in countries like Uganda, Tanzania, Kenya, Malawi and Mozambique. 10 With this in mind, South Africa has developed a new cadre of mid-level health professionals called Clinical Associates, with the aim of improving quality of health care at hospitals, revitalising primary health care at district level and universalising health coverage in the country. 10 Formation of this cadre in South Africa led to development of a 3-year Bachelor of Clinical Medicine Practice (BCMP) degree programme resulting in qualification as a Clinical Associate. The BCMP was started by three universities in South Africa, with the University of the Witwatersrand (Wits) launching its programme in 2009. 11 The BCMP course structure is based on the principle of developing a sound knowledge of medical and clinical sciences to enable students to understand medical conditions of patients and their management strategies with a patientcentred approach. The curriculum follows an integrated approach with a combination of teaching modalities delivered in the classroom, skills laboratory and district hospitals led by family medicine practitioners and clinical associate tutors. During this evaluation the third-year BCMP students were placed at various district and provincial hospitals in Gauteng and North-West provinces for clinical rotations. In 2011 the hospitals used in Gauteng included the Kopanong and South Rand District Hospitals and Natalspruit Hospital, whilst those in North-West included Taung District Hospital and Rustenburg and Mafikeng Provincial Hospitals. As per the curriculum, students needed to complete 5-week clinical rotations in the following departments during their third year in 2011: (1) Surgery, (2) Emergency Medicine, (3) Paediatrics, (4) In-patient Medicine, (5) Out-patient Medicine and HIV, and ( 6) Elective (can choose any department from the above five again or a different department). Students are allowed to remain in one hospital for more than one rotation, generally spending 5-15 weeks at one hospital. As evident from the literature, many evaluations (qualitative and quantitative) have been conducted globally to measure the academic environment of health sciences programmes (undergraduate and postgraduate) by utilising different methodologies. These evaluations include assessment of medicine, nursing, physiotherapy, dental science, chiropractic and other related programmes. 12 In order to gather information on whether the programme was meeting the expectations of students in terms of a better learning environment, and if its design was student-centred, an initial evaluation was required to guide course organisers for better development of the programme. It was therefore decided to evaluate the educational environment of the BCMP in 2011 with the rationale of quality improvement, to incorporate students' feedback in course development and provide a better academic experience for future Clinical Associate students. This study is a preliminary evaluation which was conducted with the first cohort of third-year students (placed in Gauteng and North-West provinces) who graduated in 2011 as Clinical Associates. Our objectives were firstly to understand and evaluate students' (Clinical Associates) perceptions and/or experiences of the educational environment in the BCMP programme by using the Dundee Ready Education Environment Measure (DREEM); secondly, to generate a profile of students' perceptions in terms of the strengths and weaknesses of the educational environment by exploring individual item scores; and lastly, to utilise results to draw lessons for improving the curriculum and learning environment and develop student support mechanisms in order to improve the academic experience. http://www.phcfm.org doi:10.4102/phcfm.v7i1.778 --- Research methods and design --- Study design and setting This cross-sectional quantitative survey was conducted in 2011 amongst third-year (final-year) Wits BCMP students (based in North-West and Gauteng provinces) who graduated in 2011 as Clinical Associates. Whilst the overall response rate was 88% (n = 22/25), there was an item-specific response rate where items 1, 13, 16 and 28 were responded to by 10 out of 11 students from Gauteng. Although the demographic characteristics of respondents (such as gender and age) were not captured through the questionnaire, we assumed that any differences in these did not have a significant influence on their views about the learning environment of the BCMP. In addition, we assumed that due to small sample size the exclusion of respondents' demographic information would ensure that their identity was concealed and that no information would be easily identifiable to us. 13 --- Data collection Data were collected using the self-administered DREEM questionnaire consisting of 50 items to be answered on a 5-point Likert scale ranging from'strongly agree' to'strongly disagree'. The questionnaire was administered within a classroom setting after the lecture. Before conducting the study and selecting the instrument we assumed that the learning environment as perceived by students was not restricted to the third year (final-year) of their clinical rotations when they answered the questions, but pertained to the overall environment throughout the three years of the course. We assumed that whilst answering the questions students perceived the terms in the questionnaire as follows: • 'Teaching' -overall teaching including classroom and hospital based; --- Data analysis Data were analysed using Microsoft Excel 2010 and interpreted to draw conclusions about the educational environment in the BCMP programme in the following ways: • Average DREEM score and average scores for five subscales (North-West + Gauteng). • DREEM score and subscale scores for both Gauteng and North-West student groups separately. • Mean scores for all 49 questions -average scores out of 49 and separately for two groups (individual indicators with high and low scores highlighted). • Total percentage scores: agreement (calculated by adding responses of students who'strongly agree' and 'agree' categories); uncertain (calculated by giving a percentage to all student responses which indicated uncertain/unsure); and disagreement (calculated by adding responses of students who'strongly disagree' and 'disagree'). Data were scored as per the scoring guide by Roff et al. 14 All 49 items are in the form of statements pertaining to the student's learning environment. Out of 49, the positive items (such as 'teaching helps to develop my confidence') are rated on a 5-point Likert scale scored as 4 for 'Strongly agree', 3 for 'Agree', 2 for 'Uncertain', 1 for 'Disagree' and 0 for 'Strongly disagree'. In addition to that, 9 out of 49 are negative items/worded negatively (such as 'I find the experience disappointing') and are reverse scored as 0 for 'Strongly agree', 1 for 'Agree', 2 for 'Uncertain', 3 for 'Disagree' and 4 for 'Strongly disagree'. 14 Apart from five subscale scores, all 49 items are given an individual mean score out of a maximum of 4 to 'pinpoint specific strengths and weaknesses'; 15 areas"'. 15 For example, a mean score of 3.5 for a positive statement ('teaching helps to develop my confidence') would imply that students positively perceive that teaching develops their confidence, whilst the same score on a negative item (e.g. 'teachers ridicule the students') implies that students disagree that teachers ridicule them. Similarly, a mean score of 2 or less for a positive item (e.g. 'teaching is well focused') implies that students do not perceive teaching as well focused, whilst the same score on a negative item (I find the experience disappointing) implies that students are not happy with the experience and find it disappointing. --- Ethical considerations This study received ethics clearance from the Human Research Ethics Committee, Faculty of Health Sciences, University of Witwatersrand, South Africa (Protocol M10802). --- Results Table 1 shows a summary of scores for both groups along with subscale scores, average DREEM scores and their inference. The average DREEM score for both groups was 130 out of 196, with individual scores of 129 and 131 for Gauteng and North-West students respectively. The subscale which was scored highest by both groups is 'academic selfperception' (76.6%), which inferred a feeling more on the positive side, followed by 'perception of atmosphere' (66.6%) inferring a more positive attitude. Table 1 further shows that the subscale with a lowest score is'social self-perception' (54.4%), inferring that this aspect of the programme is not too bad. After the reverse scoring of the 9 negative items (numbers 4, 8, 9, 17, 25, 35, 39, 48 and 49), mean scores of all 49 items for both groups separately and combined were calculated along with per cent agreement, percent uncertain and percent disagreement scores (presented in Table 2). In terms of per cent scores from both groups (indicated as bold in Table 2), 86% of students agree that they are encouraged to participate in teaching sessions, teaching helps to develop their competence (81%) and confidence (86%), the teachers practise a patient-centric approach (86%) and give clear examples during teaching (82%). Almost 90% or more students are confident of passing this year and agree that they have learnt a lot about empathy in their profession (96%), their problem-solving skills are being well developed (91%), they are able to ask the questions they want (96%), the atmosphere is relaxed during lectures (91%), and that much of what they have learnt seems relevant to a career in health care (96%). --- Insight into scores With regard to mean scores for individual items in the DREEM inventory, 5/49 items for the Gauteng group and 8/49 items for the North-West group were marked at 2 or less than 2 out of the maximum score of 4 (highlighted in Table 2 as bold). Most of these low scores are in'social self-perception' for both groups. In addition to this, 33/49 individual items for both groups in the DREEM inventory have been marked with an average mean score of between 2 and 3, suggesting that some aspects of the BCMP programme could be enhanced. Majority of these middle scores (falling between 2 and 3) are in'students' perception of teachers' and 'perception of atmosphere' for both groups. Lastly, Table 2 reveals that 3/49 individual items for both groups in the DREEM inventory have been marked with an average mean score of 3.5 and above out of the maximum score of 4. These are regarded as'real positive points' for the BCMP programme. All of these high scores are in 'academic selfperception'. --- Discussion The educational climate of an institution or course reflects the academic offering and contributes to development of students as practitioners. 16 Evaluation of the BCMP was done as an exercise to provide initial feedback on the programme. The results provide a profile of students' perceptions of the BCMP programme, which highlights its strengths and weaknesses. Whilst the BCMP programme is new and thus still undergoing modification, this evaluation indicates that it is positively perceived by students. The DREEM evaluation of the Wits BCMP programme shows an average score of 130/196 (with minor differences between North-West and Gauteng students), indicating a'more positive than negative' educational environment. The global DREEM score reported by medical schools in countries such as the United Kingdom (124/200), 15 India (123/200), 17 Sri Lanka (108/200), 18 Nigeria (118/200) 1 and Iran (100/200) 19 were lower than in our study. The differences could be attributed to the fact that our total DREEM score was based on a single cohort of Clinical Associate students, whilst other studies either administered the DREEM questionnaire to a large number of undergraduate students at different years of enrolment, 17 administered the DREEM questionnaire to all students in medical schools, 15 or conducted comparative cross-sectional studies in a number of medical schools. 20 There are other studies with a single cohort of students like our study which also had a lower DREEM score of 106/200 21 (Kuwait) and 118/200 (Malaysia). 22 The differences could also be attributed to the fact that the BCMP is a new programme which makes a learning environment exciting for Clinical Associates. Our DREEM score was similar to that found in a study conducted amongst medical schools in Nepal (130/200). 23 The results demonstrate a positive perception of learning by the students, which reflects that the BCMP offers a favourable learning environment (32/48). The highest individual scores in this section indicate that teaching helps to develop the students' confidence (3.2) and competence (3.1), which are important traits needed in their future professional settings. Students also believed that they were encouraged to participate in class (3.2), which is an indication of an open and interactive learning atmosphere. Contrary to our findings, a study conducted in Malaysia with secondyear medical sciences students (n = 67) found a mean score of 1.88, which indicates that teaching does not provide enough experiences to help them to have confidence. 22 One of the areas of concern perceived by students in the current study is that 'teaching over-emphasises factual learning' (45.5%). This finding is consistent with that of Arzumanet al. 22 We were unsure whether, whilst answering this question, students referred to class-based teaching (which includes factual learning) or hospital-based teaching (which is more of a hands-on experience). Hence due to lack of clarity the findings cannot be interpreted to draw a conclusion as to whether the BCMP programme stressed factual learning. Further exploration in this regard through qualitative research can provide a clear picture. Two more results from this section need to be addressed. Only 54.5% of students agreed that 'teaching is often stimulating', and only 61.8% agreed that 'teaching is student-centred', with mean scores of 2.4 and 2.5 respectively. The latter is important as the BCMP curriculum focuses on student-centred learning and problem-solving. More emphasis in the curriculum on promoting student-centred learning is suggested. The next area of the study demonstrates that educators in the BCMP programme are being perceived as moving in the right direction in terms of quality of their teaching. Students in the North-West group were more satisfied with their teachers than the Gauteng group. Best traits of teachers as perceived by students include providing constructive criticism, giving clear examples in class and not getting angry at students. This is encouraging for the BCMP programme, as the findings of a study conducted by Aghamolaei and Fazel 19 were that student perceptions of teachers were that they do not provide constructive criticism. Our study highlights two areas that need to be addressed for this section. All students perceived their 'learning atmosphere' in the category of 'a more positive attitude', with a slightly higher score for Gauteng students compared to North-West students. The best perceived aspect of the learning atmosphere is that students from both groups were able to ask questions they wanted, which potentially enhanced their communication skills with and learning from teachers. The atmosphere was relaxed during teaching, thereby promoting teacherstudent interaction and sharing of scientific and conceptual knowledge. Other studies also found that the overall learning atmosphere for the students was comfortable. 22 Areas of concern included perceptions by both student groups that the course was not well time-tabled (1.7), which may be explained by the fact that the BCMP was newly introduced and undergoing modifications. Clearly the results indicate an area for coordinators to address. Similarly, other studies found that students' perceptions of atmosphere were that the school/course was not well time-tabled. 19,21 North-West students perceived that the course was too stressful for them to enjoy, which could be attributed to their placements in distant areas far from family and friends. Although all of the students had different social environments at their places of clinical placements, their overall'social selfperception' does not stand as an indicator with good scores. Students from Gauteng perceived their social environment as 'not too bad', whilst those from North-West perceived it as 'not a nice place' -which could be attributed to distant location and being away from home for a long time. A positive perception was that students had good friends within the groups, indicating that fellow students acted as supporters. This is similar to the findings by Arzuman et al., 22 who found that students had a good social life, which was reflected by them having good friends on campus. Similar to other studies, 19 areas of concern include lack of good support systems for students who get stressed (mean score of 1.5, second lowest score), feeling lonely and lack of a good social life. It was unclear as to what type of support system they referred to (academic/social/personal). In order to improve the social aspect and ensure that apart from academics students also enjoy their social life, we need to further explore this aspect through qualitative research and take measures accordingly. Considering the nature of the BCMP, which is an innovative curriculum and combines both classroom-based learning as well as external hospital/clinic-based teaching, DREEM does have some limitations in this context. Terms such as 'course organisers/teachers', 'atmosphere' and 'learning/ teaching' present ambiguity in terms of whether the students perceived these for classroom-or hospital-based learning environments. In this regard we are considering designing instruments that are suited to evaluate the learning environment at different stages of the BCMP degree, so that they can capture students' perceptions without ambiguity. The venue used to administer the questionnaire might have influenced the students' responses about the programme. --- Conclusion The DREEM evaluation study offers a preliminary introspection into the learning environment of the BCMP programme. Despite the sample size, this study tried to evaluate the overall educational climate of the innovative BCMP curriculum at Wits. Results from this study demonstrated that the BCMP programme was perceived as a positive learning environment, contributing to the course objectives. This study highlighted strengths and weaknesses in the programme that can guide course organisers to design/modify the course. As this was the first cohort, future evaluations will need to be conducted periodically on a larger scale with an increased sample size and incorporating more variables, and a modified questionnaire better suited to the context of the BCMP learning environment, with an additional qualitative component for better exploration of students' perceptions. --- Competing interests The authors declare that they have no financial or personal relationship(s) that may have inappropriately influenced them in writing this article. --- Authors' contributions A.D. (University of the Witwatersrand) was the project manager at the Lehurutshe site, and participated in drafting

Background: An important determinant of a student's behaviour and performance is the school's teaching and learning environment. Evaluation of such an environment can explore methods to improve educational curricula and academic atmosphere.To evaluate the educational environment of the Bachelor of Clinical Medicine Practice programme as perceived by students at the University of the Witwatersrand, South Africa. Setting: This cross-sectional study was conducted with all final-year students (n = 25) enrolled in 2011, with a response rate of 88% (n = 22). Students were in two groups based in the Gauteng and North-West provinces. Methods: Data were collected using the Dundee Ready Educational Environmental Measure questionnaire, which was administered to all students. Total and mean scores for all questions were calculated for both groups.The learning environment was given an average score of 130/196 by the students. Individual subscales show that 'Academic self-perception' was rated the highest (25/32), whilst 'Social selfperception' had the lowest score (13/24). Positive aspects of the academic climate included: student competence and confidence development; student participation in class; constructive criticism provided; empathy in medical profession; and friendships created. Areas for improvement included: feedback provision to students; course time-tables; ensure non-stressful course; provision of good support systems for students; and social life improvement.Students' perceptions of their learning environment were 'more positive' than negative. Results from this study will be used to draw lessons for improving the curriculum and learning environment, improve administrative processes and develop student support mechanisms in order to improve their academic experience.

Introduction In recent years, many social websites that allow users to publish and share content online have appeared. For example Twitter, with 302 million active users and more than 500 millions posts every day, is one of the main actors of the market. These social media have become a very important source of data for many applications. Given a pre-defined need, two solutions are usually available for collecting useful data from such media: 1) getting access to huge repositories of historical data or 2) leveraging streaming services that most media propose to enable real-time tracking of their users' activity. As the former solution usually implies very important costs for retrieving relevant data from big data warehouses, the latter may constitute a very relevant alternative that enables real-time access to focused data. However, it implies to be able to efficiently target relevant streams of data. In this paper, we consider the case of user-centered streams of social data, i.e. where each individual stream fur-nishes access to the data published by a particular user, under restrictive constraints related to the number of streams that can be simultaneously considered. On Twitter for instance, data capture is limited to 5000 simultaneous streams1. Regarding the huge number of available users on this media, it requires an incredibly important effort for targeting relevant sources. Selecting relevant users to follow among the whole set of accounts is very challenging. Imagine someone interested in politics wishing to keep track of people in relation to this topic in order to capture the data they produce. One solution is to manually select a bucket of accounts and follow their activity along time. However, Twitter is known to be extremely dynamic and for any reason some users might start posting on this topic while others might stop at any time. So if the interested user wants to be up to date, he might change the subset of followed accounts dynamically, his goal being to anticipate which users are the most likely to produce relevant content in a close future. This task seems hard to handle manually for two major reasons. First, the criteria chosen to predict whether an account will potentially be interesting might be hard to define. Secondly, even if one would be able to manually define those criteria, the amount of data to analyze would be too large. Regarding these important issues, we propose a solution that, at each iteration of the process, automatically selects a subset of accounts that are likely to be relevant in the next time window, depending on their current activity. Those accounts are then followed during a certain time and the corresponding published contents are evaluated to quantify their relevance. The algorithm behind the system then learns a policy to improve the selection at the next time step. We tackle this task as a contextual bandit problem, in which at each round, a learner chooses an action among a bigger set of available ones, based on the observation of action features -also called context -and then receives a reward that quantifies the quality of the chosen action. In our case, considering that following someone corresponds to an action, several actions have to be chosen at each time step, since we wish to leverage the whole capture capacity allowed by the streaming API and therefore collect data from several simultaneous streams. With the current activity of a certain user corresponding to its context features, our task suits well the contextual bandit framework. However, in its traditional instance, features of every action are required to be observed at each iteration to perform successive choices. For our specific task, we are not able to observe the current activity of every potential user. Consequently, we are not able to get everyone's context and directly use classical contextual bandit algorithms such as LinUCB (Chu et al. 2011). On the other hand, it cannot be treated as a traditional (non-contextual) bandit problem since we would lose the useful information provided by observed context features. To the best of our knowledge, such an hybrid instance of bandit problem has not been studied yet. To solve our problem, we propose a data capture algorithm which, along with learning a selection policy of the k best streams at each round, learns on the feature distributions themselves, so as to be able to make approximations when features are hidden from the agent. The paper is organized as follows: Section 2 presents some related work. Section 3 formalizes the task and presents our data capture system. Section 4 then describes the model and the algorithm proposed to solve the task. Finally, section 5 reports various sets of experimental results. --- Related Work In this section we first present some literature related to bandit learning problems, then we discuss about some data capture related tasks and finally present some applications of bandits for social media tasks. The multi-armed bandit learning problem, originally studied in (Lai and Robbins 1985) in its stationary form has been widely investigated in the literature. In this first instance, the agent has no access to side information on actions and assume stationary reward distributions. A huge variety of methods have been proposed to design efficient selection policies, with corresponding theoretical guarantees. The famous Upper Confidence Bound (UCB) algorithm proposed in (Auer, Cesa-Bianchi, and Fischer 2002) and other UCBbased algorithms ( (Audibert, Munos, and Szepesvari 2007;Audibert and Bubeck 2009)) have already proven to solve the so-called stochastic bandit problem. This type of strategies keeps an estimate of the confidence interval related to each reward distribution and plays the action with highest upper confidence bound at each time step. In (Auer, Cesa-Bianchi, and Fischer 2002), the authors introduce the contextual bandit problem, where the learner observes some features for every action before choosing the one to play, and propose the LinRel algorithm. Those features are used to better predict the expected rewards related to each action. More recently, the LinUCB algorithm, which improves the performance of LinRel has been formalized (Chu et al. 2011). Those two algorithms assume the expected reward of an action to be linear with respect to some unknown parameters of the problem to be estimated. In (Kaufmann, Cappe, and Garivier 2012), the authors propose the Bayes-UCB algorithm which unifies several variants of UCB algorithms. For both the stochastic and contextual bandit case, Thompson sampling algorithms, which introduce randomness on the exploration by sampling action parameters from their posterior distributions, have been designed (Kaufmann, Korda, and Munos 2012;Agrawal and Goyal 2012a;Chapelle and Li 2011;Agrawal and Goyal 2012b). More recently, the case where the learner can play several actions simultaneously has been formalized respectively in (Chen, Wang, and Yuan 2013) (CUCB algorithm) and (Qin, Chen, and Zhu 2014) (C 2 UCB algorithm) for the non contextual and the contextual case. Finally in (Gisselbrecht et al. 2015) the authors propose the CUCBV algorithm which extends the original UCBV of (Audibert, Munos, and Szepesvari 2007) to the multiple plays case. To the best of our knowledge, no algorithm exists for our case, i.e. when feature vectors are only observable with some given probability. Regarding social media, several existing tasks present similarities with ours. In (Li, Wang, and Chang 2013), the authors build a platform called ATM that is aimed at automatically monitoring target tweets from the Twitter Sample stream for any given topic. In their work, they develop a keyword selection algorithm that efficiently selects keywords to cover target tweets. In our case, we do not focus on keywords but on user profiles by modeling their past activity on the social media. In (Colbaugh and Glass 2011) the authors model the blogosphere as a network in which every node is considered as a stream that can be followed. Their goal is to identify blogs that contain relevant contents in order to track emerging topics. However, their approach is static and their model is trained on previously collected data. Consequently, this approach could not be applied to our task: first we cannot have access to the whole network due to APIs restrictions and secondly, the static aspect of their methods is not suitable to model the dynamics of users' behaviors that can change very quickly. In (Hannon, Bennett, and Smyth 2010) and (Gupta et al. 2013), the authors build recommendation systems, respectively called Twittomender and Who to Follow. In the former collaborative filtering methods are used to find twitter accounts that are likely to interest a target user. In the latter, a circle of trust (which is the result of an egocentric random walk similar to personalized PageRank (Fogaras et al. 2005)) approach is adopted. In both cases, authors assume the knowledge of the whole followers/followees Twitter graph, which is not feasible for large scale applications due to Twitter restriction policies. Moreover, the task is quite different from ours. While these two models are concerned about the precision of the messages collected from the selected followees (in order to avoid relevant information to be drown in a too large amount of data), we rather seek at maximizing the amount of relevant data collected in a perspective of automatic data capture. Bandit algorithms have already been applied to various tasks related to social networks. For example in (Kohli, Salek, and Stoddard 2013), the authors handle abandonment minimization tasks in recommendation systems with bandits. In (Buccapatnam, Eryilmaz, and Shroff 2014), bandits are used for online advertising while in (Lage et al. 2013), the authors use contextual bandit for an audience maximization task. In (Gisselbrecht et al. 2015), a data capture task is tackled via non-contextual bandit algorithms. In that case, each user is assumed to own a stationary distribution and the process is supposed to find users with best means by trading off between the exploitation of already known good users to follow and the exploration of unknown ones. The proposed approach mainly differs from this work for the nonstationnarity assumptions it relies on. This work is considered in the experiments to highlight the performances of the proposed dynamic approach. 3 Data Streams Selection --- Context As we described in the introduction, our goal is to propose a system aimed at capturing some relevant data from streaming APIs proposed by most of social media. Face to such APIs, two different types of data sources, users or keywords (or a mixture of both kinds), can be investigated. In this paper, we focus on users-centered streams. Users being the producers of the data, following them enables more targeted data capture than keywords. Following keywords to obtain data about a given topic for example would be likely to lead to the collection of a greatly too huge amount of data2, which would imply very important post-filtering costs. Moreover, considering user-centered streams allows a larger variety of applications, with author-related aspects, and more complex topical models, than focusing on keywords-centered streams. Given that APIs usually limit the ability to simultaneously follow users' activity to a restricted number of accounts, the aim is to dynamically select users that are likely to publish relevant content with regard to a predefined need. Note that the proposed approach would also be useful in the absence of such restriction, due to the tremendous amount of data that users publish on main social media. The major difficulty is then to efficiently select user accounts to follow given the huge amount of users that post content and that at the beginning of the process, no prior information is known about potential sources of data. Moreover, even if some specific accounts to follow could be found manually, adapting the capture to the dynamics of the media appears intractable by hand. For all these reasons, building an automatic solution to orient the user's choices appears more than useful. To make things concrete, in the rest of the paper we set up in the case of Twitter. However, the proposed generic approach remains available for any social media providing real-time access to its users' activity. --- A Constrained Decision Process As described above, our problem comes down to select, at each iteration t of the process, a subset K t of k user accounts to follow, among the whole set of possible users K (K t <unk> K), according to their likelihood of posting relevant tweets for the formulated information need. Given a relevance score r a,t assigned to the content posted by user a <unk> K t during iteration t of the process (the set of tweets he posted during iteration t), the aim is then to select at each iteration t over T the set of user accounts that maximizes the sum of collected relevance scores: max (Kt)t=1..T T t=1 a<unk>Kt r a,t(1) Note that, in our task, we are thus only focused on getting the maximal amount of relevant data, the precision of the retrieved data is not of our matter here. This greatly differs from usual tasks in information retrieval and followees recommendation for instances. Relevance scores considered in our data capture process depend on the information need of the user of the system. The data need in question can take various forms. For example, one might want to follow the activity of users that are active on a predefined topic or influent in the sense that their messages are reposted a lot. They can depend on a predefined function that automatically assigns a score according to the matching of the content to some requirements (see section 5 for details). Whereas (Gisselbrecht et al. 2015) relies on stationarity assumptions on the relevance score distributions of users for their task of focused data capture (i.e., relevance scores of each user account are assumed to be distributed around a given stationary mean along the whole process), we claim that this is a few realistic setting in the case of large social networks and that it is possible to better predict future relevance scores of users according to their current activity. In other words, with the activity of a user a <unk> K at iteration t -1 represented by a d-dimensional real vector z a,t, there exists a function h : R d <unk> R that explains from z a,t the relevance score r a,t that a user a would obtain if it was followed during iteration t. This correlation function needs to be learned conjointly with the iterative selection process. However, in our case, maximizing the relevance scores as defined in formula 1 is constrained in some different ways: • Relevance scores r a,t are only defined for followed accounts during iteration t (i.e. for every a <unk> K t ), others are unknown; • Context vectors z a,t are only observed for a subset of users O t (i.e., it is not possible to observe the whole activity of the social network). Constraints on context vectors are due to API restrictions. Two streaming APIs of Twitter are used to collect data: • On the one hand, a Sample streaming API furnishes realtime access to 1% of all public tweets. We leverage this API to discover unknown users and to get an important amount of contexts vectors for some active users; • On the other hand, a Follow streaming API provides realtime data published by a subset of the whole set of Twitter users. This API allows the system to specify a maximal number of 5000 users to follow. We leverage this API to capture focused data from selected users. For users in O t, their relevance score at next iteration can be estimated via the correlation function h, which allows to capture variations of users' usefulness. For others however, we propose to consider the stationary case, by assuming some general tendency on users' usefulness. Therefore, we get an hybrid problem, where observed contexts can be used to explain variations from estimated utility means. Figure 1 depicts our system of focused data capture from social media. The process follows the same steps at each iteration, three time steps are represented on the figure. At the beginning of each iteration, the selection policy selects a set of users to follow (K t ) among the pool of known users K, according to observations and some knowledge provided by a learner module. Then, the messages posted by the k selected users are collected via the Follow streaming API. As we can see in the central part of the illustration, after having followed users in K t during the current iteration t, the collected messages are analyzed (by a human or an automatic classifier) to give feedbacks (i.e., relevance scores) to the learner. --- System Description Meanwhile, as a parallel task during each iteration, current activity features are captured from the Sample streaming API. This allows us to feed the pool of potential users to follow K and to build the set O t of users with observed contexts. Context vectors for iteration t are built by considering all messages that the system collected from the Follow streaming and the Sample streaming APIs for each user during this iteration. Therefore, every user whose at least one message was included in these collected messages at iteration t is included in O t+1 (messages from a same author are concatenated to form z a,t+1, see section 5 for an example of context construction from messages). Contexts vectors collected during iteration t serve as input for the selection strategy at iteration t + 1. --- Model and Algorithm This section first introduces some settings and backgrounds for our contextual bandit approach for targeted data capture, details the policy proposed to efficiently select useful users to follow, and then describes the general algorithm of our system. --- Problem setting As defined above, we denote by K the set of K known users that can be follow at each iteration of the data capture pro-cess. At each round t <unk> <unk>1,.., T <unk> of the process, feature vectors z a,t <unk> R d can be associated with users a <unk> K. The set of k users selected to be followed during iteration t is denoted by K t. The reward obtained by following a given user a during iteration t of the process is denoted by r a,t. Note that only rewards r a,t from users a <unk> K t are known, others cannot be observed. In the field of contextual bandit problems, the usual linear hypothesis assumes the existence of an unknown vector (Agrawal and Goyal 2012b)). Here, in order to model the intrinsic quality of every available action (where an action corresponds in our case to the selection of a user to follow during the current iteration), we also suppose the existence of a bias term <unk> a <unk> R for all actions a, such that finally: <unk> <unk> R d such that <unk>t <unk> <unk>1,.., T <unk>, <unk>a <unk> K : E[r a,t |z a,t ] = z T a,t <unk> (see <unk>t <unk> <unk>1,.., T <unk>, <unk>a <unk> K : E[r a,t |z a,t ] = z T a,t <unk> + <unk> a (2) This formulation corresponds to a particular case of the Lin-UCB algorithm with the hybrid linear reward proposed in (Li et al. 2010) by taking a constant arm specific feature equal to 1 at every round. Note that individual parameters could have also been considered in our case but this is not well suited for the types of problems of our concern in this paper, where the number of available actions at each round is usually high (which would imply a difficult learning of individual parameters). For the sake of simplicity, we therefore restrict this individual modeling to this bias term. In our case, a major difference with existing works on contextual bandit problems is that every context is not accessible to the selection policy before it chooses users to follow. Here, we rather assume that every user a from the social media owns a probability p a (0 <unk> p a <unk> 1) to reveal its context3. The set of users for which the process observes the features at time t is denoted O t, whose complement is denoted <unk>t. Then, since K t <unk> K and K = O t <unk> <unk>t, selected users can then belong to the set of users without observed context <unk>t. Finally, as told above, while our approach would remain valid in a classical bandit setting where only one action is chosen at each step, our task requires to consider the case where multiple users can be selected to be followed at each iteration (since we wish to exploit the whole capture capacity allowed by the API). At each round, the agent has then to choose k (k <unk> K) users among the K available ones, according to observed features and individual knowledge about them. The reward obtained after a period of capture then corresponds to the sum of individual rewards collected from the k users followed during this period. --- Distribution assumptions To derive our algorithm, we first perform a maximum a posteriori estimate for the case where every context are available, based on the following assumptions: • Likelihood: reward scores are identically and independently distributed w.r.t. observed contexts: r a,t <unk> N (z T a,t <unk> + <unk> a, <unk> a ), with <unk> a the variance of the difference between the reward r a,t and the linear application z T a,t <unk> + <unk> a ; • Prior: the unknown parameters are normally distributed: <unk> <unk> N (0, bI d ) and <unk> a <unk> N (0, <unk> a ), where b and <unk> a are two values allowing to control the variance of the parameters and I d is the identity matrix of size d (the size of the features vectors). For the sake of clarity, we set b, <unk> a and <unk> a equal to 1 for all a in the following. Note that every results can be extended to more complex cases. Proposition 1 Denoting T a the set of steps when user a has been chosen in the first n time steps of the process (T a = <unk>t <unk> n, a <unk> K t <unk>, |T a | = <unk> a ), c a the vector containing rewards obtained by a at iterations it has been followed (c a = (r a,t ) t<unk>Ta ) and D a the context matrix related to user a (D a = (z T a,t ) t<unk>Ta ), the posterior distribution of the unknown parameters after n time steps, when all contexts are available, follows: <unk> <unk> N <unk>, A -1 0 (3) <unk> a + zT a <unk> <unk> N <unk>a, 1 <unk> a + 1 (4) With: A 0 = I d + K a=1 (<unk> a + 1) <unk>a b T 0 = K a=1 (<unk> a + 1) <unk>a <unk>a = D T a D a <unk> a + 1 -za zT a <unk>a = c T a D a <unk> a + 1 -<unk>a zT a <unk> = A -1 0 b T 0 <unk>a = t<unk>Ta r a,t <unk> a + 1 za = t<unk>Ta z a,t <unk> a + 1 Proof 1 The full derivation is available at4. It proceeds as fol- lows: denoting D = ((ra 1,1, za 1,1),...(ra n,n, za n,n)) and using Bayes's rule we have: p(<unk>, <unk>1...<unk>K |D) <unk> p(D|<unk>, <unk>1...<unk>K )p(<unk>) K a=1 p(<unk>a) <unk> e -1 2 n t=1 (r a t,t -z T a t,t <unk>-<unk>a ) 2 <unk>a + <unk> T <unk> b + K a=1 <unk> 2 a <unk>a. Rearranging the terms and using matrix notations leads to: p(<unk>, <unk>1...<unk>K |D) <unk> e -1 2 <unk> T A 0 <unk>-2b T 0 <unk>+ K a=1 (<unk>a+1)(<unk>a+z T a <unk>-<unk>a) 2. The two distributions are directely derived from this formula. All the parameters above do not have to be stored and can be updated efficiently as new learning example comes (see algorithm 1). Theorem 1 For any 0 <unk> <unk> <unk> 1 and z a,t <unk> R d, denoting <unk> = <unk> 2erf -1 (1<unk>) 5, for every action a after t iterations: P |E[r a,t |z a,t ]-<unk>a -(z a,t -za ) T <unk>| <unk> <unk> <unk> a <unk> (1-<unk>) with <unk> a = 1 <unk> a + 1 + (z a,t -za ) T A -1 0 (z a,t -za )(5) Proof 2 E[ra,t|za,t] = (za,t -za) T <unk> + <unk>a + zT a <unk>. By combining equations 2, 3 and 4: E[ra,t|za,t] <unk> N <unk>a + (za,tza) T <unk>, 1 <unk>a+1 + (za,tza) T A -1 0 (za,tza). The announced result comes from the confidence interval of a Gaussian. This formula is directly used to find the so-called Upper Confidence Bound, which leads to choose the k users having the highest score values s a,t at round t, with: s a,t = <unk>a + (z a,t -za ) T <unk> + <unk> <unk> a (6) We recall that this formula can be used in the traditional contextual bandit problem, where every context is available. In the next section, we propose a method to adapt it to our specific setting of data capture, where most of contexts are usually hidden from the agent. --- Dealing with Hidden Contexts Selection scores derived in the previous section are defined for cases where all context vectors are available for the agent at each round. However, in our case, only users belonging to the subset O t show their context. This particular setting requires to decide how judging users whose context is unknown. Moreover, it also implies questions about knowledge updates when some users are selected without having been (contextually meaning) observed. Even though it is tempting to think that the common parameter <unk> and the user specific ones <unk> a can be learned independently, in reality they are complexly correlated as formula 3 and 4 highlight it. It is important to emphasize that to keep probabilistic guarantees, parameters should only be updated when a chosen user was also observed, i.e. when it belongs to K t <unk>O t. So, replacing the previous T a by T both a = <unk>t <unk> n, a <unk> K t <unk> O t <unk> (we keep the notation |T both a | = <unk> a ) allows us to re-use the updates formula of Proposition 1. However, computing the selection score for a user whose context is hidden from the agent cannot be done via equation 6 since z a,t is unknown. To deal with such a case we propose to use an estimate of the mean distribution of feature vectors of each user. The assumption is that, while non-stationarity can be captured when contexts are available, different users own different mean reward distributions, and that it can be useful to identify globally useful users. New notations: • We denote the set of steps when user a revealed its context vector after n steps by T obs a = <unk>t <unk> n, a <unk> O t <unk> with |T obs a | = n a. 5 erf -1 is the inverse error function, erf (x) = 2/<unk> x 0 e -t 2 dx. • The empirical mean of the feature vector for user a is denoted <unk>a, with <unk>a = 1 n a t<unk>T obs a z a,t. Note that <unk>a is different from za since the first is updated every time the context z a,t is observed while the second is only updated when user a is observed and played in the same iteration. Main assumptions: Note that the requirement ||<unk>|| <unk> M can be satisfied through proper rescaling on <unk>. Proposition 2 The excepted value of the mean reward of user a with respect to its features distribution, denoted E[r a ] satisfies after t iterations of the process: E[r a ] = E[z T a,t <unk> + <unk> a ] (7) = E[z a ] T <unk> + <unk> a = ( <unk>a -za ) T <unk> + <unk> a + za T <unk> + (E[z a ] -<unk>a ) T <unk> Theorem 2 Given 0 <unk> <unk> <unk> 1 and 0 <unk> <unk> <unk> 1 2, denoting <unk> = <unk> 2erf -1 (1 -<unk>), for every user a after t iterations of the process: P |E[r a ] -<unk>a -( <unk>a -za ) T <unk>| <unk> <unk> <unk>a + 1 n <unk> a <unk> (1 -<unk>) 1 - C a n 1-2<unk> a with <unk>a = 1 <unk> a + 1 + ( <unk>a -za ) T A -1 0 ( <unk>a -za ) (8) Where C a is a positive constant specific to each user. Proof 3 The full proof is available at URL 4, we only give the main steps here. Denoting ma = ( <unk>aza) T <unk> + <unk>a, <unk>2 a = 1 <unk>a+1 +( <unk>aza) T A -1 0 ( <unk>aza), Xa the random variable such that Xa = ( <unk>aza) T <unk>+<unk>a+ za T <unk> and using Cauchy Schwarz inequality, we have: |E[ra]-ma| <unk>a <unk> |Xa-ma| <unk>a + ||E[za]-<unk>a||M <unk>a Then, using the Gaussian property of Xa, we show that: P |Xa-ma| <unk>a <unk> <unk> = 1 -<unk>, with <unk> = <unk> 2erf -1 (1 -<unk>). On the other hand, with Chebyshev inequality, we prove that: P ||E[za]-<unk>a||M <unk>a <unk> 1 <unk>an <unk> a <unk> 1 -M 2 d n 1-2<unk> a T race(<unk>a) Finally, combining the two previous results proves the theorem. For <unk> <unk> 1 2, the previous probability tends to 1<unk> as the number of observations of user a increases (as in equation 5). Then, the inequality above gives a reasonably tight UCB for the expected payoff of user a, from which a UCB-type user-selection strategy can be derived. At each trial t, if the context of user a has not been observed (i.e. a / <unk> O t ), set: s a,t = <unk>a + ( <unk>a -za ) T <unk> + <unk> <unk>a + 1 n <unk> a (9) Given that <unk> > 0, and that each user a has a probability 0 <unk> p a <unk> 1 to reveal its context at each time step t, the extra-exploration term 1/n <unk> a tends to 0 as the number of observations of user a gets larger. The above score then tends to a classical LinUCB score (as in equation 6) in which the context vector z a,t is replaced by its empirical mean <unk>a, as t increases. --- Contextual Data Capture Algorithm The pseudo-code of our contextual data capture algorithm from social streams is detailled in algorithm 1. The algorithm starts by initializing the set of observed users O t from the Sample API during a time period of L. Then, for each iteration t over T, it proceeds as follows: 1. Lines 10 to 15: every user from O t that does not belong to the pool of potential users to follow is added to K, in the limit of newM ax new arms at each iteration to cope with cases where too many new arms are discovered each iteration (in that case only the newM ax first new ones are considered, others are simply ignored). This allows to avoid over-exploration for such cases; 2. Lines 16 to 20: updates of empirical context means <unk>a (according to data collected from user a at the previous iteration) and observation counts n a for every observed user a in O t ; 3. Line 22: parameters <unk> of the estimation model are updated according to current counts; 4. Lines 23 to 37: selection scores are computed for every user in the pool K. As detailed in the previous section, selection scores are computed differently depending on the availability of the current context of the corresponding user. Note also that the selection scores of new users added to the pool during the current iteration are set to +<unk> in order to enforce the system to choose them at least once and initialize their empirical mean reward; --- Experiments Classical bandit algorithms usually come with an upper bound of the regret, which corresponds to the difference between the cumulative reward obtained with an optimal policy and the bandit policy in question. In our case, an optimal strategy corresponds to a policy for which the agent would have a perfect knowledge of the parameters <unk> and <unk> a (for all a) plus an access to all the features at each step. However, in our case, in the absence of an asymptotically exact estimate of z T a,t <unk> + <unk> a, due to the fact that some contexts z a,t are hidden from the process, it is unfortunately not possible to show a sub-linear upper-bound of the regret. Nevertheless, we claim that the algorithm we proposed to solve our problem of contextual data capture from social media streams well behaves on average. To demonstrate this good behavior, we present various offline and online experiments in real-world scenarios. Our algorithm is the first to tackle the case of bandits with partially hidden contexts, the aim is to demonstrate its feasibility for constrained tasks such as data capture from social media streams. --- Experimental Setup Beyond a random policy that randomly selects the set of users K t at each iteration, we compare our algorithm to two bandit policies well fitted for being applied for the task of data capture: CUCB and CUCBV respectively proposed in (Qin, Chen, and Zhu 2014) and (Gisselbrecht et al. 2015). These algorithms do not take features into account and perform stationary assumptions on reward distributions of users. Compared to CUCB, CUCBV adds the variance of reward distributions of users in the definition of the confidence intervals. This

Social Media usually provide streaming data access that enable dynamic capture of the social activity of their users. Leveraging such APIs for collecting social data that satisfy a given pre-defined need may constitute a complex task, that implies careful stream selections. With user-centered streams, it indeed comes down to the problem of choosing which users to follow in order to maximize the utility of the collected data w.r.t. the need. On large social media, this represents a very challenging task due to the huge number of potential targets and restricted access to the data. Because of the intrinsic nonstationarity of user's behavior, a relevant target today might be irrelevant tomorrow, which represents a major difficulty to apprehend. In this paper, we propose a new approach that anticipates which profiles are likely to publish relevant contents -given a predefined need -in the future, and dynamically selects a subset of accounts to follow at each iteration. Our method has the advantage to take into account both API restrictions and the dynamics of users' behaviors. We formalize the task as a contextual bandit problem with multiple actions selection. We finally conduct experiments on Twitter, which demonstrate the empirical effectiveness of our approach in real-world settings.

the fact that some contexts z a,t are hidden from the process, it is unfortunately not possible to show a sub-linear upper-bound of the regret. Nevertheless, we claim that the algorithm we proposed to solve our problem of contextual data capture from social media streams well behaves on average. To demonstrate this good behavior, we present various offline and online experiments in real-world scenarios. Our algorithm is the first to tackle the case of bandits with partially hidden contexts, the aim is to demonstrate its feasibility for constrained tasks such as data capture from social media streams. --- Experimental Setup Beyond a random policy that randomly selects the set of users K t at each iteration, we compare our algorithm to two bandit policies well fitted for being applied for the task of data capture: CUCB and CUCBV respectively proposed in (Qin, Chen, and Zhu 2014) and (Gisselbrecht et al. 2015). These algorithms do not take features into account and perform stationary assumptions on reward distributions of users. Compared to CUCB, CUCBV adds the variance of reward distributions of users in the definition of the confidence intervals. This has been shown to behave well in cases such as our task of data capture, where a great variability can be observed due to the possible inactivity of users when they are selected by the process. At last, we consider a naive version of our contextual algorithm that preferably selects users with an observed context (it only considers users in O t when this set contains enough users), rather than getting the ability of choosing users in <unk>t as it is the case in our proposal. For all of the reported experiments we set: 1) The exploration parameter to <unk> = 1.96, which corresponds to a 95 percent confidence interval on the estimate of the expected reward when contexts are observed; 2) The reduction parameter of confidence interval for unknown contexts to <unk> = 0.25, which allows a good trade-off between the confidence interval reduction rate and the probability of this interval; 3) The number of new users that can be added to the pool at each iteration to newM ax = 500, to avoid over-exploration, especially for online experiments, when many new users can be discovered from the Sample API; --- Offline experiments Datasets In order to be able to test different policies and simulate a real time decision process several times, we first propose a set of experiments on offline datasets: • USElections: dataset containing a total of 2148651 messages produced by 5000 users during the ten days preceding the US presidential elections in 2012. The 5000 cho-sen accounts are the first ones who used either "Obama", "Romney" or "#USElections". • Libya: dataset containing 1211475 messages from 17341 users. It is the result of a three-months capture from the F ollow API using the keyword "Libya". In the next paragraph, we describe both how we transform messages in feature vectors and which reward function we use to evaluate the quality of a message. --- Model definition Context model The content obtained by capturing data from a given user a at time step t is denoted <unk> a,t (if we get several messages for a given author, these messages are concatenated). Given a dictionary of size m, messages can be represented as m dimensional bag of word vectors. However the size m might cause the algorithm to be computationally inefficient since it requires the inversion of a matrix of size m at each iteration. In order to reduce the dimension of those features, we used a Latent Dirichlet Allocation method (Blei, Ng, and Jordan 2003), which aims at modeling each message as a mixture of topics. However, due to the short size of messages, that standard LDA may not work well on Twitter (Weng et al. 2010). To overcome this difficulty, we choose the approach proposed in (Hong and Davison 2010), which aggregates tweets of a same user in one document. We choose a number of d = 30 topics and learn the model on the whole corpus. Then, if we denote by F : R m -<unk> R d the function that, given a message returns its representation in the topic space, the features of user a at time t is z a,t = F (<unk> a,t-1 ). Reward model We trained a SVM topic classifier on the 20 Newsgroups dataset in order to rate each content. For our experiments, we focus on 4 classes to test: politics, religion, sport, science. We propose to consider the reward associated to some content as the number of times it has been retweeted (re-posted on Tweeter) by other users if it belongs to the specified class according to our classifier, or 0 otherwise. Finally, if a user posted several messages during an iteration, his reward r a,t corresponds to the sum of the individual rewards obtained by the messages he posted during iteration t. This instance of reward function corresponds to a task of seeking to collect messages, related to some desired topic, that will have a strong impact on the network (note that we cannot directly get high-degree nodes in the user graph due to API restrictions). Experimental Settings In order to obtain generalizable results, we test different values for parameters p, the probability for every context to be observed, k, the number of users that can be followed simultaneously, and L, which stands for the duration of an iteration. More precisely, we experienced every possible combination with p <unk> <unk>0.1, 0, 5, 1.0<unk>, k <unk> <unk>50, 100, 150<unk> and L <unk> <unk>2min, 3min, 6min, 10min<unk>. Results For space reasons and brevity, we only report results for k = 100, L = 2min (which means that every 2 minutes the algorithms select 100 users to follow) for the USElections dataset and the four rewards defined above. For the Libya dataset, we only show the results for the politics reward. Similar tendencies were observed for different values of k and L. Figures 2 and 3 represent the evolution of cumulative reward for different policies and rewards, respectively for the USElections and the Libya dataset. For contextual algorithms, their naive versions that select observed users in priority are given with same symbols without solid line. First, it should be noticed that every policy performs better than the Random one, which is a first element to assert the relevance of bandit algorithms for the task in concern. Second, we notice that CUCBV performs better than CUCB, which confirms the results obtained in (Gisselbrecht et al. 2015) on a similar task of focused data capture. More interesting is the fact that when every context is observable, our contextual algorithm performs better than stationary approaches CUCB and CUCBV. This result shows that we are able to better anticipate which user is going to be the more relevant at the next time step, for a particular information need, given what he said right before. This also confirms the usual non-stationarity behavior of users. For instance, users can talk about science during the day at work while being more focused on sports when they come back home. Considering contexts also allows one to converge faster towards interesting users since every user account share the same <unk> parameter. Results show that even for low probabilities of context observation p, our contextual policy behaves greatly better than non-contextual approaches, which empirically validates our approach: it is possible to leverage contextual information even if a large part of this information is hidden from the system. In particular, for the Libya dataset where no significant difference between the two non-contextual CUCB and CUCBV is noteworthy, our algorithm seems much more appropriate. Moreover, for a fixed probability p, the naive versions of the contextual algorithm offer lower performances than the original ones. By giving the ability of selecting users even if their context is unknown, we allow the algorithm to complete its selection by choosing users whose average context corresponds to a learned profile. If no user in O t seems currently relevant, the algorithm can then rely on users with good averaged intrinsic quality. For k = 100, the average number of selected users for which the context was observed at each time step is 43 for p = 0.1 and 58 for p = 0.5, which confirms that the algorithm does not always select users in O t. --- Online experiments Experimental Settings We propose to consider an application of our approach on a real-world online Twitter scenario, which considers both the whole social network and the API constraints. For these experiments, we use the full potential of Twitter APIs, namely 1% of all public tweets for the Sample API that runs in background, and 5000 users simultaneously followed (i.e., k = 5000) for the Follow API. At each iteration, the selected users are followed during L = 15minutes. The messages posted by these users during this interval are evaluated with the same relevance function as in the offline experiments, using the politics topic. We also use an LDA transformation to compute the context vectors. Given Twitter policy, every experiment requires one Twitter Developer account, which limits the number of strategies we can experiment. We chose to test the four following ones: our contextual approach, CUCBV, Random and a another one called Static. This Static policy follows the same 5000 accounts at each round of the process. Those 5000 accounts have been chosen by collecting all tweets provided during 24 hours by the Sample streaming API, evaluating them and finally taking the 5000 users with the greatest cumulative reward. For information, some famous accounts such that @dailytelegraph, @Independent or @CNBC were part of them. Results Figure 4 on the left represents the evolution of the cumulative reward of a two-weeks-long run for the four tested policies. From these curves, we note the very good behavior of our algorithm on a real-world scenario, as the amount of rewards it accumulates grows greatly much more quickly than other policies, especially after the 500 first iterations of the process. After these first iterations, our algorithm appears to have acquired a good knowledge on the reward distributions, w.r.t. observed contexts and over the various users of the network. In order to analyze the behavior of the experimented policies during the first iterations of the capture, we also plot on the right of the figure a zoomed version of the same curves on the first 150 time steps. At the beginning of the process, the Static policy performs better than every others, which can be explained by two reasons: 1) bandit policies need to select every user at least once in order to initialize their scores and 2) users who are part of the Static policy's pool are supposed to be relatively good targets considering the way we chose them. Around iteration 80, both CUCBV and our Contextual algorithm become better than Static policy, which corresponds to the moment they start to trade off between exploration and exploitation. Then, after a period of around 60 iterations (approximately between the 80 th and the 140 th time step) where CUCBV and our Contextual approach behave similarly, the latter then allows the capture process to collect greatly more valuable data for the specified need. This is explained by the fact that our Contextual algorithm requires a given amount of iterations to learn the correlation function between contexts and rewards and then taking some advantage of it. Note the significant changes in the slope of the cumulative reward curve of our Contextual algorithm, which highlight the ability of the algorithm to be reactive to the environment changes. Finally, the number of times every user has been selected by our Contextual algorithm is more spread than with a stationary algorithm such as CUCBV, which confirms the relevance of our dynamic approach. To conclude, in both offline and online settings, our approach proves very efficient at dynamically selecting user streams to follow for focused data capture tasks. --- Conclusion In this paper, we tackled the problem of capturing relevant data from social media streams delivered by specific users of the network. The goal is to dynamically select useful userstreams to follow at each iteration of the capture, with the aim of maximizing some reward function depending on the utility of the collected data w.r.t. the specified data need. We formalized this task as a specific instance of the contextual bandit problem, where instead of observing every feature vectors at each iteration of the process, each of them has a certain probability to be revealed to the learner. For solving this task, we proposed an adaptation of the popular con- textual bandit algorithm LinU CB to the case where some contexts are hidden at each iteration. Although not any sublinear upper-bounds of the regret can be guaranteed, because of the great uncertainty induced by the hidden contexts, our algorithm behaves well even when the proportion of hidden contexts is high, thanks to some assumptions made on the hidden contexts and a well fitted exploration term. This corresponds to an hybrid stationary / non stationary bandit algorithm. Experiments show the very good performances of our proposal to collect relevant data under real-world capture scenarios. It opens the opportunity for defining new kinds of online intelligent strategies for collecting and tracking focused data from social media streams.

Social Media usually provide streaming data access that enable dynamic capture of the social activity of their users. Leveraging such APIs for collecting social data that satisfy a given pre-defined need may constitute a complex task, that implies careful stream selections. With user-centered streams, it indeed comes down to the problem of choosing which users to follow in order to maximize the utility of the collected data w.r.t. the need. On large social media, this represents a very challenging task due to the huge number of potential targets and restricted access to the data. Because of the intrinsic nonstationarity of user's behavior, a relevant target today might be irrelevant tomorrow, which represents a major difficulty to apprehend. In this paper, we propose a new approach that anticipates which profiles are likely to publish relevant contents -given a predefined need -in the future, and dynamically selects a subset of accounts to follow at each iteration. Our method has the advantage to take into account both API restrictions and the dynamics of users' behaviors. We formalize the task as a contextual bandit problem with multiple actions selection. We finally conduct experiments on Twitter, which demonstrate the empirical effectiveness of our approach in real-world settings.

INTRODUCTION Of the 82.4 million people forced to leave their homes (an increase of 2.9 million people from 2020), around 40 million are underaged (<unk>18) and a substantial number are in transition to adulthood. Available data show that from 1990 to 2020 there has been a radical increase in the number of young migrants (15)(16)(17)(18)(19)(20)(21)(22)(23)(24), including refugees, of almost 10 million, and an increasing number arrive as unaccompanied minors (IOM, 2021). A large proportion of refugees is internally displaced (45.7 million), and most of those who have crossed borders reside in neighboring countries. A small proportion, including unaccompanied minors and young adults, try to reach Western and other affluent countries to seek safety and possible asylum. The present situation is characterized by high-income countries' attempts to hinder refugees from crossing their borders. This policy of deterrence has gradually become the key element of many countries' policies toward refugees. Deterrence policy causes prolonged flights, more suffering and mental health problems (Hassan et al., 2016), and a higher risk of death (Borja and Black, 2021). Many refugees from the Middle East and Central Asia reach northern Europe via the Balkan route. The entry point to Europe is often Greece. The geographical location of Serbia makes it a transit hub for refugees taking that route. In March 2016, the EU and Turkey signed an agreement to reduce the number of refugees entering Europe by sending migrants crossing from Turkey into Greece back to Turkey and by increasing border security (Weber, 2016). The endeavor to close the Balkan route has only been partly successful, but it has made the flight more difficult and dangerous (Weber, 2017;Jovic, 2018). The aim of this article is to explore aspects of development among young refugees during flight and after arrival in reception countries in order to understand their developmental, psychological, and social needs and resources and how they manage developmental challenges. Furthermore, and based on this analysis, we will suggest what conditions and types of intervention may be suitable and possible. Research among refugees in Europe, North America, and Australia has shown an overrepresentation of mental health suffering among individuals with a refugee background compared to the majority populations (Fazel et al., 2005(Fazel et al.,, 2012;;Priebe et al., 2010;Sabes-Figuera et al., 2012;Bogic et al., 2015;Hocking et al., 2015;Hassan et al., 2016), even after many years in the host country (Vaage et al., 2010;Opaas et al., 2020). In addition to war-related trauma preceding flight, many refugees have suffered adverse and potentially traumatic experiences during childhood and adolescence (Opaas and Varvin, 2015). High levels of mental health problems among unaccompanied minors have been documented (Grant-Knight et al., 2015;Ferrara et al., 2016). Many refugees have been exposed to multiple, cumulative, and prolonged human-caused, potentially traumatic experiences such as childhood abuse, domestic violence, sexual and physical assault, persecution, torture, combat, and other life-threatening experiences. This may lead to a decreased threshold for later development of posttraumatic stress disorder (PTSD) and problems of personality functioning, such as affect dysregulation, relational difficulties, identity disturbance, and problematic behavior (Briere and Scott, 2015). Furthermore, childhood adversities above a certain level predict psychological and physical health problems, reduced prosperity, and premature death (Anda et al., 2006). Cumulative stress, such as exposure to multiple traumatic events, poses a risk factor for mental health among young refugees, including greater suffering and functional impairment (Stotz et al., 2015). Increased levels of depression among unaccompanied minors have been documented (Keles et al., 2016). Unaccompanied minors have high support needs on arrival in the host country (Vervliet et al., 2014), and their considerable mental and social problems often go unrecognized (Grant-Knight et al., 2015;Ferrara et al., 2016). Furthermore, many show a lack of trust in health services (Majumder et al., 2015). There is a lack of knowledge about psychosocial resources that may sustain post-resettlement psychological adaptation among unaccompanied minor asylum seekers. Oppedal and Idsoe (2015) showed in a study of unaccompanied minors resettled in Norway that they suffered from high levels of ongoing war-related intrusive symptoms and depression. They were, however, still engaged in adaptation processes normative to youth with immigrant backgrounds in terms of constructing supportive networks and developing culture competence (Oppedal and Idsoe, 2015). Psychological development is a lifelong process, and being forced to flee, often accompanied by severe traumatization, implies ruptures in developmental processes at all ages, inflicting wounds that can be difficult to heal. There are certain critical developmental phases, where ruptures and upheavals may have more serious consequences, especially when support and help are insufficient or lacking, i.e., early childhood, adolescence, parenthood, and transition to old age. There are also indications that mental health problems increase in these phases for refugees (Vervliet et al., 2013(Vervliet et al.,, 2014;;Betancourt et al., 2015;Ashwini et al., 2016). Moreover, difficulties in exile may influence developmental paths and especially identity development (Johansen andVarvin, 2019, 2020). We focus the double perspective on adolescence proposed in Erikson's psychosocial, epigenic model (Erikson, 1950), where psychobiological development is seen in interaction with the social and cultural frame for this development. In parallel with the normative developmental tasks of childhood and adolescence, refugees face other specific and unique challenges, especially during flight and in reception centers. These are challenges they usually have not been prepared for and that unfold in situations, where guidance from parents or other elders is usually lacking. In addition to managing new ecological surroundings with often extreme danger during flight, many struggle with the consequences of earlier traumatic experiences, loss, and adversity (Anda et al., 2006;Opaas and Varvin, 2015) which may make it more difficult to cope. The life of a refugee is characterized by many uncertainties which may render young people especially vulnerable. Horst and Grabska (2015) claim that uncertainty is a basic characteristic of the life of a refugee. They distinguish between radical and protracted uncertainty. Radical uncertainty characterizes more acute situations, for example, during flight, which may be extremely unpredictable due to a lack of information needed to make qualified judgements. Protracted uncertainty characterizes longer phases of the refugee experience, often lasting many years, making it extremely difficult to plan or imagine a future with possibilities (Horst and Grabska, 2015). These uncertainties have obvious negative effects on young people's ability to endure difficult situations and to activate psychosocial resources (Sagbakken et al., 2020). Erikson proposed identity formation as the primary developmental task during adolescence and consolidation of identity as important in the young adult phase, where autonomy and the abilities to form intimate relationships and plan education/work are central tasks (Erikson, 1950(Erikson,, 1964;;Blos, 1967). He explored nuances and variance in developmental processes and how social, historical, and political contexts interact with personal capacities. Central for identity development is a sense of temporal and contextual integration, often understood as corresponding to what he described as "a subjective sense of invigorating sameness and continuity" (Erikson, 1968, p. 19;Syed and McLean, 2016). Identity, then, can be understood as an integration of the numerous, possibly conflicting, aspects of individual lives. Adolescence is characterized by profound biopsychosocial restructuring, during which important biological changes produce psychical challenges to be met in a changing sociocultural context (Hauser, 1999). Different contexts offer more or less stable opportunities for the developing identity. Adolescence is the time of the second separation-individuation phase, where many earlier challenges and conflicts in the childhood separationindividuation phase (Mahler et al., 1975) are actualized in a new biopsychosocial context (Blos, 1967). The challenges of immigration, where similar themes need to be reworked, have been characterized as a third separation-individuation phase (Akhtar, 1995). Adolescence and young adulthood are developmental phases of separation from parents/caregiving persons and the process of individuation as a separate person. The aim is to achieve relative independence, while simultaneously preserving relationships with the family and the community. There are important sociocultural differences in how these developmental tasks are negotiated between community-oriented societies and individualistic societies, where adherence to and identification with the family and the group are stronger in the former (Dajani, 2017). Nevertheless, it is possible to delineate some basic features. It is a time when the individual's emergent self learns to regulate emotions, establishes more secure inner models for relationships with others, and develops unique, creative ways of meeting challenges and developing capacities that will be useful in their future lives in their cultural setting (Blos, 1967;Akhtar, 1999). Central in this process is a sense of future directedness, a concept of what lays a head, and which profession and role one is to have in the adult sociocultural context. To develop, the young person needs a future perspective on his/her development (Greene, 2021). Young refugees experience obvious obstacles to normative or culturally adapted adolescent and young adult development. Early trauma directly impacts developmental pathwayspsychological and neurodevelopmental -and is a major contributor to ongoing derailment of development and to the establishment of vulnerability to a range of psychological disorders (DeBellis et al., 1999;DeBellis, 2001;DeBellis and Zisk, 2014;Opaas and Varvin, 2015;Opendak et al., 2017;Newman, 2019), including development of PTSD (Allen and Fonagy, 2015). Potentially traumatic experiences during early childhood have also been shown to explain the severity of psychopathology after later traumatization (Opaas and Varvin, 2015). For the adolescent to develop a stable self with a sufficiently secure identity, certain conditions must be fulfilled. Under normative developmental circumstances, this means a relatively stable ecological environment, including sufficient support, recognition and affirmation from parents and elders and from the community (Bohleber, 1996). For young refugees, there are deficiencies in almost all these dimensions. The sociocultural context is alien and often extremely unreliable and dangerous, they often lack parental support and affirmation, they are exposed to antisocial circumstances that are profoundly dehumanizing, and are often subjected to human rights violations that can be traumatizing (Arsenijevi et al., 2017). This is a context which often contains both radical and protracted uncertainty (Horst and Grabska, 2015), where normative development is largely impossible, where it is difficult to develop useful strategies for the future, and where resilient strategies are hard to find. Thus, a sense of future possibilities so important for development in these phases is adversely affected (Walg et al., 2020). Due to the psychological separation from the family, many adolescents and young adults are confronted with feelings of fear, loneliness, abandonment, powerlessness, and inferiority during their refugee experience. In normative development, the "second individuation phase" (Blos, 1967), starting from the sense of being with the parents, entails a process of separation that progresses from differentiation from the parents to a temporarily narcissistic phase, with a tendency to split relationships into all good or all bad, together with devaluation and idealization of others (Streeck-Fisher, 2019). Eventually, in normative development, this process is overcome by the young person's reconciliation with themselves and their parents. In parallel, they can integrate sexuality in relationships and rework fantasies of grandeur into career perspectives (Erdheim, 2014). Cultural propositions provide the individual with dispositions that make it possible to navigate phases of development. These dispositions connect the individual with the wider community and provide ways of understanding, deriving meaning from experiences and finding strategies for action (Dajani, 2017(Dajani,, 2018)). The cultural dimension in each group and society builds on tradition and on development of knowledge that makes group life (e.g., family life) and intimate relationships meaningful (Rosenbaum and Varvin, 2007;Varvin and Rosenbaum, 2011). Thus, dislocation represents a double movement whereby the learned capacities may no longer be valid in new situations that may differ radically from wherever the learned capacities are valid. In Bourdieu's sense, the individual is dislocated from the field in which their habitus (cultural, social, and economic capital) has developed (Bourdieu and Passeron, 1990;Prieur and Sestoft, 2006). According to Bourdieu, "field" is a multidimensional space, where the agent's position is determined by their cultural, social, economic, and symbolic capital. Habitus designs the learned capacities of an individual, capacities that make the person feel and act as if at home in a certain field. It is obvious that, from the beginning, the refugee operates in fields, where his/her habitus, the ways to understand and the means to act, are not in tune with the (power) relations in these new fields. Participants in our study come from different contexts, some of which have been characterized by war and danger for long periods, and they may thus have developed a habitus more suitable for danger during flight than those who come from more peaceful societies, where upheavals occurred more recently. In situations of uncertainty, for example in long, undetermined situations during flight, in refugee centers, in hiding, on arrival or while waiting for asylum application decisions, habitus may radically fail to function as something that can preserve meaning and directedness and can guide actions. Being lowest in the hierarchy dominated by smugglers and police/border guards, refugees' social capital is too low to make any resistance of importance. Their economic status is constantly challenged because of minimal resources to cover necessities and because thefts and robberies deplete their resources. To the best of our knowledge, few studies exist that examine refugees' developmental needs and problems during and after flight. Flight represents a radical break with normative development in peoples' culture of origin, and the aim of the present study is to explore how refugees manage their developmental, psychological, and social needs and resources during flight and shortly after arrival. In this context, it is important to identify resilience-promoting and resilience-inhibiting factors on individual and contextual levels; that is, to explore which resources and deficiencies influence their lives at different stages in the flight journey. Our preunderstanding is that their refugee journeys represent radical alterations in expected development, and we expect strong variations in how they cope with ongoing adversities during flight (uncertainty, life-threatening situations, physical abuse, torture, etc.). We expect that their resilience capacities are stretched and that many may develop more resigned ways of life, losing hope, while others may develop resilient functioning. We expect developmental challenges to be extremely difficult to meet in the contexts of flight and exile. Based on our analysis, we will propose conditions and interventions that may promote resilience and better mental health. --- METHODOLOGY This study presents qualitative data from a mixed method study with participants recruited in Norway and Serbia. Results from this study have been presented in earlier publications (Gr<unk>tvedt et al., submitted;Sagbakken et al., 2020). Quantitative data related to traumatic events and indicators of mental health are presented in a separate article (Gr<unk>tvedt et al., submitted). Qualitative methodology was used to identify resilience-promoting and resilience-inhibiting factors on both individual and contextual levels for asylum seekers during their flight through Serbia and during their stay at asylum reception centers in Norway. --- Sample --- Characteristics of the Serbian Sample The participants in the Serbian part of the overall Serbian study consisted of 100 subjects in total, interviewed in two different locations: the first group was recruited in city parks in Belgrade (N = 50), where they made a brief stop before organizing the next stage of their journey. The second group was interviewed in a reception center in Pre<unk>evo in southern Serbia (N = 50), where individuals and families were residing for longer periods of time. Due to differences in the settings, we had only 10 female subjects, all of whom were interviewed in the reception center. In Belgrade, it was not possible to reach any women for reasons that are difficult to evaluate. The Serbian sample was recruited between January and April 2017. Interpreters and researchers approached the participants with oral information about the study, and written consent was obtained. Some participants, after hearing about the study through others, approached us and volunteered to participate. --- Characteristics of the Norwegian Sample The participants in the Norwegian part of the overall study consisted of 78 participants recruited between September 2016 and June 2018 at five reception centers for families and single adults in different counties, both urban and rural, the latter located in regions far away from the capital. To initiate and strengthen the recruitment process, an information meeting about the study was held in all the centers. Participants volunteered in connection with the information meeting or were recruited through center administration and staff. Participants both in Serbia and in Norway were informed that the research had no link to the asylum application process. Both samples were convenience samples, where participants volunteered for the project after receiving information (Patton, 2014). The majority of the participants in the Norwegian sample came from the Middle East and Central Asia and had fled through Serbia. The inclusion criteria for the whole group were refugees and asylum seekers during flight who were able (mentally and physically) to conduct an interview and aged above 18 in Norway and above 16 in Serbia. In Serbia, one can give informed consent from the age of 16, in Norway from the age of 18. --- The Participants in This Part of the Study We made a purposeful sampling strategy out of the total sample of 178 participants. The participants represented the age groups adolescence and young adults 1, both of which are important phases in the transition to adulthood (16-22 in the Serbian group, 18-27 in the Norwegian group). Their age was designated based on their declared age, but we indicated cases, where there were discrepancies between declared age and the age assigned to them after age testing was performed on participants in Norway. Twenty subjects -10 participants from the Serbian context and 10 participants from the Norwegian context -were selected in order to gain the best possible in-depth knowledge of the developmental process in adolescence and in the young adult phase. We thus included one group still on the move (through Serbia) and one group in reception centers after arrival in a host country (Norway). Besides representing adolescents (16-18) and young adults (18-27), they were selected according to the quality of interviews, i.e., good, deep descriptions that concerned the research questions, in order to achieve as high information power as possible (Malterud et al., 2021). Information power depends on the study's aim (whether it is specific enough to gain information needed in terms of sample size), sample specificity (whether the sample shares enough common characteristics to allow the aim/research question to be studied in depth), whether theory is specified sufficiently so that reflections on theory propositions can be made based on the material, whether the quality of researcher-participant dialog was good enough (the good interviews were selected), and whether the sample is homogenous or diverse. In this case, we ensured variation in both background and actual refugee experience at the same time, given the shared destiny of both groups as refugees/asylum seekers. We found that having 10 subjects from each context was suitable for an in-depth analysis, enabling us to discover patterns as well as variances in the different stages of people's flight journeys. This study may lay the groundwork for further research on larger or more specific groups in different contexts both during flight and in host countries. --- Sociodemographic Data --- Serbian Participants In this part of the study there were nine males from Afghanistan and one female from Iraq. Ages varied from 16 to 22 years (one was married, and one was engaged in their home country, but traveled alone). The education levels varied from 4 to 13 years. --- Norwegian Participants The participants in this part of the Norwegian study were all male. Two were from Afghanistan, four from Syria, two from 1 There are cultural and social differences in how the age periods for adolescence and young adulthood are defined. Adolescence is often from 12-13 to 18 years and young adulthood from 18 years up to 30+. In this paper, we decided to use 13-18 years for adolescents and 18-28 years for young adults. The interviews were selected based on what we judged was the participant's psychological age, partly independent of their chronological age. Iraq, one from Ethiopia, and one from Iran. The ages varied from 17 to 25 years. The four who declared their age as 17 had been determined to be 18 or more after age testing. Their education levels varied from 0 to 12 years. --- Symptoms In the quantitative part of the overall study (Gr<unk>tvedt et al., submitted), we used the Hopkins Symptoms Checklist (HSCL-25; Mollica et al., 2004), which measures symptoms of anxiety and depression, and the Harvard Trauma Questionnaire (HTQ; Mollica et al., 1992, pp. 111-116), which measures a variety of trauma events as well as emotional symptoms considered to be uniquely associated with trauma. In the overall study, the results indicate high levels of mental distress in both samples of refugees (Serbia: N = 100, Norway: N = 78). Participants in the Serbian sample reported higher levels of symptoms than the participants in Norway. Moreover, the study found that female gender, low education, refused asylum, high age, and concerns about family correlated with mental distress among the participants (Gr<unk>tvedt et al., submitted). The samples studied in this article (Serbia, N = 10, Norway, N = 10) also showed higher levels of symptoms (anxiety and posttraumatic symptoms) in the Serbian group. --- Data Production A brief, open, semi-structured interview guide was developed. The guide consisted of three main questions created to collect the participants <unk> narratives describing difficult and relevant pre-flight, flight, and post-flight experiences, including their perceived and experienced quality of life from the time they decided to leave their home until the interviews were conducted. Participants were asked to provide examples of situations, persons, or activities that had an impact on them during their refugee journey. Allowing people to tell their stories may facilitate an understanding of how people construct themselves and their relationships in a time of disruption (Frith et al., 2005). As participants may tend to describe the stressors they experienced in each period, they will be specifically prompted to outline the strengths and resources (resilience factors) they brought to bear on the situation that enhanced their coping, such as "Was there anything (or anyone) that helped you handle the difficulties you are describing?" By posing this type of question, we may identify trajectories of resilience during flight and after arrival in the host country. Similarly, to be able to identify both positive and negative experiences (e.g., the presence or absence of services), participants were asked to provide examples of situations, persons, or activities that influenced well-being, and of perceived access to any services. In this way, the participants' narratives can provide the basis for a contextualized, contiguous interpretation through their stories of specific situations (Butler-Kisber, 2018). Instruments used for quantitative data analysis have been described elsewhere (Gr<unk>tvedt et al., submitted). --- The Serbian Context Interviews were conducted with the help of translators. The interviews lasted between 1 and 2 h. The interviews were of varying quality, as participants' motivation varied. Sometimes cooperation with translators did not work well and the context of conducting interviews in a park during winter made it difficult to conduct long interviews. --- The Norwegian Context In the Norwegian context, all the interviews were planned ahead, and lasted between 60 and 150 min. They took place in a separate room/office in the reception centers, in a familiar and undisturbed place, thus ensuring confidentiality and facilitated the quality of the interviews. --- Both Contexts Interviews were conducted with the help of authorized interpreters, and the same interpreters were used during the research period to strengthen intersubjectivity between the interpreters and the researchers. If preferred by the participants, interviews were also conducted without interpreters (in Norwegian, if the participants in Norway had learned the language) or in English (both places). All interviews were audio recorded, transcribed, and translated into a language suitable for analysis (Serbian, Norwegian, or English). Because of the difficult circumstances in the Serbian part of the project, we took care to select interviews that were comprehensive and of better quality for this part of the study. The research included participant observations and field notes in the selected reception centers. In this way, we could move beyond selective perceptions and discover issues that were overlooked during interviews. This contextual knowledge facilitated a better understanding of what had been expressed during interviews (Fangen, 2010). Spending time with the residents in receptions centers and in city parks contributed to building a rapport and facilitated recruitment as well as informal and formal conversations/interviews. --- The Research Teams The research team in Serbia consisted of a psychiatrist, a psychologist, and two master students in clinical psychology. The research team in Norway consisted of two researchers with backgrounds from psychiatry and nursing/anthropology, and five master students in the field of nursing. --- Qualitative Analysis The interview guide was composed of three open questions aimed at allowing the participants themselves to construct their stories. The subsequent data analysis was inspired by the principles of phenomenological analysis of Giorgi (1985) as modified by Malterud (2017), an analytical approach that attempts to understand the meaning of events and interactions within the framework of how individuals make sense of their world. We developed an analytical approach that relied on five interconnected stages: (a) familiarization, (b) indexing, (c) identification of a thematic framework, in this study, a developmental frame, (d) interpretation and development of preliminary categories, (e) confrontation with existing developmental theory, and (f) reinterpretation of themes, contextualization and development of the final conceptual framework. The main researchers (in addition to the master students) also conducted interviews. In this way, the familiarization process started during the interview process and was followed by in-depth reading of the interviews. Identification of meaningful units and themes (indexing) was done through reading and re-reading the interviews and was conducted in a collaborative process between the authors. A preliminary network of categories was developed through an interpretive process, clustered around concepts developed based on the research questions for this study. The aims of the research were subsequently confronted with the material and refined. As it was important to see the gathered passages in context, whole interviews and transcripts from participant observations were re-read when considered necessary. The different backgrounds of the researchers/interviewers allowed for negotiation from different perspectives in the process of interpreting the material, basing our analysis on researcher triangulation. Identity (how they viewed themselves, how they perceived that others view them), relationships with family members (how they described their relationships with their families, the lack of contact with them, etc.), and conceptions of the future (what plans and hopes they had, how these were represented in their minds) emerged as categories of overriding importance as well as themes related to the contexts in which they were living in at the time of the interview. We arranged the passages relating to these themes in a table to provide us the main basis for the analysis. --- Ethical Issues The collection and analysis of data from Norway and Serbia were approved by the Norwegian South-East Regional Committee for Medical and Health Research Ethics (2016/651). All refugees in reception centers in Norway have the right to receive help from the mental health services. Refugees at reception centers can be referred to mental healthcare professionals through their general practitioner with assistance from social workers at the centers. All participants with severe mental distress during data collection received immediate help from mental health professionals. In the Norwegian context, all the participants provided written consent. Interviews in Serbia were conducted with verbal approval by the Commissariat for Refugees and Migration and in collaboration with the Center for Rehabilitation of Torture Victims (CRVT) and the International Aid Network (IAN). CRTV offers mental health and medical services to refugees, and any of them who asked for help could receive it from IAN. All the participants received information about the purpose and aim of the research from the interviewer and gave their written or oral consent to participate. Most of the minors were interviewed in the park. Those who were in the centers had guardians, and both the guardians and the minor refugees were informed about the research and gave their oral consent. None of the participants received any reward for taking part in the study, neither in Serbia nor in Norway. The interviews were anonymous, as participants did not provide names. The data in this study were stored in anonymized format in a secure database. --- RESULTS Danger and Unpredictability: No Possibilities for the Future All participants described life-threatening conditions as the main reason for their flight. Afghan refugees described family members who were killed (mostly fathers and brothers) and how they themselves experienced lethal danger, especially from the Taliban and Daesh (ISIS). Accounts of life-threatening conditions were also given in the narratives of refugees from other countries, and similar accounts were given by participants interviewed in Serbia and in Norway. However, the participants interviewed in Serbia seemed to have the horrors at home more present in their mind and gave more detailed accounts. Amir 2, an 18-year-old boy from Afghanistan, interviewed in the park in Belgrade, told how life in his province in Afghanistan became increasingly difficult. His elder brother, who studied at the university, was abducted by a Taliban commander, and killed. He and his family tried to live their life as normal afterward, but then another family member was killed: About 2 years ago, my father was also abducted and killed by the Taliban, and it was the biggest shock for the family. When my father was killed, we no longer had support, neither economically nor spiritually, apart from an uncle, but in a way, we lost hope of continuing to live in Afghanistan, we were in danger (Interviewed in Serbia). He was 16 years old at the time, and suddenly lost material support and, most importantly, what he called spiritual support, which can be understood as the necessary guidance for taking his place as a man in his family and in society. Mohammed, a 16-year-old boy from Afghanistan, gave a detailed description of the situation in Afghanistan and of the experience of unpredictability. People could be killed for no reason. He underlined the importance of feeling safe and of having worth as a human being: Many innocent people were killed by the Taliban and the Daesh. That's how it happens there, you can be innocent and be killed. Because of those problems, we left the country and went somewhere, where we can live peacefully and be respected. An uncle who worked in the military was killed by the Taliban. He was abducted and beaten, the Taliban demanded money for him, we did not have enough to pay, and they killed him. We feel safe here, we feel we are worth something. Nobody touches us. We feel respected. In Afghanistan, we were not allowed to go to school out of fear. Whenever we left the house, 2 All names are fictitious. we were never sure that we would return safely (Interviewed in Serbia). Habdeh, an 18-year-old boy from Syria who fled alone when he was around 14, expressed his reason for fleeing in a few words: There was war in Syria. There was killing, beheading, and no school. We are a lost generation (Interviewed in Norway). These quotations express a general trend in the material and show that flight represents an escape from a situation of a lack of possibilities for young people. As the last quotation demonstrates, these young people feel that they represent a lost generation, a generation without possibilities for future development. Rajab (judged by immigration authorities to be 19, but saying he was 18), from a minority persecuted by the Taliban, described both danger and lack of possibilities: There was war in my home country, so I had no possibility to go to school. The Taliban are dangerous. It was when they said, yes come with us and you get weapons, and you shall kill people who were innocent, he added. Or if you do not come with us, we will kill you (Interviewed in Norway). As indicated in the quote, many of these young refugees decided to leave when they reached a tipping point in the form of a growing sense of being in danger, or a fear that someone in the family might be killed or that the Taliban would physically come to their home to recruit young boys. Abdullah, 22 years old and from Afghanistan, explained the lethal dilemma he faced in his home country. He had lost his father and mother in the war and was living with a relative who was a member of Taliban. The relative tried to prepare him to be a suicide bomber: If we are on the side of the Taliban, we must commit suicide (by fighting) for them. And if we side with the police, the Taliban will also kill us. It was very difficult there; we could not protect ourselves anymore (Interviewed in Serbia). For the Afghan refugees, the impossibility of living in their home country was repeatedly underlined. They lived under a terror regime, where "the Taliban" was a gruesome and unpredictable force. The participants described their lives as dangerous, with no possibilities, and where normative development as young people was impossible. Being dangerous even to go outside one's house or village made it extremely difficult to go to school, to form relationships, develop interests, and find a way to develop one's identity. Many seemed to have developed an embodied sense of being endangered subjects whose main task was to be on the alert for danger. The aim of the flight was formulated by several participants as an obligation to help their families. Parents often sold their property and everything they had to invest in their journey. There was thus great pressure from the family at home to succeed; an obligation to help their families but without the necessary support to manage it. Pablo, an 18-year-old boy from Afghanistan, talked about his motives for leaving his home: My motive is to go to Europe to study, to start working after graduation, to have a better life, to bring my family and to help my family (Interviewed in Serbia). The loss of a significant supporting person was often behind their decision to flee, in a situation, where they had to take responsibility as the eldest son in the family. Sander, a 17-yearold boy from Afghanistan, told how the death of his father made him feel obliged to leave: My problem was that I left my country. My father was killed. He was the first in his tribe and was killed by the Taliban. I do not want to tell the long story of how he was killed. Because of my brother and mother, and one sister, I went to Europe to get a job and help my family (Interviewed in Serbia). To help the family seemed to be part of their identity. They felt an obligation to take on responsibility as family provider -very much as a future goal. For them, this could justify that they had escaped and left their families in difficult situations. For all the participants, the flight represented a change in their way of managing life. They had to go through dangerous situations and thus constantly be strong fighters, without the possibility to express their feelings, needs, or sense of helplessness. At the same time, many carried a heavy responsibility for the family they were obliged to help and

Most studies on refugee populations are organized around trauma-related issues and focus on explaining pathological factors. Few studies are anchored in general developmental psychology with the aim of exploring normal age-specific developmental tasks and how the special circumstances associated with forced migration can influence how developmental tasks are negotiated. This study is part of a larger mixed method study seeking to identify resilience-promoting and resilience-inhibiting factors, on individual and contextual levels, among asylum seekers and refugees on the move (passing through Serbia) and settled in reception centers in Norway. A strategic sample of 20 adolescent and young adult refugees/asylum seekers during flight in Serbia (10) and after arrival in Norway (10) was chosen from a sample of 178 refugees interviewed in depth in Serbia and at receptions centers in Norway. The sample reflects the focus of this paper, which is to explore adolescent and young adult refugees/asylum seekers' psychological and social needs and resources during flight to and after arrival in the host country, including how developmental tasks are negotiated. Through qualitative analysis, experiences associated with the developmental changes the participants experienced before, during, and after flight are contextualized. Their sense of self, their relationships with their families and their perceptions of their situation as adolescents or young adults in a highly unpredictable situation are presented in the light of relevant theory and findings from similar refugee studies. All the participants have fled from dangerous and intolerable situations in their home countries. They describe extreme dangers during flight in contexts that are unpredictable and where they feel lonely and unsupported. Most have unmet psychosocial needs and have received little support or help for their mental health issues during flight or after arrival in Norway. Suggestions for interventions and resiliencepromoting actions are given based on the findings of the study.

supporting person was often behind their decision to flee, in a situation, where they had to take responsibility as the eldest son in the family. Sander, a 17-yearold boy from Afghanistan, told how the death of his father made him feel obliged to leave: My problem was that I left my country. My father was killed. He was the first in his tribe and was killed by the Taliban. I do not want to tell the long story of how he was killed. Because of my brother and mother, and one sister, I went to Europe to get a job and help my family (Interviewed in Serbia). To help the family seemed to be part of their identity. They felt an obligation to take on responsibility as family provider -very much as a future goal. For them, this could justify that they had escaped and left their families in difficult situations. For all the participants, the flight represented a change in their way of managing life. They had to go through dangerous situations and thus constantly be strong fighters, without the possibility to express their feelings, needs, or sense of helplessness. At the same time, many carried a heavy responsibility for the family they were obliged to help and, hopefully, provide for in the future. The flight thus represented a profound identitychanging experience, an abrupt transition to an expectation of being strong young men with adult responsibilities. --- They Were Not Prepared for What Was to Come Desperation and hope drove the participants to flee, but few were prepared for what would come during their flight. The descriptions of their flight experiences were filled with danger, terror, deceit, and lost hopes, and resembled situations, where all normative developmental possibilities as young people were put on hold. Ahmad, 16 years old, from a minority in Afghanistan, stated that he decided to flee because there was no future for him in Afghanistan. However, he was not prepared for all the dangers he would encounter: We had a lot of problems along the way, we were scared. We were afraid of the police, of wild animals (in the woods during the escape), that they would lock us in a closed room and that they would ask us for money, as was the case. We had a lot of problems (Interviewed in Serbia). Many described being exploited when working during their flight to Europe. Several told of how they had worked hard (in Turkey, for example) but did not get paid or were paid far less than promised, not receiving food and the like. Violence, executed by smugglers but also by police and border guards, was frequent. Mohammed, a 16-year-old Afghan boy, talked about one of the many arbitrary stays in prison, similar to the experiences of many of the participants: There were five or six in each room (in a refugee camp in Bulgaria). The police went in groups, pulled out one and beat him in front of us, as much as they wanted, they beat him (Interviewed in Serbia). Another frequently mentioned situation filled with danger was border passing. Abdullah, 22 years from Afghanistan, told how every time he and his friends tried to cross the Bulgarian border the police would steal their few essential valuables and subject them to physical violence and humiliation: We tried to cross the border three or four times, and each time we failed we were beaten by the Bulgarian police. Bulgarian border guards took our money and all our valuables. They took sneakers and boots from us; they even took our socks. We were wearing only our undershirts and underwear, and they let us go back. It was very cold, and it was very hard; I just got beaten up and told to go back. On the third or fourth attempt, the police took everything from me, both my boots and my clothes (Interviewed in Serbia). Many of the participants had similar experiences every time they crossed a new border; being arbitrarily imprisoned, deprived of all valuables, being threatened, beaten, and left in a powerless position. Pablo, an 18-year-old boy from Afghanistan, described the following experience: The Bulgarian police took us to the police station in Sofia, they took our money and phones and beat us more. The Hungarian border guards returned us, but before that they took our money and any valuables that we had (Interviewed in Serbia). Several of the participants mentioned how the loss of their mobile phones increased their sense of being powerless and unsafe; not being able to communicate with friends they lost contact with on the way or with parents and relatives in their home country, all of which represented sources of comfort and belonging. Moreover, the smugglers represented danger in the sense of exposing the refugees to threats, harassment, and violence. Abdullah, a 22-year-old boy from Afghanistan, talked about this additional source of danger: They were very dangerous, (the) smugglers. Here and now when I talk about it, I am shaking. Very dangerous people. They physically harassed and beat us, they hurt my head, I even have a scar on my head. I do not know if my arm was broken, but it was red and swollen. I got so beaten on my head that sometimes I feel like I'm not thinking normally. I was beaten so badly in those days that I forgot how much trouble I had in Afghanistan. (...) In one room underground, a very small and dark space, one smuggler beat us every day (Interviewed in Serbia). As illustrated above, people with authority and power often use their position to abuse, maltreat, and exploit young refugees, depriving them of everything, from essential means to survive to a sense of human worth. --- Help and Support During Flight Stories about support, care, and help were sparse, and for many they were totally lacking. Some recounted experiences of getting help during their flight in the Middle East. Arriving in Greece was for some a good experience of being welcomed, helped, and supported. Others told of being helped during their journey through Europe (having train tickets bought for them and being given food). One reported getting help from health care institutions and from family when ill, and some reported how well they were received and helped by the police on arrival in Norway. Abdullah, a 22-year-old boy from Afghanistan, talked about his experience with Kurds in Iraq, where he not only was helped, but felt supported in a way he felt was empathic: Those people, the Kurds, when they saw how poor and miserable we were, they all gave us food, they were very kind. I cannot forget them. The Kurds on the border with Iran asked us where we were from. We said from Afghanistan. They gave us food and I cannot forget that. I also worked in a bakery. They were very kind, they gave us money, they gave us everything. I told my life story to the owner of that bakery. When he heard my story, he paid me more than other Afghans. He told me, "Do not go. Stay here and work for me. " I said I had to go further and get my life in order because my brother and sister-in-law are in a bad shape and I had to help them (Interviewed in Serbia). Most of the participants traveled alone or in groups with other refugees of similar age, and they could help and support each other to some degree, as exemplified by Sander, a 17-yearold boy from Afghanistan: I have a lot of friends here in the barracks and it is a good experience. I can talk to them. When I am lonely I tell them about my sadness and in some way, I am relieved (Interviewed in Serbia). However, when the participants started talking about good experiences, the bad ones often came back to their mind. Mohammed, a 16-year-old boy from Afghanistan, talked about how he felt received in a good way in Serbia before he reverted to a situation filled with violence and danger: It is a good experience that we are here, we are satisfied. People from different countries are here, they come from European countries to help us. It is a nice experience. I think that the best people in the world who have seen my eyes are the people of Serbia. (...) The worst experience was when I was in Bulgaria in demonstrations, Camp H. that is the worst. After the demonstration, more than 1,000 policemen came to the camp to beat us (Interviewed in Serbia). Most participants said that their flight was characterized by uncertainty in acute situations of danger, but also as unpredictable in the long run as they could not foresee how it would end. Most of the time during their flight through Asian and Middle Eastern countries, they had to manage without any care in the form of emotional support or health services. The interviews were characterized by a scarcity of descriptions of good experiences which one might expect from young people in this age group, including relationships with the opposite sex. They described scenes that reminded them of war zones, so they had to focus their attention on ways to survive and cope. However, they often expressed what they wanted for the future, such as the possibility to study and be able to help their families. A cross-cutting theme was concern about and longing for the family. Habdeh, the 18-year-old boy from Syria mentioned earlier, told of how he fled when he was around 14 and spent 2 years alone in Turkey working under bad conditions (long working hours, little payment). He remained in the demanding situation in Turkey, but the situation for his family seemed to worry him the most: --- I am young but I look much older because I have experienced so much (...). You think all the time. I think of my family and how they are. (...) Has something happened to them? How are they? How do they manage? One thinks of them all the time. Even if many of these young boys gave the appearance of being tough and of managing, feelings of loneliness and longing often lay beneath the surface. This impression was confirmed when two of the authors met some Afghan boys in a park in Belgrade. After telling us about their hard journey and how they had managed, we asked whether they missed their parents, and they responded by starting to cry. Many had lost one or both parents. They sometimes expressed sadness but seemed emotionally distant to the loss of loved ones. This seemed to be a necessity during the harsh conditions they had to endure, with sparse contact with loved ones. Habdeh told how difficult it was when the surviving members of his family had been spread to different countries. While waiting for an answer to his asylum application in Norway, he told of how his role in the family had changed: Habdeh had seven siblings, who all lived in Aleppo with the extended family before the war. Now they were separated, four in Syria and four in various countries in Europe. He expressed how his role as the eldest boy, replacing his father, was emotionally challenging to handle. Responses to Self: Who Am I Here, and Who Can I Be? The participants had to leave their habitual sociocultural contexts, where they could have expected a reasonably predictable course of life, where relationships with others in the community would have helped or hindered their desired development. How they were looked upon and responded to by their primary family members, peers, teachers, and others should have been guiding them in the development of their identity. Thus, their flight deprived them of the sociocultural settings, which normally would have represented a frame of reference for developing their identity. Below, we will explore the ways in which the participants expressed their views about themselves, the responses they experienced, and how these have affected their self-image. What emerged in the interviews was that many had a negative view of their life and of themselves, and that this was related to a lack of good and meaningful experiences, combined with many degrading encounters. Ahmad, a 16-yearold boy from Afghanistan, exemplified such experiences: When I was seven, I lost my mother and those were bad moments for me. Honestly, I have not had any good experiences in my life. So far, I have not had a good moment (Interviewed in Serbia). Salma, a 20-year-old woman from a minority group in Iraq, gave the following answer to the question of whether anything good had happened to her before or during her flight: --- No, they (good experiences) do not exist. I do not think anything nice happened to us, only that my mother and my brothers are with me (Interviewed in Serbia). Others, however, had preserved their wishes and dreams for the future that are normal for this age period, though they were often expressed in more general ways and often referred to the time before flight. They were, nonetheless, still active in their minds, and many expressed angry oppositions to the conditions they experienced in the asylum centers and an unwavering hope that things may get better. Amir, an 18-year-old boy from a minority group in Afghanistan, said: I loved living and studying a lot, and studying was my greatest wish. Apart from studies, I did not think of anything else. I wanted to finish high school, to enroll in college, I really wanted to enroll in some technical college to be an engineer (Interviewed in Serbia). Similarly, Samir, a 17-year-old boy from Afghanistan, talked about his earlier dreams of the future and of how these hopes were still vivid in his mind: I would rather go to school and learn the craft. I loved working with computers. I did photos with computers, technology, and stuff. I had great visions about my future, I loved sports a lot, I played football. I would like to be an athlete, to continue to play football, to continue to be a photographer and to continue and complete my studies (Interviewed in Serbia). Many participants talked about how they thought they were going to be received as refugees; of being welcomed and recognized as a human being in need. The need for recognition is normal for any human being, but especially important for young people. However, the experiences associated with this were often brutally disappointing. Ahmad, a 16-year-old boy from Afghanistan, elaborated: When I left Afghanistan, I thought when I came to European countries that they would respect and help us and look at us as human beings. We left our country because our lives were in danger. We thought that European countries would respect our human rights, but we see that here we are nothing to them and they do not respect us, as if we were not human (Interviewed in Serbia). During flight, several expressed some hope for the future, although often indirectly by talking of their plans for education when they were in their home country. The experiences of those interviewed in Norway showed that they had problems understanding how to interpret and manage their situation at the asylum centers, such as who they were, how they were perceived by others, and what developmental possibilities they had. Some clearly expressed that they were in a dead-end situation. Shadi, an Iranian man now aged 24, came to Norway 8 years ago, aged 17. His application was rejected but he had no possibility to return to his home country. He felt totally helpless and did not know what to do or how to live his life. He said simply: "I have no capacities that can help me, " describing how in his present situation he saw no future possibilities and that he had no personal resources that could help him. It expressed the confused image he had of himself: "Who am I here, what can I be?" Similarly, Hédi, a 24-year-old boy from an African country, felt he had become pacified and could not find the resources within himself to manage his situation. He had great difficulties managing his insecurity and anxiety while waiting for the decision on his application in Norway. Referring to his former life, he said he had happiness there, he could work and earn money: Here, I cannot earn anything, I have to "take" money from others. --- (...) I feel my life is meaningless. I am a person without any success. (...) What is the meaning with my life? Animals eat food and sleep. (...) What can I do? I feel hopeless... I feel hopeless. What can I do here? (Interviewed in Norway). Feelings of hopelessness and worthlessness, and for many an unclear image of who they were or could be, dominated the interviews of the participants in Norway. The context of waiting and not being able to do anything, of no school or of not being allowed to work, made life unbearable for many of the participants. Another theme that came up in connection with identity was age. Many of the participants in Norway had been tested to verify their age and in order to determine whether they were to be treated as adults (above 18) or not. If the test results concluded that a person was over 18, it implied being moved to an asylum center for adults with considerably fewer facilities and opportunities. Many described the feeling of not being believed when they gave their age, combined with a sudden change from being treated as a minor to being treated as an adult, losing sources of care and follow-up. Who they were, their identity, was suddenly changed by a decision made by external authorities. As a result, some became passive and lost hope, but others strongly opposed the decision and saw this as a problem with the system. Yousuf, supposedly 18 years old from Afghanistan, said that his application had been rejected but that he had appealed, and he described the waiting time as extremely pacifying, worse than being in a war zone. As a result of the age test, he suddenly had become another person, deprived of his life as an adolescent in the center for minors, and now he was angry and confused: --- My biggest problem is that I am placed on this adult center (...) I have undergone such an age test. (He is aware of the uncertainties of the age test). How does this machine know my age? My mother knows (Interviewed in Norway). Abdi, a boy from Ethiopia, had been tested for his age and was found to be 18 years old, and consequently was transferred from a center for children to a center for adults: First, I went to school, then I got two times rejection of my application. I was then not allowed to go to school. And then they insisted I was more than 18 years old. That was a lie, I said I was underaged. That is why I was denied permission to stay, I did not like being a human here, only living (...). And, I have simply lost my taste for life now. --- I am not looked upon as a human being (...) I do not live like a human being here. Demoralized (pause), only living. (...) It is only dark, I am only siting here. (...) There is no difference from the dictatorship at home. (...) Youth in my age do many things in their life and they are in process of changing their lives and participating in society. So I have no hope, I cannot do anything, just sit and think (Interviewed in Norway). In the interviews, the participant's view of themselves as worthless with no possibilities came repeatedly to the fore. Habdeh from Syria talked about the immense hardships he had endured during his lonely flight, which started when he was 14 years old. Now aged 18, he lived in an asylum center waiting for the decision on his application, feeling utterly lonely and with no sense of direction in his life. At the time of the interview, he was considering suicide as a possibility for change: I think about my family and how they are doing. There is worry and fright all the time. Has anything happened with them (...) you think all the time. My head works all the time. And I think about myself and my situation. I have not got A permit to stay. Maybe I will be rejected, what will happen then? (...) I try to sort my thoughts. I think about my future, but nothing is clear yet, about my future. (...) Life here at the center is very, very boring. One hate being here. One gets tired of the clothes one has. Thinks all the time. One wants to commit suicide (Interviewed in Norway). Habdeh went on to describe how he tried to give himself some hope by thinking that maybe he would get a residence permit the next day, and then maybe be able to start something. But all in all, he perceived his situation as almost unbearable, as everything was put on hold. His life in the asylum centers (he had been to several) was experienced as a lonely struggle and as a life with no care. He expressed profound confusion over how he could be active and develop, and who he could become. Since he saw no possibilities, suicide came to his mind as a solution. What we see in these quotes is young people who are in situations of serious conflict regarding how they should view themselves based on the responses they receive from external others. Some feel resignation and sink into passivity, but others, like Yusuf, manage to keep a more positive image of themselves as people who can achieve things and make something of themselves. The participant's experiences of flight and of how they were received tested their endurance. Those interviewed in Serbia were on their way and tried to cross borders and find a way in spite of great obstacles. For those interviewed in Norway, the conflict between hope and hopelessness was clearer, and many described an unbearable feeling of just waiting, just "being in wait" as one put it. Many felt pressured into a passive position with few possibilities of being able to use their resources, while some felt they no longer had the capacity to do anything. In some cases, this could lead to thoughts of suicide as the only active possibility. --- DISCUSSION The narratives in this study were related in two different situations: during flight and in reception centers after arriving in Norway. They were told by young people, most of whom were adolescents when they started their refugee journey. The stories based on experiences in their homeland were similar, describing how war, persecution, harassment, and serious losses made it impossible for them to live in their country. The recent takeover of power by the Taliban in Afghanistan confirms their descriptions of the dangers experienced in regions, where the Taliban has pursued its campaign of terror for many years. What was conveyed during flight in Serbia differs in some respects from what was conveyed in Norway. The participants in Serbia were still on their way, and even though they encountered severe difficulties and were mistreated by smugglers and police/ border guards, they had some hope of reaching their goal, as exemplified by their repeated attempts to cross borders. Among those who were in Norway and were waiting for decisions or had been rejected, the tone and quality of their stories was different, often characterized by hopelessness, confusion, worry, and a sense of dislocation. In a previous work, we analyzed how living in inconclusive life situations that unfold in the context of a bewildering, bureaucratic system that is difficult to understand or influence causes feelings of confusion and powerlessness (Sagbakken et al., 2020). Being left in a situation of insecurity and unpredictability, often for several years, distorts people's perception of time when they lose the ability to visualize their future life. This troubled perception of time was experienced as a continuation and reinforcement of unsafe and dehumanizing pre-flight and flight experiences. --- Time and Development Central in developmental processes is the ability to link the past to the present, and the present to the future. This is especially true in the adolescent and young adult phases. In fact, a central task in these phases is to imagine future possibilities and to find what grounds or capacities one has developed to manage whatever one sees as possibilities. One needs to be aware of one's desires and wishes, which often can be unclear and can provoke anxiety, and to find out which wishes and desires one is capable of realizing, which ones are acceptable, and which ones can be realized within the sociocultural context in which one lives (Erikson, 1963;Syed and Fish, 2018;Greene, 2021). As previously described, adolescence implies a major reorganization on biological, psychological, and social levels, and is a period where the need for contextual stability (family, social context) is critical for being able to manage the turmoil and internal upheaval one goes through. The psychosocial context represents a frame of reference for orienting oneself in time and space. To achieve this, one needs some stability in order to connect the past and the present with the future and to imagine what one can develop to become or how one's identity can develop. This difficulty in imagining a future with possibilities seemed to create a sense of uncertainty and meaninglessness, which caused impairment in the sense of reality, whereby cognitive organization of perceptual impulses became extremely demanding due to difficulties in distinguishing the important from the unimportant (Sagbakken et al., 2020). Thus, when the context of development collapses, the consequences may be serious for managing development tasks. Within a context, where external and alien forces during flight (cheating and unpredictable smugglers, brutal behavior by police and border guards, and human rights violations) and the bureaucratic, unpredictable jungle experienced on arrival, the time frame of development becomes extremely difficult to maintain (Christensen and Lindskov, 2000). Without reliable guidance, these young people must manage by themselves and make up for whatever the context did not provide by trying to organize their mental life in an effort to imagine and foresee what may come. In the interviews, we heard how many tried to think of education, but again and again they returned to how all hopes were dashed in their home country. As Habdeh said: "There was war in Syria. There was killing, beheading, and no school. We are a lost generation. " He expressed a general trend among our participants. They fled from extreme conditions with no or few developmental possibilities and found themselves in a situation of limbo, with the hopeless feeling of being subjects of impersonal forces which seemingly did not care for them. --- Tasks to Solve: Erikson's Stages Identity formation is seen as the primary developmental task during adolescence (Erikson, 1963(Erikson,, 1968). Erikson explored nuances and variance in these developmental processes and how social, historical, and political contexts interact with personal development (Syed and Fish, 2018). Healthy identity implies, in this perspective, sense of temporal and contextual integration and identity can be understood as an integration of the numerous, possibly conflicting, and aspects of individual lives (Akhtar, 1995;Johansen and Varvin, 2020). In Erikson's eight developmental stages, development in each stage is dependent on how successful or unsuccessful the previous stages have been. Each stage implies a psychosocial crisis in the sense that the expectations from the environment demand reorientation and an expansion of the interpersonal and psychosocial world (Erikson, 1950). There is no deterministic relationship between the stages. Healthy development may ensue even if earlier stages have been marked by much stress or trauma, especially if interpersonal relations can give support and promote healing and resilience (Erikson, 1995). Thus, the context and the person mutually influence each other (Sameroff, 2010). To manage the psychosocial crisis in each developmental stage, there cannot be too big a discrepancy between the capacities of the person and the potential of the ecological environment. At the core of the adolescent crisis lies the tension between developing a more stable identity and identity confusion (Erikson, 1963). The interpersonal relationships in which identity develops need to be stable enough to avoid identity diffusion and confusion. Early hardships and potentially traumatizing experiences in childhood may lower resistance and the ability to cope during later hardships (Opaas and Varvin, 2015). There is thus the possibility that participants in our research have been especially vulnerable to the hardships they encountered during and after flight, and that these may have contributed to the hopelessness described by, for example, Salma as follows: "No, they (good experiences) do not exist. I do not think anything nice happened to us, (...). " Salma expresses resignation, as if she had acquired the identity of a person who has been and will be treated badly. Following Erikson and others, problems in earlier developmental stages influence how later developmental crises are solved (Erikson, 1964;Lilly and Lim, 2012;Mikulincer and Shaver, 2015;Opendak et al., 2017). We do not have much specific data on early childhood in this study, but research on other groups of refugees has revealed abundant early childhood hardships (Opaas and Varvin, 2015). During later hardships and upheavals, the consequences of earlier traumatization may be actualized in the form of catastrophic anxieties, and the present may be perceived as confirmation of the early experiences of danger and unpredictability (Varvin, 2003;Rosenbaum and Varvin, 2007). Early primitive defenses may reappear, such as splitting and projective identification, leading to increased confusion, identity diffusion, and lack of any possibility to organize future perspectives in the mind (Henningsen, 2005). Many of the participants expressed hope, but it is a hope that cannot be anchored in a defined future experience. The participants experienced the waiting for something that would define their lives for better or worse, but with limited possibilities to influence the outcome: one can try again to cross a border or send a new application, but those who decide are outside their sphere of influence (Biehl, 2015;El-Shaarawi, 2015;Horst and Grabska, 2015). A fundamental resource for displaced people is hope for the future, and thus the possibility of "future-figuration" (Akhtar, 1999;Artero and Fontanari, 2021), a possibility that is severely reduced among the participants in this study. Habdeh' s comment on the lost generation represents an interesting reflection on the loss of this resource, a situation that seemed typical for many of the participants in this research: no possibilities at home, and an extremely insecure future. A future directedness as a driver for development is seriously disturbed. --- What Kind of Identities Can Develop in the Refugee Context: The Limbo Experience In a previous publication (Sagbakken et al., submitted), we described the experience of rejected asylum seekers as one of being in limbo, referring to their experiences of endless waiting. In this context, the word "wait" lost its meaning because for them there was nothing to wait for (Bjertrup et al., 2018). The term "liminal phase" is often used to refer to the stage in a "rite of passage, " where the subject is on the threshold to the new and partly unknown (Turner, 1964, pp. 4-20). It is often described as the transition phase between two stages where the next is more or less known, like a "passage of transition" between youth and adulthood (Turner, 1964). Being in limbo, on the other hand, is a situation of being stuck, with limited possibilities to move forward or develop into the next phase. To be in limbo, according to Collins dictionary, means to be "in a situation, where you seem to be caught between two stages and it is unclear what will happen next, " or to be in an "imaginary place for lost, forgotten, or unwanted persons or things" (Collins Dictionary, n.d.), which is similar to the feelings of many of the participants in our study as well as in others (de Zulueta, 2011;Nimführ and Sesay, 2019). Adolescence can be described as a liminal phase, and different cultures provide possible outcomes that offer possibilities. The participants in this research had all fled to arrive at some place with a wish for new possibilities that were not present at home. Their flight can thus be seen as an extended liminal phase, where all kind of hopes and desires were activated, but with a realization that there were limited possibilities of realizing any wished-for plans. They had as a rule been backed by their families and filled with hopes and expectations. Their will to fulfill these hopes and obligations and to help their families were apparent in many interviews. The situation of the participants was thus complex: they were in the middle of a complicated developmental phase involving bodily changes, mental reorganization, and establishment of capacities for both intimate and social relationships. At the same time, they strived to fulfill the tasks and hopes given them by their families, or what they themselves imagined was important for their kin. In other words, they strived to establish identities both as grown-up and relatively independent individuals and as members of their group or family, with the obligations this imposed on them. In other words, a complicated, multidimensional, mental, and sociocultural situation, where they have to deal with biopsychological development, relationships with their family and group, cultural pressures, and the flight/exile situation (Rosenbaum and Varvin, 2007;Rosenbaum et al., 2020): Situations of stress, crisis, or traumatization challenge one's capacity to cope and resist and creates a situation where resilience can be activated if enough early resources have been established (Alayarian, 2007). The possibility to construct meaning of experience is of overriding importance. It is in a person's relationship with cultural discourses that meaning can be constructed from experience. This gives the person reason to believe that there can be a continuation, and thus create a time dimension. It influences relationships in the group, where identity is shaped by the reciprocal relationships with others and where one gradually learns who one is and who one wants to be, based on responses from others. The cultural umbrella thus safeguards a meaning-producing dimension that helps a person endure crisis and hardship (Obeyesekere, 1990). Without these relational contexts, identity formation becomes difficult. Participants in this study have experienced major upheavals and breakdowns in their development so that to a certain degree they lack what could have served to support and guide their identity development. What the participants described was a replacement of relational contexts with peer groups (fellow refugees, some of whom were dangerous), with smugglers who were notoriously dangerous and unreliable, and with border guards/ police who looked upon them as intruders or strangers who were to be rejected or thrown out. Additionally, they described the lack of a cultural, meaning-producing system with its moral values and implications. What the participants met was a chaotic, unpredictable context and an absence of common, moral guidelines. The formation of identity is always a process and is always dependent on responses from others and being seen as an outcast or as dangerous or of no worth marks people's identity development. Thus, being a young refugee in these contexts severely disturbs both development of identity and the sense of being of worth. Follow-up studies on refugees in receiving countries show variable outcomes regarding mental health and quality of life, but many show resilient development (Vaage et al., 2010;Bogic et al., 2015). There is, however, a tendency of long-term mental health problems (Opaas et al., 2020), especially since support and treatment tend to come late and be insufficient (Pumariega et al., 2005;Opaas and Varvin, 2015). --- "To Be in Wait": Passivity vs. Aggression This expression by one participant conveys development of an identity characterized by passivity. It expresses an empty space where no one is there to see the person, affirm their existence, give some hope that waiting may have an end and to express personal hope for the person. Being in situations where feedback on who one is comes from people who want to exploit, reject or harm you creates a special situation of uncertainty, not only on a social level (Horst and Grabska, 2015), but also on an internal level. Development of one's identity is always an interaction between one's involvement in a group and the reorganization of one's internal world. Culture becomes part of the mind through these interactions, and when one lives in an extremely dangerous and aggressive world, this may be internalized, creating internal instability and often a tendency to identify with the aggressor (Hirsch, 1996). One internalizes the relationships one is in to manage and survive in a dangerous world. This is the background for what is called the externalizing behaviors that can be seen among young refugees (Fazel et al., 2012). Identifying with the aggressor protects against the helplessness one experiences at the hands of the perpetrator (Frankel, 2004). This was very clear when interviewing the participants in the park in Belgrade. It took time to build trust, the participants behaved for a long time as if the researchers were allied with the police and treated them accordingly with suspicion and animosity. Thus, we may interpret this as a passive-aggressive

Most studies on refugee populations are organized around trauma-related issues and focus on explaining pathological factors. Few studies are anchored in general developmental psychology with the aim of exploring normal age-specific developmental tasks and how the special circumstances associated with forced migration can influence how developmental tasks are negotiated. This study is part of a larger mixed method study seeking to identify resilience-promoting and resilience-inhibiting factors, on individual and contextual levels, among asylum seekers and refugees on the move (passing through Serbia) and settled in reception centers in Norway. A strategic sample of 20 adolescent and young adult refugees/asylum seekers during flight in Serbia (10) and after arrival in Norway (10) was chosen from a sample of 178 refugees interviewed in depth in Serbia and at receptions centers in Norway. The sample reflects the focus of this paper, which is to explore adolescent and young adult refugees/asylum seekers' psychological and social needs and resources during flight to and after arrival in the host country, including how developmental tasks are negotiated. Through qualitative analysis, experiences associated with the developmental changes the participants experienced before, during, and after flight are contextualized. Their sense of self, their relationships with their families and their perceptions of their situation as adolescents or young adults in a highly unpredictable situation are presented in the light of relevant theory and findings from similar refugee studies. All the participants have fled from dangerous and intolerable situations in their home countries. They describe extreme dangers during flight in contexts that are unpredictable and where they feel lonely and unsupported. Most have unmet psychosocial needs and have received little support or help for their mental health issues during flight or after arrival in Norway. Suggestions for interventions and resiliencepromoting actions are given based on the findings of the study.

uncertainty, not only on a social level (Horst and Grabska, 2015), but also on an internal level. Development of one's identity is always an interaction between one's involvement in a group and the reorganization of one's internal world. Culture becomes part of the mind through these interactions, and when one lives in an extremely dangerous and aggressive world, this may be internalized, creating internal instability and often a tendency to identify with the aggressor (Hirsch, 1996). One internalizes the relationships one is in to manage and survive in a dangerous world. This is the background for what is called the externalizing behaviors that can be seen among young refugees (Fazel et al., 2012). Identifying with the aggressor protects against the helplessness one experiences at the hands of the perpetrator (Frankel, 2004). This was very clear when interviewing the participants in the park in Belgrade. It took time to build trust, the participants behaved for a long time as if the researchers were allied with the police and treated them accordingly with suspicion and animosity. Thus, we may interpret this as a passive-aggressive attitude that helped them manage in unclear situations. The same has been observed in traumatized refugees in treatment. Long periods in the treatment process may be characterized by passive-aggressive attitudes (Varvin, 2003). The task in these treatments is then to activate something good in the internal world of the patient, a relationship with some good and caring inner object. What is attacked and often destroyed during atrocities like torture and serious maltreatment is this inner link to an empathic other (Laub and Podell, 1995;Laub, 1998), which is essential to the development of a subjective sense of a stable identity (Erikson, 1968). The mistrust and skepticism, we saw in our participants can be seen as a survival strategy during extreme circumstances (Varvin, 2003). What is needed then, to help young refugees both during and after flight, is help with establishing inner security, which happens through mobilizing inner resources in good inner relationships. Early interventions and caretaking after traumatization and hardships may thus be of importance for mental health and social functioning in the long run (Keilson and Sarpathie, 1979;Steel et al., 2011). This leads to the questions of how refugees should be cared for and what interventions and treatment they need. --- Interventions/Treatment Due to the hardships and potential traumatizing experiences during a vulnerable phase in their lives, it is to be expected that many will develop psychological problems in the long term (Henley and Robinson, 2011;Vervliet et al., 2013;Jensen et al., 2014). We have shown that there are many uncovered psychosocial needs in this group, both during and immediately after flight. As a rule, these needs are not met and any organized and planned help is scarce. Based on the findings of this research and other sources (Jensen et al., 2014;Thommessen et al., 2015;Keles et al., 2016), we hold that help should be provided as early as possible and at any stage in the flight. It should be comprehensive and flexible (i.e., it should include various forms of health care, mental health care and psychosocial interventions as well as legal protection, security measures and addressing basic needs), be easily accessible, sensitive, and tailored to trauma and torture victims, and be well designed to tackle obstacles that are well recognized in provision of services for asylum seekers, such as low awareness of services, linguistic barriers, and differences in cultural perspectives and service preferences (Satinsky et al., 2019). Interventions should also take into account that the social functioning of individuals who have experienced traumatization and torture can be impaired (Jovi<unk>, 2017). They should cut across traditional medical and mental health care and social services, be tailored to cover as wide an area as possible and have an interdisciplinary and intersectoral approach. According to IASC 3, "the composite term mental health and psychosocial support is used in this document to describe any type of local or outside support that aims to protect or promote psychosocial well-being and/or prevent or treat mental disorder" (IACSC, 2007). It is already recognized that there is a "need for a comprehensive, well-planned approach at the different stages in the refugee journey" (Varvin, 2021, p. 453) during flight, in refugee camps on the route (which are often improvised crowded shelters with inadequate accommodation), in asylum and reception centers on arrival, all the way to the (ideal) completion of the asylum procedure, settlement in a new country and new home, and adjustment to a new physical, social and cultural environment. These services will depend on the needs of refugees and should be adapted to the flight regarding safety concerns and medical consequences of physical and psychological stress. After resettlement there will be a need for comprehensive, specialized services to prevent chronic sequelae. A model for such help and support during flight was developed in Serbia in 2015 by International Aid Network, Center for Rehabilitation of Torture Victims (IAN/CRTV) for refugees passing through the country and in need of basic help and support. Mobile teams were organized comprising a medical doctor and a nurse to attend to refugees' physical and medical needs using a set of interventions known as medical first aid (MFA). The teams also comprised a psychologist, responsible for providing psychosocial first aid (PFA), and a translator (Jovic, 2018). Guidelines for brief psychosocial interventions for refugees were developed that included: 3 The Inter-Agency Standing Committee is an inter-agency forum for coordination, policy development and decision-making involving the key UN and non-UN humanitarian partners. 1. Safety, at a time when it was important to determine whether the person is at real physical risk of falling victim to peer violence, human trafficking, sexual abuse, or forced prostitution. 2. Basic needs (food, hygiene items, or shelter). 3. Legal protection (support with registration procedures, asylum applications, and their rights in both cases). 4. General health status/needs (physical health, chronic disorder, or physical trauma). 5. Assessment of mental health needs, psychological and psychosocial support, and treatment and rehabilitation. As refugees are now mainly accommodated in reception centers in Serbia and on the Balkan route, it is highly probably that their mental health needs will come to the fore, due not only to high incidence of posttraumatic responses, depression, suicidal behavior, psychosis, family problems/pathology, but also to conditions in camps that are harmful to health (Varvin, 2021). Still, organizing well-structured treatment is a challenging task on account of the very basic services provided by the health care sector and frequent changes in accommodation. Few studies have evaluated early prevention and treatment programs. Studies assessing community-based interventions, such as outreach programs, workshops, train-the-trainer models, refugee employment schemes and mentoring programs focused on self-help, inclusion, empowerment and advocacy, consistently found some reduction in symptomatology in participants (Williams and Thompson, 2011), and a long-term follow-up of early intervention showed promising results (Drozdek et al., 2014). Comprehensive preventive programs are therefore a necessity, especially for vulnerable groups such as unaccompanied minors and for adolescents in general. Based on clinical/practical experience, we believe that the above guidelines outlined by IAN/CRTV will have positive effects on refugee groups as well as on the community they function in, as indicated in review of Williams and Thompson (2011). There is thus a need for further systematic studies. --- CONCLUSION Young people experience a radical break in their development when they embark on their refugee journey. They flee from war, persecution, and often from countries where the infrastructure has broken down. They bear the double burden of solving developmental challenges, while trying to adapt to dangerous conditions during flight and incomprehensible bureaucratic systems in the country in which they seek asylum. Many feel obligated to their families to succeed and be able to support their family at home. Many have suffered severe losses and have been traumatized. Their need for safety and to able to plan is severely hampered, and their developmental needs are generally not met. Furthermore, young refugees are at risk of developing mental health problems and of having difficulties adjusting. It is therefore argued that support and help should be set in as early as possible, preferably already during flight, and that their life in reception countries should be supported and treatment (mental and somatic) provided when necessary, regardless of their status in the asylum system. They should be seen as young people in important developmental phases, and interventions and help should be adapted to this situation. Suggestions for how this could be organized are provided above. The participants in this study were selected from a larger study in which 178 people were interviewed either in a reception center or in an outside location. The focus on adolescents and young adults in the larger study facilitated a study of a typical group in the chosen contexts (Serbia and Norway) at the time. The in-depth quality of the interviews and the fact that we chose the more comprehensive interviews strengthen the validity of the findings. The specific context for the interviews, e.g., that most of the participants were in a desperate situation, also imposes limits in that it influenced how the narratives were constructed at the time of the interview. However, we took their specific situation into consideration when performing the analysis, and participant observations combined with notes from the interviews (body language, expressions of emotions, etc.) helped validate what was expressed verbally. As this is one of few studies to focus on the developmental aspects of being a refugee, we believe it may pave the way for more studies that focus on refugees as ordinary human beings who need to take their developmental health into consideration, regardless of the social context they find themselves in. --- DATA AVAILABILITY STATEMENT The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. --- ETHICS STATEMENT The studies involving human participants were reviewed and approved by Norwegian South-East Regional Committee for Medical and Health Research Ethics (2016/651), The Commissariat for refugees and migration and in collaboration with the Center for Rehabilitation of Torture Victims (CRVT) and International Aid Network (IAN). The patients/participants provided their written informed consent to participate in this study. --- AUTHOR CONTRIBUTIONS SV and MS developed the study design and conceptualization and completed interpretations of the analysis. MS, SV, and IV conducted the interviews. All authors participated in process of developing the data analysis. SV developed the main text of the paper. All authors reviewed the results and approved the final version of the manuscript. --- Conflict of Interest: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Publisher's Note: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Most studies on refugee populations are organized around trauma-related issues and focus on explaining pathological factors. Few studies are anchored in general developmental psychology with the aim of exploring normal age-specific developmental tasks and how the special circumstances associated with forced migration can influence how developmental tasks are negotiated. This study is part of a larger mixed method study seeking to identify resilience-promoting and resilience-inhibiting factors, on individual and contextual levels, among asylum seekers and refugees on the move (passing through Serbia) and settled in reception centers in Norway. A strategic sample of 20 adolescent and young adult refugees/asylum seekers during flight in Serbia (10) and after arrival in Norway (10) was chosen from a sample of 178 refugees interviewed in depth in Serbia and at receptions centers in Norway. The sample reflects the focus of this paper, which is to explore adolescent and young adult refugees/asylum seekers' psychological and social needs and resources during flight to and after arrival in the host country, including how developmental tasks are negotiated. Through qualitative analysis, experiences associated with the developmental changes the participants experienced before, during, and after flight are contextualized. Their sense of self, their relationships with their families and their perceptions of their situation as adolescents or young adults in a highly unpredictable situation are presented in the light of relevant theory and findings from similar refugee studies. All the participants have fled from dangerous and intolerable situations in their home countries. They describe extreme dangers during flight in contexts that are unpredictable and where they feel lonely and unsupported. Most have unmet psychosocial needs and have received little support or help for their mental health issues during flight or after arrival in Norway. Suggestions for interventions and resiliencepromoting actions are given based on the findings of the study.

blameworthy for their assaults (than those who were not) due to stereotypes of regarding women who drink (i.e., rape myths) as loose and deserving of victimization. In addition, survivors may feel the assault is at least partially their fault having internalized stereotypes of women who drink as "asking for it." (Edwards, Turchik, Dardis, Reynolds, & Gidycz, 2011;Hammock & Richardson, 1997;Suarez & Gadalla, 2010;Stormo, Lang & Stritzke, 1997). Survivors who engaged in pre-assault alcohol use to the point of incapacitation are more likely to disclose to informal, but not formal, support sources than are survivors who were drinking but were non-impaired survivors (Littleton, Grills-Taquechel, & Axsom, 2009). Both formal (e.g. police) and informal (e.g. family) sources might react negatively to survivors who were drinking. Whereas most survivors report receiving a mixture of both positive and negative social reactions, survivors of alcohol-related assaults report receiving more negative responses (Ullman & Filipas, 2001;Ullman & Najdowski, 2010) than those who were not drinking. In particular, women who experience alcohol-related assaults report more blaming, controlling, stigmatizing, and distracting responses from formal and informal support sources than survivors of non-alcohol-related assaults (Ullman & Filipas, 2001;Ullman & Najdowski, 2010). Thus, survivors who were drinking prior to assault may get fewer positive social reactions and face a more complicated recovery than survivors who did not drink. While there is a considerable amount of research on social reactions received by survivors of alcohol-related assaults, there is little research that examines social reactions specific to pre-assault alcohol use (e.g., alcohol-specific social reactions), and even fewer studies that examine these social reactions through qualitative research methods. This research is needed, as there are notable differences in the experiences, disclosures and social reactions of alcohol-and non-alcohol related assaults (Ullman & Najdowski, 2010). There are few qualitative studies that include an analysis of social reactions to survivors who engaged in pre-assault substance use including drinking. One study of social reactions from formal support providers such as law enforcement and prosecutors found that blaming reactions stemming from the survivor's substance use can still be interpreted as supportive if the responder also provides emotional support (Dworkin, Newton, & Allen, in press). Research to date assesses only social reactions in general for survivors disclosing alcoholrelated assaults (see Macy, Nurius, & Norris, 2006;Ullman & Filipas, 2001;Ullman & Najdowski, 2011). These results are largely mixed and fail to examine social reactions specific to survivors' pre-assault drinking. Few studies to date have examined social reactions that specifically comment on the survivor's pre-assault drinking, termed alcoholspecific social reactions (Relyea & Ullman, 2015). In a study of the larger database of survivors who completed mail surveys and from which the qualitative sample of the present study was drawn, those who drank prior to assault, who had more education and greater alcohol impairment or resistance during assault were more likely to disclose pre-assault drinking (Lorenz & Ullman, 2016b). Of women who disclosed their drinking, more educated women with more violent assaults received more negative social reactions specific to preassault drinking, especially from parents, police, or doctors. Less educated women received more positive and negative social reactions specific to pre-assault drinking. Disclosing preassault drinking in greater detail was related to positive social reactions specific to pre-assault drinking and greater alcohol impairment during assault was related to more positive and negative social reactions specific to pre-assault drinking. Other analyses of the same data showed that alcohol-specific reactions were not related to depression, posttraumatic stress symptoms, binge drinking, or intoxication, but were related to greater characterological self-blame and alcohol problems (Reylea & Ullman, 2014). The link between alcohol-involved assault and social reactions highlights the importance of exploring the correlates of reactions related to pre-assault alcohol use. By understanding the disclosure and social support received following alcohol-involved assault, researchers can begin to develop and evaluate interventions aimed at improving social responses to survivors of alcohol-involved assault that may enhance the recovery process. --- Present Study Extant research highlights the influence of social reactions on recovery (see Orchowski et al., 2013;Ullman, 2010;Ullman & Najdowski, 2011) and shows worse post-assault adjustment for survivors of alcohol-related assault receiving negative social reactions (Littleton et al., 2009;Ullman & Najdowski, 2010). Given the prevalence of alcoholinvolved assaults, it is important to establish what factors lead to receipt of certain alcoholspecific social reactions. The present exploratory study examines processes involved in dyadic matched pairs of family, friends, and significant others disclosing and receiving social reactions related to alcohol and sexual assault. While some important past qualitative research examined differences in social reactions from family, friends, and significant others in general (Ahrens & Aldana, 2012;Dworkin, Pittenger, & Allen, 2016), that work has not specifically focused on alcohol-related sexual assaults. Quantitative findings of studies that examine alcohol-specific social reactions suggest more examination of disclosure and social reactions to alcohol-related assaults are needed and in greater depth than statistical analyses can provide. In addition, no studies to our knowledge have been published using dyadic matched pairs of survivors and informal support providers in relation to alcohol related sexual assaults, disclosures, and social reactions. --- Method Participants A sample of adult female sexual assault survivors were recruited through advertisements in the Chicago metropolitan area and had previously participated in a 3-year longitudinal mail survey study (N=1,863) regarding their unwanted sexual experience and the social reactions they received when disclosing these experiences (see Peter-Hagene & Ullman, in press for study description). For the present study, women who indicated interest in being recontacted for interviews were asked to provide contact information for a friend, family member, or significant other who they told about an unwanted sexual experience. The support providers were contacted later for a separate interview that focused on their experiences helping survivors cope with unwanted sex. Interviews with survivors and support providers took place over two years, resulting in a sample of N=45 matched pairs of survivors and support providers. The present study includes only matched pairs involving alcohol-related or alcohol and drug-related assaults (n=19). Survivors of this subsample were an average of 40 years old, and ethnicity was: 42.1% African American, 31.6% White, 21.1% Other, and 5.3% unknown. Just under half (47%) had children, 89% had attended or graduated from college, 26% were in school, and 37% were employed. Of the 19 matched pairs in the subsample, the majority described their relationship as friend (n=13), followed by family (n=4), and significant other (n=2). Support providers were 39 years old on average, 68% female, and 31.6% African American, 31.6% White, 15.8% Other, and 21.1% unknown ethnicity. Half of support providers had children, 83% had attended or graduated from college, and 56% were employed. --- Procedures Face-to-face semi-structured interviews were conducted by one of three trained interviewers on the research team, with length ranging from 30 minutes to 3 hours (average 1 hour). All survivors and support providers interviewed received $30 for participation. After each interview, interviewers created brief summaries with noteworthy points, questions, thoughts, and feelings emerging from the interview. Interviews were audio-recorded, transcribed, and checked by other members of the research team. The transcription process included adding brief summaries and lists of patterns. Interviewers did the final reviews of their transcripts, and following transcription, interviewers met to discuss emerging themes and patterns later used to develop a coding scheme. --- Measures Interview protocols for survivors and support providers focused on disclosure of the unwanted sexual experience, social support provided/received, social reactions and appraisals of those reactions, appraisals of the relationship, and the impact of disclosure on survivor's recovery as well as on providers. Survivors were asked to tell the story of their unwanted sexual experience, at which point pre-assault alcohol use was disclosed to the interviewer. For the current study, we focused on just one aspect of the interviews: social reactions (positive, negative, mixed) to disclosure and appraisals of the reactions specifically related to the survivor's alcohol or substance use/abuse. We also looked at general social reactions codes (positive, negative, mixed) that may have also alluded to substance use and/or abuse in general among matched pairs where the survivor engaged in pre-assault alcohol use. We used these codes to try to get at any and all material that interviewees talked about that related to how others responded to their disclosures that included references to substance use/abuse. --- Data Analyses Each interview transcript was summarized to identify patterns and themes, and identified themes were discussed, as several trials of interview coding and refinement were used to develop a codebook covering individual interviews and themes reflecting the matched pair relationship. The codes were descriptive and summarized the primary topic of the excerpt (Saldana, 2012). We focused on substance use related social reactions codes applicable to survivors and support providers where they described the social reaction they received or gave, and how they appraised the quality of the reaction. We examined the context of codes by reviewing the transcripts and interviewer summaries when identifying relevant quotes. Atlas.ti Version 7 qualitative analysis software was used for coding and analysis. We identified codes that made the most analytic sense of the data (termed "focused" coding; Charmaz, 2006) and used them to code transcript segments. We coded the data separately and compared interpretations to achieve consensus (Eisikovits & Koren, 2010) in several phases. First, pairs of coders coded each interview matched pair using the codebook. Second, one coder in the pair reviewed both coded transcripts to identify inconsistencies in assigned codes. Third, coders met to discuss interpretations until reaching a joint consensus version (Patton, 2002). In cases where agreement could not be reached or more than one code applied, we used double coding (i.e. simultaneous coding; Saldana, 2012). Fourth, coded transcripts were reviewed by the original interviewer for any further corrections to assignment of codes, which were discussed with coders until consensus was reached. During the coding process, coders created memos within transcripts to highlight relationships or inconsistences within interviews or between interview pairs, or to capture unanticipated themes (Charmaz, 2006). Including memos allowed us to pay attention to the relationships between and among matched pair interviews. Analysis took place using an iterative process in several stages at both the individual level and matched pair level. First, queries were conducted in Atlas.ti software to identify the number of times each interviewee endorsed codes. Second, we individually reviewed quotes for each query to search for patterns. Third, we met several times to discuss identified themes and patterns, looking for similarities and contrasts within matched pairs and across the 19 dyads. --- Results --- Sample Characteristics This sample of alcohol-related assaults was based on survivors' who drank alcohol and/or used other substances prior to the assault (N = 19). Sixty-eight percent (n=13) were using alcohol at the time of the assault, and 32% (n=6) used both drugs and alcohol at the time of the assault. Forty-two percent (n=8) were not incapacitated, 11% (n=2) had difficulty speaking, 11% (n=2) had difficulty moving limbs, 11% (n=2) had difficulty walking, 5% (n=1) was asleep, 5% (n=1) were unconscious or blacked out, and three (15%) did not respond. Of survivors who were drinking at the time of the assault, 82% (n=14) of respondent/support providers knew about it and discussed this during the interview.. Of the 14 survivors with support providers who knew about their pre-assault alcohol use, only 11 survivors actually disclosed to the support provider (i.e., three support providers found out about the survivor's pre-assault alcohol use in another way). Among these 11 survivors, 9% (n=4) mentioned it briefly to the support provider, 18% (n=2) discussed it a little, 18% (n=2) talked in a general way, and 27% (n=3) talked about their pre-assault alcohol use in great detail. Ten survivors discussed during interviews that they were under the influence of alcohol or drugs at the time they disclosed their unwanted sexual experience. Thirteen (68%) survivors received social reactions specific to their alcohol use and discussed these social reactions during the interview. Twelve support providers (63%) discussed their social reactions with the survivor specifically based on pre-assault alcohol use. --- Social Reactions from different Relationship Types The sections that follow describe the social reactions (positive, mixed, negative) received by survivors in different types of relationships (significant others, family, friends). For counts of each social reaction pattern across the different relationship types, see Table 1. Significant Others-Significant others included spouses and romantic partners at the time of the interviews, or in the past, in cases of ex-husbands or ex-partners told about assault in the past. Positive reactions: Only two support providers of survivors of alcohol-related assaults were significant others. Of these two support providers, both responded to the survivor in a positive way, discussing in their interview how their past experiences allowed them to relate to the survivor and provide positive support. For example, one support provider witnessed his father abusing his mother when he was growing up. He explained in his interview that it is because of this experience that he knew how to approach his girlfriend when she disclosed her unwanted sexual experience. Another married matched pair discussed the positive support provided to one another related to the survivor's unwanted sexual experience and leaving her previously abusive relationship, but also in terms of providing support to one another for alcohol-abuse recovery. In this case, assault was viewed later on through the lens of assault risk being due to the survivor's drinking in the past, which was no longer a risk at the time of the interview. "Since I've been sober, and I kind of brought some of these things back up, sometimes, a few times, his reaction was like, 'What? You did what?' I say, 'yeah, don't you remember I told you?' (laughs) 'Remember I told you that?' and, um, what did he say, but he'll end it with 'ok, but that was in your past,' you know, so "as long as you don't...start back getting high anymore. I can, I know those things are not gonna happen." You know so." -S Mixed reactions: One survivor discussed the mixed and nonverbal reaction she received from her long-term romantic partner, where he reacted with little emotion, as the survivor recalls, and did not do or say much after the disclosure except hug the survivor at the end of the conversation. The survivor also recalled his body language communicating disgust at what she had told him. However, the support provider characterized his immediate reactions as shock, so this might explain the discrepancy between the pair. "...He was just like, and he had an unpleasant look like, like a distasteful look, like not like I'm distasteful or anything but like it's such a horror and that's when I told him I was like telling him all these things and I was like kind of remembering and just spewing off stuff and then afterwards when we got out of the car he just hugged me and I was like sorry you had to listen to that..." -S The survivor remembers being drunk during their first discussion of her victimization history and that she wanted to shock her partner with her disclosures. But she characterized him as someone who also has a history of drug and alcohol abuse and therefore he was not shocked by her disclosure. The pair has not had many follow-up discussions about the assault and the support provider expressed that he feels more discussions are not necessary. While her partner listens and asks some questions, which is positive, when the survivor becomes upset, he seems to view this with a mixture of empathy and exasperation. He tried to deflect attention away from the unwanted sexual experiences when the survivor brought them up as a way to redirect what he perceives as rumination on a subject that is stressful to the survivor. "...F (survivor) likes to talk about her experiences, in general. She likes an audience. She is an actor. And so she likes that, and I am kind of used to it. Umm, and the only times I might question her a little bit more are probably more when I think she's getting sad or upset. Ummm, which does happen from time to time. And then I might, you know, also try to change the topic." -SP Negative reactions: Many of the blaming reactions this survivor received centered on her history of drug and alcohol abuse. She was intoxicated at the time of one of the assaults and many family members blamed her for this. Her significant other, the interviewed support provider, believed he does not need to tell the survivor that the assault was not her fault because he does not think she blames herself. At the same time, her significant other recalled a conversation in which he gave the survivor a blaming reaction. He believed the survivor would be able to tell that he did not mean to blame her but that he is merely expressing surprise at her actions. "And I mean I was stunned, I was like "how the fuck could you go to C (neighborhood) at night and why would you do something like that?" I don't think I said that...But if I had, she would have, understand the context that I meant that in. You know, that's dangerous. It's not worth a bag of crack or whatever to do it. So I mean, I don't think even if I had phrased it that way, that that would have been particularly surprising to her cause she would know what I meant by that." -SP This appraisal by her significant other may or may not be the case, as when the survivor was asked in her interview if he told them about the drinking when talking to others about this experience, she said: "They knew, there's always alcohol present. I didn't stop drinking until I met [current partner] really in 2001, other than that, I was drunk every day." I: Do you think that had any impact? S: Oh yeah, they think I was drunk, put myself in that position, I got what I asked for." -S Significant others relied on their own experiences with abuse to provide positive social reactions to survivors. Blaming reactions were commonly focused on survivor substance use during the assault. These reactions could also be accompanied by supportive reactions and could be communicated (and interpreted) through non-verbal actions, highlighting the complexity of how both positive and negative reactions can actually be intermingled in the same response. Friends-Friends included current and past friends, a couple of whom had been romantic partners or spouses in the past but were now friends. Thirteen (68%) survivors of alcoholrelated sexual assault disclosed to friends. Positive reactions: One survivor described several unwanted sexual experiences, beginning when she was 18. When the survivor first disclosed to her friend about the most recent assault, the friend hugged her and responded positively. The support provider knew that alcohol was involved in the assault, and did not respond differently knowing that the survivor had been drinking. The support provider explained that she provided a positive reaction to the survivor, simply by being a person that the survivor could tell about her experiences. Another support provider discussed her reaction and, following disclosure, her seeking of religious guidance about how to best support the survivor after the survivor disclosed her history of alcohol/drug facilitated assault, CSA and IPV. "I just basically tried to listen and ummm make her as comfortable as I could. I actually started praying for God to use me to help her, whatever way I could...all I wanted her to see was that I wasn't judging her and trying to make her as comfortable as I possibly could, that was my objective and I don't know if I was successful but that's what I was trying to do." -SP This is the reaction from a support provider who is an Alcoholics Anonymous (AA) sponsor of the survivor, but they became friends. Another survivor described how she disclosed to her friend while they were out with other friends at a bar. A casual conversation about date rape drugs led to her disclosing that she had been the survivor of a drug-facilitated sexual assault. The support provider was shocked and sad (see paragraph below). This pair had a long history of friendship and the support provider seemed sad at how easily the survivor was able to discuss the assault, almost as though she was desensitized to it. The survivor picked up on her friend's genuine empathy for her and while she and her friend talked about getting revenge, it was almost light-hearted or joking. The survivor appreciated all of her friend's responses and interpreted them as the support provider caring for her in a way she had not previously realized. "Yeah, I was just like really sad. I remember wanting to cry...a lot because like, um, I don't know. Just somebody so close that happened...to, just, and she was like completely like matter-of-fact about it. I was like 'Ugh, ah, ahhhh!' Yeah." -SP The support provider does not remember his'revenge response' and even says there is no way he would have thought about revenge (not prompted by interviewer-all responses were confidential). The support provider also downplayed his response as being unremarkable and even 'filler supportive stuff.' This could be because he said he was shocked at the time, but did not treat her differently due to her drinking. Still, perhaps he felt his response was not adequate even though it had a profound positive impact on the survivor. This could be due to the context of their friendship in that they were not spending very much time together when this interview was conducted. The disclosure occurred while both people were intoxicated, so intoxication could explain why the two diverged in their accounts of the disclosure. "SP: Um, I, I'm sure I said, like the, the regular filler stuff just to like, you know, stall or pause for time or something so I could kinda collect myself, but it's like "oh, that's terrible, I'm so sorry, like. Who did blank, blank, blank, blank?" I'm sure it wasn't anything that I remember so...it must not have been that, um, remarkable. Or, it's just the regular, you know, filler supportive stuff." -SP This survivor appeared to perceive only getting positive reactions related to drinking, as when asked if anyone blamed her because of alcohol she said: "Fortunately, no and I'm really grateful for that. If someone would have said that to me I would have lost it (laughs). I would have went off [gotten angry] on them." -S Mixed reactions: Several (n=5) survivors of alcohol-related assault received mixed reactions from the friends they disclosed to. One friend described his surprise that the survivor was able to remain composed during the disclosure. "It was kind of weird how she portrayed the story to me. I was like 'wow was it really that bad?'[...] cuz I would be a little more traumatized by it." -SP Another support provider explained in his interview that he was unsure of how to react to the survivor's disclosure because they had both been drinking at the time, and the support provider was "very drunk." One friend discussed in her interview that her reaction was upsetting to the survivor because she provided a harsh response to her friend that she felt would help her in the end. "I told her somethin' she didn't wanna hear. I was like that's probably touched you because it's probably something you needed to hear. And she thought about it, oh she was so mad. She was mad at me. [...] They say the truth hurts. And it was something she needed to hear." -SP In this case, something that may have viewed negatively in the literature was ultimately interpreted positively by the survivor --even though it upset her at first, because she understood her friend was well-intentioned. Negative reactions: Five survivors of alcohol-related assault received negative reactions when disclosing to friends. There were some friends who had what they perceived were negative social reactions specific to the drinking however. In one case, a friend was even aware of the possible harm of asking her friend about the alcohol aspect of what was a violent alcohol related assault where the survivor, a teenager at the time, was choked by the offender into unconsciousness and did not know what had happened to her when she woke up in the house where they had both been partying with friends. The friend said that: "I had to get myself together to ask her the next question, cuz you know I didn't know if I should or if I shouldn't, but I did it anyways. Cuz I'm her friend and I'm concerned. And I asked her you know, was like uh, I asked her, do you remember or can you recall anything happening before you blacked out? Do you remember like coming to during? You know and she just, she was like no. She didn't ask any questions [about what happened to her]. She didn't want to even think about it. And she just picked up her stuff and left." -SP This sounds like she had good motivations of concern for her friend but also knew asking her friend such questions might not be the right thing to do given the sensitivity of alcoholrelated assault. The fact that the survivor left certainly doesn't indicate this was a positive reaction or at least not that she was willing or able to dialogue about it at the time. This friend also said in her interview that she does not try to be her friend's counselor and that the survivor is a caring, giving person, but being young did dumb things, but that that [assault] should not have had to happen. Then she said what sounded like a more positive non-blaming social reaction to the survivor: "I was trying to explain to her that no matter what you done or what's going on between you or your mother, this is not your fault because you want to go out and have a good time, you get a few drinks in you, and men take advantage of that, that's not... your fault. And I think she was having a hard time understanding that cuz it was maybe if I didn't go out, maybe if I didn't get drunk, or smoke marijuana maybe this wouldn't have happened to me." -SP Interestingly, some friends said "no" when asked by interviewers whether the alcohol part of the assault affected how they reacted, but then they would say things that showed in fact it may have impacted their reaction, or said it affects other people's reactions in general or how others responded to the survivor, but not their own reactions. According to the survivor, this one friend of many years was very supportive to her, more than anyone else in her life. Given that the survivor experienced a lot of self-blame and had negative reactions from others (e.g., police) or could not get really good support from others (e.g., family), having her friend not blame her and say it was not her fault was hopefully helpful. For example, according to the support provider: "I mean I would guess for other people in her life that probably did make a difference (overlap) in their reactions. I think, um, probably, especially like with her dad. And, um, probably with her ex-boyfriend. I would probably like sort of like illegitimize some of her part of, I mean people who have that reaction like my dad, that's like 'Well, you were drinking with a guy alone'" -SP Another friend explained how the combination of her friend's assault being by a mutual male friend of theirs and the assault being alcohol-involved led her to initially disbelieve the survivor and give a negative reaction. "I think if like she hadn't been drinking and he had been drinking though I do think that I would have believed her a lot sooner and been more supportive" -SP In her interview, the survivor explained she would not tell anyone she thought would blame her. "I feel like if I'm gonna tell somebody I was like assaulted, if I was gonna tell a person that means I trust them. And a person I trust would never be the type of asshole to be like 'oh well, you were drinking' so I would not even be friends with them. I would hate that kind of person." -S She also said that her friend (the support provider) was supportive, saying that drinking is not an excuse, that is just because someone is drinking does not mean you can assault them. She also did not think others blamed her for drinking either. She said of another guy friend: "He knew that we were drinking, yeah, he knew. And I don't think he had a problem with it. Like I don't think he was like oh, you shouldn't have been drinking. Because he knew he (the offender) was coming over and he knew that we were gonna drink because it was like summer and just sitting in the backyard drinking, like whatever. It really wasn't something; it wasn't even a concern. Like the concern was she just did this. You know what I mean? It wasn't like 'oh, well.'" -S So drinking was normative in her milieu, or the situation where the assault occurred, and was not something out of the ordinary that would have been used to blame her for being assaulted. When the interviewer asked more about discussing alcohol's role, the survivor spoke of her friend's reaction of anger but then interpreted it as caring about her: "I don't know, I've never really thought about it and now that I am thinking about it, it makes me nervous because what if he didn't bring it up because he thinks I'm an asshole. Well maybe, cuz actually, I know this is a terrible thing to do, but after that happened and I talked to him, like it was like seven o'clock in the morning where he is so I like woke him up and after I went to a bar by myself and drank a beer. And when I told him that he was mad at me for like going out and drinking alone, like again. Like not again, but just going out like right after that [the rape] happened and like drinking. So I don't, with (fiancé) I don't feel like hurt by the fact that, I'm sure he knew, but the fact that he wasn't like 'oh, so what you were drinking,' like to me obviously he doesn't care. He more cares like oh, are you okay, are you hurt, what happened, like what's going on, how are you feeling? Like that's caring more about like me than -you know what I mean?" -S Finally, another friend emphasized that the relationship was more important than alcohol to the social reactions she received, saying: "I think the relationship history really outweighed that [alcohol], so I don't really think that but, I know like, people weren't validating enough, like, you know what I mean cause of my relationship history with him they were like, invalidating, like that was what they were the worst about." -S Overall, friends' reactions to disclosure were perceived by survivors as mostly positive (n=9), even if their reactions contained elements of blame (verbal or non-verbal). More friends provided positive (n=2) than negative (n=0) social reactions specific to pre-assault alcohol. Support providers were almost always aware of the ways in which alcohol could play a role in their reaction. While not every support provider explicitly tried to mitigate the potential harm of survivors' heightened self-blame due to alcohol use prior to the assault, many tried to communicate positive support to survivors who they knew had been drinking or under the influence at the time of the assault. Family-Family included parents and now adult children, some of each who were survivors and in other cases support providers. Positive reactions: In a mother-daughter pair, both the survivor and support provider described their relationship positively and explained how they were open with one another about most aspects of their lives. The survivor experienced several unwanted sexual experiences throughout her life, many of which involved drugs and/or alcohol, as the survivor had struggled with drug addiction issues in the past. In the interview, it seems that the survivor's mother was there to help and care for the survivor throughout her life, but their relationship was better with the survivor recovered from her substance abuse issues. However, the survivor did not disclose her unwanted sexual experience to her mother until years later. The survivor described when she first told her mother about her unwanted sexual experience: "You know, she just tried to console me as much as she could...she had a really strong reaction." -S Similarly, the survivor's mother explained the pain she felt hearing about her daughter's unwanted sexual experience, but also expressed that what happened was in part due to her daughter's substance abuse issues: "It hurt me to see her like that...And then I cried about it, because I told her thank god that did nothing [more serious] happen to her...it was a very serious experience that she went through. And I hate this so bad, but there's nothing I can do, but talk to her about that and tell her don't get in that kind of situation." -SP Within another mother-daughter pair, the daughter recalled feeling angry and wanting to retaliate on behalf of her mother, but it does not seem as though she expressed these feelings to her mother based on her description below and how her mother understands the disclosure in the above quotes that she deliberately did not express those feelings to her mother. This showed remarkable presence of mind. The daughter was able to respond in the moment by stopping the conversation when her mother seemed overwhelmed and moved on to reassurance, empowerment ("don't be ashamed"), and telling her that being on drugs did not make her mother more blameworthy. The supportive, positive reactions from the young daughter to her mother fit within the larger context of their mother-daughter relationship. The mother struggled for years with drug abuse and the daughter took care of her other family members during that time, finally helping her mother stay sober for the last five years. The daughter remained supportive and loving of her mother and did not seem to have resentment toward her. There may have been elements of parentification in this pair, too, as having a drug-addicted parent is also a risk factor for children being parentified, or taking on more responsibility for their family members than is appropriate for the child's age and role within the family (Stein, Riedel,

Studies of informal support dyads are lacking to understand the disclosure of sexual assaults and social reactions informal supporters make to survivors. This study of 19 informal support dyads using interview data examined how three relationship types: significant others (i.e. romantic partners), family, and friends differ in social reactions to sexual assaults in the context of drinking or alcohol problems. It was expected that alcohol's role in responses to such disclosures would differ depending on relationship type as well as role alcohol played in the assault and/or in the survivor's life or those in her social network, including the perpetrator. Results show that alcohol has mixed effects and that alcohol-related assaults as well as contexts where survivors, perpetrators and/or their support networks have alcohol problems need further study to understand how such disclosures and social reactions occur and the impact they have on survivors, their relationships, and recovery.

but it does not seem as though she expressed these feelings to her mother based on her description below and how her mother understands the disclosure in the above quotes that she deliberately did not express those feelings to her mother. This showed remarkable presence of mind. The daughter was able to respond in the moment by stopping the conversation when her mother seemed overwhelmed and moved on to reassurance, empowerment ("don't be ashamed"), and telling her that being on drugs did not make her mother more blameworthy. The supportive, positive reactions from the young daughter to her mother fit within the larger context of their mother-daughter relationship. The mother struggled for years with drug abuse and the daughter took care of her other family members during that time, finally helping her mother stay sober for the last five years. The daughter remained supportive and loving of her mother and did not seem to have resentment toward her. There may have been elements of parentification in this pair, too, as having a drug-addicted parent is also a risk factor for children being parentified, or taking on more responsibility for their family members than is appropriate for the child's age and role within the family (Stein, Riedel, & Rotheram-Borus, 1999). "And I love my brother to death no matter where he came from, he my brother. You know, and I always told my mother don't be ashamed. You know, don't hide in the corner and ball up. You know, you didn't ask for that. That's something that somebody does to you that you didn't ask for. You know, whether if you was under the influence, under drugs or whatever happened, you shouldn't have got that done to you."-SP In another case, the son was very supportive to his mother who told him about her assault saying: "He was like, I told him I was drinking and then he was like 'that doesn't matter.' He was like, you know, even if you were drinking that doesn't give someone the right to touch you. He was like you still have rights as a human being." -S Mixed reactions: Other survivors had reactions that were more ambiguous and perhaps mixed in nature or indirectly blaming, such as a daughter who turned 21 and was drinking when she was assaulted. She had told her mother about prior assaults and did not want to tell her about the most recent one as she didn't feel comfortable talking to her. This daughter also had mental illness that may have left her more vulnerable to repeated assaults she had experienced. Regarding not telling her mother about the recent assault, she said: "You know, so but I don't think she would say anything too bad, just like you shouldn't have been drinking before you (overlap) 21, but not anything specific, just more so the same sort of thing like "oh, you know, you think she would have said you should have been at your grandma's house. Not that it was your fault that it happened to you, but you still should have been at your grandma's house." -S In the interview, discussion focused on how the expected negative reaction was not directly blaming of the survivor, yet as a form of indirect blame or double message (e.g., it's not your fault it happened but you should have been in a different place at the time) led her to refrain from disclosing. --- Negative reactions: In other cases, social reactions were not based on the assault but other factors like the survivor's drinking history. One survivor was not a drinker, so disclosing an alcohol-related sexual assault led to disbelief by most people she told, including her husband. "I told him [husband] that we had been, you know, and that was, like I said, that was one, the first reaction was, 'You don't even drink.'" -S Other questioning from her (now) ex-husband appeared to be reappraised positively by the survivor as him as being supportive and/or perhaps she was rationalizing and/or defending the support provider, now her friend. "The only thing my ex-husband said was, we had been together, friends, we had been, you know, on and off friends, only thing he said was like, 'Did he give you the alcohol? Did you drink it on your own?' You know, 'Where did you-where did the alcohol come from?' And you know, questions like that, stuff like that, crazy stuff like that, then..." -S When the interviewer queried why do you think he was asking those kinds of questions? The survivor replied: "At first I thought he was like attacking me, but then when we kept talking, I knew that he was just trying to get to the bottom of, to straighten it out in his own mind, you know, and I always respected him about, he never blamed, you know, he neverif he did, he never said it, and after that he just tried to make it easier for me." -S The survivor appears here to be struggling to make sense of his negative reaction to her, which was inconsistent with her generally positive feelings about him and that he never blamed her and was supportive afterwards. In some cases, negative reactions were avoided that survivors knew they would get from family if they disclosed, as one survivor explained. "The alcohol and the drugs didn't make a difference with the family...with him (emphasized), you know, cuz I said, I never did discuss it with them [the family], the other stuff. You know, they don't wanna know I was doing drugs and in an alley and got raped. That would make you looking worser." -S Similar to the previous types of relationships, familial pairs had mixed reactions that were mostly interpreted as positive by the survivors. Family members, who tended to have long histories with one another, were more likely than other relationship types to view the role of alcohol in the assault and their subsequent response to disclosure as linked to survivors' longer struggles with addiction and substance abuse. These long, familial relationships also change over time, thus the way survivors perceive alcohol-related social reactions is a dynamic process that unfolds over time within the context of the relationship. --- Discussion The present study is the first to examine dyads of ethnically diverse survivors and informal support persons (friends, family, significant others) they told about alcohol-related assaults. Dyadic data can yield greater insight into social reactions from both survivor and supporter perspectives, a growing area of qualitative interview research (see Eisikovits & Koren, 2010 for a review regarding qualitative dyadic research methods). This qualitative study illuminates prior work in various ways by examining alcohol-related assaults separately according to relationship between survivors and supporters (family, friends, significant others). While results showed both positive and negative reactions to these assaults, specific responses related to drinking of survivors were salient in some cases, with some asking survivors about why they were drinking, and supporters articulating the idea that if women take risks, like drinking, they can expect to be assaulted. This is consistent with societal rape myths that shift blame from the perpetrator to the survivor (Suarez & Gadalla, 2010), and suggest the presence of rape myths specific to alcohol use where survivors are blamed for engaging in pre-assault drinking (e.g., suggesting that women who drink are behaving in ways that lead to victimization; Edwards et al., 2011). Given the small n's it is not possible to determine whether friends were more supportive in their responses to alcohol related assaults. Past research shows that female friends are most supportive to survivors in general (Ahrens & Aldana, 2012), especially those who had experienced sexual assault themselves. However this was not specific to alcohol-related assaults, so more research is needed to determine whether friends, and female friends in particular, are more supportive than family and significant others in response to these assaults. Support providers in this sample were largely female friends of survivors and provided no negative reactions directly related to the survivor's pre-assault alcohol use. Conversely, family and significant others were more upset, angry, and questioning of survivors. While overt blame was uncommon, implied blame and mixed messages about blame (mixed social reactions) occurred. Significant others did not provide any reactions specific to pre-assault alcohol use, but there were only two significant others in this sample. Survivors may not disclose pre-assault alcohol use to significant others and family members as frequently as friends, perhaps over concerns of receiving hurtful responses, which was the case of one survivor who discussed her fears of being blamed by her family for having been under the influence at the time of the assault. So, while the present study found differences in social reactions based on the relationship type -in support of previous research -future research should examine relationship differences in these types of reactions using a larger sample. With friends and significant others, relationship history of the survivor and perpetrator appeared to outweigh alcohol in judgments that were made, although this was only explicitly mentioned by one support provider, Negative reactions may have resulted in that case from the survivor being friends with the perpetrator or involved with him previously. Interestingly, when drinking was normative for the survivor or their social milieu, it was not used to blame the survivor in particular or at least that was survivors' perceptions. It makes sense that contextual factors are considered in making judgments about sexual assault, as survivors and support providers attempt to make sense of what occurred and why. Unfortunately, rape myths may influence how people view whether victims are blameworthy, even though supporters and survivors themselves may not always be conscious of this occurring (Edwards et al., 2011). On the positive side, survivors may give the benefit of the doubt to their close relationship partners a break and positively appraise what appear to be negative reactions in light of their knowledge that the person cares for them and wants what's best for the survivor. In some cases support providers mentioned this explicitly in framing their negative reactions as not being harmful because the survivor knew what they meant or knew they had good intentions towards the survivor. In close relationships, the relationship history and inferred intentions of the support provider are likely to overwhelm the actual social reactions they make, as survivors look at them in light of all they know about their loved one or close friend (Dunkel-Schetter & Skokan, 1990). Family dyads, that were mostly parent-child relationships in this study, were particularly illuminating in that children with parents who were survivors struggled to deal with the impact of the assault on themselves and to put on a strong face and be supportive to their parents, which appeared to lead to parentification or role-reversal in some cases. In cases with parent supporters of their adult children who had been assaulted, they struggled to support their daughters but often also made judgements of their drinking behavior and/or coping post-assault, so support was perceived as mixed at best in some of these relationships. More research is needed to understand not only long-term effects on parents and children of disclosure and reactions to sexual assault, but also how this traumatic event may alter the dynamics of these relationships in ways that has deleterious effects on both survivors and supporters. Some negative reactions were reappraised as positive by survivors, confirming past research showing that survivor's perceptions of the support provider's intentions and/or simply seeing negative reactions in a positive light was common (Ahrens & Aldana, 2012). It is possible that reappraisals occur on an ongoing basis as disclosures unfold over time and more than one discussion occurs following the assault, which may also be influenced by how the survivor feels about herself and the assault and whether she was to blame. Given that we know that self-blame generally declines over time in sexual assault survivors (Koss & Figueredo, 2004), which then drives recovery (i.e., symptom reduction), we need further research on the process of disclosures, social reactions, and appraisals of both to better understand the dynamic, unfolding process over time within the context of different interpersonal relationships. Some support providers talked about other people's attitudes about alcohol-related rape and rape myths while not endorsing those attitudes themselves. Sometimes, it was hard to know how accurate they were about their social reactions to survivors, especially when survivors told us there were negative reactions that their support provider did not discuss in their interview. This finding makes sense in light of Davis and Brickman's (1996) research showing that significant others and rape survivors rated levels of supportive behavior (from significant others to survivors) the same, but survivors rated unsupportive behavior as higher than significant others' ratings of their own unsupportive behavior. Some of the survivors seemed to get negative reactions and/or engage in self-blame due to their own drinking history or alcohol abuse, which was also discussed as being part of what led to some assaults. In such cases, there may be a double stigma with survivor blame by others including support providers and survivor self-blame for such assaults being compounded for survivors. This was often rather subtle, but could be seen when reading between the lines in some of the interview transcripts. For example, a survivor would be told 'it's not your fault that the assault happened', but was also told they should not have been at the bar or been drinking, implying her behavior did make her culpable for the assault. Likely the experience for these survivors is worse or at least different than for others, especially as they may have also experienced more assaults, due to their higher risk behavior. One spouse was particularly angry at his partner who told him about her assault but then went right back to the bar after being assaulted, as drinking was normative for her but also her way of coping with the assault. Anger at survivor's risk-taking and coping behaviors may have been evoked in those close to survivors where support providers lacked control over the survivor's behavior that increased their risk of being assaulted. This is consistent with research showing that how survivors are coping and how symptomatic they are affects how they are treated by social network members (Silver, Wortman, & Crofton, 1990;Winkel & Koppelaar, 1991). --- Limitations and future directions While survivors were drinking, it is not clear in all cases how impaired survivors actually were when assaulted, so level of impairment should be assessed to understand how that affects survivors' memories both of their assaults, disclosures, and social reactions. This is vital as past research has shown from survivors (Littleton et al., 2009) that some may simply not recall what happened, including who they told and how they reacted due to poor memory, even if they no longer drink and/or because it was long ago. It is important to say that just because memory is fuzzy or absent, does not mean survivors are not traumatized by being sexually assaulted, and some data suggest more self-blame and PTSD over time for those who drink prior to assault (Kaysen et al., 2010;Read, Bachrach, Wright, & Colder, 2016;Peter-Hagene & Ullman, in press). Appraisals of social reactions may also change and be reflected in discussions of the assault over time between the survivor and support provider, as some interviewees alluded to. This is part of the process of recovery but may also reflect her relationship with the support provider and feelings about the person and the relationship at the time of the interview. This is part of why collecting dyadic data is so important to provide a more dynamic portrait of disclosure and reactions in the context of relationships that are dynamic over time (Eisikovits & Koren, 2010). This study was limited by being a small, volunteer, convenience sample of dyadic matched pairs that most certainly were more positive in their relationships, disclosures and social reactions by virtue of still being in relationships and willingness to do an interview on this topic. Support providers were contacted through survivor-provided contact information to do an interview about the survivor's sexual assault experience and the disclosure and reactions she received from others, so the interview was primarily about their relationship with the survivor and her experience and not centered on their own experiences beyond helping survivors, given the nature of how they were recruited. Future research is needed of dyadic matched pairs of survivors and supporters recruited simultaneously with more balance of asking supporters about their own experiences. Additionally, only one support provider was interviewed, but survivors often told multiple sources and may have received different reactions specific to their pre-assault alcohol use. However, limiting interviews to one support provider allowed for the survivor to provide more depth in responses related to the single support provider. This study is also limited by the retrospective design. Lapses in memory/recall could have influenced accounts of social reactions received and provided by matched pairs. Time elapsed since the assault and participating in this research may influence how women interpret these reactions and how support providers recall reacting to the disclosure. Most assaults occurred quite a while ago, with an average of 12 years since the assault (M=12.3; SD=11.7). While survivors and support providers were willing to talk about social reactions related to drinking prior to sexual assault and/or survivors' problem drinking, this is a loaded topic and some interviewees, typically support providers, were likely highly sensitized to even the appearance of blaming the survivor or perpetuating rape myths. Given that they knew they were being interviewed about survivors' experiences and brought in via the survivor, they may have also been more apprehensive and cautious about what they said, even though confidentiality was promised to both survivors and supporters. Naturally, some pairs may have talked to each other prior to or after interviews and participating may have affected their future discussions about sexual assault and their relationships. However, given that talking about sexual assault appears to be helpful to survivors with few reporting negative effects of participation and some positive effects on help-seeking (Kirkner, Relyea, & Ullman, in press), such a study may be an opportunity for dyadic matched pairs to reflect on this topic as well as their relationships. --- Author Manuscript Ullman et al. Page 20

Studies of informal support dyads are lacking to understand the disclosure of sexual assaults and social reactions informal supporters make to survivors. This study of 19 informal support dyads using interview data examined how three relationship types: significant others (i.e. romantic partners), family, and friends differ in social reactions to sexual assaults in the context of drinking or alcohol problems. It was expected that alcohol's role in responses to such disclosures would differ depending on relationship type as well as role alcohol played in the assault and/or in the survivor's life or those in her social network, including the perpetrator. Results show that alcohol has mixed effects and that alcohol-related assaults as well as contexts where survivors, perpetrators and/or their support networks have alcohol problems need further study to understand how such disclosures and social reactions occur and the impact they have on survivors, their relationships, and recovery.

Introduction There are currently around 350,000 people registered either Severely Sight Impaired (SSI) or Sight Impaired (SI) in England, Scotland and Wales (NHS Information Centre, 2011, National Statistics, 2010, National Statistics, 2012). However, research suggests that more than one in ten people aged over 75 in England, Scotland and Wales have a level of visual impairment that would make them eligible for registration (Evans, 2005). If these were all registered the number of people aged over 75 registered in England and Wales alone would increase to over 500,000 (based on 2011 Census data). There is particularly strong evidence for the impacts of visual impairment on older people, for whom it has been shown to impact negatively on quality of life (Mitchell & Bradley, 2006); to be associated with reduced levels of social participation (Alma et al., 2011) and social support (Burmedi et al., 2002a, Verstraten et al., 2005); and to be an important risk factor for falls (Legood et al., 2002). There is also a well-established link between visual impairment and depression (Burmedi et al., 2002b, Evans et al., 2007, Chou, 2008). Although there is less evidence for the impacts of visual impairment on working-age adults, a similar picture seems to exist as for older people (Nyman et al., 2010). Registration, particularly as SSI, can be the key to accessing vital financial and social support to keep people active, healthy and engaged. To be registered, an individual must first be formally assessed and certified by a consultant ophthalmologist. At the point of diagnosis or certification, they may be directed to an Eye Clinic Liaison Officer (ECLO) (also referred to as a Sight Loss Adviser) or other early intervention support worker who can provide initial support and advice, including information about other sources of support. The role of ECLO was first established by the Royal National Institute of Blind People (RNIB) in the mid-1990s (Subramanian et al., 2011) but has expanded significantly in recent years, so that there are now approximately 120 ECLOs and about the same number of similar early intervention posts in hospitals throughout the UK (RNIB, 2015a), although these are thought to cover only 30% of eye departments in the UK (RNIB, 2015b). However, these services have not been established as a statutory element of eye health care, and there is great variation in the way they are set up; from full-time professional, fully funded positions to part-time posts held by volunteers. Furthermore, with most ECLO posts reported to be funded on an annual basis, there is ongoing concern for their future existence (RNIB, 2014). Once an individual has been certified, their statutory Social Services Department (SSD) is notified and they should then be contacted by a social worker asking them if they wish to be registered as SI/SSI. They should also be asked about their support needs, and offered an assessment. Following registration, or referral to Social Services, individuals may be offered a range of low vision (LV) services including rehabilitation; low vision aids (LVAs); advice about welfare benefits; and emotional support. These forms of support can be vital in enabling people to maintain their independence and quality of life. However, although there is agreement about what should be included in such services, there is no standard model of provision (NHS Eye Care Services Programme, 2007). Thus there is considerable variation in the level of service offered and in the ways in which services are configured in the UK (Dickinson et al., 2011). This paper explores this landscape of complex service provision through the experiences of people with a visual impairment. It is based on research conducted in North West England and North Wales by researchers from the University of Liverpool. The aims of the research were to explore the changing support needs of people with a longstanding visual impairment from the clinical, visual function and quality of life perspectives. --- Methods The research was conducted in two phases; the first in 2007, the second in 2010. Findings from the first phase have been published separately (Thetford et al., 2009;2011). This paper draws primarily on the qualitative findings from the second phase, focusing particularly on participants' experiences in the three years between the two phases of the study. In the first phase, qualitative data were derived from narrative interviews (Wengraf, 2005, Wengraf, 2001); in the second phase, from semi-structured interviews. Separate ethical approvals were obtained for each phase: for the first from a NHS Research Ethics Committee (REC); for the second from a university REC. Thirty seven people with a visual impairment were interviewed for the first phase of the study; all were living with sight loss that was uncorrectable with glasses or contact lenses. Participants were recruited via local voluntary organisations working with people with a visual impairment. Consent was sought from participants to be contacted about taking part in a follow-up interview: all consented to this. In 2010 participants were invited to take part in a follow-up interview, with the aim of finding out how their needs had changed in the intervening three years. Twenty one (57%) of the original sample of 37 consented to be re-interviewed. Six people declined to take part; eight had died; and two were untraceable. Informed consent was taken, either written or audio-recorded, at the start of each interview. The interviews were digitally audio-recorded, transcribed and anonymised. A thematic approach to analysis of the data from the second phase of the study was developed by the research team. This approach needed to be broad in scope, to allow the data to be explored from a number of different angles relevant to the study. It also needed to allow the findings from the second phase of the study to be integrated and compared with those from the first. In order to build 'inter-rater reliability' into the analysis (Armstrong et al., 1997) three members of the research team independently coded two of the transcripts, identifying segments of text that seemed to carry interest or meaning in relation to the research topic. This process showed there to be a high degree of convergence across the three coders. Differences tended to be in the breadth of coding adopted, for example, whether to code bereavements, major illnesses and changes in domestic circumstances under separate codes or under a single code of 'life events/changes'. It was decided to adopt a broader coding strategy, given the relatively small sample size, and the fact that visual impairment was the central issue around which other issues were being examined. The codes developed as a result of this process were then mapped onto the codes used in the original study and a thematic framework developed (Spencer et al., 2003). The first author then used this thematic framework to code all the transcripts, using the NVivo qualitative software package. The coded data were then reviewed, refined and sorted into a set of final thematic areas that addressed different aspects of the study. --- Findings --- Participants Participants were initially recruited by advertising the study through voluntary organisations providing services to people with visual impairments in Merseyside, Cheshire and North Wales. The final sample was drawn from a mix of urban and rural localities across these areas. Although we do not claim that participants form a statistically representative sample of people living with a visual impairment, they do reflect a range of characteristics and experiences not untypical of this population. The gender composition of participants in the second phase was less balanced than in the first, (13 women, 62%, compared to 19 women, 51%, respectively). The mean age of participants in the second phase was 72, which, allowing for the elapse of three years, was similar to that in the first (68). In the first phase, 34 of the 37 participants classed themselves as White British, compared to 20 of the 21 participants in the second. The employment status of participants was similar in both phases, with 16 (76%) of the 21 participants in the second phase describing themselves as retired, compared with 28 (76%) in the first. Findings related to a number of thematic areas: changing visual health; the changing context of people's lives; and accessing support. Full findings from the study are reported elsewhere (Hodge et al., 2011). In this paper we report findings relating to participants' access to formal systems of health and social care provision. --- Clinical services One of the main issues highlighted in the first phase of the study was, for many people, the lack of regular, ongoing eye clinic appointments once they had received an initial diagnosis. Recent developments in the treatment of eye conditions, particularly wet age-related macular degeneration (AMD), have improved the outlook for many patients. However, at the time that most of the study participants were diagnosed, the options were much more limited. Many, particularly those with wet AMD, were discharged at the point of diagnosis, left with the message that nothing could be done for them. By the time of the follow-up interviews little had changed: with some exceptions, those who had been attending regular eye clinic appointments at the time of their first interview were still doing so; those who were not then were still not. One exception was Maureen who, at the time of the first interviews, had not seen an ophthalmologist for several years. However, she had recently experienced a medical condition which had affected her sight temporarily and as a result she now found herself back 'in the system': I seem to be like caught up back in the system, because they wanted to discharge me last time I went, well the registrar did. And then when the consultant came over he said 'oh no, we'll ask for more tests to be done' or whatever. And when I spoke to a lot of the people in the walking group they were saying don't let them discharge you because things like, you know there's new tests coming up all the time... Maureen, 58 Like Maureen, other participants recognised the importance of staying 'in the system', in order to continue having their vision monitored. Even when there was nothing further that could be done to improve their sight, those who were receiving regular monitoring wanted to continue doing so in case new treatments became available, and, more basically, in order not to be abandoned by services: I imagine once I've had both these cataract operations done I'll be discharged and that will be the finish of it. (...) I'd much prefer to still be on their books and even if it's 12 months ahead have another appointment to go back. I think it's partly knowing that you still belong. If you're discharged you're thrown out, you're finished with, you're put on the scrap heap. And that's very depressing. Isabel, 84 What these findings reflect is a sense of the system of ophthalmological care being an important source not only of hope, but also of affirmation of the individual's social and personal worth. Another important function of clinical care is that it can enable access to other services such as rehabilitation. However, in the first interviews it was clear that not only had many people been discharged with the message that nothing could be done for them clinically, but they had not received any information about other services that could help them (Thetford et al., 2009;2011). By the time of the follow-up interviews, there had been one positive development: the establishment of ECLO posts at all three hospitals used by participants in the study; at the time of the first interviews only one of the three had an ECLO. This was commented on positively; however, it was perceived to be primarily targeted at newly diagnosed patients: I believe they now have somebody in their clinic two days a week that's there, well for support to give like new patients to tell them about [voluntary organisation] and other things like that. But I've only heard about this second or third hand, I've not actually been told about it. The clinic, I think they still fall down by not giving you information, access to information, telling you about things that are going on. Isabel, 84 --- Community optometry Another key element of clinical care is community optometry. One of the key roles optometrists play is in identifying secondary, and potentially treatable, eye conditions, such as cataracts and glaucoma. Where people had a good relationship with their local optometrist, this could be a valuable source of advice, as in the case of Jill, who developed a detached retina: I was still going to the optician's yearly and of course I got frightened and to be honest with you I didn't quite know what to do so I thought well there's nothing they can do till the morning in any case so I rang the optician up and she said go straight to [town] hospital. Jill, 86 Community optometrists can thus be a valuable element of the clinical care pathway. However, whilst some participants' experiences of optometry services were very positive, others reported more mixed experiences: I used to go to [high street chain of opticians], great, and then all of a sudden they started doing it like sort of mass produced, I'm not even sure you see an actual optometrist, I don't know who the hell you see but it's a bit like being in a factory you know (...) They gave me glasses that I couldn't read with, they couldn't seem to get the right sort of distance. Maureen, 58 --- Registration Registering as SI or SSI can be the key to accessing social care services in the UK. However, findings from the first interviews showed that not everyone who was eligible for registration was registered (Thetford et al., 2011). By the time of the follow-up interviews little had changed in terms of participants' registration status, with the majority registered as either SI (n=11) or SSI (n=7), and only one individual moving from SI to SSI. However, becoming registered does not necessarily mean that people are automatically directed to the services and benefits that this entitles them to. Many years after first being registered there were still examples in our study of people only recently finding out about benefits that they would have been entitled to years before, for example: 'I don't know how people sort of like find like the Disability Living Allowance. I may have been entitled to that 12 years ago when I first registered as blind but I didn't know about it...' (Maureen, 58). There was also evidence that some participants' registration status needed to be reviewed, as their level of impairment had worsened. Frank had been registered SI for over a decade, but his eyesight had deteriorated significantly: 'I really should go for a reassessment. I can barely see anything now'. He had been discharged by his ophthalmologist some years before, so in order to be reassessed and registered as SSI he would need to be referred back. However, he had other physical health problems which meant that his sight loss was not his main concern. He was, however, in regular contact with his GP, who should have been aware of his deteriorating vision and could have referred him back to an ophthalmologist. --- Social services Local authority SSDs are the main statutory bodies responsible for ensuring social care is provided to people registered as SI or SSI. Most people in the second phase of the study were registered as SI or SSI and so would have had some contact with either a social worker or a rehabilitation officer at some point. However, models of provision varied considerably across the localities covered in this study, as across the UK. In two areas, SSDs provided assessment and rehabilitation services, but in others these were contracted to voluntary sector providers. In the first interviews there was a high level of dissatisfaction with the accessibility of SSDs, with a variety of problems highlighted; three years later the same problems were reported. These included long waiting times, not knowing who to contact, limited contact time, limited follow-up and lack of resources:...this liaison officer I saw at the hospital, she wants me to see a social worker again. Because I haven't seen one since I was first diagnosed. But she phoned me up last week and she's got an eight week waiting list. So, it will be well into the New Year before she can come. But, I won't qualify for help in the house or anything like that, because my savings are still too high. Teresa, 72 For Teresa the problem was mainly the long waiting time to see a social worker, but also the fact that her main need was for help with cleaning that she was unlikely to be able to access from Social Services. Frank similarly needed home help but had also found that his SSD did not provide this. He also expressed a wish for some ongoing contact and emotional support. However, the only contact he had had with Social Services had been a one-off visit to arrange for an adaptation to his bath because of his other health problems: It would be nice to have someone to talk to and let all your problems out wouldn't it? I did have a lady here from Social Services, but I think she was rather keen to get away you know, she arranged to have a chair lift in the bath, but I haven't seen her since. Frank Similarly Eva had had no contact with Social Services since she was first registered:...they do advise you but obviously they've got so much to do that unless they really -they obviously can't be here, there and everywhere at the same time can they, so I, I mean I know at least I know there is somebody there that I could call upon if I really were in need. Eva, 75 Eva was increasingly dependent on her husband, and her quality of life and emotional well-being were badly affected by her deteriorating eyesight. Although she was aware that she could contact Social Services she did not feel that her support needs were significant enough to warrant it. Several participants described how, after an initial contact when they were first registered, it was left up to them to make contact should they need anything in the future. Interestingly, two participants in the follow-up interviews had received valuable help from Social Services since their first interviews, but in both cases this was not because of their visual impairment. In his first interview Andrew described being badly affected by his situation and felt as though he had been abandoned by statutory services, being eligible only for registration as SI, not SSI, limiting his access to benefits that he felt would have helped him. However, by the time of his follow-up interview Andrew's situation had improved significantly, largely because he was now receiving support from a Local Authority mental health service; he had received no further support specifically related to his visual impairment. Rita had also benefitted from input from Social Services since her first interview, but again, not because of her visual impairment, but because she had had surgery on her foot. She explained that Social Services had provided a range of aids and adaptations that had never been offered on the grounds of her visual impairment, despite being registered SSI: So we had Zimmer frames, high toilet seats and toilet frames and commodes and you name it, and they even came and put a hand rail at the top of the stairs, one going down the two stairs at the far end of the landing, one going all the way round to the wash house, but none of this because I was blind, because I'd had an operation on my foot! Rita, 72 Two individuals were unusual in receiving visual impairment related support from Social Services, although in both cases the support provided was to some degree 'unofficial'. Richard had established a good relationship with his social worker over a number of years, and despite the fact that the initial reason for the support (learning Braille) no longer applied, he was continuing to receive support: 'Oh I still speak to the social worker yes. I mean it technically hasn't stopped, she's still technically teaching me Braille but we don't seem to do much Braille' (Richard, 59). Peter's support came from a rehabilitation officer: 'I was put in touch with [Social Services] when I first was registered. I was partially sighted but I still had a rehabilitation officer, actually it wasn't Bob it was somebody else, but Bob's been there a long time now' (Peter, 64). The reason for Peter's support was mainly because the rehabilitation officer who supported him was a former colleague. --- The voluntary sector For the majority of participants, the voluntary sector figured much more prominently than the statutory sector in the networks of social support they had established. This was partly due to the fact that across the area covered by the study the voluntary sector played a significant role, with two large organisations involved in providing services to people with a visual impairment. Furthermore, study participants had been recruited via these organisations. Each provided a similar range of services including assessment and rehabilitation, training, equipment and activities. However, although both organisations provided LVAs, in the area covered by one of them they were free of charge whereas those living outside this area had to pay for them. Voluntary organisations were also an important source of information e.g. about benefits and new treatments: '...the Macular Group in [city] are quite good, they get somebody out, certainly every couple of years to talk about benefits and what is available' (Isabel, 84); 'I'm a member of the RP [retinitis pigmentosa] Society, so if things come out you know I do hear sort of, yes through them usually if there's anything happening in the way of treatments...'(Richard,59). This practical information and advice also came through involvement in support groups: We just play games, have a cup of tea and a bit of a chat and a biscuit, it's only for a couple of hours but it gets you out, doesn't it. You listen to their problems and they listen to yours. One of the most important aspects of the voluntary sector was the opportunity it provided for mutual support which, for many people, had been crucial in helping them to adjust to their visual impairment: I loved people and I loved going out but you're completely lost then, you're really lost and then my daughter seen a cutting in the paper for the [club] in [town]. And I was in that for quite a long time, for partially sighted and blind people, and I found that got me seeing other people in similar circumstances as myself. And I think that's the biggest shock is losing your sight and accepting it is the bigger part isn't it? Lily, 82 Four of the participants in the follow-up study also worked as volunteers, either providing support or campaigning for improvements in services and policy. For those who were no longer able to continue in paid employment because of their visual impairment, being able to have an active role as a volunteer was important to their sense of identity. --- Networks of support A key theme in both phases of the study was the importance of the networks of formal and informal support that participants built. This often involved the individual going to considerable efforts to manage a complex web of arrangements. For example, Isabel was in her 80s and at the time of the first interviews was living with her husband, but by the time of the follow-up interviews her husband had died. Although she reported having struggled on her own at first, she had eventually developed a good network of support. She used a home delivery service for shopping and a community transport service to get her to regular social activities. Her daughter lived nearby and although she received some informal support from her, she was keen not to be dependent on her. In one way, stories such as Isabel's are positive; there are services and resources that can enable people to live fulfilled and independent lives. However, in order to access them, the individual needs to be resourceful, motivated and informed. For many people, the complex network of agencies involved in delivering services is confusing and overwhelming. Several individuals in the follow-up study were not accessing the services they needed, because they did not know how to, were unwilling to seek out further support or because those services simply did not exist in the area they lived in. --- Discussion The findings reported here create a picture of a system of variable support for people with visual impairments, access to which is largely dependent on the initiative and good fortune of the individuals themselves. Participants in the study described networks of support involving varying degrees of input from formal health and social care services, informal sources such as family and friends and voluntary sector organisations. The voluntary sector was particularly important, both in facilitating mutual support and opportunities for people to use their skills and experience as volunteers and as a provider of essential support services and information. Local authority SSDs generally played a much less significant role in participants' support networks. Where participants reported positive examples of good practice, these tended to be due to the particular circumstances of the individuals themselves, and especially to their relationships with individual workers who were able to exercise their discretion as'street-level bureaucrats' (Lipsky, 2010) to provide valuable, but not necessarily formally sanctioned social care. The key element in the system of health and social care provision for most people with a visual impairment is the clinical care system. Not only can this be a source of hope and personal validation; it also represents the gateway to other forms of support through the registration system. Yet the disparity between the numbers of people registered and the numbers actually living with a visual impairment suggests that this system is not functioning properly. The declining numbers of people registered is a particular source of concern (Barry & Murray, 2005, Boyce, 2012). This may be due partly to the fact that there are several points on the route to registration at which people might'slip through the net', and to the fact that all the processes involved can be slow and difficult (Boyce, 2012). However, as a starting point, it is important that people are made aware of the benefits of registration. Because access to services is so dependent on the individual agency of the person with a visual impairment, information plays a vital role. One of the key findings from the first phase of the study was the lack of information given to people when they were first diagnosed or registered (Thetford et al., 2009). A similar finding has emerged from research by Beverley et al. (2011) which shows that people with a visual impairment obtain most of the information they need not from health care professionals but independently from a range of sources. The rolling out of ECLO services has been an important development that should help to improve access to information about the registration process and sources of support, although evidence of their impact is as yet limited (Conway et al., 2012, Subramanian et al., 2012, Subramanian et al., 2011) and concerns exist about their future funding (RNIB, 2014). However, they are targeted primarily at people newly diagnosed. What the findings from this study highlight is the importance of ensuring that those who were diagnosed before these services were in place are made aware of their existence, and indeed that all patients with a visual impairment attending outpatient appointments are reminded about them. This is because diagnosis is often a shocking event (Thurston et al., 2010, Hodge et al., 2013), as is the point of certification/registration (Boyce, 2012), and not always the time at which people are ready to receive information and support. If anything, the growth in early intervention services reinforces one of the key messages from the first phase of our study; the need for people to be given multiple opportunities to receive information and access services, as they adjust to having a visual impairment, and as their support needs change (Thetford et al., 2009). The follow-up phase of our study provides further evidence of those changing support needs. The first phase of our study, looking back over the whole of people's experience since they became visually impaired, showed that access to rehabilitation and social support had been limited. The second phase shows that little had changed. Those who had access to support in 2007 were still accessing it, but those who had received little formal support by 2007 had still not received any. This may be part of a wider issue, which is the lack of uptake of health and social services by older people. The reasons for this include the perceived lack of accessibility, appropriateness and affordability of services; as well as a tendency to view the effects of ageing as inevitable, which may deter people from seeking help (Howse et al., 2004). Thus there is a risk that those who do not actively seek out support are left to their own devices. However, an appearance of coping can mask a situation in which the individual is having increasing difficulty with daily living tasks, becoming socially isolated and, potentially, depressed. Certainly in the three years between the first and second phases of this study, people's needs had changed, with some participants finding certain daily living tasks increasingly difficult as their sight deteriorated and their social circumstances changed. For some this meant having to rely more on support from relatives and friends, but many were drawing on a mix of formal and informal support. However, negotiating their way around the network of services available involved a significant degree of resourcefulness that not all those in our study were in a position to muster. Reinforcing the message of previous research, this study highlights the importance of people's support needs being regularly reviewed (Percival, 2003). As a minimum, we suggest that SSDs, or agencies holding the SI/SSI registers, should make annual contact with those registered to provide them with updated information about services and ask if their support needs have changed. Where this is done it is reported to have been experienced positively (Douglas et al., 2008). GPs also have a role in ensuring that people's eye health is just as carefully monitored as their physical health and that people are referred back into secondary services where appropriate (Sinclair, 2014). One of the key routes of entry back into secondary services is via community optometrists. However, one of the issues highlighted in the first phase of our study was that, where participants had been diagnosed with an 'untreatable' condition such as AMD and discharged from an eye clinic with the message that nothing could be done for them, they had also stopped having eye tests (Thetford et al., 2008). The lack of information and support provided at the point of diagnosis meant that people often had little understanding of their condition and of the importance of continuing to have their eye health monitored. Again, the rolling out of ECLO services could be invaluable in highlighting the importance, particularly with those newly diagnosed, of attending for regular eye checks-ups. However, although some of those in our study reported very positive experiences of community optometry services, others reported negative experiences. So although there is value in establishing a relationship with a good community optometrist, there can be difficulties in accessing adequate services. Finally, it should be noted that due to the recruitment strategy used for this study, all the participants had had some contact with voluntary sector organisations at some point; it is very likely that there are many more individuals living with a significant level of visual impairment in the community whose access to services is even more limited. This underlines the need to make information about sources of support widely available, not only through specialist and non-specialist health services but also through other information channels such as newspapers, social media, places of worship and community centres. --- Conclusion Visual impairment is associated with high levels of unmet need. In order to have their support needs met, individuals often have to build complex networks of formal and informal support for themselves. Service providers must take a more proactive role in reaching out to individuals and improving access to services for all who need them. The rolling out of ECLO services represents an important step, particularly in informing people about sources of support, the benefits of registration and the importance of having regular eye check-ups. However, it is also important that this information is directed not just at people who are newly diagnosed but at all those attending eye clinics with a visual impairment. Social Services Departments should make regular contact with everyone on the SI/SSI registers to invite them for a reassessment of their support needs. Finally, general practitioners should have a more proactive role in considering the eye health of patients as part of their overall health needs, referring into secondary eye care where appropriate. If implemented, these changes would go some way to preventing people with visual impairments from being left to find their own way around the complex systems of support that they currently have to navigate if they are to have their support needs met. --- Wengraf, T. (2005) The Biographic-Narrative Interpretive Method (BNIM) shortguide version 3.1, Unpublished.

The systems of support for people with a visual impairment in the UK are complex and can be difficult to access, involving input from a range of health and social care services. In this paper we report qualitative findings from research looking at people's experiences of accessing health and social care services. These highlight the inconsistencies and variability in existing systems of support for people with a visual impairment, and show that access to services is largely dependent on the agency and initiative of individual service users in establishing networks of support for themselves. This means that those who are less able to do this may find themselves without the support necessary to maintain or improve their quality of life. We argue that health and social care agencies need to be more proactive in addressing the support needs of people with a visual impairment beyond the point of initial diagnosis or registration.

Introduction In post-conflict and conflict-ridden societies prevalence rate of mental illness, specifically post-traumatic stress disorder can be much higher [1]. Based on DSM-5, post-traumatic stress disorder is characterized by increased stress and anxiety following exposure to traumatic event and associated with intrusive recollections of the events such as recurrent distressing memories, dreams, flashbacks, hyper-arousal symptoms of irritability and hypervigilance, problems with concentration and sleep, persistent avoidance of stimuli associated with the event and negative alternation in cognition and mood are often part of the clinical picture of the illness [2]. Our continent Africa accounts for almost 88% of the world's conflict-related morbidity and mortality as well illustrated by Virgil Hawkins's stealth conflicts map [3]. Starting from the past three decades, more than 75% of African countries have been through warfare [4] results in countless losses of lives, destruction of infrastructure and causing untold suffering for hundreds of millions of Africans and more than 70% of the casualties have been non-combatants [5]. Studies on the National and regional prevalence of posttraumatic stress disorder in Sub-Saharan Africa is estimates up to 74% and pooled prevalence of war-exposed regions was 30% [6]. The burden of untreated PTSD is enormous since it causes prolonged morbidity, impairment in day-to-day activities and poor quality of life in all dimensions including health, productivity and social interaction regardless of age, gender or socioeconomic status [7]. Based on the pooled PTSD disability ratings from the World Mental Health Survey studies, 30% of PTSD cases fall into the severe disability range [8]. In different parts of the world, the extent of post-traumatic stress disorder varies with the nature, severity, and length of traumatic events. A recent study carried out following the Russia's war operations in Ukraine and uses a convenience sample of 314 adult civilians in Ukraine revealed that 37.3% of adult Ukrainian were diagnosed with PTSD [9]. A study conducted among war affected residents in Palestinian among 139 adults exposed for political war trauma, violence at home, neighborhood and/or school showed that 53.4% of them met the diagnostic criteria for PTSD [10]. A study conducted in Syria during wartime revealed that 36.9% of adults reported full PTSD symptoms [11]. Consecutive studies were conducted on the prevalence of PTSD among post-war population of Lebanon civilians and those with witnessed traumatic event, imprisonment, serious injury and being in combat situations in three different period quantified magnitude of PTSD ranges from 17.9% to 29.3% [12][13][14]. Following the 2011 Conflict in Libya, Post-conflict mental health burden were studied through population based study across different regions and PTSD prevalence was estimated at 12.4% [8]. In another study in Uganda, seven years after the conflict in three districts, (The Wayo-Nero Study) revealed, respondents with previous negative life events (familial, physical and psychological events) had prevalence of PTSD found to be 11.8% [15]. A study among victims of Boko haram terrorism in north-eastern Nigeria, more than 63% of the respondents were diagnosed with PTSD [16]. Following a three year civil war in Northern Nigeria a cross sectional study using a multistage sampling technique was study in northern Nigeria among 200 older adults, confirms 59% for full PTSD [17]. Cross-sectional study among 1200 waraffected South Sudanese adults were studied on prevalence of PTSD and founds 28% of participants with exposure to war-related traumatic events fulfilled the diagnostic criteria for PTSD [18]. In Ethiopia, PTSD among Koshe landslide survivors in Addis Ababa, shows a prevalence of 37.3% [19]. Following inter-communal violence along the borders of the Gede'o zone southern Ethiopia reveals prevalence of PTSD was 58.4% [20]. In Northwest Ethiopia, a study was conducted following Mai-kadra Massacre revealed a prevalence of PTSD was 59.8% [21]. Multiple factors contribute with the occurrence of post-traumatic stress disorder following exposure to traumatic events. sociodemographic factors like, being female [13,19,22], being divorced/separated or widowed [16,19], clinical factors including depression [20], anxiety [21] and previously diagnosed for common medical disorders [14]and trauma-related factors such as having a history of exposure to traumatic events [12,20], destruction of personal property, witness murder of family members/others [17,20], loss of a loved one [9] and having low social support [19] are prominent and frequently reported factors in association with PTSD. Therefore, this research was conducted in Northeast Ethiopia, Woldia town (Zonal city of North Wollo) following armed conflict between Ethiopian National defense forces(ENDF) and Tigray forces in Northern Ethiopia and later expanded to neighboring region of Amhara. Woldia town was highly affected by the war. Due to its strategic location for war, closeness to the border areas of Afar and Tigray regions, serious military surge and a five-month encroachment and several war related human right violations for relatively longer period of time than nearby towns. This study was done at a community level by using post-traumatic stress disorder checklist (PCL-5), not used previously and no published data on PTSD were done on the area. Therefore, knowing the magnitude of PTSD and its predictors in woldia town will reveal sound and reliable picture to fill information gap, used as input information to identify affected groups and add clarity to the scope of war-related mental health burdens. Moreover, this study helps the area through providing insight for stakeholders/voluntary organization to focus on post-war psychosocial rehabilitation for war affected woldia town residents. --- Materials and methods --- Study design and setting A community-based cross-sectional study was conducted from May-15 to June 15-2022 in Woldia town, Northeast Ethiopia. Woldia town has a total population of 180,000, of whom 81,750 are men and 98,250 women. The town contains ten kebeles (the smallest administration unit in Ethiopia). The area is one of most affected area during war-brake between Federal government of Ethiopia and Tigray regional forces, leads to displacement, destruction of properties, infrastructures and injury on civilians. --- Study population The study population includes adult residents of Woldia town and all are aged 18 and above and stays on the town for the minimum of six months. Exclusion criteria: Individuals who/ family member/care givers reported that they were severely ill to give an interview for data collectors, those with medical certificate or related document that brief the current illness of the patient and those with observable severe physical illness/verbal communication problem were decided as not eligible for the study. --- Sample size determination and procedure The sample size was determined by using a single population proportion by taking the prevalence of post-traumatic stress disorder (PTSD) 59.8% from a study conducted in Northwest Ethiopia, following Maikadra Massacre [21] with; 5% margin of error, 95% confidence interval as follows; n 1<unk>4 Z 2 p<unk>l <unk> p<unk> d 2 Where d = margin of error = 0.05 Z = level of confidence (95%) = 1.96 p = population proportion = 0.598 n 1<unk>4 <unk>1:96<unk> 2 <unk>0:598 0:402<unk> <unk>0:05<unk> 2 N = 369.4 Since, Multi-stage sampling technique was employed, by taking account design effect, the sample size was multiplied by 1.5. So, 369.4 x 1.5 = 554, adding 10% non-response rate total number samples becomes 609. Multi-stage systematic random sampling technique was applied to collect the required amount of sample size. Initially, four administrative unit (kebeles) was selected from the ten kebeles by lottery method and total number of households (sampling frame) in each kebele was obtained from kebele administration. Then, we allocate samples to each of Administrative unit (kebeles) proportionally depending on the total number of households under each kebele. (see Fig 1). In a selected household, if there were more than one individual who fulfills the requirements of the study a lottery method was used. For an eligible participant who was not found at their house, the interviewers revisited the area at another time during the data collection period and if the residents were not avail during the data collection time, they were checked three times and lastly either preceding or succeeding household were nominated. --- Data collection instruments Data was collected using interviewer administered structured and pretested questionnaires. It was collected by four BSc Psychiatry Nurses and regularly supervised by two Psychiatry professionals. The questionnaire was designed in English and translated to Amharic and back to English to maintain consistency. Data collectors were trained on how to interview participants and explain unclear questions and the purpose of the study. PTSD was assessed by the post-traumatic stress disorder checklist (PCL-5) [23] was a 20 item self-report, measures the 20 DSM-5 symptoms of post-traumatic stress disorder. A total score is computed by adding the 20 items, so that possible scores range from 0 to 80 with a five likert scale (0 = Not at all, 1 = A little bit, 2 = Moderately, 3 = Quite a bit, 4 = Extremely) with a cut-off point of 33(PTSD). Validity and reliability of the PCL-5 had been tested and proven in a different country, for example, Zimbabwe (Cronbach's alpha = 0.92) with sensitivity and specificity of 74.5% and 70.6% respectively. Exposure to traumatic event was assessed from the list of 16 items from Harvard trauma questionnaire (HTQ) [24] for the past one year and participants respond 'yes' for at least one traumatic event was considered experience a traumatic event. Previously adapted for use in South Sudan and has 0.87 Reliability coefficient Cronbach's <unk> for the HTQ [18]. Cumulative trauma [20,25] was assessed to examine the dose-response association between exposure to different types of traumatic events and the development of PTSD. Exposure categories were derived from various sources, including single events, two to three events, and four or more traumatic events. Depression was measured by patient health questionnaire item nine (PHQ-9) [26], a multipurpose instrument for screening, monitoring and measuring depression with the minimum cut-off point of 10. This is a highly reliable scale with a sensitivity of 80.8% and a specificity of 79.5% to assess depression [27]. Anxiety was measured by the GAD-7 scale [28], a seven-item instrument that is used to measure or assess generalized anxiety disorder (GAD). Each item is scored from 0 (not at all) to 4 (nearly every day). GAD-7 has adequate psychometric validity, i.e., convergent validity, diagnostic validity, factorial validity, internal consistency, test-re-test reliability in various populations. Social support was measured using the Oslo 3-items social support scale [29] with scores ranging from 3 to 14: 3-8 = poor social support; 9-11 = moderate social support; and 12-14 = strong social support. Socio-demographic, substance use history and clinical factors were operationalized according to different literatures [20,21,30]. --- Data processing and analysis The collected data were coded, entered and cleaned using Epi Data Software version 4.6.02 then it was exported and analyzed by using STATA version 14. every variable was checked in bivariate logistic regression analysis to get variables that had an association with the dependent variable, then variables with a p-value of 0.2 were included in multivariable logistic regressions for further analysis. Finally, P-value 0.05 in multivariate logistic regression was considered as statistically significant and the strength of associations was determined using adjusted odds ratio (AOR) at 95% CI. --- Ethical consideration and consent to participation All procedures undertaken during data collection in accordance and ethical clearance was obtained from University of Gondar CMHS, school of medicine review committee with the ethical approval number SOM/1540/2022 and additional supportive letter was taken from department of psychiatry and taken to Woldia town Mayor Office. Participants were informed about the aim of the study and no identification or names was recorded so as to maintain confidentiality. The study participants were informed of their right to refuse or stop participating at any time during the interview. Data collection was done by trained data collectors by visiting each selected household and for those agreed they proceed to interview by putting their signature on the informed consent sheet and for few of those participants who were not literate, data collectors read the information sheet and explained to them when necessary, and assisted them to give finger print to confirm their consent before initiating interview. The principal investigator of the study supervised the data collection process on daily base and performs a routine check on the questionnaires for consistency and completeness. --- Result --- Socio-demographic characteristics of the respondents From a total of 609 samples, 597 participants were included in the study with a response rate of 98%. The mean age (<unk>SD) of participants was 33.2%(<unk>10.48) years, with a range from 18 to 81 years of age. Among participants 329(55.11%) were male and 261(43.72%) of them were in the age range of (25-34) (see Table 1). --- Trauma exposure of participants The majority of participants 460(77.05%) reported exposure to Noise of explosions during wartime, 208(34.84%) participants lost/face serious injury to their family members, friends or others important and 129(21.61%) individuals report witnesses of murder of either of their family member, friend or unknown others. In terms of cumulative traumatic events, 43.4% of respondents had experienced two to three traumatic events, and 92 (15.4%) out of 238 participants had been exposed to four or more traumatic events (see Table 2). --- Clinical, social support and substance related factors Out of the total participants, 47(7.87%) of them were found to have previous history of mental illness and half 315(52.76%) participants were positive for depression and 259(43.38%) of them were positive for anxiety. On social support, 284(47.57%) had poor social support and 73 (12.23%) had a strong social support. Regarding substance-related factors, 287 (48.07%) and 125(20.94%) of respondents had ever used history for alcohol and khat respectively. Regarding current use of a substance, 240(40.2%) of them currently use alcohol. --- Prevalence of post-traumatic stress disorder The study reveals that, prevalence of post-traumatic stress disorder (PTSD) was 56.28% with (95% CI, 52.2, 60.2). The estimated prevalence rate was 53.6% and 46.4% among males and females respectively. --- Factors associated with post-traumatic stress disorder In Bivariable logistic regression analysis variables like age group, experience the traumatic event on themselves, destruction/looting of personal property, lack of food and water, murder/ injury of a family member /friend, a witness of the murder of family member /friend/other, made to accept ideas against will, unlawful imprisonment, poor and moderate social support, depression and anxiety were to have a p-value of less than 0.2 and these variables fulfills the requirements for further analysis in multivariable logistic regression analysis. In multivariable logistic regression analysis, factors Destruction/looting of property, murder/ serious injury of a family member/friend or others important, a witness of the murder of family member/ friend/others, unlawful imprisonment, experience traumatic event on themselves and having a poor and moderate level of social support, depression and anxiety were statically significant for post-traumatic stress disorder with p-value less than 0.05. The odds of developing post-traumatic stress disorder were 1.6 times higher among participants with the destruction/looting of their properties as compared to their counterparts (AOR = 1.6, (1.11, 2.47)). the odds of having PTSD among those with murder/ injury of their family/friend or other had two-fold higher as compared to those who didn't have (AOR = 2.1, (1.37, 3.22)) and the odds of developing PTSD among individuals witnessing murder was 1.6 times higher as compared with those didn't witness (AOR = 1.6, (1.01, 2.71)). Individuals with unjudicial imprisonment against their will had odds of 1.7 times higher as compared to their counterparts (AOR = 1.7, (1.06, 2.74)). In comparing way of exposure to traumatic event, individuals exposed to traumatic events through learning, the odds of having PTSD among those experience on themselves is twofold higher (AOR = 2.0, (1.22, 3.58)). The odds of developing PTSD were more than threefold in participants with anxiety as compared to their counterparts (AOR = 3.3, (2.26, 4.97)) and for those with depression, the odds of having PTSD were two times higher in compared to their counterparts (AOR = 2.0, (1.37, 2.93)). As compared to those with strong social support, those with poor social support have an odd of 3.1 times and those with moderate social support have odds of 3.0 times higher for developing PTSD (AOR = 3.1, (1.60, 6.04)) and (AOR = 3.0, (1.56, 5.87)) respectively (see Table 3). --- Discussion The study quantifies that, the prevalence of post-traumatic stress disorder (PTSD) was 56.28% with (95% CI, 52.2, 60.2) among residents of Woldia town and this finding goes in line with the previous studies conducted following the Mai-kadra massacre, Northwest Ethiopia 59.8% [21], Gede'o zone, Southern Ethiopia 58.4% [20], Northern Nigeria 59% [17] and Northern Uganda 57% [31]. While comparing, the result of this study was lower than other studies conducted in North-Eastern Nigeria 63.7% [16], and Siri Lanka 68% [27]. The possible justification could be, variation in the study population, this study includes Woldia town residents aged 18 and above, but in a study conducted in, North-Eastern Nigeria and Siri Lanka participants were only 15 to 35 years, school-aged, children and adolescents respectively. Another reason for the variation could be the time gap from a traumatic event to the study, in North-Eastern Nigeria a study was conducted among Internally displaced populations after six weeks and as the time of study is close to the event participants might have a high potential to be symptomatic. Also existing geographical and socio-cultural variations contribute for the differences. On the other side, the finding of this research is higher than other studies conducted in South Sudan 28% [18], Northern Uganda(The Wayo-Nero Study)11.8% [15], Ukraine 37.3% [9] and southern Lebanon 29.3% [13,32]. The discrepancy was due to variation in the assessment tool, in South Sudan MINI International Neuropsychiatric Interview (MINI) was employed and in Southern Lebanon, PTSD was assessed by checklist from DSM-IV-TR PTSD symptom criteria. In contrast, this study utilizes post-traumatic stress disorder checklist (PCL-5 with extended Criterion A of DSM-5) assessment tool with a better internal consistency to measure PTSD [20]. Another reason is due to the difference in psychometric properties of tools, a study conducted in South Lebanon shows a sensitivity of 78% [14], but the tool we use has a sensitivity of 81.55% results more detection of true positive cases, so increases the prevalence of PTSD. More than 66%(398) of participants were in the age range between 25-44 and among those identified positive for PTSD 68% of them are members of this age group. Those in this age range are at imminent risk for exposure to war-related traumas since adults are at the right age to join military camps and participate in combat fields [33]. Destruction/looting of personal property was 1.6 times higher to develop PTSD agreed with studies conducted in Gede'o zone, Ethiopia [20] and South Sudan and Liberia [34] such devastating situations are shocking, difficult to accept, adjust self to previous and puts individuals and family future life at risk and acute stress is a typical response [35]. Participants who lost their family member/friend were twice as likely to have PTSD, it was supported by studies done in Northern Nigeria [17] and Mai-kadra, Ethiopia [21], this is due to their chance of exposure to traumas in some extent and perception of subsequent injury and distress for later traumas, also had negative intrusive thought of revenge, payback and impact on emotional well-being of victims family members [20]. Those study participants who had a direct witness to murder of family members/ others were 1.6 times more likely to have PTSD, agreed to studies conducted in South Sudan and Liberia [34], and Gede'o zone, Ethiopia [20], This is due to traumatic events especially direct witnessing having a higher tendency to be captured in mind and later reprocessed in the form of nightmares, flashbacks, hyperarousal, psychological distress and later PTSD [19,21]. Individuals, who pass through unlawful imprisonment had 1.7 times higher to develop PTSD as supported by studies conducted in Northern Nigeria [17], It might be due to the absence of law enforcement bodies and this leads way for mal treatment of civilians, threatening, manipulation, physical abuse, torture and being secluded from families and social contact increase the likelihood of trauma exposure and leads to mental distress and PTSD [35]. Participants with anxiety were 3.3 timed higher odds of developing PTSD as compared with others didn't have, as supported by other studies done in Turkey [36], northern Cameron [33] and Mai-kadra, Ethiopia [21]. Even if, casual relation between anxiety and PTSD couldn't concluded, comorbid cases of anxiety disorders make persons more vulnerable to develop PTSD [35]. Additionally, it was discovered that anxiety symptoms could independently predict PTSD and there was a strong association between exposure to trauma and anxiety symptoms [20]. The odds PTSD were twofold higher among respondents with depression as supported by studies conducted in Southern Lebanon [12], Gede'o zone and Addis Ababa, Ethiopia [20,22], the onset of depression is associated with preexisting vulnerabilities and subsequent psychosocial stressors, whereas the onset of PTSD is associated with stressors related to disaster-related events [14] and up to 86% develops co-occurrence [37]. Also level of depressive symptoms was high as the level of exposure increased and likelihood of having comorbid disorders of PTSD and depression was elevated following war related traumatic events [18,31] and association of symptoms between PTSD and depression was observed [15]. Participants with poor and moderate social support had 3.1 and 3times higher odds to develop PTSD as compared to those with strong social support. This finding also goes along with other studies conducted in South Lebanon [14], Turkey [36] and Koshe landslide survivors [19]. This is possibly due to the fact that poor mental health might result from traumatic injury and peoples with strong social support allows people to express their worries, feel secure, and have a sense of belonging, social networks and strong social support systems may be able to lessen the impact of stressful life events. Individuals who directly experience the traumatic event on themselves had two times higher odds of developing PTSD as compared with those learning the traumatic event from close family or close friend, this finding agrees with studies conducted in Netherlands [38] and Northern Uganda [31]. This is because, direct experiences have physical, psychological and emotional effects comparable to and possibly worse than those of witnesses or other forms of exposure [35]. --- Conclusion Since the study was conducted in a short period after armed conflict in the area and beyond the infrastructural destruction and physical disability, psychological and mental health-related crisis is potentially high, as this study quantifies, more than 50% of participants develop posttraumatic stress disorder(PTSD). Especially, individuals on whom trauma happens, witnessing trauma in person, murder/injury of a family member /friend, destruction/looting of property, unlawful imprisonment, depression, anxiety, poor and moderate social support were significantly associated with post-traumatic stress disorder(PTSD). The finding implies the association of exposure to different war-related traumatic events for the development of PTSD, lends a strong theoretical and practical significance for considering strengthen mental health care services for the community. Furthermore, depression and anxiety disorders contribute as a factor for PTSD and this finding necessitates a planned intervention and making informed clinical assessment of PTSD and possible co-morbidity and renders psychiatric and post-conflict psychosocial rehabilitation in the community. The association between poor and moderate level of social support for PTSD needs the strengthening and improvement of social network of the community. Specifically, by using existing self/others help programs, then engaging community leaders in the social gathering and facilitate in restoration of their relationship and interaction. --- Strength and limitation of the study As a strength, such study is not done previously in the area and its timely and reveals the current impact of war on the community and done by adding important factors not included in previous studies and applies a standardized tool, PCL-5, post-traumatic stress disorder checklist (PCL-5) and Harvard trauma questioner(HTQ) not used previously. As, a drawback the utilization of items that needs remembering of previous events in the questionnaire may invite recall bias and we were not able to verify whether the depressive and anxiety symptoms preceded or followed the PTSD. Sensitive traumatic events like rape and sexual violence is not assessed because of social desirability issue and under reporting and we strongly encourage to conduct such focused studies through different methodologies. --- Supporting information --- S1 Data. (XLS) --- Author Contributions Supervision: Demeke Demilew, Biruk Fanta, Haregewoin Mulat, Dawed Ali, Jemal Seid, Abiy Mulugeta, Jerman Dereje. Writing -original draft: Abenet Kassaye. Writing -review & editing: Abenet Kassaye.

Post-traumatic stress disorder is marked by increased stress and anxiety following exposure to a traumatic or stressful event. Events of conflict and war-related traumas were commonly reported situations and people who have undergone through have a higher tendency to develop PTSD Woldia town had been under a serious military surge and a five-month encroachment, so the expected destruction in property, impact on physical, social and mental health of civilians was potentially high. More importantly, there is no study that investigated the significance of association between war-related traumatic events and posttraumatic stress disorder in the area. so this study aims to assess prevalence of post-traumatic stress disorder and its associated factors among war-affected residents in Woldia town, North East Ethiopia, 2022.A community-based cross-sectional study design was employed by using a multi-stage systematic random sampling technique from May-15 to June-15/2022. A total of 609 participants were enrolled. PTSD was measured by the post-traumatic stress disorder checklist for DSM-5 (PCL-5). Data were entered by Epi data version 4.6.0.2 and analyzed using STATA version 14. Bivariable and Multivariable logistic regression analysis was done to identify associated factors to PTSD and P-values less than 0.05 were considered statistically significant.The overall prevalence of PTSD was 56.28%. Destruction/looting of property (AOR = 1.6,95%CI,1.11-2.47), murder/injury of family member (AOR = 2.1,95% CI,1.37-3.22),

Background Due to the changing demographic structure of the population in many countries, the proportion as well as the number of older people with care needs will continue to increase (European Commission 2009;Hoffmann and Rodrigues 2010). This development portends a growing need for long-term care services, a challenge for most countries due to limited resources (European Commission 2008). Family carers play a strategic role in providing long term care (Kro <unk>ger 2003;Winqvist 2010). In Europe approximately 80 per cent of care hours for older people are provided by family carers (European Commission 2008;Marin et al. 2009;Hoffmann and Rodrigues 2010), and many of them are spouses, who care for impaired or disabled husbands or wives (McConaghy and Caltabiano 2005;Larsson 2006). Despite the positive aspects inherent in care giving, there are several negative aspects as well. The intensity of caring tasks, often with need for constant presence at home, leads many carers to feelings of social isolation, anxiety, burdensome responsibilities or depression (Ekwall et al. 2004;McConaghy and Caltabiano 2005;Hoffmann and Rodrigues 2010). Although interventions and support services for older family carers exist, there is still a need for more effective ways of supporting them (Schulz 2001;Lamura et al. 2008). Moreover, the means by which such support is offered does not always match the carers' needs. Caregiver support is often applied in standardized care activities and programmes, thereby being non-flexible and non-personalized (Clare 2002;Lauriks et al. 2007). The use of information-and communication technology (ICT) as a tool to support the family carers of older people is increasing (Powell et al. 2008;Schmidt et al. 2011). ICT services may contribute to improve the flexibility of caregiver support, and is considered an important factor in meeting future care needs. These services also have the potential to help older family carers increase their quality of life, ease their burden, stay healthier and improve the quality of care provided. A key characteristics of ICTbased support is the ability to provide the help needed, when needed (Schmidt et al. 2011). A range of ICT-services are currently being used throughout Europe, including telecare (Jarrod and Yeandle 2009) online support groups, online training and monitoring systems (Schmidt et al. 2011). ACTION (Assisting Carers using Telematics Interventions to meet Older peoples' Needs) is an example of a support system that offers a combination of multimedia programs and videophone. It was developed in collaboration between six European countries, aiming to support frail older people and their carers in their own homes, via ICT (Magnusson et al. 2002). Research has shown ACTION to be both user-friendly and beneficial. Frequent contact and support from staff and group meetings with fellow participants have been essential parts of the service. Evaluations have shown the group meetings to be valuable as they provide the opportunity to continuously meet with peer carers (Magnusson et al. 2002;Magnusson et al. 2005). Older family carers living in rural areas may have difficulty participating in supportive activities and group meetings due to geographic issues and long distances. In many rural areas across Europe, the proportion of the population 65 years and older is greater than in the more densely populated areas (Goll 2010). Rural areas are also changing their social infrastructure with depopulation and loss of local facilities such as health-and social-care provision and public transportation. ICT-based support services have the potential to make support to family carers in rural areas more available, since it can be provided and accessed regardless of geographic distances (Schmidt et al. 2011). However, due to the current rural reality of limited services, it is likely that a support system like ACTION cannot be organised in the same way in rural areas as it functions in more populated areas or in those with shorter distances. Neither can family carers in rural areas expect the same intensity in group meetings or in personal contact with support staff. This raises the question as to whether ICT-based carer support, like the ACTION programme, can be equally successful for older family carers in rural areas. Research with quantitative designs can be successful in exploring the prevalence of the use of ICT technology in informal care as well as in clarifying the different effects of using ICT as support for caregivers. There is, nonetheless, a lack of knowledge about the deeper meaning of ICT use among older caregivers in rural areas. For example, it is important to understand how such informal care providers perceive ICT support, and what it means for them. Such insight requires qualitative enquiry. The aim of the study described in this article, therefore, is to illuminate the meaning of ICT-based carer support as experienced by older family carers living in rural areas, by means of qualitative analysis, --- Method In this study a qualitative, descriptive design was used to obtain a starting point regarding older family carers' experiences of receiving ICT-based support in rural areas. The qualitative enquiry was part of a larger study based upon both quantitative and qualitative methodologies. According to Polit and Tatano Beck (2008), the open ended interviews that are employed in qualitative research can make it possible to obtain rich, detailed information about the phenomenon under study. --- Setting and intervention Older family carers living in 15 different local municipalities in vast rural areas in the northern parts of Sweden were invited to participate in an intervention, in which they would receive ICT-based caregiver support from their local authorities for a period of 3 years. These rural areas are mostly large tracts of mountains and forests where small villages and very few towns are situated. Public services are concentrated within the few towns, leading to long distances to such service for those living more remotely. The area also has long, cold, snowy winters (Nov-Apr) and public transportation is limited or non-existent. The intervention consisted of a modified version of the ACTION-service (Magnusson et al. 2005) including the provision of a computer with Internet access, information and educational software, webcam, and access to a secure social community with other family carers as well as call centres. The technical equipment was given to the family carers through each local authority at no cost to the carer during the project period. The information-and educational programmes covered such themes as caring in daily life, coping, planning, respite care, claims and benefits. Specific topics in the educational programmes regarding caring in daily life were incontinence, dementia, stroke, pressure sores, end of life care and emergency situations. The purpose of the internet-based secure social community was to enable safe and easily accessible contact with peer family carers as well as encouraging them to contact the call centre via internet and webcam. Each municipality had a call centre with professional support nurses. The call centres were located within existing municipal support service offices for family carers and were accessible during office hours. Prior to the start of the intervention, family carers were educated in using the technology. Training and education was given by support nurses in the family carers' homes. --- Participants and recruitment Participants were consecutively included in the study as they came in contact with support service offices. The larger study included a total of 63 family members who became users of the intervention. Inclusion criteria were: to be the primary family carer providing care to a co-habiting spouse, 65 years or older and living in a rural area. After 1 year of using ICT, 33 of the family carers were randomly selected and invited to be interviewed. The interviewees were evenly distributed over the 15 municipalities in the same proportions as the total number of participants. Two of the family carers declined being interviewed, because they had not used the support. Demographic data of participants are presented in Table 1. As shown, the 31 interviewees in the qualitative phase of the study did not differ from the 32 other study participants on the demographic variables. --- Data collection Background information was collected in a structured questionnaire handed out personally by the support nurse who recruited the participant, along with written information about the study and a stamped addressed envelope. Semi-structured interviews were conducted with the 31 family carers. A topic guide was used and the interviewees were encouraged to speak freely about the topics. The interviews were carried out over the internet, due to the large geographic area, using webcam. Webcam interviews with older people might appear to be an unusual method for gathering research data. In order to seek quality as well as avoiding unnecessary mishaps which could negatively affect the interview situation, a strategy for the interview procedure was developed (Fig. 1). The strategy was based on knowledge regarding techniques for telephone interviewing (Denscombe 2010) and reflections on communicating with older people (Eide and Eide 2009). Richness in detail in the strategy was considered relevant to make it replicable. The interviewer initiated the call from a computer using webcam, built-in microphone and speakers. Interviewees were located in their homes, at an equally equipped computer. The time for the interview was scheduled in advance by the interviewee. All interviews were audio recorded using a digital recorder and were transcribed verbatim. Some of the interviewees spoke with heavy local accents and in those cases transcriptions were made in standard Swedish (Malterud 2009). --- Ethical considerations The study was approved by the Research Ethics Committee of the Medical Faculty at Umea <unk>University, Sweden (Dnr 2010-187-31). Participants were guaranteed confidentiality and informed of their right to withdraw their participation at any time without giving a reason. They were also assured, in writing, that withdrawal from the study would not exclude them from access to further support. Family carers were guaranteed that they would be allowed to keep the technical equipment and have full access to services included in the intervention for the full project period, at no cost even if they decided to withdraw from the interview study. --- Data analysis Data analysis was performed using qualitative content analysis inspired by Graneheim and Lundman (2004), in which the analysis is performed in several steps, starting with reading through every interview several times to get a sense of the text as a whole. The text was then divided into meaning units, containing one or more sentences that were condensed. These shorter sentences were abstracted and labelled with codes and grouped together in categories, which constitute the manifest content. During the process of analysis there was a movement back and forth between the whole and parts of the text. In order to ensure the reliability of the categories, their in-depth grounding in the text and a logical structure, the categories were constantly refined and rearranged until consensus among the authors was achieved (Patton, 2002). Through a process of reflection and discussion the latent content of the categories was formulated into two themes. --- Results The analysis resulted in the themes Adopting new technology with help from others, which consisted of three categories (active use of ICT, carers need encouragement and support, and personal development), and Regaining social inclusion, which consisted of two categories (increased security and recapturing a position in society). For an overview of the themes and the categories, see Table 2. --- Active use of ICT Usability was considered to be good. Some carers had experienced technical problems in the beginning, but felt that support nurses had been quick to solve problems. A majority of carers used the computer frequently and felt it had given new meaning to everyday life. There was consensus that one must give it priority and actually use the technology, not just wait around for someone to call. Some carers stated that they had not yet experienced benefits from their ICT-support. The common denominator for not using the support was lack of use. Woman, 79 years: 'At first I was uncomfortable. I decided to do it anyway. I understood it wasn't going to work by itself'. Woman, 66 years: 'I only used it a few times and haven't figured it out yet. I really want to experience all those good things, but I still have friends and prefer to see them than sitting by the computer when I have time for myself'. --- Carers need encouragement and support Those who were encouraged by friends, family members and fellow family carers used the ICT-service frequently and found it useful. In the beginning, many of the family carers were rather afraid of the technology and felt hesitant about using it. Receiving praise from others for being bold and modern worked as a motivator to continue even for those who struggled at first. Rewards, such as feeling less lonely, increased security, growing self-esteem and being considered 'cool' by the grandchildren made family carers take on new challenges and develop their technology skills to a wider use of ICT than what was included in the intervention. For example downloading public programmes for webcam communication and chat, buying printers, USB-sticks and digital cameras. Technical support and follow ups from call centres were important to get started with regular use of the computer. Those who had neither continuous follow ups with support nurses, nor encouraging family members did not use their computers much. However, family carers who had one of these supports became frequent users. One participant who had looked forward very much to talking to people in the community encountered personal problems as the spouse turned out to be jealous and gave her a hard time when using the computer. She kept the computer hoping the situation would improve but ended up using it only a few times. Lack of support and motivation constituted obstacles to using ICT. Woman, 77 years: 'The support nurse is supposed to come here and show me more, but it keeps being postponed and I have a hard time getting to it on my own'. --- Personal development The majority of the participants had no previous computer experience and felt unsure before starting the intervention. Taking on the challenge to learn modern technology was sometimes disapproved of by grown children who tried to talk the old parent out of it. After learning how to handle the computer, and proving the doubters to be wrong, family carers experienced great satisfaction and increased self-esteem. Woman, 81 years: 'I feel like my life has come to a new level. Everybody speaks about computers. I didn't have a clue, but now I know, and it feels like I have risen from being a nobody to being equal'. Using webcam and network media to meet with peers was an important feature. Besides the satisfaction of sharing experiences with someone in the same situation, family carers also found others asking them for advice. Being acknowledged as a competent carer gave a sense of satisfaction and boosted self-confidence. After learning to use the computers, family carers also became creative in finding ways to use it. They described how they used to keep themselves busy with caring-and household tasks, while the cared-for-spouse did not have much to do during Eur J Ageing (2013) 10:191-199 195 the day. Now they were able to find innovative ways to use the computer, enriching life for themselves as well as for the spouse. For example, spouses suffering from stroke were able to express themselves through the computer and those with dementia used the computer to practice their cognitive abilities. Woman, 73 years: 'There has definitely been a change since he started doing those exercises on the computer. He still forgets things and all that, but it is different somehow, I think this is very good for him'. Man, 75 years: 'Now she can sit with the computer and it allows me to get some free time'. --- Increased security Knowing there were people close at hand, both professionals and others, available for webcam contact at the moment of need, contributed to increased feelings of security. This was particularly important as participants in this study lived in vastly remote areas, where they otherwise could not expect immediate help or support. Man, 75 years: 'If I worry, or something suddenly happens, I push a button and another person comes into my living room. It is such a security. Otherwise it takes them an hour to get here'. A majority of the family carers had cared for their spouse for several years and were self-taught. Few had received education about the spouse's condition or how to handle daily life on these terms. The computer gave access to information programmes, websites and streamed lectures, where family carers often found that their ways of handling matters were confirmed. Man, 77 years: 'I've been doing this for many years, without really knowing anything about how one should do. Now I found out I have been doing it right, it feels very good and relieving'. Access to information programmes and websites with facts about specific conditions and illnesses helped provide family carers with a higher feeling of security-not so much for the current caring situation, but for broadening their knowledge about health issues in general and making them more prepared for what was to come. They also felt they could have been spared plenty of trouble throughout the years if they had this information earlier in their caring career. One thing pointed out as particularly beneficial was to virtually attend seminars by health professionals, which contributed to carer competence. Man, 81 years: 'I don't have to worry anymore about doing the wrong things. Not only do I feel competent, I know I am'. Woman, 78 years: 'If I didn't have the computer I would never have been able to learn all this, living out here everything is too far away'. --- Recapturing a position in society Apart from experiencing social-and e-inclusion through chat, e-mail and webcam-conversations with friends and family, the family carers found that the internet provided them with a feeling of recapturing a position in society and 'being somebody'. Ordinary things like listening to radio and watching TV-programmes had previously contributed to the feeling of exclusion as programs often refer to websites for recipes and other detailed information. With Internet they became fully worthy radio-listeners. Over half of the interviewees read newspapers on the internet. Access to long gone services was described as very valuable. Besides facilitating everyday life, it also gave a sense of independence. Man, 78 years: 'I can now go to the bank, the chemist's, and different shops for clothes and this and that, on the Internet. I never had such a selection before. For many years one place after the other has closed down. Now I can access all of this, from my home, when I have the time, without needing to arrange for a sitter for my spouse'. All interviewees expressed feelings of loneliness and isolation of various extents. These feelings originated from difficulties in leaving the spouse, friends who had vanished over time and family members living far away. Using ICT gave new opportunities to break isolation and reduce feelings of loneliness. Woman, 83 years: 'Just having the computer here, knowing that if I would want to, I can just press a button and someone will practically be here with me in this room. It may sound silly, but it makes me feel less lonely'. New acquaintances were made through the social community with peer carers. Knowing that everybody in the contact list had agreed to the conditions of the service encouraged family carers to make contact with persons they did not previously know. Those who felt too shy to initiate contact were happy when others contacted them. Picture quality was considered good and gave the feeling of actually meeting someone in their own home. Man, 76 years: 'One becomes rather isolated in this situation and this is an excellent way to be able to make contact and talk for a while, communicating with others in the same situation. They are right here on the list, and you know they want contact, too. I haven't had anyone to talk to before, so this has made a big difference'. Keeping up a close relationship with children and grandchildren living far away had previously been seen as difficult. Computer with webcam enabled closer contact and strengthened relationships. In particular, grandchildren played an important role in that they encouraged their grandmother or grandfather after they had started using the technology, giving them technical and emotional support. Communication between grandchildren and grandparents changed from making dutiful phone calls on birthdays to keeping in touch on a weekly or daily base, mostly through chatting. The family carers felt more valuable as the young ones asked them for advice regarding baking, stain removal and other household chores. The older persons appreciated grandchildren being online most of the day and felt secure when seeing the on-line symbol in the chat programme they used. Woman, 76: 'I turn the computer on first thing in the morning. Then I can see who is ''home'' or online as they call it. It's such a great company. When the little symbol is lit and I know they (my grandchildren) are there, I don't feel alone'. Communication with children living far away took place mostly through e-mail and chatting. The family carers appreciated taking a greater part in the children's lives than earlier, as via webcam they could see their new haircuts, interiors from their homes and other things. Communication via webcam added a new dimension compared to phone calls and gave a feeling of actually being together. Using ICT-based carer support also strengthened the relationship with the cared for spouse. Couples who used the computer together found it to be a pleasing asset affecting their relationship in a positive way. Through the computer they experienced fun and joy together. They participated in sing-alongs, played games, explored the world through internet and used webcams to talk to other couples. Both family carers and cared for spouses said that daily life felt more meaningful since they started using ICT. Woman, 74 years: 'We have a lot of fun together these days, and keep ourselves busy here with the computer. Before, he sat in the TV-room and I spent most of my time in the kitchen, both of us sulking. We didn't talk much, because there wasn't much to talk about. Now we laugh a lot'. --- Discussion ICT-based carer support can be a valuable tool for older family carers in rural areas, as it contributes to improve quality in daily life in a variety of ways (Schmidt et al. 2011). The findings of this study suggests that to fully experience the benefits family carers must adopt the new technology and make the effort to use the ICT systematically and not only occasionally, since the mere access to ICT-based carer support cannot by itself have a positive influence on daily life of family carers. This condition must be kept in mind by providers of caregiver support, in order to avoid the misconception that ICT-based support is a solution for everyone. In this study, carers who were able to get out of the house gave priority to that instead of using the ICT-based support. Most participants in this rural study were unable to attend real life group meetings with peer carers and professionals. Nevertheless, the results suggest that online interaction and access to online services is of meaning for the interviewees and opens up a possibility to regain social inclusion. Suitable training along with encouragement and support from others (interaction) are known triggers motivating family carers to take on the challenge of technology (Magnusson et al. 2004;Schmidt et al. 2011). Participants in this study emphasized the significance of ICT, providing flexible, personalized support when they needed it. Previous research with informal caregivers in rural and remote settings has shown that use of formal support services is generally low, due to gaps in service provision, and that available services do not always meet the needs of rural family carers (Innes et al. 2011). Furthermore, geographical distance between themselves and other family members result in difficulties accessing wider family support (O'Reilly and Strong 1997). Family carers in our study pointed out how ICT helped to strengthen valuable relationships, for example with grandchildren. Contact with grandchildren brings pleasure and joy (Wenger and Burholt 2001). Contact frequency between grandparents and adult grandchildren is usually low, particularly when it comes to time-consuming face-toface visits. The relationship often evolves into occasionally meeting at family events, such as birthdays (Geurts et al. 2009). Webcam contact with peer family carers provides a way of meeting people even when unable to physically leave the home or transportation is unavailable. Sharing experiences with someone who understands the situation of being a family carer helps in coping with one's own situation and builds self-confidence and self-esteem (Magnusson et al. 2002). Interviewing older family carers through webcam was a challenge for the researchers but following the formed strategy helped in supporting the structure as well as the overall quality in the interviews. None of the family carers felt uncomfortable with the technological web-cam interview situation. The number of interviews was fairly large for being a qualitative interview study, but was decided upon to increase trustworthiness and transferability of data. After 20-25 interviews, saturation was obtained and the last interviews served to confirm the information of the others. Another way of enhancing the trustworthiness of the results was to have a constant and open discussion between the coauthors in the analysis of the data, until reaching consensus. The results from this study are mostly positive, which may provoke questions about selection bias. We should clarify, therefore, that in order to participate in the study the participants only had to be willing to use the technique and to have it installed in their homes. A majority of the participants had little or no previous experience of using computers (which could have lead to more practical problems than actually occurred). When technical problems were encountered, support nurses were quick to provide solutions. The most serious limitation of the study is possibly its limited survival time. The rapid development of technical solutions related to ICT as well as increasing prevalence of computers with Internet access in older people's homes makes this a constantly changing arena. At the same time this is one of the strengths of the study too, since the results show that the participants do not only use the services of the intervention but also expand it by making use of the extended capabilities in relation to have access to high speed internet. In that sense the intervention itself expands. From a rural perspective, moreover, access to internet, was expressed by the interviewees as a factor facilitating daily life. ICT-based caregiver support can be a way to introduce internet to older people who would not try it on their own, guiding them to make use of online services which may support them in daily life, with advantages both for the individual as well as the society. Health-and social-care provision has traditionally been weaker in rural areas and the ongoing changes may result in older people in rural areas becoming even more vulnerable due to increased loss of essential services (Phillipson and Scharf 2005). Internationally, the mean age of residents in rural areas goes up faster than in urban areas (Huber et al. 2008). The costs for provision of home care are higher in rural areas, partly attributed to greater transport costs (Wilson and Fernley 2007), which is likely to lead to an increased reliance on family caregivers providing support to those with dementia (Innes et al. 2011). One concern before the study was that it might be difficult for older family carers in rural areas to experience the same benefits from ICT-based caregiver support as in previous studies in more urban contexts. The results showed the opposite. A majority of family carers in the study had the ability to adapt and to self manage, in line with Huber et al. (2011) new definition of health. With mostly long-distance support from support nurses and without regular group meetings, the participants took initiatives of their own, finding ways to meet their individual needs. Instead of relying on support nurses as major support providers, family carers found peers and relatives to be supportive. Webcamera communication within the secure community was included in the intervention, but family carers extended the means of communication by also using chat-and video-communication programs available online, often initiated by their grandchildren. In doing so they were able to keep in regular contact with family members living far away, who did not have access to the secure community. Summarising the results, we conclude that ICT-based carer support appears to be a way for older people living in

The aim of this intervention study was to illuminate the meaning of ICT-based caregiver support as experienced by older family carers living in vast rural areas, caring for a spouse at home. In order to access, the support service participants were provided with a computer and high speed Internet in their homes. Semi structured webcam-interviews were carried out with 31 family carers. A strategy for webcam interviewing was developed to ensure quality and create a comfortable interview situation for the family carers. Interviews were analysed using content analyses, resulting in the themes: Adopting new technology with help from others and Regaining social inclusion. The results indicate that ICT-based support can be valuable for older family carers in rural areas as it contributes to improve quality in daily life in a number of ways. In order to fully experience the benefits, family carers need to be frequent users of the provided support. Adequate training and encouragement from others were essential in motivating family carers to use the support service. Access to Internet and webcamera contributed to reducing loneliness and isolation, strengthening relationships with relatives living far away and enabled access to services no longer available in the area. Use of the ICT-service had a positive influence on the relationship between the older carer and adult grandchildren. It also contributed to carer competence and promote feelings of regaining independence and a societal role.

Introduction The sinking of the Sewol ferry, also referred to as the 416 Sewol ferry disaster, occurred in the South Korean waters near Jeollanamdo on April 16, 2014. The ferry was carrying 476 people, including passengers and crew members. Most of the passengers were students from Danwon high school who were on a school trip. Of the 325 students on board, 250 died or went missing, and only 75 survived. The captain and most of the crew fled the ship after it capsized; thus, no proper evacuation orders were issued. Investigations revealed that unsafe operation by Chonghaejin Marine Company Ltd., the lack of disaster countermeasures, and the government's delayed response were also main causes of the disaster. Yet, it should be noted that delayed rescue operation by the Korean Coast Guard and the Ministry of Oceans and Fisheries, in particular, and subsequent bureaucratic and political cover-ups exacerbated the situation [1]. The Sewol ferry disaster was a social disaster that can be regarded as caused by incompetence and fatal errors of social or socio-technical systems [2]. The accident claimed the lives of 304 individuals, including more than 200 high school students on a school trip. The wreck of the Sewol seemed also to traumatize citizens who were merely exposed through media, maybe because the tragedy appeared to have been so preventable. Generally, a disaster is a condition where damage or loss of life is too massive for humans' capacity to cope [3], and such an event can devastate the affected individuals and further extend its damage to corresponding and nearby areas [4]. A disaster can vary in its magnitude, extent, and further impact. It is generally divided into two categories: a natural disaster and a manmade disaster. While the former covers natural hazards such as earthquakes, floods, droughts, or landslides, the latter corresponds to: i) engineering or technological hazards such as the collapse of bridges and buildings, industrial accidents, or transport crashes; and, ii) purposeful incidents such as massacres, terrorist acts, or wars [5]. The social impact of a disaster is generally affected by if it is a natural or manmade disaster. A natural disaster and a manmade disaster may occur simultaneously and interact with each other in various ways [6]. Manmade disasters seem to result in far more severe consequences than natural disasters [7]. If a disaster's causes can be attributed to certain individuals or a particular group of people, the fear and anger of the affected people will be directed toward them. Victims and their families may suffer from strong emotional distress and may have difficulty dealing with rage, distrust, self-criticism, or guilt, eventually resulting in severe mental health problems in the long-term [7][8][9]. A social disaster can impact the victims' social relationships as well as their mental health. Picou and Marshall found, in their study of damage by a hurricane in 2005, that the victims not only went through loss of family and homes but also were exposed to suffering from discrimination, conflict, and contempt during restoration [10]. When they were relocated to a new site, the victims had to endure distrust and conflict with their hosts and adjacent communities. Similarly, it was reported that after a disaster, victims were exposed to rumors, competition, and conflicts that exacerbated their anxiety and fear [11], generational conflicts [12], and significantly weakened communal bonds [13]. For the individuals affected by a social disaster, damage can be extensive and prolonged and can result in various social adjustment problems, at times undermining existing social systems and cultural structure. Therefore, it is vital to account for significant changes in social relationships when a disaster happens, as well as for the consequences of those changes (e.g., secondary trauma) [14]. However, existing research on social disasters has mainly studied psychological trauma and symptoms, overlooking issues of social relationship change for victims. Moreover, existing studies have rarely addressed victims' interpersonal experiences or changes in social relatedness in their daily life by looking at their own voices. Looking into victims' internal experiences helps develop effective interventions for them [15], and a qualitative approach can suitably inform understanding their own perspectives regarding such specific situations [16]. Also, given that the loss of a child can result in mourning far more intense and prolonged than any other loss [17][18], we can assume that victims' families' experiences in the Sewol disaster may show different characteristics compared to those from other social disasters. In this study, we adopted a qualitative approach to grasp changes in social relationships. For affected individuals, trauma is not a mere set of events; it can be perceived differently depending on how they make sense of the incident individually and socially, and how they react to it [5]. This study is intended to examine how the families' social experiences have changed since the tragedy. We conducted detailed interviews with family members of the Danwon high school student victims, exploring how their social relationships have changed since the Sewol ferry disaster. Looking into their subjective experiences can contribute to the establishment of effective and realistic interventions for victims and their families through the development of proper coping and social support resources as well as of relevant rehabilitation policies. --- Methods Participants Study participants were immediate family members, mostly parents, of the high school student victims. If a student victim had lived with a grandparent or other relative as a caretaker, he or she was also included in the study. The interviews were conducted from February to June 2016. The committee for special investigation of the Sewol ferry disaster provided a register of the student victims' families who had signed a consent form for release of information for research purposes. The committee and this study's research team contacted the 127 families who consented, and 64 participants agreed to participate in the interviews. While preparing for the interviews, 10 participants left the study due to scheduling conflicts or personal issues. We conducted semi-structured interviews with the remaining 54 participants, comprising of 27 fathers, 24 mothers, 2 grandparents, and 1 aunt. The interviews lasted from 2 to 2 1<unk>2 hours. --- Participants interview guide development We developed a first set of in-depth interview questions based on: i) reports of physical and mental health of the Sewol victims and their families, reports on human rights issues regarding the 416 Sewol disaster, and other pertinent materials, and ii) results from a pilot interview with four student victims' families. The in-depth interview questions largely consisted of an introduction, interview content, and an ending. The introduction covered the purpose of the interview, introducing the interviewer, signing a written consent form, explaining about recording and follow-ups, and offering a gift as compensation. The interview content included questions about aid and support provided by the government, organizations, and individuals, their social relationships from April 15, 2014 to the current date, and lastly changes in family relationships and other relationships outside the family. This set of interview questions was revised two times through discussions among the researchers. A pilot study was conducted with the revised set of questions. After interviewers and researchers, via two meetings, revised and modified the in-depth interview, based on the pilot study results, the final set of in-depth interview questions was produced. --- Interview procedure The research team consisted of four researchers with doctoral degrees in counseling or clinical psychology, one researcher with a master's degree, and one with a bachelor's degree. In addition to the research team, 10 more interviewers joined to conduct the in-depth interviews. All interviewers were female and had a registered license in counseling or clinical psychology with at least three years of clinical experience working at a public health organization or community mental health center. We performed each interview based on the set of in-depth questions to establish consistency and reliability across interviewers and to prevent response bias from face-to-face interactions between interviewees and interviewers. We implemented an education and training plan for interviewers. We provided educational meetings and several training sessions to interviewers two months prior to the actual interviews. An interview manual summarizing a re-questioning technique in the event of uncertain responses and a technique for conducting follow-up questions was introduced and used to maintain consistency and accuracy across interviewers in obtaining interview content. This study, as qualitative research, has fully followed the guidelines suggested by Lincoln and Guba [16], consisting of truth value, applicability, consistency, and neutrality. Before launching the project, this study was approved by the Ajou University Medical Center Institutional Review Board (SBR-SUR-15-519, SBR-SUR-15-520). We prepared a secure place in Ansan [The high school is located in Ansan] [15] convenient for participants' access to the interviews. The interviewers scheduled the interviews with each participant by phone. During these phone conversations, the participants were told the reasons and purpose for the interview and the qualifications of the interviewers, and were given an opportunity to ask questions. During the introduction, a written consent form was obtained regarding audio recordings and follow-up interviews when necessary. Accordingly, if specific information was found to be missing after the interviews, the interviewers arranged meetings for additional interviews or contacted the participants by phone to obtain the missing information. The semi-structured interviews based on the pre-established set of questions were performed, allowing additional questions whenever needed. Two participants were severely distressed during the interview; after obtaining their consent, they were referred to a nearby mental health center. Compensation was considered for their time and effort for the interview and a gift was offered as approved by IRB. --- Analysis This study aimed to draw common themes and implications from interview content, rather than to apply concepts or hypotheses from existing theories or studies for analysis. To grasp changes in the relationships established prior to the disaster, the victims' families were asked about such changes in their immediate family and other relationships. The research team attempted to draw themes and concepts based on the families' responses to those questions. This method is useful to capture meanings and subjective experiences involving participants' perceptions, understandings, and conceptions about themselves and their surroundings under certain circumstances. For this, a complete transcription of each in-depth interview was performed and cross-checked so that the participants' expressions and thoughts were accurately transcribed. Then, the interview data were sorted by research questions. Applying a thematic analysis method [19], we attempted to identify recurring, meaningful themes and grouped categories and concepts emerging out of the raw data. We applied a coding process without attempting to fit the data into any pre-established frame, so that the identified items were solely data-driven. A thematic analysis method searches for themes, focusing on similarities and differences among the participant's experiences and their meanings. This exploration seeks to 'discover' new themes for certain situations. Through this process, we aimed to capture in-depth meanings, perceptions, and the various feelings that the student victims' families had been experiencing. Five researchers with doctoral degrees in psychology with previous research experience in thematic analysis and two researchers with master's degrees participated in a thematic analysis of the responses to the interview questions. Seven researchers separately reviewed and sorted the data according to each research question, and performed coding independently. As each analysis was completed, all researchers gathered to discuss coding differences; for three months, the researchers worked out coding disagreements. A unified codebook was finalized after five modifications. Based on the codebook, the researchers recoded the interview data. For each item in the codebook, corresponding interview content from the data was written down. Similar content was regrouped and re-sorted repetitively via comparison analysis. From these processes, the raw data was summarized and trimmed to 15 pages, which would serve as the basic materials for interpretations. A frame of categories, subcategories, and concepts was established, and the details of the contents were composed. The researchers examined whether the frame and the content were valid and finalized the frame for interpretations. --- Results --- Changes in social relationships Relationships with their immediate family members. The student victims' families stated that they have experienced significant changes in their relationships with their immediate family members in daily life. The most salient change was seen in communication, with either sudden decreases or increases in conversation. For example, some had barely talked to one other, avoiding mentioning anything related to the Sewol incident. And they no longer celebrated holidays or anniversaries together: We rarely have time together for supper. You mean, together with family. When <unk> was with us, we always had supper together. Since <unk>'s gone, we have never gone out for dinner or something. His sister is now in college and comes home on Fridays. Yet, we don't talk like before, laughing. She's just in her room looking at her cell phone. We don't talk any more. Years ago, we used to go out together at least once or twice a month or drive somewhere near just to talk and eat. But, we don't do that anymore. We rarely go out together for dinner or something (A 49-year-old father who lost a daughter). Additionally, the families reported that they had become sensitive to other family members' psychological changes and had experienced changes in their families' roles or structures. They often felt that some family members, such as children or old grandparents, were being neglected. Parents realized that the siblings of the lost child were experiencing certain changes in their own lives. For example, some parents were aware that the remaining children often had to find employment while being traumatized by losing their sibling, or had to drop out of school due to being discouraged about their uncertain futures. As a result, the parents were concerned about the other siblings, as to whether they would adjust in their social lives or get hurt from a privacy breach. They had become concerned with their children's struggles with adjustment in school. Consequently, they expected intervening efforts for siblings of the student victims, for example support groups or a government-led rehabilitation plan for siblings of victims. Relationships with co-workers, friends, and relatives. According to the study, since the Sewol ferry disaster, the victims' families have experienced feelings of isolation and differences in their relationships with coworkers, friends, and relatives. Most struggled with feelings of discomfort and frustration when communicating with people they had known before the incident. They stated that it was difficult for them to express their feelings freely in front of people other than fellow victims' families and that they avoid people who knew their circumstances: More than anything else, we found it's difficult to talk to other people. Actually, I could have tried harder, but just found myself avoiding it. You know, it's even hard to talk to my family, then how is it not difficult talking with them? Together? (A 50-year-old father who lost a daughter) They tried to tune into others around them in trying to get along. Yet the victims' families reported feeling distant and that their acquaintances often approached them in nongenuine ways or showed mere sympathy. They often felt supported when the people around them demanded a thorough investigation of the Sewol incident, but felt discouraged when they were advised to go back to their homes and workplaces and stop their protests. Table 1 shows the changes in social relationships for the student victims' families. Relationships with the people related to the school. The victims' families stated that they experienced much conflicts with people associated with the school. They had arguments over maintenance of classrooms, and also felt disheartened by surviving students and their parents who declined to participate in the victims' families committee. They became reluctant to communicate with surviving students because their contact might pressure them: We had another trauma coming. It's because of the classroom. I moved to this town, decades ago, in 1994 right after I got married. So I know them really well. These people... they had a meeting once. I got there. Then, those people I've known for ten years, they shut me out and kept me from even entering the room. It's shocking to me, honestly....(A 52-year-old mother who lost a son). Although the families of student victims were struggling with the people associated with the school, they were also concerned about the well-being of the surviving students. Relationships with the community. The victims' families often felt bonded with their neighbors, but felt wary of them at the same time. They experienced a sense of being different and discomfort in their relationships with neighbors, so they often ended up mostly avoiding them: If someone I know is coming this way, I lower my head and take a detour. It's because I don't want to see them. It was much worse at that time, than now. Shampoos, soaps, and something like that, they all were donated, and those things were very useful. It's still the same. Yet, now I need some clothes or something, I just make an order on TV home shopping and have them delivered (a 47-year-old mother who lost a daughter). --- Relationship with citizens. While the victims' families reported that they felt supported by the public and also sought such support, they noted experiencing conflict and discomfort. They reported that civic groups, volunteers from the island Jindo, and oriental medical doctors had supported them enormously: (The solidarity 416) I saw them on Facebook and got a chance to get there, Gumi last Saturday. It was so helpful just looking at how they worked (A 46-year-old mother who lost a son). They sought support from the public and also tried to maintain a sense of solidarity with the public by holding meetings for public discussion and moderating their demands, taking into account public opinion. Despite their efforts, the families stated that they have struggled tremendously. For example, they reported incidents where they were slapped on the cheek and felt hurt by undervaluing statements about the Sewol ferry disaster. They reported uneasiness in that people seemed to misunderstand their compensation requests. They also felt sorry and worried about whether they were burdening other people or making them uncomfortable in general: It's not about monetary compensation. But, even yesterday, someone from the parents' association talked to me. An old woman asked me, why we keep this tent and not stop protesting now that we got so much monetary compensation. She said we got that much money and should move on. Those kinds of words hurt us most, like cutting our heart out. We don't accept the compensation money. We told them this so many times, but they don't believe us. They said, "well, you got that much money; then why don't you move on?" If you lost a child, can you forget and move on? It's not them who lost a child. And they say things so easy (A 52-year-old mother who lost a son). Relationships with the Sewol victims' families committee. Many victims' families reported that they struggled in daily life and had difficulty recovering their previous social relationships. Instead, they relied on a new community, the victims' families committee. It was reported that among the student victims' families, a strong bond and solidarity was formed so that they were able to share each other's' pain by supporting and caring for each other. They reported that, while working for the committee, they had shared personal tips for psychological coping and worked through their own suicidal urges and attempts: Right! It helps each other. When we see other families, they're the one who went through the same pain as ours. They are holing up too. I still think about killing myself. It's still same that I got to follow (the child). But now, I have another strong feeling that I shouldn't disappear like that (A45-year-old mother who lost a son). The families had built up positive experiences by working for the families committee, but were also faced with conflicts regarding reparation and compensation for the incident. For instance, they often were discontented with and unconvinced about the work done by the committee, which they thought had made little progress in representing their plights and requests. Relationships with government and society. The grieving families stated that they have struggled with feelings of anxiety about people and the world, and of bitterness about human nature and the unjust world, as a consequence of their changing social experiences after the Sewol ferry disaster. Some families were always anxious about the safety of their remaining children and nervous about seeing other people. Furthermore, the families felt shame about their dysfunctional government: It's even frightening to see other people, honestly. So, I don't see anyone, though I have known them for decades. My brothers and sisters don't come. Well, even though they come to see me, I try not see them. Because I can't say nicely like before. Now I can't be near anywhere people I know are around (A 48-year-old mother who lost a son). Because of their changing social experiences after the incident, the families have experienced distrust in the government or people and growing worries about dysfunctional government. Their distrust had often developed due to suspicions about the official activities of the Korean national intelligence service or the Ministry of Oceans and Fisheries. For example, some had become paranoid that the government might track their locations or tap their cell phones. As one parent stated: Really, even one little thing (doesn't make sense). If the ferry didn't leave port in such heavy fog, nothing would happen. Why they let it leave port, why let only the Sewol ferry leave? Why can't the government investigate that? We're really disappointed and frustrated. If the accident didn't happen, we could just pull the ship up and find out why. If the government investigates how the ferry left port, everything will be clear. You know, honestly, I'm so frustrated about the government of South Korea, that can't investigate anything. The way the government responded makes us more upset and angrier than the accident itself. Even if I went there [to Cheongwadae, an official residence of the South Korean president], would I shoot someone or do something bad? What would I do? Why don't they just let us vent, instead of making us worse? They kept us from going forward in everywhere (A 46-yearold mother who lost a daughter). --- And as another said: The government doesn't do anything to find out how it happened. In fact, it's not that they can't. It's more like they don't do it on purpose. Rather, the government, along with Cheongwadae and all institutions, is obstructing, covering up, and manipulating the truth. It's really... (A 51-year-old father who lost a son). Although they felt fear, shame, and distrust in the government and society, they have reported their increased care about people in pain since the Sewol ferry incident: Now that I'm hurt, I could see people who were hurt. I could see the comfort women [the girls and women who were forced into sex slavery in the Japanese military during World War II. They were abducted from their home or lured by a promise of a job] in the Japanese military... Though it's their pain, I could see their hurting now that I'm hurt. I would like to join them, and if there's someone who's hurt, I like to band together with them.... like I have been received. I don't know what kind of job I will have, but I would like to help them by doing my job (A 22-year-old brother who lost a younger brother). --- Discussion The analysis of the changes in social relationships for the Sewol victims' families showed that they have experienced a variety of changes in their social relatedness. Changes in their social relationships were largely divided into relationships in the immediate/proximal environment and relationships in distal environments. The former included subcategories such as immediate family, coworkers, friends, relatives, surviving students, and the victims' parents as well as concepts corresponding to each subcategory. The latter involved subcategories such as neighbors, other citizens, the victims' family committee, government, and society as well as concepts corresponding to each subcategory. First, the roles and characteristics of the immediate families have undergone considerable change since the Sewol ferry disaster. Whereas communication among family members had decreased in general, for some families, communication had increased. They stated that they have become more sensitive toward their own families. Parents expressed concerns about their remaining children and about how those children had changed. The grieving families struggled with the task of social adjustment due to recent significant changes in their surroundings, and with coping with multiple losses simultaneously. According to Dyregrov [20], when an individual loses a family member, he or she tends to struggle in daily life while being overwhelmed by long-lasting, strong feelings toward the lost one. Therefore, the remaining family members are likely to face several changes in their practical roles and the relationships associated with their roles and life structures. That is, the remaining family should search for their own ways of adjusting, while admitting that their loved one is gone. Their life goes through a long process in which they struggle to adjust to the new environment without the deceased. While interacting with their close coworkers, friends, and relatives and also with the neighbors or general citizens, the victims' families experienced feelings of being different, isolated, uneasy, and disappointed, although at times they felt supported by those same people. Moreover, it was found that they experienced various conflicts including arguments with surviving students and their parents over maintenance of the classrooms. According to Picou et al. [11], conflicts and competition tend to occur among victims, who are then likely to be exposed to rumors and to suffer from severe psychological anxiety and fear. In this study, it was reported that the victims' families realized they were living in 'a different world' in their proximal and distal relationships and that such feelings had led them to avoid any acquaintance if possible. Normally, when an individual lacks interaction with his or her family, neighbors, friends, or people in a small network, a sense of isolation can occur [21]. Moreover, a parent who has lost a significant object of attachment, for example, a child, can suffer from difficulty relating with others [22], which in turn aggravates a sense of isolation. The grieving families felt frustrated with the surviving students' parents whom they had once believed to be the most understanding. The conflicts with those parents again led the victims' family to feeling isolated, and having the vicious cycle of avoidance and the following sense of isolation continue. It was reported the victims' families experienced fear, shame, and distrust in the government and society, and were also concerned the government was not working. As Erickson [23] indicated, if the party responsible for the disaster excuses themselves from culpability or evades their own accountability, victims and grieving families are likely to suffer from frustration and distrust of society and the world, which in turn can lead them to feel shame and humiliation. And their surrounding communities can be shattered into pieces beyond repair. Their painful experiences have led them to feel close to people who were hurt like they were. Such an increase in social consciousness for other marginalized groups serves as an example of how a sense of solidarity is constructed among the victims of social disasters during the course of mistreatment by the government. Also, it was found that the grieving families found their greatest support from relationships with their fellow victims' families who had also lost a child in the Sewol ferry disaster. The victims' families had organized the families committee and built a strong bond, caring for and supporting each other. Such support shared by the families seemed to have helped them maintain mental health. Although at times they had conflicts with other families regarding coming up with a collective demand for compensation, they still were the most powerful resources for each other. It is known that social support can extensively moderate the impact of trauma, serving as a vital, protective factor for posttraumatic stress disorder [24][25]. Particularly in a collectivistic society, where social relationships and appraisals are highly appreciated, supportive interpersonal experiences can soothe the impact of a social trauma [26]. Similarly, the Sewol victims' families depended on other victims' families' as offering the most reliable support each other. The implication of this study's findings can be summarized as follows. First, we examined the daily interpersonal experiences of the Sewol victims' families, exploring how the disaster had changed their social relatedness. The results show that in the relationship with immediate family, parents were trying hard to be the parents of the lost child and of the remaining ones despite a significant change in their family morale and life styles. By contrast, it was found that in their general social interactions, except for those with the families committee, they experienced a sense of'uneasiness' and 'conflict.' According to Doka [27], the mourning process would be stuck until the survivors could recognize the reality of or make sense of the loss. When parents who lost a child are unable to understand the reason and cause of the death, in their mourning they can suffer from great difficulties like severe adjustment problems [28]. It is likely that when the parents cannot make sense of the child's death, the child's psychological presence vividly goes on with the parents, while his or her physical presence disappears [29]. Although more than two years have passed since the disaster, the cause of and the incompetent rescue and search operation of the incident has not been fully investigated. Considering this predicament, the grieving family may have had difficulties accepting their loss and had developed a strong sense of distrust in society itself, which can bring about conflicts with various social relationships within the community. Second, we focused on the social experiences of the parents of the Sewol student victims. Grieving the loss of a child entails intense emotional, behavioral, cognitive, and sociopsychological reactions [30]. Deeper exploration into such reactions is indispensable as effective rehabilitation from a manmade disaster needs to tap into subjective and individual feelings and perceptions of the disaster and of the reparation process, as well as the disaster's explicit consequences. Third, this study is intended to shed light on how to implement supportive social environments and aid for victims and their families suffering from a manmade disaster. According to the findings, the relationships with 'the families committee' made the victims' families feel supported and positive, through their ability to share their painful experiences and feelings during the incident with the fellow families [31], offering mutual support to each other. LaCapra [32], who studied the shooting at a Norwegian youth camp on the island of Ut<unk>ya explained that the victims and their families were able to work through the pain of grief because the community as a whole shared their grief and painful memories while reflecting on the disaster altogether. Facilitating a family committee or small support groups can effectively promote experiences of mutual support and aid among victims and their families in that they can share common feelings and experiences about their loss and grief, with less concern about not being understood. --- Limitations It should be pointed out that this study was conducted two years after the Sewol ferry disaster and that the families' reported experiences may have changed over time. Therefore, a systematic assessment based on more detailed timelines could contribute to the establishment of long-term rehabilitation plans for the victims' families. Also, there were some parents of the school student victims who did not participate in this study. It is likely that they might be more traumatized than other families. Those missing families are a clear limitation of this study. Also, the Sewol ferry victims' families are classified to two groups, the families awaiting recovery of missing bodies and the families who found the dead bodies. Although the research team gathered interview data from the families with missing children, the difficulties with those families were determined to be a significant subject requiring a whole different approach; therefore, the data was not used in this study. A future study can focus on the families of children of missing bodies, particularly on their coping with social conflicts, or explore further differences in changes in social relationships between the family who found the bodies and those awaiting recovery of bodies. --- Supporting information S1 Appendix. Semi-structured interview guide. (PDF) --- Author Contributions Conceptualization: Sun Mi Cho, Ansuk Jeong, Jung Hee Ha, Eun Young Kim. Data curation: Sun Mi Cho, Ansuk Jeong, Jung Hee Ha, Eun Young Kim. Formal analysis: Ansuk Jeong, Jung Hee Ha, Eun Young Kim. --- Funding acquisition: Sun Mi Cho. Investigation: Sun Mi Cho, Ansuk Jeong, Jung Hee Ha, Eun Young Kim. Methodology: Ansuk Jeong, Jung Hee Ha, Eun Young Kim. --- Project administration: Sun Mi Cho. Supervision: Sun Mi Cho, Jung Hee Ha. Writing -original draft: Sun Mi Cho, Ansuk Jeong, Jung Hee Ha, Eun Young Kim. Writing -review & editing: Sun Mi Cho, Ansuk Jeong, Jung Hee Ha, Eun Young Kim.

The Sewol ferry incident on April 16, 2014 in South Korea claimed the lives of 304 individuals, including about 250 high school students on a school trip. The majority of South Korean citizens were watching live updates on the capsized Sewol ferry, anxiously watching on TV how the vessel fully sunk over time. They were desperately hoping for the rescue of the survivors inside. However, their anxiety had become shock, anger, and helplessness, and the disaster has become a daunting, collective trauma, not just to the victims and their families, but also to the citizens who were exposed only through the media. In this study, we interviewed victims' families two years after the incident. We explored how they have experienced changes in their social relationships. We conducted semi-structured interviews of 54 family members of the student victims. We qualitatively examined the data applying a thematic analysis. Changes in their social relationships were largely divided into the relationships in the proximal environment and the relationships in distal environments. The former included subcategories such as immediate family, coworkers, friends, relatives, survived students and their parents, and concepts corresponding to each subcategory. The latter involved subcategories such as neighbors, other citizens, the victims' family committee, government, and society, and concepts subject to each subcategory. Based on these findings, rehabilitation plans for trauma victims and their families should take into account the significant changes in their social relationships and the further consequences of those changes.

INTRODUCTION Adolescence, here broadly defined as being between 10 and 19 years old, is characterized by a series of physical, psychological, hormonal and behavioral changes associated with the process of maturing (Inchley, 2016). Factors such as the need for social affirmation, influence of peers, greater independence from the family, associated with failure to perceive long-term risks properly, make adolescence a phase with many new challenges (Patton et al., 2016). During this phase, the family assumes a more important role as the main source of individuals' socialization, and parental figures, be they parents or guardians, have a strong influence on the well-being of adolescents in the short, medium and long terms, by building the base for their development in the transition from childhood to adulthood (Loke and Mak, 2013). Specifically, the family structure, regarding the conjugal status of the parents, is indicated as a factor with impact on individual formation. Youths who have experienced the divorce of their parents or the absence of a parental figure generally suffer strong emotional stress, with potential to develop behavioral disturbances, triggering future physical and mental health problems (Butters, 2002;Loke and Mak, 2013;Schenker and de Souza Minayo, 2003). This emotional baggage together with the typically rebellious behavior that is characteristic of the phase not only can influence the early consumption of psychoactive substances, but also increase the risk of abusive use of these substances throughout adult life. Substances such as alcohol, tobacco and illicit drugs can have serious negative impacts on those who consume them, so the earlier people start using them, the greater will be their negative effects on future health, development and welfare (Aarons et al., 1999;Gruber, 2009). These substances can also detract from academic performance, in turn affecting the professional life, placement in the job market, earnings and quality of life over the long run (Inchley, 2016;Lima and Santos, 2016;Malta et al., 2010). The literature on the effect divorce on the behavior of children and adolescents has noted the difficulty of isolating the causal effects of the huge number of factors that can lead to behavioral disturbances, such as family structure, the causes of the divorce, time gap since the divorce, socioeconomic level after the divorce, how the divorce occurred (peaceful or traumatic for the child, including degree of judicial intervention) and relationship of the parents with the child after the divorce. The interplay of these and other factors is very hard to unravel, and can result in biases in estimating the effects (Amato and Keith, 1991;Amato and Anthony, 2014). Besides unobservable characteristics that can affect adolescents related to the rupture of the family structure, other past factors can affect changes in behavior, such as some type of abuse or an adverse situation suffered during early childhood. As astutely observed by Felitti et al. (1998), children who experience adverse situations in infancy, such as domestic violence (against the child and/or the mother), or family dysfunctions (judicial or criminal problems of the parents), tend to be more likely to engage in behavior risky to health in adolescence. With regard to evidence about health risk behavior, the statistics produced by the Brazilian Institute of Geography and Statistics (IBGE) stand out, obtained from the National Student Health Survey (PeNSE). According to information gleaned by the survey, in 2015, 22.9% of Brazilian students between the ages of 13 and 17 years had smoked a cigarette on at least one occasion. Of them, 12.4% had first experimented with smoking tobacco with age of 13 years or under. With regard to consumption of illicit drugs, the statistic was 12%, with 4.2% having done so when they were 13 years old or younger. The prevalence of alcohol consumption among Brazilian adolescents is even more worrying, given that 61.4% of the survey's respondents reported having experimented with drinking, with 30.6% having done so for the first time in early adolescence. In light of this scenario and recognition of the importance of the family structure on the behavior of adolescents, we designed this study to evaluate the effects of family structure on the health risk behavior of adolescents aged from 13 to 17 years as reflected in the PeNSE for 2015. Our working hypothesis was that being raised in a single-parent household rather than one with two parents can contribute to negative behaviors of adolescents, increasing their exposure to health risks due to a greater likelihood of consuming tobacco, alcohol and illicit drugs. For this purpose, we used propensity score matching (PSM) to mitigate biases and make inferences about the effects of family structure on the risk behavior of Brazilian adolescents based on data from the PeNSE. The choice of this method was mainly due to the availability and format of the PeNSE database, as well as the mentioned ability to mitigate biases brought by observable characteristics and the absence of common support. To complement the study, we also evaluated the existence of heterogeneous effects between students in elementary school (through 9th grade) and high school (through 12th grade), given the ample evidence that the propensity for risky behavior tends to increase with age among youths. We believe the evidence presented here contributes to the literature on the influence of family structure on the risky behavior of adolescents, and can support proposals of policies focused on reducing such behavior. Besides this introduction, the article has five more sections. In the second, we present a review of the literature on the interplay of family background, the health risk behavior of adolescents and unobservable variables that should be considered in measuring the causal effect under analysis. In the third section, we describe the data collection and the empirical strategy to estimate the influence of family arrangements on the health risk behavior of Brazilian youths. The analysis and discussion of the results are presented in the fourth section, followed by brief considerations on public policies formulated to address the problem. Our final considerations are in the sixth section. --- LITERATURE REVIEW --- Family background The nuclear family is defined as a private institution that, irrespective of its composition, has the main function of primary socialization of children, besides the choice of friends so as to build healthy ties among the members, in particular to instill standards in offspring considered socially adequate (Schenker and de Souza Minayo, 2003). Within the nuclear family, the parental figure or figures assume the principal role of setting examples for habits and behavioral patterns. The parents (or other caregivers) thus serve as role models for the offspring, as well as taking measures to protect them from behavioral disturbances that can in some way affect their welfare in the short, medium and/or long term, and provide the foundation for the transition to adult life (Loke and Mak, 2013). Specifically regarding behavioral disturbances, such as consumption of psychoactive substances during adolescence, some family characteristics can positively or negatively influence this consumption. The use of these substances by one or both parents or older siblings is the most direct route of negative influence. In turn, the conjugal situation of the parents and their relations with the children exert strong influences on the feelings of adolescents and hence on their behavioral problems (Schenker and de Souza Minayo, 2003;Loke and Mak, 2013;Butters, 2002). Furthermore, a problematic family structure, gauged by the conjugal situation of the parents, can be considered a stress factor. Amato and Keith (1991) performed a meta-analysis of the effects of a family rupture on children and youths. They observed a lower level of wellbeing among youths who had experienced separation/divorce of their parents in comparison with peers who had not experienced such family turmoil. They also found that the reduction of well-being was more intense when experienced in the early phases of development. The reflections can be ascertained in various ways, such as lower academic performance, behavioral disturbances, psychological disturbances, altered self-perception, and changes in relations with friends and parents. Further according to Amato and Keith (1991), the exit of a parental figure from the home can cause a reduction of aggregate household income, with a negative effect on the standard of living of adolescents. In extreme cases, this can prompt youths to drop out of school so as to help financially support the family. Besides this, the fact of living in a single-parent household after a divorce usually means a reduction of routine contact with both parents, which can trigger a series of complications for youths' development and mental health, among other factors. Lima and Santos (2016) observed, based on panel data on Brazilian students, a negative effect of single-parent households on school performance, providing evidence that the stress provoked by a change in the family structure really affects school results. Various other secondary factors can cause stress, leading to lower well-being of youths whose parents are divorced or who never counted on the presence of both parents. The move to a new home, the efforts to adapt to a new school and new social context, ongoing conflicts between parents, and remarriage of one or both parents can all serve as triggers to start consuming psychoactive substances. According to Butters (2002), in a study among young students in Ontario, Canada, the feeling of family disruption when a parent leaves the dwelling can result in a negative feeling in adolescents, making them more likely to consume marijuana and also to use it more intensely. According to the author, the feeling of reduced stress caused by this psychoactive substance leads adolescents to use it as a way to cure, so to speak, the negative feelings and stress caused by parents' divorce. Although the isolated effect of the family structure was considered low, the author also alerted that when observing this together with the effect in school, the overall effect can be considered strong. In other words, youths with disruptive family structures tend to have more problematic school performance, with greater isolation from classmates, among other problems at school, leading to increased consumption of cannabis, generating a vicious circle. Another perspective was presented by Loke and Mak (2013) in a study among adolescent students in a district of Hong Kong considered to be economically vulnerable and with high prevalence of youths with health risk behavior. They found that besides the behavioral and affective characteristics of the family, such as an authoritative parenting style with little emotional support, the students in the survey from single-parent households had a higher probability of developing the habits of smoking cigarettes and drinking alcohol. Among the conclusions, the authors indicated that this pattern can be due to reduced parental contact, such as the absence of a father, as well as the smaller family income earned by a single parent. In longitudinal studies with more than 600 Scottish students between 13 and 18 years old (divided into subsamples with ages of 13-14, 15-16 and 17-18 years), Shucksmith et al. (1997) and Glendinning et al. (1997) observed similar results regarding the analysis of family and socioeconomic factors and their relationship with the regular consumption of cigarettes and alcoholic beverages. The results of both studies indicated that the socioeconomic status of the parents had little or no effect on the regular consumption of the two psychoactive substances in question. On the other hand, the family structure, parenting style and students' relationship with their parents had strong influences on the evolution of these habits, making consumption of the substances considered regular à posteriori by the students. With respect to family structure, the authors analyzed students from single-parent, two-parent and reconstituted families (with the presence of a step-parent). Students from two-parent households tended to smoke less than those from single-parent households or those with reconstituted families. They also observed a lesser prevalence of regular consumption of alcohol among students living with both parents. More specifically, among students reporting consumption of alcohol more than once a week, the prevalences were 9%, 10% and 18% for two-parent, reconstituted and single-parent families, respectively. Nevertheless, in both studies, the authors stressed the importance of the link created between parents and offspring according to the parenting style as one of the characteristics with the strongest impacts on the healthy development of adolescents. Shucksmith et al. (1997) divided the family parenting styles into authoritarian (little emotional support from parents associated with strong control levels), permissive (strong emotional support and little control), and neglectful (weak control and low emotional support). 1 Based on these styles, the authors stressed that the levels of dialog with parents (support) and their control have a greater effect on consumption of alcohol in proportion to older age ranges of the sample, mainly when this consumption becomes more frequent 2. In general, adolescents of older age ranges had a greater consumption pattern from families where the parents were authoritarian (i.e., with little dialog and strong control). On the other hand, albeit with lesser effects, in younger adolescents the authors observed greater consumption when the parents were considered neglectful (little dialog and little control). This means to say that extreme parenting styles perceived by the adolescents, both authoritarian and neglectful, had a direct effect of increasing the consumption of alcoholic beverages (Shucksmith et al., 1997). Glendinning et al. (1997) also observed similar results regarding smoking and different parenting styles. They analyzed the same family characteristics and parenting styles as in the study by Shucksmith et al. (1997), finding a greater influence of the relationship with parents of older adolescents regarding the frequency of smoking cigarettes. Additionally, the consumption of cigarettes was higher in students who defined their parents as being authoritarian (low dialog and high control), as well as neglectful (little dialog and low control). --- Family structure and causal effect in the literature Important problems have been raised by many researchers who have investigated the family structure and the effects caused by family disruption on children and adolescents, namely the impossibility of conducting randomized experiments and the difficulty of preventing numerous unobservable characteristics involved in the development of children from generating endogeneity, and thus interfering in the conclusions obtained regarding the expected causal effects (Amato, 2010;Amato and Anthony, 2014). Amato and Anthony (2014) stressed that among the difficulties of isolating effects are questions that are almost always impossible to observe. For example, many of the factors that prompt parents to separate also directly or indirectly affect the child, resulting in a spurious relationship between the results of divorce and the effects observed on the child. When the object of study is adolescents, this difficulty can be even greater, since adolescents already carry baggage from their childhood, which also can undesirably interfere in the results, and thus hamper obtaining solid results. For example, we can mention the study by Felitti et al. (1998), based on the medical assessment of adults with a comorbidity. They found that adverse situations experienced during childhood (such as various types of abuse, often occurring concomitantly) significantly increased the chances of exposure to risky behaviors and factors such as consumption of psychoactive substances, obesity and suicidal tendencies, among others. In other words, when studying the effect of family disruption on adolescents and the relationship with health risk behaviors, a possibility exists of an alternative hypothesis according to which part of this effect is due to some form of abuse suffered during childhood. Another finding reported in the literature is the fact that many of the effects of divorce observed on children and adolescents also can likely be blamed on problematic relations within the family. Among these are the presence of many conflicts between the parental figures and their difficulties in establishing healthy relationships with offspring (Amato, 2010;McLanahan et al., 2013). The questions leading to divorce can also play a role in the matter of endogeneity and hamper isolating the results of the effect itself. An economic crisis can lead to unemployment, in turn reducing the quality of life of all family members, hence increasing the likelihood of divorce and directly impacting the child. Furthermore, marriage between adolescents can increase the probability of later conflicts between parents and of divorce or separation, with strong negative effects on offspring. Besides these factors, infidelity, domestic violence and high levels of dissatisfaction in the marriage are other hard-to-observe variables that can lead to divorce. with a strong impact on the offspring (Amato, 2010). Because of these potential methodological drawbacks in this area, researchers have been seeking methods to overcome them and hence eliminate the bias caused by unobservable variables. Among these are growth curve models, individual or sibling fixed effect models and the propensity score matching (PSM) method. When longitudinal data are available, one of the ways to mitigate this undesired effect is to apply fixed-effect models, with the aim of eliminating all time-invariant differences among the individuals analyzed. Amato and Anthony (2014) applied this method in two databases (elementary school children and high school adolescents) considering periods before and after divorce. This enabled them to capture the negative effects on the students from broken families, such as lower reading and math scores, and difficulty of self-control and externalizing problems. Nevertheless, when the database does not permit controlling for/observing specific characteristics (of the family, the individuals' past experiences or the process of divorce of parents) that can influence the results, PSM has been employed in the literature as one of the most effective ways to circumvent this question. Frisco et al. (2007) analyzed the possible impact of marriage dissolution on adolescents' academic performance. They compared the application of two methods, PSM and traditional ordinary least squares regression analysis, to investigate the influence of endogeneity and selection bias on the results. With both methods, the authors found lower performance of students, mainly in math scores, in the short run, but the results estimated via PSM were more robust based on the common database used. Likewise, Hussey et al. (2016) also used PSM to control for the effect of unobservable characteristics in a longitudinal study to assess the impact of divorce on a broad range of outcomes, such as academic performance, unemployment, risk behavior during adolescence and mental health issues, in a period of 1 to 14 years after the event. Among the conclusions, they observed that students submitted to family separation had lower academic performance in all intervals studied, more precarious mental health in the short and medium run, and were more likely to develop health risk behaviors. --- Health risk behaviors during adolescence There are various theories about the factors that can lead adolescents to start consuming psychoactive substances and thus pose risks to their physical and mental health. Among these are the influence of peers, the various physical and psychological changes due to puberty, and stress caused by psychological traumas, among others (Gruber, 2009;Almeida and Ara<unk>jo J<unk>nior, 2016;Inchley, 2016;Malta et al., 2010). According to data from the World Health Organization in a survey investigating the health risk and protection factors of European youths (Health Behaviour in School-Aged Children -HBSC), from 2013-2014, over 14.6 million youths stated they had consumed cannabis at least once in their lives. Besides this, the data indicated that smoking cigarettes began on average at the age of 13 years or younger, with no differences between girls and boys. With respect to alcohol consumption, the survey for those years pointed to a worrying increase in the proportion youths aged between 13 and 15 years throughout Europe who had been drunk at least one time in their lives (Inchley, 2016). Of equal concern, in the United States one-third of the students interviewed in the Youth Risk Behavior Survey (YRBS) stated they had smoked cigarettes in the 30 days before being questioned. Furthermore, among elementary school students (up to sixth grade), 75% stated they had tried an alcoholic beverage at least once, and over 60% of high school students stated they had been inebriated at least once (Gruber, 2009). Among students who were already regular smokers in the USA, Gruber (2009) also found that they underestimated the probability of being addicted as adults: 56% stated they would not continue smoking for more than five years. In turn, Harrison et al. (2019) studied "first-time offending court-involved, non-incarcerated" (FTO-CINI) youths and found evidence that recent smoking was higher among these youths than the average of their peers in the general population, and they had a high probability of developing behavioral disturbances such as delinquency, especially problems associated with anxiety. In Brazil, based on analysis of the first version of the PeNSE ( 2009), Malta et al. ( 2010) observed that the prevalence of smoking among ninth-graders (24.2%) was relatively low compared with the data from studies at the global level. With regard to drinking, the rates were high and differed between the sexes (67.6% for boys and 71.9% for girls). These high rates were explained by the continuing social acceptance of teenage drinking and the heavy advertising of alcoholic beverages in the media. The authors urged policymakers to take measures to reduce these levels so as to preserve the future health of today's youths. Malta et al. (2014a) compared the prevalences of the main health risk behaviors and measures to protect against non-communicable chronic diseases among students in Brazilian capital cities based on the 2009 and 2012 versions of the PeNSE. They reported a reduction in the percentage of adolescents who had tried cigarettes, from 24.2% to 22.3%, but no change in the percentage of regular smokers (6.0%). With regard to illicit drugs, they observed an increase from 8.7% 2009 to 9.6% in 2012, while regarding alcohol, the experimentation rate remained the same at around 70%. Finally, they found that 27% of the respondents reported having consumed alcohol in the previous 30 days, in both versions of the survey. Therefore, the authors concluded there had been only slight changes in the patterns of health risk behavior as reflected in the two versions of the PeNSE. Malta et al. (2014b) studied the consumption of alcohol and other drugs using the database of the PeNSE for 2012 and reported evidence of an equally worrying scenario. In particular, they found very young ages for first consumption of alcohol (12-13 years old), possibly explained by the facility in Brazil of buying alcoholic beverages at bars, restaurants and supermarkets, among other establishments. With regard to smoking, when analyzing 9th-grade students using data from the PeNSE for 2012, Barreto et al. (2014) observed that one in five students interviewed had tried smoking. Among these, one-fourth were defined as regular smokers. Besides this, they found that students with the highest risk, mainly regular smokers, more often were gainfully employed or came from single-parent households. Finally, Horta et al. (2014) investigated the prevalence of the use of illicit drugs using the same database. Besides identifying a positive relationship between the frequency of consumption and the age of the respondents, they observed earlier start of consumption. Furthermore, they also found no significant differences between the sexes, indicating the possible occurrence of homogenized consumption among boys and girls. --- METHODOLOGY --- Description of the database This study is based on data from the PeNSE carried out by the IBGE together with the Ministry of Health (MS). The specific objective is to monitor and evaluate the health risk and protection behaviors of adolescent Brazilian students. Among the risk factors are matters such as sexual and reproductive behavior, smoking, consumption of alcohol and illicit drugs, diet, sedentary lifestyle, mental stress and some types of violence. The survey covers elementary and high school students in Brazilian private and public schools, which are selected from the list schools registered with the An<unk>sio Teixeira National Institute of Educational Research and Studies (Inep). Our focus is on 9th-grade students between the ages of 13 and 15 years. We chose this age range in accordance with the recommendations of the World Health Organization (WTO), whereby students older than 13 years have the necessary skills to fill out a self-applied questionnaire, and mainly for being the age range when adolescents are considered susceptible to exposure to various risk factors. The PeNSE used in this study is that applied in 2015, when the IBGE decided to add a second sample to enable comparison at the international level. In this respect,, the age range was expanded to between 13 and 17 years old (roughly corresponding to 6th to 12th grades, depending on how old they were when entering 1st grade and possible grade repetition). This second sample was obtained by interviewing 16,608 students from 371 Brazilian schools from all five of the country's geographic regions (North, Northeast, Midwest, Southeast and South). The sample is relatively balanced as to gender, with 49.95% girls (8,269 students) and 50.05% boys (8,287 students). Furthermore, 60.37% were in elementary school and 39.63% were in high school at the time of the survey. The PeNSE database comes from a transversal sample, so it has some limitations, mainly regarding observation of characteristics and occurrences in the respondents' lives before the application of the questionnaire. Another drawback is that it is a recent survey3 in comparison with other similar surveys used as models for its creation. Among these are the Youth Risk Behavior Surveillance System (YRBSS), which involves multiple surveys applied since 1991 among adolescents in all American states, and the Health Behaviour in School-aged Children (HBSC) survey, which began in 1983 in 5 European countries and in its latest version (2017-18) encompasses 45 countries from Europe and North America. [I consulted the website and discovered these numbers.] Table I below identifies the variables used in the estimations. The result variables pertain to the three risk factors: consumption of alcoholic beverages, consumption of cigarettes and consumption of illicit drugs. The treatment variables designate the control and treatment groups, along with the covariates. --- Propensity score matching It is not possible to conduct a randomized study to investigate the impacts of a family structure on the descendants. Indeed, it would be a violation of ethical questions to randomly designate which adolescent would undergo certain experiences (Amato, 2010;McLanahan et al., 2013). This leads to the inability to isolate the causal effects of the event from other unobservable variables that can influence the results obtained, causing selection biases and endogeneity (Caliendo and Kopeinig, 2008;Caliendo et al., 2017;Ichino et al., 2008). Therefore, since our database came from a transversal survey with determined observable characteristics of Brazilian students, we applied the PSM of (Rubin, 1974) and Rosenbaum Note: Prepared by the authors based on PeNSE survey. a According to the glossary of the PeNSE (2015), in this item students were questioned about whether they had used an illicit substance at least once in their lives, such as marijuana, cocaine, crack, organic solvents, ecstasy, oxy, etc. and Rubin (1983a;b). 4 We believe this is the best strategy to circumvent these problems, and thus to minimize the biases caused by unobservable characteristics, seeking to obtain more robust results (Caliendo and Kopeinig, 2008;Li, 2012). Thus, we applied the PSM method to estimate the effect of a single-parent family structure on the health risk behavior of adolescents related to consumption of alcohol, tobacco and illicit drugs. The PSM method basically involves evaluating the differences between individuals receiving a certain treatment from those not receiving it. In this case, the treatment is coming from a single-parent household. Nevertheless, it is not possible simply to observe this difference based on an individual in two different scenarios, nor is this possible by means of the average result of individuals from two-parent households. Hence, the problem arises of selection bias, where individuals that participate and do not participate in the treatment can differ. This difference precludes the desired comparison (ROSENBAUM; RUBIN, 1983a). The PSM technique aims to overcome that problem by the process of matching. This involves the identification of two subsamples, here according to the students' family structure. According to the terminology of Heckman et al. (1997), the first group, called the treated group, is composed of students living in single-parent households, while the second group, called the control group, is composed of students from two-parent families. Considering the two subsamples, the idea is to identify individuals from the two groups that are as similar as possible based on determined characteristics X (vector of covariates), thus enabling a counterfactual exercise (Caliendo and Kopeinig, 2008). According to Rosenbaum and Rubin (1983a), PSM is based on the key assumption of strongly ignorable treatment assignment, meaning that the distribution of the covariates between the two groups is the same due to the propensity score matching carried out. That score is estimated as a function of the observable characteristics and indicates the probability of an adolescent's belonging to a single-parent family, given the observed characteristics, denoted by the vector X, satisfying the following proposition: X <unk> D | b(X),(1) were the treatment status of the individual is represented by the dummy variable D, which assumes value of 1 for students from single-parent families and D i = 0 for those from twoparent families; and b(X) represents the propensity score. Besides the hypothesis of ignorability, the data for PSM also must satisfy the condition of unconfoundedness, basically stating that since the covariates are the same between individuals belonging to the control and treated groups, this implies that systematic differences observed between the two groups can be attributed to the treatment (Caliendo and Kopeinig, 2008): Y 1, Y 0 <unk> D | X (2) These potential results were observed in relation to the three types of health risk behavior of adolescents: consumption of cigarettes, alcohol and illicit drugs. The effects of the family structure were estimated for each of the risk behaviors. Thus, in this setup, Y 1i represents the risk behavior of individual i (i = 1,..., N) if D i = 1 and Y 0i if D i = 0. A second key condition for PSM is overlap, which indicates the characteristics that are common between the two groups (same value of X ), and the consequent probability of being caused by the same treatment: 0 <unk> Pr(D = 1 | X) <unk> 1 (3) After realization of the propensity score matching, the average effect of the treatment on the treated (ATT) can be defined as the difference between the average effect observed in the treated groups (students from single-parent families) and the average effect observed in the control group (students from two-parent families) (Li, 2012;Imbens and Wooldridge, 2009): AT T = E<unk>Y | D = 1, b(X)<unk> -E<unk>Y | T = 0, b(X)<unk> (4) We also separately estimated models with subsamples composed of elementary and high school students, and students living in rural versus urban zones. The last step of PSM is to certify that the hypothesis of ignorability has not been violated. In our study, this meant verifying that the unobservable variables did not cause biased results. In other words, through sensitivity analysis, we detected the robustness of the results by considering, for example, variables such as the causes of a family rupture in which the relationships forged between parents and offspring suffered, or of belonging to a family with only one parent from the outset, had the least possible influence on individuals' assignment to the treatment group (in this case the type of family structure), and thus did not influence the health risk behaviors (Becker and Caliendo, 2007). This step was carried out by considering the Rosenbaum limits and the Mantel and Haenszel (1959) test. 5 Hence, we considered the upper limits (Q+mh), which indicate overestimation of the effect, and the lower limits (Q-mh), which indicate underestimation of the effect, along with the gamma factor (<unk>). 6 --- ANALYSIS AND DISCUSSION OF RESULTS --- Descriptive analysis As can be seen from the descriptive statistics in Table II, the national disaggregation of the PeNSE sample was 5,752 students from single-parent families (37.25%) and 9,689 from twoparent families (62.75%). The differences were small regarding the family types from the five macro-regions, with the leaders being 21.24% single-parent families in the Midwest region and 20.56% in the Northeast. Besides this, in the national sample, 51% were girls and 49% boys, and the average age of all students was 14.16 years. Additionally, 86.63% of the students from single-parent families said they lived with their mother. With regard to mother's education, there were no significant differences, with the overall average being complete elementary school (through ninth grade). However, with respect to the schooling level of mothers of single-parent families, slightly over 84% had high school diplomas (with or without complete college) and only 15% had college degrees. That breakdown was slightly different for mothers of two-parent families, with corresponding percentages of 80% and 20%. Nearly 84% of the students from single-parent families attended public schools. The most common parenting style among students from single-parent families was permissive (5%), while for those from two-parent families, equal portions came from families with strict parents and from those with permissive parents (4.8%). A minority of these students stated their parents were authoritarian (0.6% of students from single-parent families versus 0.4% of students from two-parent families). With regard to health risk behaviors, students from single-parent families were more likely to consume psychoactive substances. This pattern was the same both for initiation (try for the first time) and repeated consumption (measured by frequency). The majority of students from single-parent households stated they had tried alcoholic beverages (64.05%), 8.18 percentage points higher than students from two-parent families (55.87%). Besides this, 23.7% of the students from single-parent families admitted having consumed a psychoactive substance on more than one occasion in the 30 days before the survey. That percentage was lower among students from two-parent families (19.06%). Although low, the proportions of students who reported smoking and consumption of illicit drugs were worrying in terms of public health. All told, 28% of the students from singleparent families had tried smoking at least once in their lives, in contrast to 19.5% of students from two-parent families. Furthermore, the proportion of students from single-parent families who admitted having experimented with an illegal substance at least once in life was 14.74%, versus 9.86% of students from two-parent households. Nearly half of these students who had experimented with psychoactive substances came from single-parent families (44.39%) while 45.16% of students from two-parent families stated they had consumed such substances at least once in the previous 30 days. Further according to Table II, 30% of the youths from single-parent families stated that the household head (parent or guardian) smoked, about 10% [percentage points?] greater than those from two-parent families. Therefore, we sought not only to control for the risk behavior of tobacco use, but also for the risk behaviors associated with alcohol and illicit drugs. According to Gilman et al. (2009), smoking by one or both parents is significantly associated with the risk smoking by adolescent offspring. Finally, Malta et al. (2014), working with data from previous versions of the PeNSE, found similar statistics as ours. With respect to the risk factors analyzed here, they found for 2012 that the most common behavior was consumption of alcohol. The percentage was even greater then we observed: 70.5% of the students in 2012 stated they had consumed alcoholic beverages. Besides this, that year more than 22% of the students stated they had tried smoking cigarettes, while 9.6% stated they had experimented with illicit drugs. The levels were higher among public school students, but there were no significant differences between the genders. --- Results of the estimations We estimated the propensity score for the complete sample (elementary and high school) for each health risk behavior through a logit model. The intervals established for the common support of the scores between the treated and control groups were, respectively, for the complete sample [0.14570534, 0.54987186]; elementary school [0.11929106, 0.70930107] and high school [0.1409808, 0.56212328]. Furthermore, the propensity score balance before matching was satisfactory for all the teaching levels. After defining the propensity score and de

The purpose of this study was to estimate the varied effects of family structure on the health risk behavior of Brazilian adolescents and to analyze this effect according to the levels of education (elementary school and high school). The analysis used data from the National Student Health Survey (PeNSE) for 2015, which provides information about health behavior among 13-to 17-year-old Brazilian students. The sample was subdivided into Elementary School (through 9 th grade) and High School (through 12 th grade). Estimations were made using the propensity score matching (PSM) method and the risky behaviors considered for the estimation were consumption of alcohol, cigarettes and illicit drugs. The results suggested that being raised in a single-parent household negatively affects the health risk behavior of adolescents when analyzed according to substance abuse, mainly regarding consumption of alcohol and cigarettes.

(2014), working with data from previous versions of the PeNSE, found similar statistics as ours. With respect to the risk factors analyzed here, they found for 2012 that the most common behavior was consumption of alcohol. The percentage was even greater then we observed: 70.5% of the students in 2012 stated they had consumed alcoholic beverages. Besides this, that year more than 22% of the students stated they had tried smoking cigarettes, while 9.6% stated they had experimented with illicit drugs. The levels were higher among public school students, but there were no significant differences between the genders. --- Results of the estimations We estimated the propensity score for the complete sample (elementary and high school) for each health risk behavior through a logit model. The intervals established for the common support of the scores between the treated and control groups were, respectively, for the complete sample [0.14570534, 0.54987186]; elementary school [0.11929106, 0.70930107] and high school [0.1409808, 0.56212328]. Furthermore, the propensity score balance before matching was satisfactory for all the teaching levels. After defining the propensity score and delineating the sample (based on the common support), we performed the matching through three methods: nearest neighbor, kernel and radius. 7However, it was important to verify the balance of the distribution of the covariates, of both the control and treatment groups, since we utilized the propensity score to carry out the matching. That required some tests. Thus, we applied the t-test and verified the pseudo-R2, Rubin's B and Rubin's R statistics 8 to test the balance. Moreover, for each risk behavior of the adolescents, the ATT chosen was that with the best balance of the covariates, based on the Pseudo-R2 value, which indicates how well the regressors explain the likelihood of participation in the treatment group. Thus, after the matching, there should be no systematic differences in the distribution of the covariates between the two groups, so the Pseudo-R2 should be very low (Caliendo and Kopeinig, 2008). Tables B.I-B.VIII in the Appendix B present these statistics, and confirm that our estimations were balanced. Table III below reports the effects of the risk behavior of the complete sample (elementary and high school students) in function of the family structure (one or two parents). Adolescents from single-parent families presented greater health risk behavior, thus jeopardizing their wellbeing (Amato and Keith, 1991;Loke and Mak, 2013;Butters, 2002). The average effects of the risk behavior of a student from a single-parent family were significant for consumption of cigarettes and alcoholic beverages. With regard to elementary and high school students, living in a single-parent family structure was associated with a 5.85% increase in the percentage of students who had tried smoking. The respective effects for the two school levels separately were 6.16% among elementary school students and 7.01% among high school students. Thus, in our sample, as youths progressed in school, the effect of living in a single-parent family on risky behaviors increased. of the treated and untreated groups. According to Rubin (2001), B should be lower than 25 and R should be between 0.5 and 2 for the samples to be considered sufficiently balanced. An asterisk is placed after the sample values of B and R that are outside these limits, indicating they are not sufficiently balanced. Similar effects were found with regard to alcohol consumption. The risk behavior of students from single-parent families was approximately 6.84% higher than those from two-parent families. Analysis of the data disaggregated by grade level revealed that elementary school students (high school students) from single-parent families on average consumed 8.37% (5.26%) more alcohol than their peers from two-parent families. The propensity to use illicit drugs also was higher among students from single-parent families, by 3.98%. Regarding the effects of the school levels, high school students had 5.3% higher propensity than their elementary school counterparts from single-parent households, for whom consumption of illicit drugs was 2.92%. These results corroborate those described by Butters (2002) in a study of adolescents in Ontario (Canada), regarding use of cannabis. The author also emphasized negative effects of the association of this risk behavior with the stress caused on youths by family instability, mainly when leading to family breakup. She also mentioned that other factors besides family instability influence the use of cannabis over the long term. The results for the elementary school students in our study also were not statistically significant regarding frequency of consuming tobacco and illicit drugs (this last result was not significant for any of the groups analyzed). A possible explanation is the fact that because elementary school students are younger, not only are they less likely to have experimented with psychoactive substances, but also they are less likely to honestly answer the questionnaire. In other words, younger students are more likely to give imprecise and incorrect responses. As described by Tavares et al. (2004), when responding to a self-applied questionnaire, despite the guarantee of anonymity and the classroom setting, information bias can still occur, i.e., the tendency to under-report illicit behavior. --- Sensitivity (robustness) analysis The estimates obtained for the risk behavior (tobacco, alcohol and illicit drug use, including frequency of consumption) by the PSM technique minimized the biases resulting from the observed characteristics of a single-parent family structure. However, there was no guarantee of the robustness of the estimates of the unobserved heterogeneity (Rosenbaum and Rubin, 1983a;1985;Becker and Caliendo, 2007). Table IV presents sensitivity test results of the Rosenbaum limits to verify the influence that unobservable characteristics can have on the estimated risk behavior factors of the total sample (elementary and high school).9 For the samples estimated, the risky behaviors of the adolescents did not present bias due to unobservable characteristics for <unk> = 1, while for <unk> = 2, it can be inferred that no bias existed that could interfere in the chances of a student belonging to a single-parent family, according to the values of the statistics (Q + mh ) and (Q - mh ). Therefore, based on this robustness test, despite the possible influence of unobserved variables on the risky behavior of students from single-parent households, the estimates found can be interpreted as reflecting the causal effect of the risky behavior of the samples used in the estimates. --- DISCUSSION AND CONSIDERATIONS REGARDING PUBLIC POLICIES The occurrence of divorce or being raised from the outset in a single-parent household can be a source of strong emotional stress. Hence, to soften or even prevent this impact that affects Note: In the sensitivity estimates for the health risk behavior, we did not consider the frequency of consuming illicit drugs, given that the effect of a respondent's belonging to a single-parent household did not have an effect on this risk behavior. Prepared by the authors based on the PeNSE survey. various aspects of the development of a child/adolescent, it is important to improve the wellbeing not only of the youths, but of an entire generation. In Spain, the psycho-educational program called Egokitzen accompanies couples going through divorce, focusing on minimizing conflicts between them, in turn easing the stress caused on their offspring. It plays an informative role about the divorce process for the parents, as well as paying attention to the impact this can cause on the children, seeking to protect them from stress and teach them how to deal with the conflict in a healthy manner (Mart<unk>nez-Pampliega et al., 2015). According to the authors, the Egokitzen program has helped reduce the perception of conflicts by children, also improving the relationship and communication among all the family members. It also increases the parents' awareness of the negative impacts that can be suffered by the children. The couples that participated in the treated group fought less and more effectively conveyed to the children that they were not at the center of the conflict. In the early 1970s, in response to the growing number of single mothers who wound up depending on government assistance in the United States due to the absence of financial support, as well as the lack of emotional support to the offspring from fathers (Allen et al., 2011), several programs were created. Among them were Child First (Kramer and Washo, 1993); Children in the Middle (Arbuthnot and Gordon, 1996); New Beginnings Parenting Program (Sigal et al., 2012); and Assisting Children through Transition (Pedro-Carroll et al., 2001). Kramer and Washo (1993), and subsequently Jewell et al. (2017), observed positive impacts of the Children First program, created in the 1980s, one of the pioneer psycho-educational programs. It was created to counsel parents and children undergoing family problems. In particular, Jewell et al. (2017) found a positive effect of the second and third versions of the Children First program, applied in some regions of Illinois, on the understanding by parents of the negative impact on their children. Among the results were better resolution of conflicts and adaptive communication between parents and children, as well as greater engagement in coparenting. One in four families with children in Ireland has a single parent. Due to this high rate, the Plus-Parenting When Separated program was created. It has been found to be effective in counseling parents and children undergoing family disruption. In comparison with the control group, formed by parents on the program's waiting list, the treated group showed significant improvements regarding the process of dealing with conflicts, both intraparental and between parents and children. In particular, there was lower stress among children whose parents had concluded the program. Therefore, there is strong evidence in the literature of the benefits of social assistance programs, with the most effective ones having a psycho-educational character, seeking to minimize the conflicts and possible sources of stress for the children of parents undergoing family disruption. The focus should be on making parents aware of the negative impacts on children and adolescents caused by disputatious divorce processes, mainly through triangulation where the offspring are in the center of the discussions, because they generally feel pressured to choose sides in the conflict, and thus suffer emotional stress. --- FINAL REMARKS This study examined, using data from Brazil's National Student Health Survey (PeNSE) for 2015, the effect of the family structure on the health risk behaviors of adolescents, more specifically in relation to the consumption of psychoactive substances (illicit drugs, alcohol and tobacco). In methodological terms, we used the propensity score matching (PSM) technique to compare students from single-parent and two-parent households, considering three samples: i) elementary school students; ii) high school students; and iii) students of both levels together. To investigate robustness, we applied sensitivity analyses to compute the Rosenbaum limits, to minimize the influence of unobservable characteristics. According to the results, adolescents from single-parent families were more likely to consume all three psychoactive substances. These effects among students at both levels were 3.82% for consumption of illicit drugs and 6.22% each for tobacco and alcohol. When the sample was divided into high school and elementary school students, the results were similar. However, this rough equality could have been caused by underreporting by younger students in the survey, since many studies have shown that the initial experimentation and frequency of consuming psychoactive substances is higher among older students. Based on these results, confirming the high prevalence of experimentation and consumption of psychoactive substances by Brazilian students, it is necessary to establish psychoeducational public policies that have broader scope than the school setting or individual level. The resulting programs need to be mulitsectorial, with the combined actions of educators, social assistants, psychologists and healthcare professionals, working with adolescents and their parents to ameliorate the negative impacts of family instability and improve the well-being of family members of the present and future generations. --- APPENDIX A: DESCRIPTIVE STATISTICS ACCORDING TO SCHOOL LEVEL Note: (1) For Elementary School students, the sensitivity estimates for health risk behavior (frequency of smoking cigarettes and frequency of consuming illicit drugs) were not considered, given the absence of effect among youths from single-parent households for these risk behaviors; (2) For Elementary School students, the sensitivity estimates for the health risk behavior (frequency of consuming illicit drugs) was not considered, due to the absence of effect among youths from single-parent households for this risk behavior. Prepared by the authors based on the PeNSE survey.

The purpose of this study was to estimate the varied effects of family structure on the health risk behavior of Brazilian adolescents and to analyze this effect according to the levels of education (elementary school and high school). The analysis used data from the National Student Health Survey (PeNSE) for 2015, which provides information about health behavior among 13-to 17-year-old Brazilian students. The sample was subdivided into Elementary School (through 9 th grade) and High School (through 12 th grade). Estimations were made using the propensity score matching (PSM) method and the risky behaviors considered for the estimation were consumption of alcohol, cigarettes and illicit drugs. The results suggested that being raised in a single-parent household negatively affects the health risk behavior of adolescents when analyzed according to substance abuse, mainly regarding consumption of alcohol and cigarettes.

refereed acording to the usual TOCS standards and accepted for this Special Issue of TOCS devoted to SAGT 2012. We would like to thank all the authors that responded to our invitation and went through a quite severe refereeing process. Our special thanks to all the reviewers for their careful work that has certainly achieved its purpose. We also want to thank the staff of TOCS and its editor-in-chief Alan L. Selman for making this special issue possible and for their support. Barcelona, July 2015

The annual meeting of the International Symposium on Algorithmic Game Theory (SAGT) is a worldwide meeting whose aim is to bring together researchers from Computer Science, Economics, Physics, Biology and Mathematics to present and discuss original research at the intersection of Algorithmics and Game Theory. It is intended as a forum covering the diverse areas comforming Algorithmic Game Theory, such as Solution Concepts in Game Theory; Efficiency of Equilibria and Price of Anarchy;

Introduction Recruiting a diverse population for qualitative research, who have both experienced the phenomenon of interest and meet specific characteristics, can be challenging, particularly groups that are socially and/or economically disadvantaged, vulnerable or 'hidden' [1,2]. We define 'disadvantaged' as those experiencing socioeconomic disadvantage (i.e. at risk of inequality related to employment, education, income, access to healthcare/ resources) and those at risk of social disadvantage (e.g. young parents, lone parents, migrants, those from ethnic minorities, people with disabilities). Those who are vulnerable due to inequality may experience an array of unique circumstances that results in their exclusion from health promotion research, enhancing the issue of'seldom heard voices' [3] and potentially further exacerbates health inequalities [4,5]. The term 'easy to ignore' rather than 'hard to reach' has been used to describe such groups, given these complexities [6], and a need for transparent accounts of researchers' experiences of engaging with such groups for research has been identified [7]. --- Open Access --- BMC Research Notes *Correspondence: [email protected] 1 Health Promotion Research Centre, National University of Ireland, Galway, Ireland Full list of author information is available at the end of the article We describe our experience of recruiting parents for qualitative health promotion research, and considerations for future recruitment for research aimed at similar populations. We recruited disadvantaged parents, with an infant aged 3-14 months, for focus groups to explore the barriers and facilitators for following infant feeding recommendations, the results of which are published elsewhere [8,9]. --- Main text --- Methods We considered the evidence that social support during parenting predicts successful child outcomes [10][11][12] alongside the social determinants of health [13], in outlining our eligibility criteria (full criteria in Additional file 1: Table S1). The research team planned three recruitment strategies (Fig. 1). The first involved recruitment through local community organisations utilising a collaboration agreement that included researchers at a range of academic institutions, with links to the local community, and detailed a commitment to supporting the project. This facilitated additional key recruitment contacts such as local support workers involved directly with groups in deprived areas. Our strategy was to reach out to key gatekeepers within such organisations and, with their permission, provide the study materials to potential participants. We also searched online for key community organisations in the country that engaged specifically with disadvantaged parents, with the aim of employing purposive and snowball sampling. When organisations were identified, we contacted them via email or phone, and provided information about the study. Many of these organisations focused on parenting skills, baby care or support and personal development for single parent families. For online recruitment of study participants, we shared study information within Facebook support groups for parents and families. The social media strategy included posting an e-flyer with details of the study and the eligibility criteria, asking parents to get in contact via phone, text message or email if they were interested in taking part, on multiple local and national parenting Facebook group pages and Twitter. They were subsequently invited to a focus group if interested. To overcome documented barriers for attendance at focus groups related to location, timing, childcare responsibilities, format and structure, and cost implications [14], we informed gatekeepers that the focus group would be held at a community venue that is local and familiar to the parents. We also made clear that they would be welcome to bring along their baby, it would be an informal environment with refreshments provided, --- Social media Searching Facebook and Twitter for local and widely followed pages/groups aimed specifically at new parents Permission through the page administrators to post an electronic flyer detailing the study and contact details for taking part --- Identifying additional organisations Online searching for and well-known programmes and websites of additional organisations Reaching out directly via phone or email with information about the study --- Collaboration agreement Committment from groups with links to community organisations to help with recruitment, prior to commencing study --- Introduction to local gatekeepers via collaborators Fig. 1 Three main recruitment strategies and where needed, participants would be reimbursed for travel costs. We described the data collection as a 'chat' to convey a sense of informality. We also gave a small gratuity voucher (€20/£20 multi-store gift cards) after focus groups to thank parents for their time and a booklet containing up-to-date recommendations on infant feeding relevant to the local jurisdiction. We gathered demographic and participant characteristic data using questionnaire items from a variety of sources including past censuses for each jurisdiction, and previous studies such as Growing up in Ireland [15]. The questionnaire used in ROI is available to view in supplementary material (Additional file 2), while the focus group topic guide is published elsewhere [16]. --- Results In the ROI, we recruited 46 parents/carers (45 female, one male) for participation in 11 focus groups, of which nine took place in urban and two in rural areas. These took place over seven months. A total of 37 mothers were recruited and took part in 8 NI focus groups, six of which took place in urban areas and two in rural. Focus groups spanned over a 4-month period. There were no fathers/ male carers recruited to the NI sample. In both NI and the ROI, the first focus group served as a pilot (NI n = 5, ROI n = 4), but, due to its wealth of data, contributed to the final sample. Forty organisations were contacted in total by email with follow-up phone calls (if a phone number was available). Of the non-responders (n = 19), all had been contacted by email/through a contact form on their website only. A further nine organisations responded to initial communication and expressed being unable to help. Of those who could not help, six provided a reason. Reasons included: perception by the gatekeeper that parents would not be interested/willing, ongoing research involvement in other projects, unsuitable time of year and perception that participants not eligible. Seven organisations proceeded with one or more focus group(s) via their organisations, while one attempted to arrange one but received very little interest. Table 1 conveys the data collection broken down by recruitment strategy, excluding one pilot focus group which consisted of a convenience sample. In NI, the team contacted 34 organisations by email with follow-up call where possible. Two organisations responded to initial emails but they could not help as their policy didn't allow them to be involved in research. Five asked to be forwarded the study material in case they had any parents interested in taking part, but they did not respond subsequently. Twelve organisations sent positive replies and a visit by a member of the research team was arranged; eight of them proceeded with a focus group. A breakdown of the sociodemographic characteristics of the participants who attended focus groups in ROI and NI can be seen in Table 2. Additional file 3: Table S2 provides an overview of the number of indicators of disadvantage met by participants. Most communication with parents was via the gatekeeper and, when it involved many steps, it was difficult to maintain contact. Moreover, when describing the process to gatekeepers, the additional stage of screening questions for participants (i.e. sending the questionnaire to participants, follow up, determining eligibility and communicating same to parents), may have been too burdensome, perhaps explaining the number who did not get back in contact. Importantly, the possibility of excluding some parents within their group or programme did not appeal to some organisations. Thus, we decided to recruit via invitation to pre-arranged focus group (as opposed to scoping out potential participants for screening and then making arrangements). This meant asking parents to self-screen and allowed for a reduced burden for gatekeepers, enabling parents decide whether to attend or not. While this process risked attendance by parents outside of the target population, it was more practical and time efficient for both the gatekeepers and the research team. In NI, the recruitment approach was adapted to avoid overlap with another study with the same target population, to avoid research fatigue, which facilitated recruitment. Some challenges included services finishing up for summer, and in one case, a query around governance. Taking note of parents' logistical preferences wherever possible allowed the research team to work around the majority and maximise attendance. We found that giving as much information as concisely as possible about the study was key when contacting community organisations (gatekeepers) for help with recruitment. Speaking by phone rather than email also proved more productive (seen in the high number of non-responders being those contacted by email only), and allowed researchers to build rapport more easily. Buy-in and interest from gatekeepers were deciding factors in whether or not information about the study would reach parents. It was sometimes clear that gatekeepers felt it important to protect their potentially vulnerable group from what they may have perceived as a risk of feeling inadequate or the perception of being accused of doing something 'wrong'. Programmes more familiar with participating in research were usually more willing to engage. Some gatekeepers sought resources from the research team in the form of information sessions or workshops around infant feeding or child nutrition in return for informing parents about the study, which helped with forging a relationship and a sense of reciprocal input. Arranging groups to coincide with events (such as workshops) already planned for parents helped attendance. Table 3 summarises the barriers to recruitment in this population that we encountered and describes solutions we found and suggests potential solutions for future research. --- Discussion Our findings demonstrate that using as wide a variety of communication tools was central to successful recruitment, and often direct communication by phone was best for reaching busy community groups. Social networking sites, whilst widely utilised among young people and a well-established recruitment avenue [25], were not a straightforward outlet for recruitment to health promotion research, particularly for capturing a specific target population such as ours. If research budgets allow, it may be worthwhile to pay for targeted advertising on social media, which has been shown as a useful research recruitment tool [26]. Some literature suggests that social media use is not representative of the general population [27] and this further emphasises the need for targeted approaches for recruitment online. Recruiting participants from deprived populations through community organisations is recommended in the literature [14], and was efficient for accessing large groups of parents. It also added a sense of legitimacy to the study when being pitched to potential participants through someone with whom they were familiar. However, recruiting through such initiatives risks over-burdening gatekeepers and potentially service-users, and in research fatigue [28]. This may help explain, at least partly, the poor response rate to initial contact. Another consideration is that such organisations tend to exist in densely populated areas, and this is reflected in our low number of focus groups in rural areas compared with urban areas. Parents experiencing marginalisation who live rurally may be further isolated by living further from amenities. --- Limitations We almost certainly encountered missed opportunities to include the voices of those who might not have the capacity or desire to attend community initiatives. Future research should explore additional strategies to --- Table 2 Participant characteristics a An indigenous ethnic minority group b Asked in ROI only; Full medical card: a means tested entitlement to reduced cost or free medical care for a wide range of services used as an indicator for low income c Asked in NI only; Healthy Start: a means tested UK government food welfare scheme for low income or at risk families used as an indicator for low income Some questionnaire items differed slightly in each jurisdiction as they were taken from either the UK or Irish censuses (Additional file 2) --- Parent characteristic N (ROI) % (ROI) N (NI) % (NI) Place include those from marginalised populations who are not in touch with local organisations, whilst being mindful of ethical considerations. Further, almost a third of participants did not meet eligibility criteria, resulting from our amendment to have potential participants self-screen and asking gatekeepers to invite participants to a prearranged focus group. This reduced our ability to manage how eligibility criteria were communicated to parents. No participants availed of travel cost reimbursement. It is possible that some may not have had the means to pay for transport up front, or that asking for reimbursement risked feelings of shame or stigma. A consideration for future studies may be to offer to arrange transport in advance. For the most part, we used proxy measures for indicators of disadvantage such as eligibility for income support, however this was in order to reduce the burden on participants. Among the research team, there was consensus that developing a questionnaire item to assess income was potentially intrusive and unnecessary, when eligibility for a medical card (in ROI)/Healthy Start (NI) was a good indicator. For some of the other indicators however, the relationship was less clear. We asked people to report their ethnicity and country of birth using items from the Irish census, however these alone do not give a clear picture of disadvantage. With the exception of one male participant, this study did not capture father-driven data. This is an important factor to consider for future research, as family dynamics are extremely complex and all perspectives are important [29]. The help-seeking behaviours of male caregivers may differ, in that the services utilised by mothers/female caregivers may not be accessible or attractive to fathers for a variety of reasons [30]. In summary, key facilitators for recruitment were building rapport with gatekeepers, simplifying the recruitment process to reduce the burden for gatekeepers, and being flexible by amending the recruitment strategy in response to barriers encountered throughout the process. Specific recruitment strategies aimed at fathers and rural parents should be considered. Avoiding too many steps in the process and pre-empting logistical barriers [20]: consider providing the inclusion criteria with details of pre-arranged focus groups, allowing participants to self-screen. Demographic questionnaires during data collection can be used to measure eligibility Difficulty recruiting via social media compared with other routes: difficult to find appropriate groups/pages to target, with correct demographic and sufficient reach Consider whether sponsored advertisements on social media may be helpful, and build this into research budget Approaching group/page administrators to post material the group to engender sense of legitimacy and relevance [21] Identifying social media 'champions' who could assist in online dissemination [22] In areas with limited organisations to contact for recruitment (such as NI), there was a risk of research fatigue among those who are regularly approached for research Consider whether collaboration with another research study to combine data collection for answering multiple research questions is feasible, while carefully assessing burden on the participant Seasonality of services and participant availability e.g. parent group closing for the summer Make a list of organisations and their schedules early on so approaching those who are seasonal can be prioritised [23] Approaches used successfully recruited female parents/carers but did not result in recruitment of fathers/male carers Additional and specific recruitment efforts for recruiting male parents should be researched and planned in advance, where fathers/male carers are explicitly invited [24] --- Availability of data and materials Not applicable. --- Abbreviations --- Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s13104-021-05653-1. Additional file 1: Table S1 --- Declarations Ethics approval and consent to participate This study was conducted in accordance with the Declaration of Helsinki. Ethical approval was obtained from National University of Ireland Galway (ref: 16/ May/10) and Queen's University Belfast (ref: 16/40) research ethics committees. Written informed consent for participation was received by all participants. --- Consent for publication Not applicable. --- Competing interests The authors have no conflicts of interest to declare. • fast, convenient online submission • thorough peer review by experienced researchers in your field --- • rapid publication on acceptance • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year --- • At BMC, research is always in progress. --- Learn more biomedcentral.com/submissions Ready to submit your research Ready to submit your research? Choose BMC and benefit from:? Choose BMC and benefit from: --- Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Objective: Marginalised populations are less likely to take part in health research, and are sometimes considered 'easy to ignore' . We aimed to describe our approach and results of recruiting parents who experience disadvantage, for focus groups exploring infant feeding on the island of Ireland. Upon receiving ethical approval, we implemented recruitment strategies that included building rapport with community organisations through existing networks, targeting specific organisations with information about our aims, and utilising social media groups for parents.We approached 74 organisations of which 17 helped with recruitment. We recruited 86 parents/carers (one male) for 19 focus groups (15 urban/4 rural). Seventy two percent met at the eligibility criteria. Most participants were recruited through organisations (91%), and the remainder on social media (9%). Recruitment barriers included multiple steps, research fatigue, or uncertainty around expectations. Factors such as building rapport, simplifying the recruitment process and being flexible with procedures were facilitators. Despite comprehensive, multi-pronged approaches, the most marginalised parents may not have been reached. Further alternative recruitment strategies are required for recruiting fathers, rural populations, or those without the capacity or opportunity to engage with local services.

Introduction --- Background of the study Strategies to ensure the safety of a pregnant mother/unborn baby is having an individualized birth preparedness plan introduced in the ANC model by WHO (1). This involves; preconception care, danger signs recognition, planning for skilled birthing, identification of a preferred health care facility for skilled birth, plans for transport, funds set aside for emergency, plan for a caretaker while the woman goes to deliver, identification of a birth companion, blood donor, decision maker in case of emergency, involvement of the community in supporting pregnancy and male partner involvement at an early stage. The Kenyan government has added the Linda Mama package where all pregnant mothers have to enroll to benefit from free maternal care services (FMC) (2). Pregnancy and childbirth complications cannot be predicted therefore; need for birth preparedness (4, 5, 6). There is no data within Kajiado County on male involvement in birth preparedness and no forums for health facility/community dialogue through maternity open days. Study questions: What is the health system factors that affect male involvement in birth preparedness, what are the nomadic cultural community factors that influence birth preparedness among male partners in Kajiado County? Broad objective: To evaluate male involvement in birth preparedness through health facility community dialogue for maternal health care services in public government health care facilities within Kajiado County-Kenya. --- Specific Objectives 1. To explore the health system factors that influence male involvement in birth preparedness through health facility community dialogue in Kajiado County Kenya 2. To explore the nomadic cultural community-related factors that influence male involvement in birth preparedness through health facility community dialogue in Kajiado County-Kenya Describe the nomadic cultural community-related factors influencing birth preparedness among male partners in Kajiado County-Kenya. --- Methodology The study employed an explorative-descriptive qualitative research design, using a narrative approach to explore and conceptualize the male partner involvement in birth preparedness through health facility community dialogue/maternity open days. There was little information on this topic, as the researcher wanted to understand the phenomena in its natural setting (nomadic culture) to answer difficult why, what and how questions. Study variables were: dependent variable (male involvement in birth preparedness), Independent variables were: ( (26). It's semi-arid, temperature ranges between 20-30 0 C with per annual rainfall of 500-1250mm. Pastoralism is the main economic activity (7). The target population was households with married male partners, aged 18 years and above with children aged 0-2 years. Community health volunteers(CHVs) were interviewed as key information informants (KIIs) to provide information on health facility community dialogue for open maternity days on birth preparedness. Inclusion-Male partners aged eighteen years and above, with children aged zero-two years, CHVs who came from the area of their jurisdiction. Exclusion-Male partners whose babies had died during delivery or were mentally challenged. For CHV if you did not come from the area where the research was being conducted. Sampling technique-Purposive sampling technique used to select households; a health facility used as a point of reference with the help of the CHV aware of house-holds with potential male partners. Sample size-Three groups were used, consisting of six members, a total of eighteen participants, supported by Krueger and Morgan on the use of 3-6 groups for focused group discussions (FGDs). I maintained the quality of data, had the right respondents, did in-depth interviews, maintained the principle of saturation, ensured active even participation, careful wording, neutral attitude, and appearance summarized the events fairly. Stimulated and guided the discussion by examining thoughts, feelings, and how ideas developed hence gained indepth discussions. An open atmosphere was enabled until saturation of information was reached exploring the group opinions and thoughts, data audio/video taped and documented. --- Method of Data Collection- --- Participant's Consent and Board Clearance Participation was on a voluntary basis and participants were free to withdraw at any stage of the study. Privacy was ensured by interviewing the male partners and anonymity was ensured by not having any form of identification on the data collection tools. Respect for the respondent, freedom, and right to self-determination upheld. Voluntary informed consent of participation before administering the questionnaire with no threat or undue influence. Confidentiality is assured by storing all the questionnaires collected in lockable cabinets accessible only to the researcher and research team. A password is used to protect electronic data in the computer. The participants informed that information obtained solemnly be used for the purpose of the research. Data was analyzed but no name was included in the study. On the informed consent, the researcher added an additional paragraph with details on issues of confidentiality, the form read as follows: To the participants, this is a focused group discussion and the researcher will take every caution to ensure confidentiality of data. The researcher reminded the participants to keep whichever information revealed by fellow participants in the focused group discussion private and should not disclose to others whatsoever. This is followed by a non-disclosure statement, individually signed by all the participants. It read as follows: As a participant in the focused group discussion, I agree that I will maintain the confidentiality of information discussed by all participants and researchers in this Study. Participant Signature................................. Date..................................... --- Results --- Data Analysis Data was analyzed simultaneously with data collection, transcribed verbatim information audio taped. Narrative analysis was deployed, transcripts were read repeatedly, and words with similar categories were grouped into themes and subthemes. --- Themes The themes for the study were; health system factors and nomadic cultural related community factors that affect male involvement in birth preparedness. --- Health System Factors --- Accompaniment of female partner to a health facility Few agreed to have accompanied their spouses for Ante Natal Care (ANC) at least once while the mother was almost due for birth, which formed the first and only ANC attendance. A few reported having taken their wives for skilled birth because it is a government directive for birth certificates for a child enrollment to pre-primary grade one (PP1). If it were not for the directive for every child having a birth certificate their wives would have given birth at home. --- Respondent one: (Nyumba Kumi Elder) "My wife had a home delivery, when I went to pick an introductory letter in order for me to get a birth certificate as a requirement when enrolling the child in PP1 our area chief was very strict and quarreled with me, he told me to always take my wife for hospital birth". The Ministry of Education uses the national education management information system (NEMIS) to track students'/pupil data to ensure a hundred percent transition from primary to secondary school through the free primary education policy introduced in 2013. Schoolchildren enrolling for PPI (pre-primary grade 1), he/she needs a birth notification. This can only be obtained after skilled birth in the hospital which leads to obtaining a birth certificate. This is a mandatory document (40). --- Reception at the health facility For the few who had taken their wives for ANC (antinatal care) or skilled delivery, reported that the reception was good, their spouses were seen first though other women were already on the que. --- Respondent one "During ANC I took my wife once in the hospital when she was almost delivering, we were served first on time though other mothers were on the que but I was left at the waiting area so I don't know the services that my wife received." --- Male Partner Birth Preparedness Few who had ever taken their female partners for maternal health services said their partners were taken in for examination, but for the male partner, they were only called in for Human Immune Virus (HIV) testing. --- Respondent one "When I took my wife for the ANC, I was left outside at the waiting area, my wife went in to be examined. I was called in for HIV testing, though I had fear of getting positive results". --- Respondent two "My wife used to bleed while pregnant; she was given herbs we did not go to the hospital". --- Infrastructure/supplies in the hospital Few male partners, who had taken their wives for skilled birth, reported that the room is small with some women laboring/birthing naked, so they are not allowed in the laboring room. Respondent two "I took a basin, soap, sanitary pads, and tea for my wife in the hospital after birth, the health facility did not have water so I had to take her home for a bath." Respondent three "I took my wife to a hospital for the birth, nurse quarreled with her about why she had not enrolled for Linda mama/ not attended four ANC visits, while we at the waiting area were getting the conversation. I had another mother being asked about her HIV status, and whether on drugs". --- Complications during pregnancy or after The majority responded that their partners had complications while pregnant/after delivery but they did not know they were danger signs. --- Respondent two "My wife convulsed while pregnant, had a continuous headache, she took herbs and she felt well, but convulsions reoccurred again". Strategies to improve male involvement in ANC/ Labour/ PNC "Maternity units /boma initiatives/mobile clinics should be near where our cattle are or within migratory corridors, support boy child education, offer males services together with ANC". --- Overall satisfaction with maternal health care services Respondent one "The government health centers are far, with one nurse, always closed on weekends/night, so we rely on private hospitals which are expensive." --- Nomadic Cultural Community factors --- Socio-Cultural Value / Belief Maasai communities are pastoralists, their culture does not allow a discussion of unborn children, birth is viewed to be natural. --- Respondent one "We do not know what birth preparedness is because our culture does not allow discussion of the unborn child, we perceive maternal health to be women's affairs and birth to be natural if you are seen on the road with your pregnant partner, you are viewed to be so weak and controlled by your wife". Respondent two "Maasai men do not discuss women issues, it's our responsibility as morans to take care of community security and protect our families and cattle". --- Health facility community dialogue The Department of Health through community health strategy with the support of good neighbors has mapped and assigned CHV 53 households each, to link the community to the health facility. There are no maternity open days in the county where the health facility invites the community to the maternity units, shows them services offered, demystifies myths/misconceptions, teaches the pregnant mother and communitabouton their reproductive health rights, solves any disputes, and lets the community be aware of which services are offered and how the community can support to improve the services hence increase utilization. This is attributed to, a lack of budgetary allocation for maternity open days, on a quarterly basis annually. --- Respondent CHV one "We educate the community on the importance of ANC/ hospital delivery, though it's hard to get the male partners and educate them, nomadic culture perceives maternal health to be women issue it's hard to involve them in maternal health". --- Respondent CHV two "There are no forums for maternity open days for allowing the community to visit maternity units be educated on their reproductive health rights or taken round the maternity unit." --- Discussion --- Health system factors Accompaniment of male partner to health facility for ANC, Skilled delivery, PNC Few reported to have accompanied their partners for ANC or skilled delivery, but none had taken the mother and baby for PNC services. Fourth ANC attendance in Kenya is at 52% nationally, this tends to agree with the sentiments from the participants that their wives managed to attend at-least one ANC visit when almost due for birth (KDHS, 2014). A few reported having taken their wives for skilled delivery due to the Ministry of education-Kenya policy of child enrollment to PP1 to have a birth certificate for allocation of NEMIS number (40). This shows that they attached skilled birth to obtaining a birth certificate. The respondents reported that it is only during PMTCT time that they were called in for HIV counseling and testing. This tends to correspond to previous research done in Cameroon and Rwanda, which showed that male partners were left outside at the reception area, so they feel not part of birth preparedness (9,10,11). They also reported fearing taking their wives for ANC because it's mandatory they will be tested for HIV yet they fear the outcome of the results which tends to concur with studies done in Ethiopia, Uganda and Nepal (12,13,14,15,16,17,18). None of them took their spouses/newborns for PNC. --- Respectful maternal care The few study respondents who have ever accompanied their spouses for maternal health care services reported that the reception was good, they were served first, and their partners were given services in a respectful manner (19,20). --- Male partner involvement in birth preparedness On birth preparedness, the respondents reported that they were left seated outside at the waiting area. This study's results tend to concur with previous studies which agree that sometimes healthcare providers see it as a burden involving male partners in birth preparedness due to the high volume of clients seeking maternal health care (21,22,23,24,25,26,27,28,29,30). --- Health system Infrastructure In terms of space, confidentiality, and privacy, male partners felt the health systems need to increase spaces in ANC/labor ward in order to accommodate them. They were never allowed in labour rooms because mothers will be laboring/birthing naked. This concurs with studies done previously which found out that male partners are never allowed in labour rooms due to lack of space, privacy and confidentiality (31,32). Adequacy of beds/beddings is a challenge in sub county/county referral hospitals due to high volume of pregnant mothers seeking skilled services. This has led to congestion, leading to early discharge of postnatal mothers before forty-eight hours are over. The WHO and Kenyan ministry of health policy on post birth discharge recommends after 48 hours because most maternal deaths occur after 24 hours due to secondary PPH (33,34,35). Mothers are supposed to be offered a hot beverage after birth, due to bleeding and depletion of blood volumes most mothers tend to shiver. Warm fluids stimulate milk production given that most mothers may not be well hydrated before/during labour, the study respondents reported having been told to bring tea or food for their partners. Birthing rooms are cold unlike home delivery where the room is lit by firewood for warmth. Health facilities do not have warm water/ showers with some health facilities water bought through vendors, therefore, mothers are required to shower with cold water immediately after birth or go home for shower. These findings tend to disagree with ministry of health directives and policy on maternal health care that recommends delivery rooms to have warm showers, wall heaters, mothers to be provided with hot beverages, and rooms to be pre-warmed before birthing (36,37,38,39). --- Complications during and after pregnancy Most of the respondents reported that their partners had complications before and after delivery of which they did not know were danger signs, this tends to concur with other studies done in Ethiopia, Malawi, Uganda and also Kenya confidential inquiry into maternal deaths (CEMD) which shows that families did not understand danger signs in pregnancy (40,41,42,43). --- Nomadic cultural community factors --- Cultural value-belief Majority of the respondents agreed that the nomadic culture affect male partner involvement in birth preparedness, they believe/perceive there should be no preparation for unborn child. Which tends to concur with a study done in Kajiado County on utilization of free maternal health care services (23). Nomadic culture views maternal health as a women affair (23,27,32,34,38). Male involvement is critical in case a pregnant mother needs proper feeds; investigations needed like point of care ultra sound. When a woman goes to seek health care services, she needs to seek permission from the male partner for support/bus fare, thisresults concurs with studies done in Uganda, Ethiopia and Ghana (1, 2, 3, 6, 22, 32, 35, 45, 46) --- Conclusions They were few male partners who had ever taken their female partners for ANC or skilled delivery but none had taken their spouses nor their babies for postnatal care. For the few who had taken their partners for the ANC/skilled birth, they had inadequate knowledge on what services their spouses received nor any knowledge on birth preparedness. --- WON IDEA AS A RESEACHER The study was able to inform policy on the need to strengthen male involvement in maternal newborn health in order to achieve the ministry of health objective of ICPD25 commitments of: Zero preventable maternal/perinatal mortality, zero unmet need for family planning and zero harmful cultural practices. The study proved some policy weaknesses and gaps in terms of implementation. Some policies like open maternity days that forms a forum for communities to be empowered on their reproductive health rights and engage male partners involvement in maternal new born health issues at Kajiado county in order to improve on maternal indicators like antenatal care, postnatal care, skilled birth, maternal/perinatal mortality and reproductive health rights not being implemented. Another policy weakness/gap witnessed, was implementation of guidelines like availing of water in all health facilities, availing of hot beverages, maternity pads to mothers in labour and after birth, ensuring labour rooms are pre-heated before birth and during birth. I presented the results at division of reproductive and maternal health Kenya ministry of health headquarters that deals with policy development and implementation, guidelines and standard operating procedures development and technical assistance at county level on maternal newborn health. As a researcher, I am privileged to work at policy level at ministry of health division of reproductive and maternal newborn health, therefore; I used the results to strengthen the already existing policies on maternal/newborn health and come up with maternal newborn health standards. The ministry is working on capacity building of health care providers on respectful maternal/newborn health. As a researcher, I understood the need to respect and understand how culture influences health-seeking behaviors that later affect maternal/newborn health indicator nationally. --- Further Research The researcher recommends further studies on strategies to improve male involvement in birth preparedness within Kajiado County. The researcher recommends further studies on the impact of maternity open days on maternal newborn health indicators within Kajiado County Kenya. The researcher recommends further studies to explore why nomadic culture view maternal newborn health as a woman centered affair, males should not be involved and women should give birth the way their animals do without support. The researcher recommends further studies to understand why ministry of health policy guidelines on quality of care for mothers who have given birth to be offered warm birth by having heaters in labour room, warm beverage to stimulate milk production, being offered warm baths, health facilities having adequate running water in place in government health facilities. This fosters respective maternity care, attends to reproductive health rights of women and newborns as enshrined in the Kenyan constitution (2010) Act (43) and encourages skilled birth. The researcher recommends the county government of Kajiado to budget for maternity open days forums every quarter of the year to capture the male partner and communities on maternal newborn issues. Kajiado County offers an enabling environment for through political goodwill for male involvement in maternal newborn health. --- How The Person Focused On, Became Interested The 18 participants who comprised of community leaders amongst others gave their views as it is on the ground. They had no interest in whichever way the data they gave would influence their lives and maternal newborn health. The only thing that they wanted to change is for the governor of Kajiado County to bring hospitals near their reach. In addition, Kajiado County ensures that they open health centers on weekends and at night. --- Other people who helped me in my achievements were My supervisors from Kenyatta University: Dr. Priscilla Kabue and Dr.Justus Ngatia. Research assistants Tracy Morinua, Jack Lugalia and Wycliff Bitengo. Others were the county education officer and health director Kajiado County-Kenya because there was some information that I had to verify from the office. --- History of the topic before I became involved A study had been done by Karanja S et al, 2018 (21) on determinants of nomadic community utilization of free maternity services within Kajiado County. The researchers found out that irrespective of free maternity care as a government policy inaugurated on June 1 st 2013, the utilization was minimal due to male partners viewing maternal newborn health as a woman's issue and also the nomadic culture never allowing discussion or any preparation for unborn baby, therefore, there was no birth preparedness amongst male partners. The researchers recommended exploring birth preparedness amongst male partners. As a researcher, I developed an interest to understand how the health facility involves the nomadic communities in birth preparedness because there is a government policy on maternity open days that directs health facilities and communities to meet every quarter and discuss maternal newborn health for community ownership of health thus advocating for primary health care. Through observation, one could rarely see a male partner accompanying a female partner for maternal newborn health services in a health facility. --- On data analysis Data that was coded, conceptualized, put in patterns and coded into themes and subthemes was what was being collected from the study participants through focused group discussions. --- How author's findings affected the field being discussed? It has fostered policy change at national level, because male are decision makers in African society, are educated compared to women and are economically empowered. Therefore, every program in order to achieve its objective in maternal newborn health has to involve the males in order for the country to achieve its indicators like zero preventable maternal/perinatal mortality, zero unmet need for family planning, zero harmful cultural practices like

Kenya's maternal mortality is 362/100000 live births which are below the WHO recommendation of maternal mortality of 147/100000 live births and the sustainable development goal (SDG) of less than 70/100000 live births. The national infant mortality rate is 39/1000 live births, the under-five mortality rate of 52/1000 live births, and the newborn mortality rate of 22/1000 live births (3). Kajiado County has maternal mortality of 462/100000 live births and infant mortality of 40.2% above the national levels (3).

Introduction Many scholars envisage a progressive, inevitable convergence of the offline and the online realities or, as others put it, of the physical and the virtual realms. The debate on this issue often fails to OPEN ACCESS recognize that, from a philosophical point of view, the difference between the real and the virtual dimension of reality precedes-and does not entirely depend on-the distinction between offline and online reality. This means that the differentiation between reality and virtuality is already insular to the offline reality. However, this distinction also concerns the online reality. From this perspective, the convergence between offline and online reality is more a revised interpretation of the role of computers (which are no longer viewed as interfaces but as tools cooperating to reontologize our reality in informational terms [1]) rather than the description of the actual possibility to overcome the differentiation between reality and virtuality that belongs both to the offline online realities. This differentiation is important for the construction of personal identity. We can speak of construction because our personal identity necessarily exists across time and is constituted, at the same time, of what we are (i.e., the "actual self" made of information describing our features) and of what we would like to be or what we ought to be (i.e., the "ideal self" made of information describing our expectations or the social expectations about us). In other words, personal identity is always constituted of both reality and imagination (virtuality). It is central to understand in this respect that the "ideal self" is not a regulative idea, in a Kantian sense, to which the actual self tends to conform itself. On the contrary, it is already a constitutive part of the self. Our personal identity is an "open text", the lines of which are made out of the intertwinement of reality and ideality. Those who affirm and those who refuse to acknowledge that social networks (or the Internet) are online spaces where people are given the possibility to endorse a second life or a new personality are both wrong, because they do not realize that social networks (or the Internet) constitute only a different way to intertwine reality and imagination in the construction of personal identity. Placed against the backdrop of the intertwinement of reality and ideality, the Self experiences itself as a dynamic reality that has to be constructed within the different and particular networked contexts of communication that form such backdrop. We should state from the very beginning how social networks enable people to construct (at least in part) their personal identity in a way that is different from what happens in the physical reality. In the physical reality, we construct our identity (narratively understood as an open text) within different environmental and social constraints, i.e., within different contexts, that are mainly already structured and given and cannot really be fashioned by our narratives. In contrast, in the digital reality, social networks for instance are platforms (as in the case of Facebook) that enable users to take advantages of networked affordances in order to construct, in informational terms, not only their identities but also to participate in the co-constructions of their networked contexts of communication. We should not forget, in this respect, that the construction of identity does not stem from the scratch but occurs in a situated context: identity is always part of a context. This is essential in order to understand the text that is formed within it. However (and this point is crucial), the context cannot be reduced to an empirical situation, to a particular networked architecture or technological platform. In the digital space, the context is, according to Dey [2], a sum of information that characterizes a specific situation (on the basis of the technological affordances of a particular architecture or platform). Hence, the following statement stems from our main point: both personal identity and the context of communication can be conceived and understood, at different levels of abstraction [3,4], in informational terms. However, our informational presence in the net is in no sense only a "virtual" presence. On the contrary, it does produce consequences in real life. As autonomous agents in fact, we are called upon to account for the consequences of our actions. The intersubjective regime of accountability (which entails moral responsibility or legal imputation) is a measure of the way in which our presence in the net is "kept in touch" with reality, to the extent to which we are called upon to justify, in real life, what we have done in the net. This makes it necessary to translate the virtual dimension of our online existence into the empirical terms of our offline existence [5]. In this sense, accountability is not only a key part of our personal autonomy but, when referred to the online world; it can also serve as a principle of reality. This regime of accountability can be set either by means of moral or legal norms, which are established across time by tradition or which are laid down by social or political institutions. Accountability can also be developed spontaneously through trustful relations. Trust can be actually viewed as a way to make someone accountable to someone else for what the trustee has been entrusted for by the trustor. In this respect, online trust cooperates to set a systemic context of communication between trustful agents and, therefore, it is meant to mold our networked reality, that is, the reality within which we can construct and project our own online personal identities. For this reason, people need to trust other people in the networked world, not only to delegate them tasks and to achieve specific goals but, first and foremost, in order to share with them the construction of a meaningful context where they are be able to disclose and recognize themselves. Disclosing information in the net is thus likely to have consequences in real life. In particular, disclosing personal data can make people infringe privacy rights, and this violation can in turn impair the construction of personal identity. Hence, there can be, in the online construction of personal identity, a tension between trust and privacy, which seems to belong to the notions of trust and privacy themselves (such as online reputation [6]). At the same time, there is a connection between trust and privacy in their competing tendency to structure the conditions and limits of the online identity construction. The issue is thus to understand the theoretical and practical tension between trust and privacy in the light of that online identity construction. This construction is not based on self-transparency (as for the Cartesian subject) nor does it make us transparent to others, since it requires us to shape the conditions and limits in which we are "revealed" to others. In this respect, a philosophical notation is to be made: as suggested by the Latin origins of the word, to reveal ("re-velare"), means both to unfold (i.e., to remove the veils) and to hide something (i.e., to multiply the veils). This signifies that the identity construction is not simply determined by the disclosure of information (that bears on trust) but is revealed by a multifaceted selection of information (a combination of disclosure and closure) that bears on trust and privacy. In other terms, the construction of personal identity is always the result of a competition (both internal with ourselves and external with others) between what is disclosed and what is hidden about us: we do not have to forget that a society cannot exist without some spheres of secrets. From this standpoint, personal identity is to be understood as the unending result of a selection of information that forges a meaningful difference (i.e., the Self) between what we wish to unfold and what we wish to keep secret. In this respect, we should reaffirm that the possibility to keep a secret from a group of people is not only a part of that construction but, primarily, a basic condition of liberalism [7]. At the same time, we have to recognize that, in the networked Information Society, it is unrealistic to prevent people from being a deliberate or unintentional source of information, whereas people should not be deprived of the power to rectify the false/incorrect information concerning them or to have this information cancelled. --- Trust I will start my analysis by focusing my attention on trust, in order to highlight some of its aspects that concern in particular the structure of intersubjectivity and the disclosure of information that a relation of trust may entail. Trust is one of the basic social concepts "that helps human agents to cope with their social environment and is present in all human interactions" [8]. Without trust in agents, infrastructures and organizations, human interactions are undermined and, ultimately, almost no social cooperation or relation would be possible. Trust is a context-dependent concept in that trust is essential when dealing with situations that display a certain lack of certainties such as: social environment (we rely upon incomplete information and subjective beliefs); someone else's behavior (we never fully rely upon predictable actions in order to achieve a goal); organizations or infrastructures (we rely upon organizations or infrastructures where their functioning or settings are not completely visible, known or manageable). Social networks, for instance, require all three of the above forms of reliance upon lack of certainties. If a context is made of information that characterizes a specific situation, the context of trust is made of incomplete information, namely, of lack of certainties, which necessarily exposes the trustor to some risk. In this sense, trust is meant to exist insofar as there is a risk [9] or something that escapes from our cognitive faculty to have control over. On the one hand, the level of trust we are willing to display depends on the perceived embeddedness in the specific situation where the relation of trust takes place. On the other hand, it depends on the perceived lack of certainties that is peculiar to such a situation, which is always relative: the context of trust is not a context of total ignorance. In fact, to trust implies to have some information (except for the case of blind trust) and to share this information in the circle of those who partake in the relation of trust. Placed against the backdrop of complex social systems, trust is a way to reduce uncertainty [9] or, in other terms, to transform an individual risk into a collective one [10]. Furthermore, the relation between trust and the context where the relation of trust takes place is more complex than it may appear prima facie. Not only does trust depend on the context in which it is embedded but it tends reflexively to reshape that context by creating a trust atmosphere or a web of trust relations filled with new information. For that reason, and this is a crucial point, the context of trust is not only one formed by incomplete information but, primarily, it is a context of communication where an environment is created in which meaningful relations can take place or, in hermeneutical terms, in which it is possible to share an "horizon of meaning" [11]. This is a cognitive dimension in which our lives can be mutually interpreted and understood. This idea is consistent with the socio-cognitive interpretation of trust displayed by Castelfranchi and Falcone [12][13][14], which we refer to hereafter. Trust consists of three elements [14]: (1) a mental attitude, a predisposition towards another agent: this attitude is a belief, the strongest cognitive element, constituted by the evaluation of trustworthiness and a prediction regarding the agent's willingness and ability to produce some effect; (2) a decision to rely upon another person: the intention to delegate the production of a desired outcome, which exposes the trustor to a risk and makes him/her vulnerable; (3) a behavior: the effective act of entrusting another agent, which entails a practical relation between the parties. The relation of trust is thus conditioned by the representation of a double context: the context of the trustor and that of the trustee. In order to trust [14], the trustor should have positive expectations both about (a) the trustworthiness of other agents (internal attribution) and about (b) the external conditions that might influence the trustees' actions (external attribution). The internal attribution of trust depends on the evaluation of the trustees' qualities and defects that define trustworthiness and ground the basic beliefs of trust. They are crucial factors since they constitute specific reasons to be trusted like concern for common goals, friendship, altruism, motivation, reciprocity, and cooperation. External attribution of trust depends on the trustor's perception and evaluation of the appropriate environmental conditions that might influence the trustees' actions. External trust is related to two types of systemic conditions: positive conditions relating to the presence of opportunities and resources; negative conditions relating to the absence of interference and adversity. The trust attitude (TA) consists of a set of mental beliefs through which the trustee is evaluated. The cognitive evaluation has degrees: as already mentioned, the level of trust is relative to various factors concerning both the trustee's reliability and the context in which the relation of trust is embedded. These factors can be rationally quantified: they represent what is known by experience or perceived in the situation of risk and uncertainty the trustor is exposed to. The trust attitude consists of a scalable evaluation: degrees of trust are related to rational beliefs and this allows the trustor to evaluate the risk and the advisability of trusting. A calculation is made, so that we can say that trust may be attributed in a stronger or weaker manner. However, this should not allow us to lose sight of the fact that the decision to trust is a clear-cut decision: when the trustor is faced with the possibility of trusting, he decides either to trust or not to trust. The decision to trust stems from a calculation but it is not a calculation in itself. In the trust decision (TD), the otherness of the agent (i.e., whether or not the agent comes to betrayal) is not reducible to a mere calculation: the trustor has to enter into relation with the otherness of the agent. When entering into relation, the trustor sets a communication with the trustee. This communication is contingent upon the context in which the trust relation takes place. More analytically, communication is conditioned by the perception of a double context: the context of the speaker (i.e., the trustor) and the context of the receiver (i.e., the trustee). The speaker can perceive those contexts as being either distinct or integrated. When the speaker realizes that the context is integrated (namely, it is the same for both of them), the perceived sameness attunes the dimension of otherness of the receiver, who is understood as part of the same horizon of meaning of the speaker. This fusion of horizons [11] is a means of smoothing the progress of their communication, since it enables the parties to establish a trustful relation within a single context of communication. The perception of a bond, through which agents form a single unit, influences both trust and privacy. The more the context of speakers and receivers is perceived as integrated, the more the ensuing context of communication can be taken as private or at least as semi-public. As a result, the more the context of communication is viewed as private or semi-public, the lower the concern for privacy and, accordingly, the higher the trust in the communication. Here we can state a crucial point of our analysis: the same social space or technological platform (social networks) can host different contexts of communication. The intensity of the trust relation and its possible trade-off with privacy depends, first and foremost, on the specific contexts of communication where the trust relation takes place, and not only on the design or settings of the technological platform in which it is embedded. The technological choice for assuring privacy by design or by means of autonomous selection of privacy settings is, undeniably, crucial but it cannot entirely account for all the possible contexts of communication growing out of the technological platform where a trust relation is embedded. Furthermore, the decision to trust sets a social (informational) norm between the trustor and the trustee: the trustee's actions (e.g., disclosures of information) will be judged according to the context of communication in which the decision to trust take place. In other terms, it is within a particular context of communication that the trustee's actions are expected and, therefore, evaluated. The peculiarity of trust resides precisely in the fact that the social norm regulating the trust relation does not precede but grows out of the communication process between the trustor and the trustee. The normative dimension of trust (i.e., the parties' expectations) is not previously fixed but grows out of the concrete process and context of communication: "The impossibility of enchaining trust within a legal disposition does not entirely exclude trust from the domain of norms" [9]. So trust displays a normative dimension, which is not forged by legal norms or by technological devices but has a cognitive status made out of shared intersubjective expectations. The process and the context of communication, where those expectations are shared, are not given and determined in advance. On the contrary, they are fashioned by a dynamic dimension, since they influence, reflexively, each other. During this form of communication (i.e., a trust relation), both trustors and trustees are progressively brought to reveal themselves. They fill their identities with meaning through their trustful relations, and this meaning (as every meaning) is contingent upon the context of communication in which it occurs. Placed against the backdrop of communication, trust is not only the way we can provide the parties of the trust relation with meaning (reduction of uncertainty), but it is also the way we can co-construct the context of communication in which that meaning will be appraised and understood (systemic reduction of uncertainty). In philosophical terms, trusting is not only aimed at entering into relation with another person but also at seeing who the other person is, which is always a necessary step, in order to recognize who we are. This means that the trustful relations of communication, where we decide to enter and where thereafter we are embedded in, participate in the construction of our personal identity. Trust is peculiar in what it enables us to give shape both to social relations and to the context of communication in which those relations are appraised and understood. By means of trust we can construct our personal identity in the sense that we delimit the sphere of what we can expect from others and of what others can accordingly expect from us, by structuring the conditions in which those expectations can be shared and evaluated. Since most human problems are problems of communication and control, as Norbert Wiener [15] has taught us, trust is also affected by problems of communication and control. As remarked, trusting is a complex act that joins together what is subject to evaluation and calculation (trustworthiness) and what remains beyond control (the otherness of someone else's behavior): trusting never involves a choice between control and lack of control, but always both of them at the same time. Trust involves a risk that can endanger our identity construction for three main raisons: it requires us to disclose and to expose ourselves to others (people cannot be prevented from being sources of information); it seems to establish an opt in/out regime for the use of personal data (dissent should be explicit, while consent is presumed); the information disclosed in a specific, trustful context of communication can be easily disembedded and re-contextualized. This is the reason why trust is meant to be at variance with privacy. It is time, therefore, to focus our attention upon the issue of privacy. --- Privacy In the past, the idea of privacy has been closely based on the notion of space. In the legal doctrine of privacy, this idea of space has always been thought of through the opposition between a private and a public sphere. From a philosophical standpoint, such opposition, still at play in some interpretations of privacy, could be misleading if not properly understood. The space of privacy is not shaped by a given content, that is, by a determined set of data that would belong to a private sphere. A set of data is not to be defined private because they would belong, ontologically, to a private space. The space of privacy is fashioned by the human power to refer something (i.e., a set of data) to a unique point of reference. We do not understand much about privacy if we do not recognize in its construction the human attempt to pave a way towards uniqueness, i.e., the regime of insubstitutability, by using materials that belong to the domain of homogeneity, i.e., the regime of substitutability [16]. In other words, what can distinguish the public from the private sphere of life is not a different content (a peculiar type of data), but a different reference of the same content to a specific point of reference: the Self (both a source and a sum of information). The so-called "private" dimension (sphere, space or data) is referred to a point to which there is not the same reference in the public. It is this power to refer to the Self as a unique point of reference which explains privacy, rather than the simple delimitation of a more restricted domain within a wider space. This means that the protection of privacy is not viewed, in this perspective, as the protection of a delimited space (however such space is conceived: the house, the person, the body, the personal data, etc.) to be interdicted. On the contrary, the protection of privacy requires us to preserve a human capacity to act, that is, our capability to refer something to ourselves: i.e., the power of self-identification, which is part of the human relations in the sense that it is both an individual and a relational activity, namely, an activity that we never achieve, entirely, by ourselves. Philosophically, this implies that there are no data, per se, which are ontologically private or public. There are no private data that cannot be turned into public and vice versa. The private and the public worlds, which are ontologically coextensive, do not exist by themselves but, on the contrary, they are the corollary of a human activity. For that reason, people are constantly troubled with tracing the mobile, unsettled line between the private and the public. What is crucial to understand is that this activity does not consist, nowadays, in isolating the individuals from the society but, on the contrary, in placing them inside it. The space of informational privacy is the space in which we project and construct our own personality: this activity, however, never occurs in a void but necessarily inside society. This is the reason why, if our personality is made out of the sum of information concerning us, we have also to realize that this is not a static but a dynamic account of who we are. In fact, the space of privacy is not constituted only by that amount of information. Rather, the space of privacy is given by the contexts of communication connecting us to the rest of society in terms of negotiated identities. In such contexts, to trace the line between the private and the public sphere of life is a matter of power. People seem to take the definition of power for granted. They tend to associate power with some capability of acting, deciding or having influence over someone else's will, and they reduce it accordingly to large availability of money or to great means. Many associate power more directly with violence, force or, at best, with authority. Others associate power, in a more sophisticated way, with the ability to persuade or to produce the relevant information upon which people take decisions or behave. Few have a gnoseological conception of power. In this philosophical perspective that goes from Heidegger [17] to Foucault [18,19], power is also a means of self-knowledge: human relations are relations of power aimed to devise who we are. From this standpoint, we do not only make use of knowledge as a form of power. We also make use of power as a form of knowledge: what I am depends on what I am capable of doing both as regards to myself and to others. In this sense, it is not my identity that brings about my capacities. Rather, my capacities determine my identity. We understand thus why privacy, understood as a way of constructing our personality, is viewed in our perspective as a power to refer something to a unique point of reference (a form of empowerment of the Self). We realize also why this power is always competitive, polemic, since the relation with others can represent an intrinsic limit to my power of self-construction. The process of referring data to one's account is not in fact an activity the subject can achieve by themselves. Our identity has always been made out of data that come from multiple and distributed sources, and often from sources that are potentially conflicting with each other. At all times, there is also a plurality of interpretation of those personal data (which co-construct the identity of individuals), and there is often a competition among these different sources of interpretation of personal data, according to the diverse contexts of communication in which those data are selected. However, this competition is not what threatens privacy but what makes privacy possible as a meaningful construction of human personality, since the meaning of data necessarily grows out of a "group phenomenon", that is, a collective process of communication embedded in some specific contexts. However paradoxical it may be, through, the commonly shared reference to a unique point of reference, we can delineate what is personal from what is impersonal. We never accomplish it by ourselves, since any reference must be shared, in order to be meaningful: there are no private linguistic games. Hence, the space of privacy is also a context of communication. This notation is crucial to us since it connects privacy and trust, when understood as constructions of personal identity. The data we refer to ourselves have to be conceived as meaningful pieces of a personal history that must be, nevertheless, interpreted and understood in a shared context of communication. The power to refer data to the self entails, thus, a society and a bundle of human relations. This means that the individual power to trace the difference between private and public communications is already the sign of the existence of the (real or virtual) community defining the shared context of meaning where such communications can be interpreted and understood. To protect privacy, i.e., to construct a personal identity, is to protect this power of self-identification, which, however, always exists within the relation of mutual implication between me and others in a determined context of communication. The power of self-identification is confronted with the power of being identified by someone else. Self-identification cannot be defended as a prerogative of an absolute subject (like for the Cartesian subject) that would be able to construct its identity in the blank, but only in the endless communication with others. In this perspective, the right to privacy is endowed with a double content: a positive content, i.e., the power to refer data to a unique point of reference, and a negative content, i.e., the power to rectify false/incorrect information, to have this information erased, and to prevent people from abusing of someone else's personal data. As we will see it in the last part of the paper, adolescents tend to understand privacy with regard to its positive content (as a construction of personality), while they happen to underestimate or to be not fully aware of its negative content (as a protection of personal data). Furthermore, the dialectics between the positive and the negative content of privacy strengthens at least three problems that are able to threaten the individual right to construct a personal identity in the Information Society, where the informational representation of human identity consists exactly of a set of "data that stem from multiple, distributed sources, in addition to traditional centralized storage devices" [20]. We have to focus our attention on these three problems. --- Exportation vs. Importation of Personal Data We have to remark that the question of privacy can no longer be confined to the risk of the exportation of data from the personal towards the public domain. On the contrary, the problem of privacy concerns more and more the risk of importation of data within the domain of the self by the multiple, distributed sources that can "speak" on behalf of us. We refer to all cases in which the power to mine, collect and distribute data can give an account of ourselves. In the Society of Surveillance [21], this goes up to the point that individuals can no longer keep a secret from others, since, because of the importation of data, they can learn about themselves what they could not have known before. This form of reference can be either concrete or abstract. It is concrete when people substitute us in the identity construction by referring to us data that detail our personality. It is abstract when people aggregate data by creating abstract profiles (for instance, statistical but not only), within which we are entirely or partly subsumed [22]. This can be achieved also without human intervention, by means of software agents or autonomic systems [16]. In this case, the process of self-identification is accomplished by means of an automated importation of data, which no longer requires a relation between me and others. In this case, the automatic importation of data displaces any shared construction of a context of communication. Here, the construction of identity cannot reconcile trust and privacy, because that impersonal construction no longer depends on the joint elaboration of a trustful context of communication that people could share as the horizon of meaning where their narrative identities can be interpreted and understood. --- Explicit vs. Implicit Communication of Personal Data Cognitive sciences and political philosophy remind us, from different standpoints, that, in our Society of Information, attention is becoming more and more a crucial issue [23]. The overload of information makes it decisive to select the relevant information to which we have to pay attention [24]. Attention to the information is a key question as regards to privacy, since the level of protection we expect for our privacy in a context of communication is correlated to the level of attention we pay to the information we disclose in it. The higher the attention we pay to the information disclosed, the higher the level of protection we expect for that information. Cognitive attention is a scalable phenomenon, graduated according to the interest we have for the information disclosed or, more properly, to the emotions we attach to that information. In this perspective, the information concerning us deserves, normally, more attention than the information concerning others. People are thus likely to be more concerned with privacy when they disclose information about themselves rather than when they disclose information about others. Of course, this is quite counterintuitive, since the probability of violating privacy is higher when we disclose information about others. This is an elementary cognitive reason why we should pay greater attention to the information disclosed about others. What is more, things become more intricate, from a cognitive standpoint, if we replace the distinction between information about us and about others with the distinction between explicit and implicit communication of information. In fact, it is not simple to perceive what is implicit ("not-said") in what we communicate: frequently people speak about others through what they say about themselves. People do not pay the same attention to what is explicit and to what is implicit in their communication, since the "not-said" is the result of an hermeneutical comprehension of what has been said, the meaning of which depends on the context of interpretation of those who make the implicit communication explicit. As we will see later on, this is one of the problems of the youth's communication by social networking, since young people trust in the information disclosed, the interpretation of which can, however, bring to light what remains just implicit in their communications (thus resulting in a privacy violation). --- Space vs. Time Conditions in the Communication of Personal Data It has often been said that, in western societies, we live more and more in an eternal present. The speed of life, the pace of technological development [25] and the pretended end of history [26] force us to focus our attention only upon the present time: all these factors seem to out-shadow the depth of past and the indeterminate future. This cultural attitude has important consequences over the issue of privacy. In fact, people (and young people in particular) are brought to undervalue the consequences of their decisions in the long term. In particular, they can underrate the effects of their implicit consent to the disclosure and use of personal information, since they do not always perceive that such information will be evaluated, across time, within different contexts of interpretation. In the Society of Information, the contextual integrity and the coherence in the construction of identity is more subjected to menaces that come from the future rather than to menaces that come from the outside. The right to be forgotten is, intrinsically and hermeneutically, put into risk by the idea that past and future are just derived forms of temporality, when confronted to the reality of the present time. The right to be forgotten is logically endangered, if the idea of the eternal present becomes the norm that structures the interpretation of our personal communications. This hermeneutical canon is indeed at variance with the legal construction of privacy as a right of personality, which is not subjected to the passage of time. The level of privacy protection will be framed by the competition between the consent, implicitly given in the present and "for" the present, and the endless right of personality. Needless to say that also the interpretation of the different contexts of communication, where a relation of trust can take place across time, is displaced or endangered by the hermeneutical doctrine of an eternal present. These problems about privacy tend to show that the identity construction in the Society of Information always grows out of a competition between different sources of data. In the following paragraph, we have to explain why this competition is, actually, intrinsic to any construction of personal identity and why it can represent, despite the problems outlined, the locus where trust and privacy can be, at least to some extent, reconciled. --- The Construction of Personal Identity In the age of information, the construction of personal identity is achieved by means of narration. In other terms, personal identity is made up of information (out of the distinction between offline and online) [1], which is provided with meaning through a coherent narrative construction of personal identity in a shared context of communication. The construction of personal identity is a polemic activity [16]. This competitive dimension becomes evident when the notion of identity is related to that of narration. Narration is, in its formal structure, the expression of a conflict that marks the passage of time. In this perspective, Arnaldo Momigliano, the great historian, affirmed that war has always been the main topic of the historical tale [27]. Needless to say, in addition, that even the philosophical identity of becoming is polemic in its nature, because expressed by the continuous passage between being and non-being, which is by itself warlike. The competitive construction of a personal narrative identity can be outlined by means of a mental experiment concerning autobiography, which we have expounded elsewhere [16] and we can briefly recall here. Imagine two persons that are invited to write about the personal history of just one of them. The first one is invited to write an autobiography, whereas the second is invited to write a biography. They are both confronted with the same objects, a set of personal data. Unlike the improbable situation in which they write the same tale, they are likely to produce two stories that differ from each other on several points. Hence, the two accounts of the same life are in competition: they both seek to be considered the best account of what has been narrated. Furthermore, this competition is not necessarily biased in favor of autobiography, since there is no reason in principle to make it prevail over biography as the best account of the narrated self. Concerning that account, we may discern two narrating selves: The autobiographical Self that stems from the process of self-reference. Autobiography is a narrative scheme that seems to reassure us, since it gives us the idea that the whole process of referring personal data to ourselves (collection and communication) is under our control. As we will see it, autobiography is also a competitive form of narration that entails a risk requiring from us a certain degree of trust; The competitive Self that stems from the process of hetero-reference. Biography is a narrative scheme that does not reassure us, since it tells us that the identity construction is always negotiated, i.e., it is at all times the outcome of a dialectics between me and others. This tension requires us to construct a trustful context of communication, where our narrative identities can be interpreted and understood. In this case, the identity of the narrated self grows out of the competition between the two tales and, hence, between the two selves that we might wish to reconcile. For this reason, the personal identity of the narrated self is able to emerge when two conditions are met: firstly, when the competition is fair, i.e.

Constructing a personal identity is an activity much more complex than elaborating a series of online profiles, which are only digital hints of the Self. The construction of our personal identity is a context-mediated activity. Our hypothesis is that young people are enabled, as digital natives and social network users, to co-construct the "context of communication" in which their narrative identities will be interpreted and understood. In particular, the aim of this paper is to show that such "context of communication", which can be seen as the hermeneutical counterpart of the "networked publics" elaborated by Danah Boyd, emerges out of the tension between trust and privacy. In other terms, it is, on the one hand, the outcome of a web of trustful relations and, on the other, the framework in which the informational norms regulating teens' expectations of privacy protection are set and evaluated. However, these expectations can be frustrated, since the information produced in such contexts can be disembedded and re-contextualized across time. The general and widespread use of information technology is, in fact, challenging our traditional way of thinking about the world and our identities in terms of stable and durable structures; they are reconstituted, instead, into novel forms.

of self-reference. Autobiography is a narrative scheme that seems to reassure us, since it gives us the idea that the whole process of referring personal data to ourselves (collection and communication) is under our control. As we will see it, autobiography is also a competitive form of narration that entails a risk requiring from us a certain degree of trust; The competitive Self that stems from the process of hetero-reference. Biography is a narrative scheme that does not reassure us, since it tells us that the identity construction is always negotiated, i.e., it is at all times the outcome of a dialectics between me and others. This tension requires us to construct a trustful context of communication, where our narrative identities can be interpreted and understood. In this case, the identity of the narrated self grows out of the competition between the two tales and, hence, between the two selves that we might wish to reconcile. For this reason, the personal identity of the narrated self is able to emerge when two conditions are met: firstly, when the competition is fair, i.e., both the autobiographical and the competitive self are given at least equal opportunities in narrating the true story of the self; secondly, when the competition occurs in a shared context of communication. Consider now the different case in which just one person is invited to write about her personal history: the case of autobiography (this case is not absorbed by the previous case, since in this circumstance the readers are not confronted by a double account of the self). The identity of the narrated self is yet again constituted by a set of data that derives, necessarily, from a selection of data. This selection is by itself significant, because the author has to choose the most relevant and trustworthy personal data, in order to give a meaningful and trustful account of the self. Of course, the omission is symmetrically significant, since also the omitted data give shape, in negative terms, to the self. From a hermeneutical or psychoanalytical standpoint, what is omitted may be, actually, more significant than what is selected. Also in this case, we can discern two narrating instances of the same Self: An autobiographical Self, whose tale is significant for what the author chooses to select and to narrate: the set of relevant and reliable information. The author is inevitably concerned with tracing a sharp line between what data have to remain private and what data can become public. A competitive Self, implied in the autobiographical Self, whose negative tale is significant for what the author excludes. The narrating instance of the competitive self is driven by strong forces, which are not always entirely governed by the author. Often, these forces are at play to prevent people from narrating what is most painful or harmful in their personal history. As we will see below, young people in particular tend to reveal only what they consider harmless to them. In this respect, what they narrate is a patchy, variable map of their relative certainties and fragilities, whilst their competitive self prevents them from reporting what can really endangers the integrity and the consistency of their self-narration. A competition in fact is at play between the autobiographical and the competitive self, since one story cannot be simply reduced to the other. The identity of the narrated self grows out of the competition between two selves or two instances of the same self. This means that both in the first case, concerning the structure of intersubjectivity, and in the second case, concerning the structure of subjectivity, the identity of the narrated self is always polemic, since it emerges out of the struggle between competing instances of the self. Nevertheless, this competition is not a blind dispute but implies a communication. It entails both communicating with others (intersubjectivity: [28][29][30]) and communicating with the other that inhabits each of us (subjectivity: [31][32][33]). Here lies my point: even when people speak about themselves (i.e., disclose personal data), they are involved in the construction of a trustful and shared context of communication, where their narrative identity is expected to be interpreted. This means that not only epistemic trust (i.e., to trust that) but also relational trust (i.e., to trust in) are deeply and necessarily intertwined with privacy in the identity construction, since the semantic understanding of data requires a society and a bundle of human relations. The construction of personal identity is, accordingly, a competitive activity, with regard to privacy. As Vittorio Mathieu points it out, "privacy does not constitute an object but a relation, which cannot at its turn be objectified, of an object with an origin and an intention [...]. In other words, privacy is, as well as law, a protection of freedom, but as freedom of and not only as freedom from" [34]. For this reason, the violation of privacy is not the violation of an inner self that would be deprived of personal data or of the control thereof. A violation of privacy deprives the self essentially of a capacity, that is the active power to take the initiative of referring a content to the Self [22,34], namely, of developing a coherent story. The tactics of privacy is no longer defensive but goes to the point of constructivism, included the construction of a shared context of communication, which becomes more feasible in a world in which a context is conceived in informational terms. This power is thought of in the horizon of the autonomy of the Self, but is neither exclusive nor uncontroversial, since the identity construction, irreducibly grows out of both a trustful cooperation and a competitive narration. This means also that the identity of the narrated self and thus the relation between trust and privacy are well founded and assured inasmuch as the relation between cooperation and competition is fair, i.e., the narration of the self is given by both cooperative and competing instances according to fair conditions. Cooperation (i.e., the mutual and trustful construction of a context of communication) and competition (i.e., the polemic narration of personal identity) between different agents and instances (e.g., between different sources of data) can never be wholly excluded because they are inherent to the construction of the self. What matters is that that construction is based on fair, reciprocal conditions, and on the setting of a shared context of communication between those instances. This brings about two consequences. Firstly, fairness is included among the philosophical premises of privacy, along with autonomy and dignity. The impact and the design of ICTs, the evolution of which can alter the conditions of cooperation and competition, are to be evaluated in relation to their potential standings of fairness, i.e., their procedural capacities to comply with the conditions of a fair cooperation and competition in the disclosure of information. The minimal condition of fairness is assured when all instances are given at least equal opportunities to shape a true account of the Self. Secondly, fairness is displaced when people do not have a grip upon the conditions of the technological communication, which in turn has a tight hold upon them. The relation of mutual implication, according to which "if I make technologies; they, in turn, make me" [35] should not be eliminated. A fair and balanced relation between trust and privacy is assured not only when agents can technologically protect their identities but, primarily, when they participate in the construction of the ethical backdrop of the value-sensitive information technologies. In other terms, it is true that the evaluation of the design of ICTs requires us to unfold the values (the standings of fairness) imbedded in the design. However, such values cannot be really appraised and given meaning if we do not co-construct a trustful context of communication, where the meaning of those values can be interpreted and understood. --- The Case of Social Networks: Facebook In the present paragraph we do not propose an empirical survey on the use of Facebook and its related problems. We construct our argument on the basis of some empirical studies on the subject and we try to understand some of their results in the light of what has been said in the present essay. Among these studies, the analysis displayed by Danah Boyd deserves special attention as regards to the networked affordances of social networks sites "that shape how people engage with these environments" [36]. Afterwards, we recall some of our previous considerations and measure their soundness in relation to the case of Facebook. Danah Boyd develops the pivotal notion of "networked publics" that is to some extent complementary to our analysis of the networked construction of identity within a "context of communication", that is an informational space structured by the tension between trust and privacy. "Networks publics are publics that are restructured by networked technologies. As such, they are simultaneously (1) the space constructed through networked technologies and (2) the imagined collective that emerges as a result of the intersection of people, technology, and practice" [36]. The idea that networked publics are both a space and an imagined collective is very perspicuous and interesting: it explains the complex dynamics according to which the networked affordances of the architecture of a particular environment enables people to develop an image of themselves that is projected, understood and evaluated within the space constructed -in informational terms -by the interactions among users. Danah Boyd explains well how this dynamics is structured: "Networks publics' affordances do not dictate participants' behavior, but they do configure the environment in a way that shapes participants' engagement. In essence, the architecture of a particular environment matters and the architecture of networked publics is shaped by their affordances" [36]. Danah Boyd is very clear in her analysis and we will take advantage of her analysis of the networked affordances of social networks sites that easily apply to the case of Facebook. However, we would like to briefly bring up a point that seems of great interest, even if it will be impossible to deal with it in full here. Networked affordances enable but "do not dictate participants' behaviors". This explains why participants can behave in a certain way in some specific environments but does not actually explain why they do behave in such way. The easiest way to deal with it would be to state that technological affordances let humans be free whether or not to endorse the possibilities they afford. In this perspective, the relation between technological affordances and actual behaviors would be only a probabilistic one. Even if we subscribe the idea that this relation is a non-deterministic one, we believe that people-knowingly or unknowingly-endorse possibilities enabled by technological affordances, when such endorsement gives them new "powers" [24], which are connected with their own desires. In the case of social networks, notably young people are given the possibility (and therefore the power) to co-construct, in informational terms, a meaningful "context of communication", where their narrative identities can be formed, enacted and evaluated, even if this informational context can easily collide with the different empirical contexts that people belong in real life and are also represented online: this collision is a direct expression of the informational tension that exists between trust and privacy. Such informational dimension is clearly grasped by Danah Boyd when she underlines that "the properties of bits regulate the structure of networked publics, which, in turn introduces new possible practices and shapes the interactions that take place" [36]. Having in mind our general remark on the point, it is useful to see what are, according to Danah Boyle, the features of social networks sites that function as technological affordances. There are four such features: "profiles, Friends lists, public commenting tools, and stream-based updates" [36]. Let us briefly summarize what Danah Boyle says for each of those features, in order to connect them to our previous analysis: a) Public or semi-public profiles: "profiles are a place where people gather to converse and share. Conversations happen on profiles and a person's profile reflects their engagement with the site. As a result, participants do not have complete control over their self-representation" [36]. This consideration is crucial and applies to what remarked in the fourth paragraph as regards to the competitive construction of personal identity. This feature is a chief character of Facebook's communication, which is backed by three more specific elements that delineate Facebook from other social networks: (1) Facebook requires users, in principle, to give their true identity (which, however, allows the possibility for users to adopt a pseudonymous or an eteronymous. It would be interesting to distinguish between the two types of profiles: the first one is the case of someone who wants to conceal his/her true identity, whereas the other is the case of someone who wants to have his/her identity to coexist with another fictive identity); (2) Facebook is meant to offer an extended level of customization that enables users to knowingly and creatively craft their profiles and to set the limits of their visibility (in relation to what and to whom is displayed), in order to reach a truly public or a semi-public audience [36]; (3) Facebook allows users to construct their own profiles using "plain text", which is a lowest common denominator assuring independence from specific programs of encoding or formatting; b) Friends list: "on social network sites, participants articulate who they wish to connect with and confirm ties to those who wish to connect with them" [36]. Danah Boyd interprets this feature as a way to articulate an "intended public" that defines the site of communication in terms of an homogeneous or heterogeneous social context (the former being characterized by ties to those already belonging to our social context and the latter by ties to those belonging to "different" social contexts). This interpretation applies to Facebook and is consistent with our conception of social ties (expounded in the third paragraph) understood in terms of trust relations intended to widen or narrow the sphere of our social context of communication. There is a further point to be stressed that is connected with people's activism in shaping the context of communication and where they decide to project their identities: to "confirm" ties gives them a flavor of voluntarism (in terms of expressed choice) that seems to be absent or just implicit in offline human connections. This aspect is even quantifiable in social networks as Facebook and becomes a synonymous of our popularity and scalability (that is the "potential visibility" [36] of our contents on a large scale). c) Tools for public communication: "most social network sites provide various tools to support public or semi-public interactions between participants. Group features allow participants to gather around shared interests" [36]. In that respect, Facebook displays several tools with which users may interact: 1) the Wall, that is a space on the user's profile page where friends can post messages and comments on the profile for the user to be seen and commented on; 2) Photos, that is a tool by means of which users can upload photos (which happen to raise privacy issues in what they may convey about other friends or persons in general and for what concerns the metadata of a photo); 3) Pokes, that enables users to send each other a notification (a virtual poke) that tells users that they have been "poked", that is to say that they have been alerted that other users would like to have access to the user's profile page. All the mentioned interacting tools are means through which users can articulate the relation between privacy issues and trust relations. In fact, on the one hand, tools for public communication and interaction function on the basis of previously determined privacy settings; on the other, they can ease the formation of trust relations since they allow users to gather around "shared interests": as mentioned elsewhere [10], one of the key element of trustworthiness is parties' shared concern for common interests or goals. d) Stream-based updates: "status updates" is a tool that "allow participants to broadcast content to Friends on the sites" [36]. Friends are informed, in this way, about users' actions, whereabouts, messages or comments. Updates can be, in some cases, re-displayed on a user's profile page for other users to comment on. We would like to highlight some comments offered on this point by Danah Boyd: "in doing so, participants get the sense of the public constructed by those with whom they connect" [36] and, more broadly, "social networks sites are publics both because of the ways in which they connect people en masse and because of the space they provide for interactions and information" [36]. These comments are crucial in order to understand three key points: (1) the notion of personal identity displayed by social networks is no longer concerned with the description of who we are but rather with the progressive construction and updating of who we would like to be: it falls upon us to decide what (including who) we want to select and what (including who) we want to give up; (2) this construction is a co-construction that expresses the tension between what we want to include (trust) and what we want to exclude (privacy) by our social context of communication; (3) this co-construction is shaped by networked affordances in the sense that technologies are no longer perceived as means directed towards determined ends but as an environment, i.e., a space to inhabit, whose structural properties based on information (bits, data, etc.) differ from those of the physical space based on things (atoms, objects, etc.). These technological affordances thus shape the environment where the online construction of personal identity by means of social networking grows out of the tension between trust (openness and inclusion) and privacy (closeness and exclusion). Against the backdrop of what we have noticed so far by summarizing Danah Boyd's interpretation of social networks, we can now recall in five points the nature of the trade-off between trust and privacy as the framework of the online identity construction. After this framework is sketched out, we will measure to what extent the general interpretation of social networks' features and the analyzed trade-off between trust and privacy applies to the specific case of Facebook. --- The Trade-off between Trust and Privacy 1. Trust and privacy are, prima facie, at variance with each other, because trust seems to involve a certain disclosure of personal information whereas privacy seems to involve a certain closure of personal information: there would be, thus, an irreducible trade-off between trust and privacy. 2. This trade-off relies on two undisclosed premises that have to be made more precise. Trust is not only concerned with the communication of personal data but, above all, with the construction of a trustful atmosphere and a web of relations. The protection of privacy is not a passive attitude concerned with the delimitation of a private space but, on the contrary, it is an active dimension concerned with taking the initiative of referring data to a personal account across time. 3. Trust and privacy are different sides of the same coin: they are both involved in the construction of identity. Trust aims at the elaboration of a specific space (a shared context of communication where to include others) where the self-narration can be interpreted, appraised and understood. Privacy is concerned with the elaboration of a specific time, in which the integrity and coherence of the narrative identity is assured across time (this implies excluding others from our narration). This is likely to be the possible cultural danger of some social network sites, like Facebook, that in a sense strengthen the dichotomy between inclusion and exclusion. 4. The identity construction is put at risk and endangered by: (a) the (automated) importation of data that pierces the context of communication from the outside and causes danger to the contextual integrity of self-narration (also because of the "replicability" and "searchability" of networked content [36]); (b) the implicit communication of someone else's personal data that pierces the context of communication from the inside and alters the contextual integrity of someone else's narration (also because of the "persistence" and "scalability" of networked content [36)]; (c) the understanding of time as an eternal present that endangers both the coherence of the narration of personal identity across time and the right to be forgotten (which is the legitimate counterpart of the persistence of online data automatically recorded and archived); 5. The construction of online personal identity is a complex process in which trust and privacy can be reconciled to the extent to which cooperation and competition between narrating instances of the Self are balanced according to standings of fairness. Those standings cannot be simply and automatically embedded in the technological solutions afforded by design and architecture but they need to be construed and evaluated according to a shared context of communication, which requires users also to share a common horizon of meaning. Behaviors can be, thus, facilitated or discouraged by technological settings but the interpretation of behaviors and the concrete choice of technological settings is still guided by the mental attitude to give meaning to specific notions as trust, privacy, identity and so forth. After these remarks, we can focus now our attention at the ways youth in particular (but not only) deal with trust and privacy notions and issues in the co-construction of online personal identity through the use of a social network like Facebook. --- The Meaning of Youth for Digital Natives First of all, empirical studies tend to mold the interpretation of the use of social networks, especially in relation to the issue of privacy, according to the different ages of users [37][38][39]. This is correct, and a necessary methodological starting point for two main reasons: (1) different ages bring about different levels of maturity and self-awareness along with different patterns of social and psychological needs; (2) the evaluation of the protection of privacy rights is linked to the explicit or implicit consent to the treatment of personal data, which depends on the major age. However, this approach seems to let apart or to out-shadows what is the key aspect of the problem: how young people implicitly interpret the definition of "youth". For digital natives, the use of ICTs, especially social networks, is a distributed, collective self-interpretation of what "youth" is and means in the Society of Information. This selfinterpretation stems from the detailed empirical studies over the teens' use of Facebook, if read from this standpoint. "Youth" is "participation". Participation is not to be understood according to the distinction between passive destination of information and active creation of content. It is not a mere synonymous of user-generated content, at least not when this expression is literally and reductively understood as meaning that content is created by an author. If such expression is correctly understood (as in [40]), it gives us a clue to understanding the meaning of participation. The networked generation of content is not a private, individual activity but a "group phenomenon" [40], according to which teens create contents which are already directed to someone else: there is always an "expected audience" [41] or "networked publics" [36]. Young people implicitly seem to know that meaning is a function of a context. For this reason, as digital natives, teens tend to interpret youth as a shared participation in the construction of the context of communication ("youth") in which they expect that their narrative identities will be meaningfully experienced, interpreted and understood. They wish to produce a web of trustful relations with their "expected audience" or "networked publics" (we have already remarked in this respect that this form of participation, this web of relations, can hide a dynamics that is ultimately based on the relation between inclusion and exclusion). This leads us to a second central point. --- The Co-construction of the Informational Context of Communication Social networks, like Facebook, are technological platforms where people experience a new form of communication, by strengthening previous ties or creating new ones, in order to sort out the sense of anonymity that has characterized the industrial society and still affects the information society. In this way, people, and young people in particular, tend to construct their online personal identity or at least to provide themselves with a re-appropriated sense of the Self. Of course, this entails a communication of personal data according to the structural affordances of online communication, which are based on the properties of "bits" instead of those of "atoms" [36]. What are those specific structural affordances and thus what are the specific features of such communication? We have already made reference to those structural affordances but we briefly recall them here, as formulated by Danah Boyd, who believes they emerge out of the properties of bits [36]: 1) Persistence: online expressions are automatically recorded and archived. 2) Replicability: content made out of bits can be duplicated. 3) Scalability: the potential visibility of content in networked publics is great. 4) Searchability: content in networked publics can accessed through search. This explains, in terms of structural affordances, a point which we have insisted on from the beginning: in the digital reality the construction of a text is not an activity truly separated from the construction of a context since both of them are made of bits that are characterized by the same structural affordances. The identity of those entering into communication grows out of the process of communication and it is part of the context of communication people jointly elaborate thanks to those structural affordances. In that sense, social networks are "contexts of communication" in which young people tend to form their identity more than contexts in which they communicate on the basis of an already established identity. What is innovative is that both identity and context are made out of a sum of information. Identity is a sum of information that gives content to the narration of the Self, whereas context is made out of a sum of information that characterizes a specific situation [2]. It is the informational nature of the context of communication that represents the originality of such form of communication. This originality is, thus, both hermeneutical (in terms of informational contexts) and technological (in terms of networked and structural affordances emerging out of the properties of bits). When communicating through social networks, young people do not only aim to bridge the discrepancy between the ideal and the actual self [42], but they implicitly and, to some extent, unknowingly co-construct their identities and the context of communication where such identities will be interpreted and understood. Similarly, Boyd remarks that "knowing one's audience matters when trying to determine what is socially appropriate to say or what will be understood by those listening. In other words, audience is critical to context" [36]. The disclosure of information young people accomplishes in the social networks is therefore aimed both at constructing "their" online identities and setting "their" shared context of communication in which such information is, normatively, to be comprehended and evaluated. This can lead teens to restrict their "social network profile to private, making it inaccessible to anyone outside their group of friends" ([43] that also refer to [37,44]). If we look to this phenomenon from this point of view, i.e., the effort to create a sphere of life where to be understood and validated by their peers, we understand why young people, according to their psychological needs, may or may not change their privacy settings (there is in fact evidence, in the literature, in both directions: see [43]). In other terms, for the first time, young people are given the possibility to co-construct the contexts of communication, i.e., the "horizon of meaning" [11], in which their self-narrations are given meaning. And they can do that in a ground-breaking way, since both the narrated identities and the context of communication are made out of the same malleable materials: information (made out of data made out of bits). "The computational rendition of reality has far-reaching implications in the sense of recapturing a growing proportion of the physical and cognitive landscape of contemporary life into the medium of permutable and recombinable information" [45]. Teens trust in social networks, like Facebook, because they are mutually involved in the elaboration of their context of communication made out of the "liquid" digital information characterizing the specific situation in which they interact by social networking: with of course all the problems that go along with the possibility of re-contextualization of such information (i.e., the collision of contexts underlined also by Danah Boyd [36]). In this, way trust plays a decisive and complex role in the analysis of social networks, and this leads us to a third point. --- The Risks of Systemic Trust and Familiarity Young people's trust, displayed in social networks, is not only a form of personal trust that depends on an intersubjective act of delegation. Rather, it is a form of systemic trust [9] that is concerned with a set of generalized expectations of stability towards a world or, in our own terms, towards a certain context of communication. The web of trustful relations (i.e., teens' mutual expectations) fill this world with an atmosphere of "familiarity" [9], which is recursive in the sense that it is a precondition for establishing further relations of trust. This brings about three consequences: (a) familiarity enables us to step out of our natural diffidence towards others, caused by the unpredictability of their future behaviors (their otherness). However, familiarity is capable of helping us overcome our diffidence, not because we get to know other users (which is true only in some cases), but since we share with them the same context that we have contributed to build: familiarity is the familiarity of the same world, and not of the same profiles. Online profiles are poles of attraction only because they are grasped and interpreted within the same universe of meanings, symbols, references; (b) familiarity widens the scope of trust since it makes trust to become a distributed attribute of the system more than a punctual quality of personal relations. This means that familiarity has the tendency both to immunize the system from single delusions, i.e., from single failures of trust relations, and to weaken the requirement of motivation, which is needed in order to establish a personal trustful relation [9]; (c) familiarity is apt to reinforce our trust in the system and thus to lessen our defense against specific aggressors: this is typical of all the regimes of inclusion, which makes us represent risks as something that comes from the outside. In contrast, it has been remarked, for instance by Helen Nissenbaum, that betrayals come "from those who are allowed within our spheres of safety, within our safe zones" [46]. As a result, systemic trust in social networks can expose adolescents to more risks, since it becomes less evident to them where risks come from. In a context built on systemic trust, it is easier to develop trustful relations but it becomes more difficult to have control over those relations, since who trusts in the system depends, eventually, on the complex functioning of the system as a whole, which makes it difficult to detect and to have power over single unpredicted behaviors. On the one hand, building a trustful context of communication reinforces trust and privacy as identity constructions, because it enables people to structure the context in which their identities are expected to be interpreted and understood. On the other, the systemic trust, which seems to be a mark of social networks, is likely to expose young people to uncalculated risks that endanger their privacy, understood as a form of control over personal information. This leads us to a further point concerning the idea of privacy. --- The Co-construction of Personal Identity Young people that use social networks, like Facebook, do not really (i.e., knowingly) pursue a balance between trust and privacy. The trade-off between trust and privacy could be misleading for the reason already outlined: both trust and privacy, experienced in social networks, are concerned with the identity construction. Rather, teens pursue, implicitly, a balance between a traditional (based on settings and norms) and an informational idea of privacy (based on structural affordances): like adults, young people pay attention to disclosing information that construct their online personal identities, even if, unlike them, they are rather inclined to believe that privacy, understood as a legal institution aimed at protecting us from unwanted intrusions, is pretty outdated in the networked Society of Information. In this sense, there might be a trade-off between the construction of personal identity and the limitation over access to personal information: "Disclosure thereby becomes an aspect of identity construction, and that construction is linked with popularity: the people that are most popular are those whose identity construction is most actively participated in by others. As a result, the risks of limiting access to personal information become greater than the risks of disclosure, because, when limiting access, the individual also limits the potential for identity construction and thus reduces his or her popularity" [47]. It has also been remarked, nonetheless, that "there is no correlation between providing personal information online and a lack of concern for privacy" [43], and, in addition to this, "students instead managed audience concerns through privacy settings and obfuscating nicknames" [48]. The relation between disclosure and limitation over access to information seems, thus, to be more a question of self-perception than one of concern for privacy violations. This is the reason why we have insisted upon a socio-cognitive conception of trust, because it is crucial to understand how teens perceive and represent the world, i.e., the entrusted context of communication they live with and interact in. When the identity construction is concerned, not only adolescents need to entrust their peers, in order to share information with them as a "token of friendship and trust" [43], but, more fundamentally, they need to be trusted, that is, they need that their self-narration is integrated by that of their peers, who are crucial instances of the competitive Self (see paragraph 4). This is backed by several studies on social media [41,47,49,50], as it has been remarked: "since much of peer socializing among young people goes on via social media, young people's conduct, both offline and online, is shaped by a general desire to be validated by their peers" [43]. Teens perceive their world and the resulting collective self-perception "forged when people make judgments based upon the mosaic of information available about us" [51, see also 36], as the necessary outcome of a co-constructed identity in a shared context of communication. This leads us to consider a final point. --- Accountability The ways teens make use of social networks, like Facebook, tend to support our doubts (see paragraph 3.3) as to what threatens the construction of personal identity via social media. Firstly, "aggregation and analysis involving large databases are increasing the possibility that individual privacy may be invaded in new and more substantial ways" [43], because "personal information is not only outside our control but also subjected to a bureaucratic process that is itself not adequately controlled" [51]. Secondly, "it is common for people who live their lives mediated by digital technologies to disclose, knowingly and unknowingly [we underline], personal information online. Once digitized, such information is virtually irretrievable and may be intercepted or purchased by commercial entities, governments, or individuals for marketing or other sinister purposes" [43] that refer also to [51][52][53]. Thirdly, "scholars claim that young people will be the first to experience the aggregated effect of living a digital mediated life, with the corresponding creation of various identities and digital dossiers over a long period of time (we underline). Solove describes modern 'architectural problems' related to privacy, which involve 'the creation of a risk that a person will be harmed in the future'" [43]. Fourthly, "networks publics force everyday people to contend with environments in which contexts are regularly colliding. Even when the immediate audience might be understood, the potential audience can be far greater and from different contexts. Maintaining distinct contexts online is particularly tricky because of the persistent, replicable, searchable, and scalable nature of networked acts. People do try to segment contexts by discouraging unwanted audiences from participating or by trying to limit information to make searching more difficult or by

Constructing a personal identity is an activity much more complex than elaborating a series of online profiles, which are only digital hints of the Self. The construction of our personal identity is a context-mediated activity. Our hypothesis is that young people are enabled, as digital natives and social network users, to co-construct the "context of communication" in which their narrative identities will be interpreted and understood. In particular, the aim of this paper is to show that such "context of communication", which can be seen as the hermeneutical counterpart of the "networked publics" elaborated by Danah Boyd, emerges out of the tension between trust and privacy. In other terms, it is, on the one hand, the outcome of a web of trustful relations and, on the other, the framework in which the informational norms regulating teens' expectations of privacy protection are set and evaluated. However, these expectations can be frustrated, since the information produced in such contexts can be disembedded and re-contextualized across time. The general and widespread use of information technology is, in fact, challenging our traditional way of thinking about the world and our identities in terms of stable and durable structures; they are reconstituted, instead, into novel forms.

Once digitized, such information is virtually irretrievable and may be intercepted or purchased by commercial entities, governments, or individuals for marketing or other sinister purposes" [43] that refer also to [51][52][53]. Thirdly, "scholars claim that young people will be the first to experience the aggregated effect of living a digital mediated life, with the corresponding creation of various identities and digital dossiers over a long period of time (we underline). Solove describes modern 'architectural problems' related to privacy, which involve 'the creation of a risk that a person will be harmed in the future'" [43]. Fourthly, "networks publics force everyday people to contend with environments in which contexts are regularly colliding. Even when the immediate audience might be understood, the potential audience can be far greater and from different contexts. Maintaining distinct contexts online is particularly tricky because of the persistent, replicable, searchable, and scalable nature of networked acts. People do try to segment contexts by discouraging unwanted audiences from participating or by trying to limit information to make searching more difficult or by using technologies that create partial walls through privacy settings" [36]. All these problems are different and, accordingly, they are not likely to be encountered in the same way. Each of them suggests a different solution that should be, nevertheless, understood as a part of a complete picture. This picture is dominated by the requirement of accountability. The first problem, concerning the automated collection and storage of data, suggests, according to [43], that "companies that collect and store personal information 'have an obligation to build secure systems, and they ought to be held accountable [we underline] under the law if they don't'". The second problem, as to the implicit communication of a third party's personal information, needs a solution based on education. However, education is not to be aimed at scaring teens away from social media [43,54,55], by exaggerating the risks involved in the use of social networks. Education should be intended in a way that makes young people sensitive to what is implicit (not-said) in their communication: teens should be made more accountable to whom might be harmed by their disclosure of personal information, when the infringement of a third party's privacy is the unintended consequence of their self-narration. The third problem, regarding the effects of adolescents' actions across time (the disclosure of information and the implicit consent to the use of personal data), needs, however, a solution based on limiting their accountability. Young people deserve, like anybody else, a "right to be forgotten": they cannot be held accountable for what can emerge from a digital dossier "over a long period of time". They cannot be unfairly enchained to the past, once they have totally left the context of communication in which personal data have originally been disclosed. People should be made accountable towards other people because of the harm they have caused to them. They cannot be made unreasonably accountable to their own past: "'profiles on social networking sites won't always show up in a search engine query, but they will appear when members of those services track down the data subject' [56]. In the past, oral gossip could tarnish a reputation, but it would fade from memories over time. People could move elsewhere and start anew. Being shamed in cyberspace, however, is capable of becoming a 'digital scarlet letter' [51]" [43]. As observed, people should always be able to take the initiative of narrating their history, even if such narration is never made out of their exclusive creation [57]. The fourth issue is difficult to deal with, "because, with the audience invisible and the material persistent, it is often difficult to get a sense for what the context is or should be. [...] In networked publics, contexts often collide such that the performer is unaware of audience from different contexts, magnifying the awkwardness and making adjustments impossible" [36]. This final remark is important since it enables us to stress that the online construction, segmentation and shaping of contexts are, on the one hand, a new and vital character of online communication and, on the other, are subject to the structural affordances that emerge out of the properties of bits and allow contexts to be pierced and collide each other both in space and across time. --- Conclusions The relation between trust and privacy is traversed by several contradictions. At a first glance, trust and privacy seems to necessarily entail a trade-off: for instance, young people share personal information on social networking because they already trust each other or in order to trust each other, with the result that they endanger their privacy and that of their peers. However, as understood from the standpoint of the identity construction, both trust and privacy point to the same end, even if in different manners. The concern for trust is directed at setting a context of communication, made up of trustful relations, in which young people's identities can be validated by other peers. The concern for privacy is aimed at suggesting what personal information it is fitting to reveal and to distribute in that context, by connecting the social expectations of privacy protection to the informational norms of contextual integrity [58], which are not previously fixed but depend on the sum of information that characterize a situation, i.e., a social network, in terms of a trustful context of communication. Creating a trustful context of communication requires social networks users to develop a systemic trust, in Luhmann's terms [9], or to create a supra-agent, in Floridi's terms [59], which is a union of agents that form a single unit based on confidentiality. This raises some problems, as already pointed out, that depend on the systemic or supra-individual dimension of trust. From a technological standpoint, it has been correctly remarked that "it is extremely difficult for the average citizen to keep up with the pace of technological change" [43]. It is true that "few people would be knowledgeable enough about digital technologies to have an effective sense of what information they are sharing is publicly accessible and what is private" [60]. This means that people are more exposed than before to the "information risk" [61] that derives from the invisible dimension [62] of their digital operations. From the standpoint of confidentiality, the following consideration is significant: "Confidentiality is a bond that is hard and slow to forge properly, yet resilient to many external forces when finally in place, as the supra-agent is stronger than the constitutive agents themselves. (...) But it is also a bond very brittle and difficult to restore when it comes to betrayal, since the disclosure, deliberate or unintentional, of some personal information in violation of confidence can entirely destroy the privacy of the new, supra-agent born out of the joining agents, by painfully tearing them apart" [59]. More generally, teens' expectations of privacy protection or, to put it differently, teens' expectations of identity construction, which are set and developed in the shared and trustful contexts of communication they have cooperated to create by social networking, can be frustrated, since the information produced in those contexts can be easily disembedded and re-contextualized across time, because of the already mentioned collision of contexts and consequences of the structural affordances of the networked acts in terms of persistence, replicability, scalability and searchability. In this sense, the disaggregation of data, their "mobility, transferability and combinability" [45], is a structural feature of the digital fabric of the Internet reality. In fact, the pervasive use of information technology is challenging our traditional way of thinking about the world and our own identities in terms of stable and lasting structures, which are more and more reconstituted into novel social and individual forms. Young people, as digital natives, are likely to be more inclined to conceive the informational construction of personal identity as a competitive activity which is characterized by a specific and original dimension: both the identity and the context of which identity is part are formed by the permutable and recombinable information based on the paradigm of computation [45]. On the one hand, this enables people to mold their self-narration and the context of communication in which this narration is displayed. On the other, the transferability and the combinability of re-contextualized information is able to alter the meaning of their information, thus endangering the integrity of their self-narration. In conclusion, the perspective we have chosen for analyzing the relation between trust and privacy, i.e., the construction of personal identity, allows us to make a final, brief remark. This remark is concerned with the supposed transparency of the personal data displayed by social networking. As mentioned at the start of this paper, the meaning of the world "revelation" is twofold; the disclosure of data, at the same time reveals personal information (transparency) and hides the conditions of transparency (which never appear). In this sense, we have to observe that adolescents' concern for privacy includes, primarily, a strive for privacy from adults, from their surveillance. Parents should be aware of the fact that the struggle for the construction of personal identity (the polemic activity of the competing instances of the autobiographical Self, seen in paragraph four) is a constitutive ingredient of a personal identity, maybe the most important one, since it is the dimension in which young people recognize the patchy, relative map of their certainties and fragilities. However, this map is consigned to the written in a way that is always more selective than is usually expected. Secret diaries, both offline and online, are most often governed by the capacity of the competitive Self to leave out of the narration what cannot be confessed to anyone, namely, what is both most painful and harmful. In other terms, even the most secret messages, are written, offline and online, as if they could be one day looked for and read by someone else. Parents who love to look in private diaries or to intrude in social networks will find out, quite often, only what is already waiting for them.

Constructing a personal identity is an activity much more complex than elaborating a series of online profiles, which are only digital hints of the Self. The construction of our personal identity is a context-mediated activity. Our hypothesis is that young people are enabled, as digital natives and social network users, to co-construct the "context of communication" in which their narrative identities will be interpreted and understood. In particular, the aim of this paper is to show that such "context of communication", which can be seen as the hermeneutical counterpart of the "networked publics" elaborated by Danah Boyd, emerges out of the tension between trust and privacy. In other terms, it is, on the one hand, the outcome of a web of trustful relations and, on the other, the framework in which the informational norms regulating teens' expectations of privacy protection are set and evaluated. However, these expectations can be frustrated, since the information produced in such contexts can be disembedded and re-contextualized across time. The general and widespread use of information technology is, in fact, challenging our traditional way of thinking about the world and our identities in terms of stable and durable structures; they are reconstituted, instead, into novel forms.

Introduction Hemoglobinopathy (HbP) is a genetic mild to acute anemia. The two common forms, thalassemia and sickle cell anemia, occur in Africa, the Mediterranean Basin, and Southeast Asia. Carriers of HbP have an evolutionary benefit in these areas where malaria is endemic. A simple blood test reveals that if both partners are carriers, they have a 25% chance in each pregnancy for a newborn with HbP. If (future) parents are at risk, they can prevent the birth of a HbP baby, or they can decide not to have more children, or consider prenatal or preimplantation diagnosis. Strategies to offer HbP carrier screening differ worldwide. Some Middle Eastern countries have nationwide premarital carrier screening programs [1][2][3][4][5][6][7][8]. In Northern Cyprus, with a high prevalence of thalassemia, premarital carrier screening is mandatory since 1980 [9]. This screening is considered effective in terms of the reduction of children affected with HbP; after 1984 the number of newborns with thalassemia in Northern Cyprus dropped from 18-20 to 6-7 annually. Since 1991 only 5 babies with thalassemia were born [9]. Prevalence of HbP has increased in Northern Europe because of migration [10]. This raises questions as to whether it is beneficial to offer carrier screening and whether ethnically targeted screening is also an option in The Netherlands [11]. HbP carrier screening is not routinely offered to high risk groups even though there are sufficient options to offer neonatal, prenatal, or even preconception carrier screening in The Netherlands [12]. A recent Dutch study showed feasibility and desirability of a combined preconception ancestry-based carrier screening for cystic fibrosis (CF) and HbP [13,14]. Since 2007 all Dutch newborns are screened for sickle cell anemia in the neonatal screening program. The method used also reveals carriers. Parents have to choose to receive the result. In the first year of the extended screening, 64 newborns were identified with sickle cell anemia (prevalence 0,035%) [15]. In addition, 806 carriers of sickle cell anemia (prevalence 0,4%) were found. Only few parents did show up for genetic counseling in the regional clinical genetic centers at Amsterdam and Rotterdam, the cities with the largest migrant populations in The Netherlands [16]. The information on neonatal carrier for sickle cell anemia and the implications for (future) parents was suboptimal and parents also are occasionally ill-informed [17,18]. Furthermore, the disclosure of a newborn carrier may lead to confusion, guilt, anxiety, and early stigmatization and it may even reveal nonpaternity [19]. Ethical issues are at stake regarding timing and extent of disclosure towards the affected child and other family members [20]. The experience with Dutch HbP carrier screening shows the importance of correctly informing high risk individuals and taking into account their familiarity with HbP. This study with data of 2007 aims to explore (a) the familiarity with HbP and HbP carrier screening among high risk groups in The Netherlands; (b) the social-psychological determinants of the intention of these high risk groups to participate in HbP carrier screening; and (c) their preferences regarding HbP carrier screening. --- Materials and Methods --- Subjects. Structured interviews by a standard questionnaire were performed in 301 Dutch subjects living in or around Rotterdam of Turkish (n = 100), Moroccan (n = 100), or Surinamese (n = 101) ethnicity (see Table 1 for demographics). These interviews lasted 29-156 minutes (depending on the questions asked by the subjects) and were held at home or in a nearby community center by an interviewer of matching ethnicity in Dutch and their native language. All subjects aged 18-45 participating in a community network of a multicultural research center were invited. Respondents were asked for names to recruit new potential participants within Rotterdam. Subjects were selected to evenly represent males-females, various age intervals, and those living in different parts of town (SES equivalent). New applicants were entered until we reached three groups of 100 participants. All participants received a gift certificate after the interview of 20C. --- Interview. The interviewers followed a constructed questionnaire with the following issues: (i) knowledge on HbP (true or, resp., false/yes or no), carrier screening, and perceived carrier risk (5 scales of (dis)agreement), based on the Precaution Adoption Process Model [21]; (ii) a short explanation of HbP and its hereditary nature; (iii) reaction to various HbP carrier screening situations, by asking intentions and hypothesized actions in explicit scenarios (see Figure 1); ( 14) 59 ( 23) 58 (14) (3) "Imagine your unborn child is diagnosed with HbP and you're both carriers"; (iv) items on determinants of carrier screening behavior, based on the Theory of Planned Behavior (TPB) [22] and Health Belief Model (HBM) [23]. These assess risk perception, attitudinal beliefs, social norms, selfefficacy, and intentions. Positive and negative anticipated affective reactions to HbP carrier screening which may influence the respondents' intention to participate in carrier screening (see Table 2), as shown in other studies on screening behavior [24,25] (all answers with 5 scales of (dis)agreement); (v) preferred screening conditions: respondents were asked eight times during the interview to select one of the two hypothetical screening situations (two (1) Background information: age, gender, partner (married, living together, no partner), kids (yes, no), and educational level (finished with a certificate). (2) Determinants measured with one or two items: profession or faith, perceived importance of faith, perceived importance of family (ties), perceived importance of adapting to the Dutch culture, perceived importance of maintaining own culture, perceived importance of health, perceived responsibility for own health, perceived importance of doctor's advice, perceived importance of fate/destiny, and perceived importance of a higher power/God which influences health. (3) Multiple item scales within scenario 1 ("Imagine you participated in HbP carrier screening and it shows you are a carrier for HbP"): knowledge about HbP and carrier (screening) (5 items), anticipated positive feelings (2 items, Cronbach's alpha = 0.49), anticipated negative feelings (3 items, alpha = 0.52), attitudinal beliefs about HbP carrier screening (10 items, alpha = 0.73), perceived social norm of HbP carrier screening (7 items, alpha = 0.88), self-efficacy (trust in own abilities to participate in HbP carrier screening; 2 items, alpha = 0.65), and risk perception of being a carrier or having a child with HbP (4 items, alpha = 0.81). 2 Backwards multiple regression analysis (Pin = 0.05 en Pout = 0.055) with the significant determinants of the univariate analysis. Adjusted R 2 = 0.27; only the significant determinants in the model are displayed. vignette cards each representing five specific conditions requiring HbP carrier screening; choose appropriate card). The conjoint analysis method forces respondents to make tradeoffs between different attributes of hypothetical situations. We distinguished five attribute categories and varied these for each vignette card: ( We used a reference level that resembles foreign screening programs and is suitable to be introduced within the Dutch health system. Half of the participants (55%) correctly answered that even if both parents are carriers a healthy baby is possible, 44% identified the statement "carriers get sick" to be incorrect, 43% identified the statement "mainly women transfer genetic diseases to their children" to be incorrect, and only 25% correctly identified the statement "HbP is more common in my country of birth. " Those of Surinamese background scored higher on knowledge (5.84 of a scale 0-10) compared to ethnic Turkish or Moroccan participants (4.76, resp., 3.78). --- Results --- Subjects. Demographics of the 301 participants ( --- 3.3. Behavior. More than one-third (38%) of the participants expressed intention to participate in carrier screening, 3% were tested, and 59% were in doubt or disregarded testing. The intention to participate was higher among females (46.6% versus 30.5%). A flow chart displays the responses by the three different scenarios (Figure 1). While 103 respondents intend to participate in prenatal screening, only 34 of them opt for abortion if HbP is diagnosed. The average negative attitude towards abortion is scored 1.97 (on a scale from 1, negative, to 5, positive). After the first trimester of pregnancy this score is even 1.52. 2) were added in the multiple regression, and the question "Will you plan to participate in future HbP carrier screening?" was a dependent variable. In the model four significant (P <unk> 0.01) variables were obtained (R 2 0.27). These were (1) beliefs about (the importance of) HbP carrier screening (P <unk> 0.001), (2) risk perception to be a carrier or to have a child with HbP (P = 0.01), (3) anticipated negative feelings (P = 0.02), and (4) it is relevant for your health to listen to a physician and follow his advice (P = 0.04). --- Intention to Participate. All determinants for intention to perform a HbP carrier test (footnote Table --- Preferred Screening Circumstances. Conjoint analysis showed that the respondents preferred screening before pregnancy instead of during pregnancy (screening during pregnancy versus screening before; odds ratio = 0.78, 95%, confidence interval = 0.71-0.87). Those with Turkish ethnicity preferred screening earlier, at school age. Screening free of charge or for less than 50 euros is preferred. The current screening test in The Netherlands costs 50C and a full diagnostic test 100-500 C. These costs are not provided by standard health insurance. --- Other Aspects. Respondents disagreed with the statement "I dislike HbP carrier screening by reference to ethnicity" (mean score of 2.16 on a scale from 1, totally disagree, to 5, totally agree). They agreed with the statement "I am more intended to get screened if I know HbP to be common in my country of birth" (mean score 4.14). The open-end questions finally revealed that important motivations to participate in carrier screening are "perceived relevance for future children" and "importance of health issues;" while reasons not to participate are "feeling healthy, no pregnancy wish, and no family member with HbP. " Ninety-two percent of the respondents would inform their family ("they have a right to know") and 98% would inform their partner ("right to know, to get tested too, and matter of honesty"). Seventy-eight percent of the respondents agree that HbP carrier screening may be offered routinely. Most respondents prefer information on a leaflet sent by mail in the Dutch language or to be informed by general practitioners. They appreciated being informed by the interview, and some stated: "it is interesting, I learned something. " --- Discussion Less than half of those from Surinamese, Moroccan, and Turkish ethnicity in The Netherlands are familiar with HbP (45%) and fewer with carrier screening (27%). Half expressed intention to be tested (41%). Those from Surinam were relatively more familiar and knew someone with sickle cell anemia. Only 25% of subjects knew HbP to be endemic in their country of birth. A recent study found that 56% of Dutch recently married couples had a positive intention to participate in CF carrier screening [26]. Van Elderen et al. [27] showed a higher intention (83.5%) to participate in HbP carrier screening when planning for pregnancy, among 109 ethnic Turkish females in The Netherlands. We found lower intention to participate in those aged 18-45 (38%), although intention was higher among the women (46.6%). Our subjects could also score "maybe yes/maybe not"; 29.3% of the women choose this answer category. Eventually 26% of the women had no intention to participate in carrier screening. Van Elderen et al. [27] found that one-third (30.3%) of the women intended to opt for prenatal screening after carrier screening and we found a similar score (33%). --- Determinants of the Intention to Participate. We also examined social-psychological factors that influence the intention to participate in carrier screening. Respondents who had a more positive intention towards participating acknowledged the risk of being a carrier and were more inclined to listen to a physician's advice. Anticipated negative feelings with regard to carrier screening appeared to be another relevant factor. In other words, the more negative feelings are anticipated, the less one intends to participate in carrier screening. This result is consistent with previous findings on screening behavior, indicating that both cognitive and affective processes play a role in the decision to participate [25,28]. --- Preferred Screening Circumstance. Our subjects show a clear preference for HbP carrier screening before pregnancy, profoundly shown in their choices with the situational cards (vignettes). Screening before pregnancy is chosen over screening during pregnancy. Those of Turkish ethnicity preferred screening even earlier: during school age as is customary in, for example, the Jewish communities in Israel, NY, and Australia [29]. While 91% of the respondents with an intention to participate opt for screening during pregnancy if both partners would be carriers, only a third of them would be willing to consider termination of pregnancy if the prenatal test results in a HbP baby. This finding is consistent with the UK, where one-third of identified "at high risk couples" chose to have prenatal diagnostic tests [30]. We did not ask why they would consider participating in prenatal screening if they are not willing to consider termination of pregnancy. However, the responses at the end of the interview showed that they would like to know their unborn child's status, mainly to be prepared for a sick child after birth. A recent qualitative study by Gitsels-van der Wal et al. [31] among pregnant Muslim women showed that beliefs associated with their religion played a significant role in decision-making regarding antenatal screening, particularly regarding termination of pregnancy. In our study we focused on the intention for prenatal screening during the hypothetic scenarios. There are other possibilities besides pregnancy termination, such as not having children, adoption, changing partners, and premarital screening. The results of this study show the importance of early prenatal testing to allow at risk couples to consider more options during the first trimester of pregnancy and importance to inform parents-to-be in the preconception phase. --- Limitations and Strengths. This study measured intention to participate in HbP carrier screening in hypothetical scenarios, but "intention" may not match "actual behavior. " Also, most subjects had not previously heard about HbP and therefore had limited time to consider the pros and cons of such behavior. Because the snowball method was used to recruit participants, we obtained more highly educated respondents. Regression analyses in this study however showed that educational level was not significantly associated with the intention to participate in HbP carrier screening. Because of the matching ethnicity of respondent and interviewer, it was possible to directly answer and ask for further clarification in the native language. Finally, the sample used in this study warrants some concern, as our sample seems not entirely demographically representative for the larger population of Turkish, Moroccan, and Surinamese people in The Netherlands. Therefore, we need to be cautious in generalizing the results from this study to the total Turkish, Moroccan, and Surinamese population. --- Implications. For public health education these results imply the need for tailored health education to high risk groups (those of Surinamese, Moroccan, and Turkish ethnicity) on HbP carrier screening. Knowledge on HbP heritability, on the risk of being a carrier, on consequences, and on implications of a positive test result is of importance. Our subjects indicated that they would be more intended to screen if they knew that HbP is more common in their country of birth. Only 25% knew this to be the case indeed. It is therefore relevant to mention that HbP is endemic in the country of ethnicity in health education. Also the information that other diseases, such as CF, are more common in Caucasians is relevant. In health education the knowledge that most Middle Eastern countries introduced national screening programs to detect carriers and the fact that Islam fatwa allows abortion in the first trimester for medical reasons like HbP [32] are also important issues for health education. For those who are willing to participate easy access to screening should be facilitated. Recent research by Lakeman et al. [13,14] showed that it is possible to offer a combination of HbP and CF carrier screening in The Netherlands, based on ethnicity. However, currently no population based carrier screening for HbP is offered in The Netherlands and most carriers are found due to unexplained mild anemia, during pregnancy of after neonatal screening. People who receive these carrier test results should be referred to clinical genetic centers that provide good quality information. --- Conclusion These findings show the importance to inform those with high risk with tailored health education. For those willing to participate easy access to this type of screening should be facilitated. Because it is important for high risk groups that both health education and carrier screening will take place before pregnancy, health education should be focused during preconception. With a standard HbP carrier screening implemented, tailored health education, genetic counseling, and decision aids should receive attention. These should be developed in concordance with the level of knowledge and the preferences of those at high risk.

Objective. To explore factors that influence intention to participate in hemoglobinopathy (HbP) carrier screening under Dutch subjects at risk, since HbP became more common in The Netherlands. Method. Structured interviews with 301 subjects from Turkish, Moroccan, or Surinamese ethnicity. Results. Half of the participants were familiar with HbP, 27% with carrier screening. Only 55% correctly answered basic knowledge items. After balanced information, 83% percent of subjects express intention to participate in HbP carrier screening. Intention to participate was correlated with (1) anticipated negative feelings, (2) valuing a physician's advice, and (3) beliefs on significance of carrier screening. Risk perception was a significant determinant, while respondents were unaware of HbP as endemic in their country of birth. Respondents preferred screening before pregnancy and at cost < 50C. Conclusion. These findings show the importance of informing those at risk by tailored health education. We propose easy access at no costs for those willing to participate in HbP carrier screening.

Introduction As the result of rapid urbanization, there were 245 million migrants in China by 2016. These migrants have made an indelible contribution to the development of various cities in China, but their subjective well-being (SWB) and mental health have not received much attention. [1][2][3] SWB refers to how people evaluate their quality of life according to their emotional response and cognitive judgments. 4 A low ratio of SWB can result in job stress and turnover. 5,6 SWB differs from mental health, as mental health includes both subjective and objective psychological feelings. Former studies have shown that the mental health and SWB of Chinese internal migrants were worse than those of urban residents. 7 This situation may be caused by the strict household registration system in China. Many migrants cannot obtain registered permanent residence in their working cities. 8 They cannot get the same health care as local residents [9][10][11] and do not integrate into the local community, 12 which may affect their SWB and mental health. It is of great significance to study the influencing factors of SWB and mental health and take some intervention measures against the influencing factors in order to improve the health of migrants. Previous studies have shown that life satisfaction, emotions, personal value orientation, 13 personal adaptation to the environment, 14,15 fear of victimization 16 and general selfefficacy 17 can all affect SWB, while length of residence, 18 language speaking, employment and general health and income 19 are influence factors of mental health. In addition, social capital may influence the SWB 20,21 and mental health [22][23][24][25][26] of migrants. Social capital refers to the state of interrelation between individuals and members of society, including norms, authority and social network. 27 It can be divided into cognitive and structural social capital. Cognitive social capital refers to subjectively shared values and emotions, such as trust and values. Structural social capital is an objectively generated social network, including related roles and organizations, that builds connections between people and groups. 28 Cognitive social capital can be represented by social cohesion, which is a social force that can adjust or coordinate the contradictions, conflicts and disputes between different factors in society. 29 Takeuchi et al. 30 divided structural social capital into two dimensions: membership in social organizations and frequency of social participation. The study of Agampodi et al. 31 showed that individual-level cognitive and structural social capital can promote the SWB and mental health of pregnant women by providing focused care. Cramm et al. 32 found that social capital influences the well-being of older adults in The Netherlands. Our previous study found that social cohesion and social interaction are positively related to SWB. 33 Therefore cognitive and structural social capital may affect the SWB and mental health of migrants. Worldwide, the gap between mental health needs and mental health care is enormous. 34,35 Studying the relationship between social capital and mental health and SWB is important to help bridge that gap. However, there were few existing studies examining the effects of social capital on mental health among migrants. In the present study, we aim to investigate the associations of three dimensions of social capital with SWB and mental health among migrants in China. --- Materials and methods --- Design and participants This cross-sectional study was conducted in five cities (Shanghai, Zhengzhou, Xinzheng, Xingyang and Baoji) from June 2017 to April 2018. We used a two-stage sampling method to sample participants 15-75 y of age in all five cities. In total, community residents in 41 districts (all 16 districts of Shanghai, all 7 districts of Zhengzhou, 6 random districts of Xinzheng, 7 random districts of Xingyang and 5 random districts of Baoji) were included in the survey. We randomly selected 5% of the population 15-75 y of age from each district. In total, 20 953 community residents were randomly selected to complete the questionnaire. Data were collected by an anonymously selfadministered questionnaire. A subsample of 3067 internal migrants (who have rural Hukou but currently work and live in cities for at least 6 months) were retrieved to examine the associations of social cohesion, membership in social organizations and social participations with SWB and mental health. --- Measurement Social capital In the present study we used social cohesion to measure cognitive social capital. Membership in social organizations and the frequency of social participation were utilized to measure structural social capital. 30 The Chinese version of the Neighborhood Scale was used to assess social cohesion (see Appendix Table A1). 36 The rating scale ranges from 1 (strongly disagree) to 5 (strongly agree) and the subscales were added together to get a total score. The Cronbach's <unk> was 0.93. Membership in social organizations was assessed by enquiring whether or not the respondents were members of six social organizations in their area (see Appendix Table A2). Only a response of participation resulted in a score on these items and addition of the six items yielded the total score. 30 The Cronbach's <unk> was 0.84. The frequency of social participation was assessed by enquiring how often respondents participated in eight different activities (see Appendix Table A3). 30 All items were scored on a scale of 1-5 (1=no participation, 2=several times a year, 3=several times a month, 4=once a week, 5=two or three times a week) and addition of the eight items yielded a total score. The Cronbach's <unk> was 0.94. The scores of the three dimensions of social capital were then divided into high and low by median. --- SWB SWB was assessed by the Chinese version of the Personal Wellbeing Index (see Appendix Table A4). 33 All items ranged from 0 (strongly dissatisfied) to 10 (strongly satisfied). The Cronbach's <unk> was 0.92. Likert scale data were normalized to the unit of the largest scale (% SM) on a 0-100 distribution and used the equation - - <unk> X K K 100, min max min where X is the fraction to be converted, K min = 0 (the smallest possible score on the scale) and K max = 10 (the largest possible score on the scale). 37 --- Mental health Mental health was assessed by the World Health Organization (WHO)-Five Well-Being Scale (see Appendix Table A5). 38 All items were scored on a scale from 0 to 5 (5=all the time, 4=most of the time, 3=more than half of the time, 2=less than half of the time, 1=sometimes, 0=never) and addition of the five items yielded the total score. A score <unk>13 indicated poor mental health. The Cronbach's <unk> was 0.94. --- Self-rated health Self-rated health (SRH) was assessed by asking respondents what they thought about their own health. The scores of all items ranged from 1 (very good) to 5 (poor). --- Health behaviours --- Smoking status We used two questions to evaluate smoking status: 'Have you smoked 100 cigarettes or more until the present moment?' and 'Have you smoked in the past 30 days?'. Only a yes response on both questions resulted in identification as a smoker. --- Problematic drinking Problematic drinking was measured by the CAGE 4-item questionnaire (see Appendix Table A6). 39 Among the four items, an answer of yes on one item indicated an alcohol-related problem, while two or more yes responses indicated problematic drinking. The Cronbach's <unk> was 0.73. --- International Health --- Physical activity Physical activity was assessed by the International Physical Activity Questionnaire (see Appendix Table A7). 40 A response of never indicated no physical activity, while answering three or more times on the first item and 30 or more minutes on the second item simultaneously indicated regular physical activity. All other answers indicated irregular physical activity. The Cronbach's <unk> was 0.75. --- Sociodemographic characteristics The study included the socio-economic factors of gender, age (range <unk>29-<unk>70 y), education (ranging from elementary school to university), marital status and employment status. --- Statistical analyses We first used a <unk> 2 test to do a univariate analysis with Stata, version 14.0 software (StataCorp, College Station, TX, USA). In order to evaluate the associations between social capital and SWB, generalized linear regressions were used after controlling for self-rated health, health behaviours and sociodemographic characteristics. The associations between social capital and poor mental health were assessed by binary logistic regressions after controlling for self-rated health, health behaviours and sociodemographic characteristics. The p-value was set at <unk>0.05 (two-tailed). --- Results After removing the migrants with incomplete questionnaires (n=29), 3038 migrants were included in the current study. There were 1568 (51.6%) males and 23.4% were <unk>30 y of age. More than 50% had a senior high school or higher education. Only 16.5% of migrants were unmarried and 78.7% were employed. The median level of SWB was 75.7 (interquartile range 62.9-85.7) and the prevalence of poor mental health was 10.0% (Table 1). The univariate analyses showed that the factors of age (30-39 y), education (a junior or senior high school education), married status, a high ratio of SRH, non-smoking status, irregular or regular physical activity, a high level of social cohesion, a high level of membership in social organizations and a high level of social participation were positively associated with a high ratio of SWB. An elementary school education, a low ratio of SRH, no physical activity, a low level of social cohesion and a low level of social participation were positively correlated with poor mental health (Table 2). The results of the generalized linear regressions demonstrated that compared with migrants with an elementary school education, those with a junior high school education (<unk>=2. 3). have a low ratio of poor mental health (Table 3). --- Discussion Since a high level of SWB and good mental health can comprehensively improve the health status of the residents and promote the harmonious development of society, 41,42 exploring their determinants is very important. To the best of our knowledge, few studies have examined the impact of social capital on the SWB and mental health of immigrants. Our current findings show that high social cohesion is significantly associated with a high level of SWB, which is consistent with previous studies. 43,44 Migrants often do not have personal relationships as extensive as those of local residents. They usually cannot get needed information or timely help from relatives or friends if they are in trouble. However, the cohesion of community members makes it easier for them to exchange ideas and obtain needed knowledge. 44,45 In the current study we found that high social cohesion and high social participation were significantly associated with a low ratio of poor mental health. This finding is consistent with previous studies. 43,46 Many migrants cannot afford to buy a house or obtain a registered permanent residence in the city in which they work. 8 It is hard for them to integrate into the local community. If a community has strong cohesion and gives migrants more opportunities to participate in community activities, people will achieve a greater sense of security, thus promoting the mental health of migrants. In addition to the impact of social capital on SWB, we also found that SRH is positively associated with SWB. These findings are consistent with previous studies. 45,47,48 The impact of SRH on SWB is especially prominent in those with poor health. The explanation is that the physical pain and psychological pressure caused by physical diseases often become an important obstacle in the human pursuit of SWB. 49 Many researchers believe that physical activity can improve an individual's SWB. [50][51][52] In our study we found that physical activity, especially regular physical activity, is significantly positive in the SWB of migrants. Prior studies in China suggested that physical activity (three times a week for 30-45 min each time) can increase SWB. 53 Physical activity can reduce psychological pressure, improve physical function and cardiopulmonary function and results in improved SWB. 54,55 --- Limitations We used cross-sectional studies to explore factors that influence SWB and mental health. It is difficult, however, to determine the causality of mental health status. Moreover, our study did not find an association between membership in social organizations and mental health. Penley and Tomaka 56 found that becoming a member of a social organization can prevent future depression, relieve depressive symptoms and prevent the recurrence of depression. Future studies could use longitudinal studies to explore these associations. In addition, as we used a self-administered questionnaire to conduct the survey, we cannot determine the respondents' level of understanding of the questions or attention to the surveys. However, our study includes urban residents in the eastern, central and western regions of China. The survey population is relatively comprehensive and may represent the current situation of Chinese migrants. --- Conclusions Our study shows that self-rated health, physical activity and cognitive social capital are positively associated with SWB while cognitive and structural social capitals can reduce poor mental health in migrants. Based on these findings, further insights can be expected from exploring the relationships between other aspects of social capital and SWB and mental health. Taking measures to improve social participation and social cohesion may improve the SWB and mental health of migrants in China. --- Supplementary data Supplementary data are available at International Health online http://inthealth.oxfordjournals.org Authors' contributions: HF and JG developed the concept and design. JG acquired the funding. JG, YZ and XN conducted the survey. HF, JG and JD provided project administration and resources. HF and JG supervised and validated the item. YZ and JG wrote and revised the manuscript. All authors read and approved the final manuscript.

Background: We aimed to examine the associations of cognitive and structural social capital with subjective well-being (SWB) and mental health among internal migrants in five cities in China.We conducted healthy cities surveys from June 2017 to April 2018. In total, 3038 migrants submitted surveys. Generalized linear regressions for SWB and logistic regressions for mental health were used to examine the associations of social capital, SWB and mental health after controlling for covariates. Results: The median level of SWB was 75.7 (interquartile range 62.9-85.7). Self-rated health and physical activity were positively associated with SWB. The SWB level among migrants who perceived high individual social cohesion was much higher than that of their counterparts (unstandardized coefficients, β=7.01 [95% confidence interval {CI} 5.82 to 8.21]). The prevalence of poor mental health was 10.0%. High social cohesion (odds ratio [OR] 0.32 [95% CI 0.24 to 0.44]) and high social participation (OR 0.77 [95% CI 0.57 to 0.97]) were significantly associated with a low ratio of poor mental health when compared with their counterparts. Conclusions: Cognitive social capital is positively associated with SWB, while both cognitive and structural social capitals were negatively associated with poor mental health. It is beneficial to migrants' mental health and SWB to promote social participation and social cohesion.

Background Adolescence and youth are considered periods of development when individuals construct their own identity [1]. These periods include life events (LEs) and transitions [2], which can be either desirable or undesirable. LEs can be stressors and demand a special readjustment to reorganise daily life and they might influence child's development [3][4][5]. Frequent exposure to LEs during adolescence and youth has been shown to be associated with worse health related quality of life (HRQoL) [6][7][8][9], psychosomatic complaints [10][11][12], poor physical functioning, higher risk of disabilities, and greater use of health services [13]. Most studies on the effect of LEs have been carried out in convenience samples which cannot be considered representative of the general population [14][15][16], limiting their ability to make inferences based on the available observations. Studies based on general population are deemed needed [13,17], to establish whether genderbased differences exist in such samples. Importantly, the associations between LEs and health related outcomes have been generally assessed in cross-sectional design studies [14][15][16]18]. This design does not allow taking into account health status previous to the LEs suffered, resulting in possible biases. For instance, girls usually report low levels of HRQoL. Lacking a previous measure of perceived health may lead to overestimate the effect of LEs experiences on girls. Gender differences in exposure and reactions to LEs have been widely discussed [14,15,19,20] and seem to have an effect onto mental health and functional outcomes stronger among girls than boys [15,19,[21][22][23]. But more recent studies have failed to observe gender differences or fewer negative associations. The latter results have reported changes in the traditional gender differences of coping with LEs, might be explained by modifications in social resources, and gender role expectations [2,[24][25][26]. This "buffering hypothesis" would predict a significant interaction effect for social support and life events in relation to psychological outcome, which seems to be the case of girls, which perceive higher levels of social support in recent studies [2]. In addition, although it is known that exposure to LEs can have health consequences, we are not sure about which ones (desirable or undesirable events, or other categories as family or extrafamily) are most influencing [3,4,27]. The aim of the present study was to investigate the impact of LEs on HRQoL using data collected with a longitudinal design, in a general population sample. Additionally, we investigated whether there were gender-based differences in the association between LEs and HRQoL [28]; and whether desirable and undesirable events and those which occurred within the family context versus extra-family impacted differently on HRQoL. We had anticipated that no gender differences in the association between LEs and HRQoL would be found, and that undesirable events and those related to family life would be more strongly associated with HRQoL [3,4,27]. --- Methods --- Sample and data collection The Spanish KIDSCREEN baseline sample was recruited between May and November 2003 as part of the European KIDSCREEN fieldwork [29]. The target population for the KIDSCREEN study was children and adolescents aged 8-18. The aim was to recruit a sample that was representative by gender and age in each participating country according to census data. Telephone sampling was performed centrally from Germany, and was carried out using a Computer Assisted Telephone Interview with random-digital-dialling. Households were contacted by telephone and asked to participate by interviewers who had received study-specific training. If the family member contacted agreed to participate, the questionnaire and other study materials were mailed to the requisite address together with a stamped, addressed envelope and informed consent for return of the completed questionnaire. A telephone hotline was used to provide further information about the survey. Two reminders were sent in cases of non-response (after two and five weeks). More details are provided elsewhere [28,29]. Between May and November 2006, follow-up questionnaires were posted by mail to all adolescents and youths and their parents who had previously agreed to participate in the Spanish KIDSCREEN follow-up study (n = 840 of 926 participants at baseline; 91%). Data collection at follow-up took place 3 years after baseline, a period which was considered a sufficient interval to allow for substantive changes in participants' health status. The fieldwork followed the same methodology applied at baseline. Postal reminders were sent four and eight weeks after the first mailing to those who had not returned their completed questionnaires. A third reminder was sent after twenty weeks and any remaining nonrespondents were contacted by phone. --- Measures --- Life events Stressful LEs were measured using the Coddington Life Events Scales (CLES) [3,4,27], which measure the occurrence of 53 LEs. Respondents indicate for each item describing a specific LE and the number of times the event has occurred in the last 3 months, 4-6 months earlier, 7-9 months earlier, or 10-12 months earlier. The frequency of occurrence is taken into account in the calculation of Life Change Units (LCUs) which also reflect the amount of stress inherent to the event and how long ago it happened. We used the original LCUs, which were obtained from ratings provided by teachers, paediatricians, and child psychiatrists [3,4,27]. A total LCU score can be calculated for each respondent as a weighted sum of all the LCU scores (range of LCUs for one LE: 5-216). We used the Spanish version of CLES that has been found to be valid and psychometrically equivalent to the original [30]. Life events were classified according to two different typologies: desirable (e.g. "Graduating from high school) vs. undesirable events (e.g. "Divorce of parents) and family-related (e.g. "Loss of a job of your father or mother") vs. extra-family events (e.g. "Going on the first date") [3][4][5]27,31]. Each LE was classified accordingly into one of the two categories in each typology, except nine which were classified in only one LE typology (e.g. "Being hospitalized for illness or injury" was undesirable but was not classified in the 'family' typology because it was neutral with regard to that particular typology). Health Related Quality of Life HRQoL was measured using the KIDSCREEN-27 [32], which was administered at baseline and follow-up to the adolescents and youths (self-reported) and to their parents (proxy-reported) with a recall period of 1 week. The KIDSCREEN-27 has 5 dimensions: Physical Well-being (PH, 5 items); Psychological Well-being (PW, 7 items); Parent Relation & Home Life (PA, 7 items); Social Support & Peers (PE, 4 items) and School Environment (SC, 4 items). We also calculated an overall index score (KIDSCREEN-10) based on selected items [32]. The KIDSCREEN items use 5-point Likert-type scales to assess either frequency (never-seldom-sometimesoften-always) or intensity (not at all-slightly-moderatelyvery-extremely). Rasch scores are computed for each dimension and for the overall score and are transformed into T-values with a mean of 50 and standard deviation (SD) of 10. The T scores refer to the mean values and SD from a representative sample of the European general population so that scores over (or under) 50 indicate better (or worse) HRQoL than the general EU population. The Spanish version of the questionnaire has demonstrated acceptable validity and reliability [33,34]. In this study, only responses from adolescents and youths on the KIDSCREEN questionnaire were used. --- Pubertal development Pubertal development was measured in order to adjust for possible differences between boys and girls in terms of pubertal status [1] and because of its demonstrated relationship with HRQoL [35]. We used the Pubertal Development Scale (PDS), a self-reported measure with acceptable levels of validity [36], which assesses pubertal characteristics. Subjects respond to each item on a 4point ordinal scale (no development = 1, development barely begun = 2, development definitely underway = 3, and development already completed = 4). An extra response category was included in each item of the PDS to determine whether development had been completed before the baseline assessment. The menarche item was scored 1 if the girl was pre-menarche and 4 if menstrual periods had already begun. Item scores were summed to produce an overall continuous score ranging from 5 to 25. Higher scores reflect a greater degree of pubertal development. Pubertal development was only measured at follow-up. --- Sociodemographic variables Other variables collected in the present study to characterize the sample were family socio-economic status and parental level of education. Socio-economic status was measured using the Family Affluence Scale [37], which includes questions on family car ownership, having own unshared room, the number of computers at home, and how many times the family went on holiday in the previous 12 months. FAS scores were categorized as low (1), intermediate (2), and high (3) affluence level. Parental level of education was collected from the adult respondent and included the highest family level of education according to the International Standard Classification of Education (ISCED). Categories were: low (a maximum of lower secondary level, ISCED 0-2), medium (upper secondary level, ISCED 3-4), and high (university degree, ISCED 5-6) [38]. --- Statistical analysis Differences between boys and girls in relation to LEs and HRQoL were tested using independent two-sample t-tests for continuous variables and chi-square tests for categorical variables. P-values were adjusted with the Hochberg method in order to address the multiple testing problem. The decision rule is to reject the null hypothesis when the adjusted p-value is less than <unk> = 0.05 [39]. To investigate the first aim of the study, bivariate analyses of the effects of LEs on HRQoL dimensions for boys and girls were performed. We calculated the effect sizes of changes in HRQoL (difference between followup and baseline scores divided by baseline standard deviation) in three different groups of increasing LEs impact. The three categories were defined based on tertiles of LCU distribution (i.e. 0-67 LCU (low), 68-160 LCU (medium), and 161 LCU or more (high)). Two way ANOVA was used to determine whether gender differences were statistically significant. Multiple linear regression [40] models were tested to investigate the second aim of the study of whether LEs typologies impacted differently on HRQoL. The dependent variables were KIDSCREEN dimensions and overall scores at follow-up; independent variables were the LCU scores of global LEs and typologies. A <unk> 2 test to assess non-linearity between LCUs and HRQoL was not significant, and LCUs were therefore introduced as a linear variable in the model. Models were fitted to estimate the relationship between LEs and HRQoL controlling by baseline HRQoL, pubertal development, and age and stratified by gender. P-values were adjusted with the Hochberg method in order to account for the analysis of multiple end-points. We tested gender differences in the association between LEs on HRQoL by running similar models where the gender and the interaction between gender and LCU scores were included and evaluating the significance of the interaction effect with a two-sided significance test at <unk> = 0.05. Coefficients in the multiple linear regression indicate the units of change in the dependent variable which is associated with 1 LCU suffered by respondents. To give more interpretable results, we selected four examples to illustrate the magnitude of the effect of the coefficients. To do so, the regression coefficients were multiplied by different LCUs values that correspond to selected amounts of LCUs (i.e. 113 LCU, 165 LCU, 235 LCU and 281 LCU) to assess the direct effect on KIDSCREEN. In addition, to determine minimally important differences and moderate important differences between respondents (LCUs necessary to have a change in HRQoL of 0.2 and 0.5 SD, respectively) [41]. This transformation was applied only to undesirable events due to their special impact. We did not attempt to evaluate age groups differences due to insufficient sample size after distributing participants by LE typology and gender. --- Results and Discussion At follow-up, 454 families were re-assessed (response rate: 54%). A total of 423 adolescents/youths with complete data were included in the analysis. Mean age was 15.4 (SD 2.84) years and 51.8% were girls (Table 1). When compared with non-respondents at follow-up, respondents were younger with a slightly higher parental level of education. KIDSCREEN scores at follow up were lower (poorer HRQoL) than at baseline for all dimensions. Girls reported significantly lower scores than boys (p <unk> 0.05) in the Physical Well-being dimension both at baseline and at follow-up. Boys reported a mean of 5.3 LEs in the previous 12 months compared to 5.7 for girls (P = 0.98) (Table 2). Desirable events (a mean of 3.1 among boys and of 3.5 among girls) were more common than undesirable events (2.0 and 2.1), respectively. Extra-family events (4.4 in boys and 4.7 in girls) were more common than family events (0.9 in both boys and girls). The mean of Life Change Units was 127.2 (SE 8.15) for boys and 139.2 (SE 8.04) for girls (P = 0.88). Girls tended to have higher LCU scores in all LEs categories, though differences in scores were not statistically significant. Table 3 shows KIDSCREEN scores at baseline and follow-up and effect sizes for each of the three LEs categories (tertiles). In general, HRQoL deteriorated over time in all KIDSCREEN dimensions for both boys and girls. However, in the group of boys that reported the fewest LEs the decrease was small, with effect sizes Multiple linear regression analysis indicated that LEs tended to affect more HRQoL dimensions in boys than in girls, though gender differences were not statistically significant (Table 4). After Hochberg adjustment for multiple comparisons, the strongest associations were seen between undesirable events and HRQoL on the Psychological Well-being and School Environment dimensions and the Overall score in boys, and on the School Environment and Physical Well-being dimensions in girls. In the case of desirable, family and extrafamily events, none of the coefficients were statistically significant. The impact on HRQoL of an increasing amount of LCUs stemming from undesirable events is presented in Table 5. Data illustrate the magnitude of the effect when several LEs combinations are lived. Two undesirable LEs lived by 16.3% of respondents involve a decrease of 2.72 and 2.14 in KIDSCREEN scores in boys and girls, respectively. While one more undesirable LE is lived, the impact increases considerably. To calculate the minimally important difference (MID) on the KIDSCREEN we calculated the LCUs necessary to have a difference of 0.2 SD, and a moderate important difference established at 0.5 SD. In the case of undesirable LEs the MID of 0.2 is achieved when at least 75 LCUs are lived in boys [weighted life event units 75*(-0.024))/ 8.58]. A score of 200 LCUs stemming from undesirable events (200*[-0.024]/8.58) would be associated with a decrease of 4.8 points in overall HRQoL in boys which involves a decrement of 0.55 SD in the KIDSCREEN. This value corresponds to a moderate important difference to detect respondents more in risk of health consequences, as reported by previous studies [41] which is the 6.5% of the study sample. In this case, the effects were slightly higher in boys despite differences were not statistically significant, though girls should experience a higher amount of LCUs to have the same effect. In this representative sample of the general youth population of Spain, we observed a negative association between LEs and HRQoL, especially on Physical Wellbeing, Psychological Well-being and School Environment, but also on Overall HRQoL. The decrements in HRQoL associated with a higher number of LCUs tended to be greater among boys, though no statistically significant differences were observed between genders. Whereas undesirable events were associated with decrements in HRQoL, desirable events and family and extrafamily events were not associated with a corresponding increase or decrease in HRQoL. In particular, the existence of a combination of undesirable events summing to at least 200 LCUs was associated with a sizeable decline in HRQoL (SD 0.5). Our results do confirm the importance of undesirable LEs in HRQoL and suggest that it is not differential by gender, which put forward the importance of a longitudinal design of the study and changes in the traditional gender differences. These results should be interpreted taking into account several study limitations. First, the response rate at follow-up was 54%. This figure is quite standard for postal surveys [19,[42][43][44] and, importantly, the sample was shown to be representative of the Spanish population in terms of age and gender when compared to census data [28]. A second limitation may arise from the fact that the CLES use an extensive recall period. Although it is conceivable that there may operate a recall bias, we tested the instrument and the recall periods in a pilot study that showed that they were feasible and acceptable to respondents [28]. Thirdly, data on important mediators such as personality and coping styles was not collected and the association described here could be confounded by a number of unmeasured variables. In addition, such confounders could act differently among the different age groups and their inclusion could modify the results. Thus, they should be considered in future studies. Finally, sample size was limited to test age differences. We performed an analysis stratified by age and it showed differences in life events experiences and LCUs scores among males (worse for the older group) but not among females. Also, worse scores in the KIDSCREEN dimensions were observed in older ages, especially in Physical Well-being, School Environment and the KIDSCREEN 10 in females and males. Stratified data analyses are not presented, but are available upon request. Our study has several strengths. Findings regarding the effect of LEs on HRQoL have almost always been presented from clinical samples [16,45,46], and the present work is one of the few to study the association in a general population sample. Contrary to previous crosssectional design studies [2,14,47,48], a longitudinal design makes it possible to assess the association between occurrence of LEs and change in HRQoL [14]. In addition, the present work complements a previous manuscript that studies the effect of changes in mental health on HRQoL [49]. There has also been relatively little research into the effects of intra-family and desirable LEs on HRQoL, with most studies focussing on relationships outside of the family, which was previously suggested by T.M. Damush et al [14]. However, we didn't find any difference between intra-family and extra-family events. Despite these results, when the Hochberg method for multiple testing is not applied, as authors like K. Rothman suggest [50], not only undesirable LEs are statistically significant, also family events. Despite of the different recommendations found at the literature, we considered to include Hochberg corrections to have more precise results. There were HRQoL decreases at follow-up in both gender groups. Girls had lower scores at follow-up [35], but the assessment of the impact of LEs on HRQoL showed no gender differences. These results suggest important differences with previous literature [15,19,[21][22][23]. Previous studies suggested that the burden of demands and limitations on girls was greater due to their role in society which in turn may have made them more vulnerable when adversities are experienced [47]. Previous reports [51,52] found no evidence of differential item functioning (DIF) by gender in the overall KIDSCREEN sample. Thus, gender differences described here must be attributed to real outcome differences rather than biases due to DIF. On the other hand, studies that found less gender differences pointed out that girls may have more sources of support upon than boys than before [25] and experience social reinforcement by turning to friends when they have a problem. Whereas males may experience criticism for not dealing with problems independently [26]. Our observation of higher scores on the social support and peers dimension among girls, although not statistically significant, would support this hypothesis. The differences we had hypothesized about the effect of the different LEs based on typologies were only partially confirmed. Undesirable events had the most important effect on HRQoL, but contrary to our hypothesis, family events did not affect any of the dimensions, when Hochberg method for multiple testing was applied. Our results are consistent with the literature [53] suggesting the effect of LEs derive primarily from their undesirability. In other words, the negative impact on HRQoL of an undesirable LE with a weight of 50 LCUs will be greater than the corresponding positive effect on HRQoL of having a desirable event with the same weight. However, it is important to note that the association of LEs with HRQoL is moderate and their effect is considerable when several important LEs are lived. In our sample, 16.3% of respondents have a LCU sum score that involves a minimal important difference [41] to consider for clinical interventions (SD 0.2). In comparison with the junior and senior high school of the validation study of the CLE scales [27], the number of LEs reported was similar (5.5 in the previous 12 months). However, in that study, the general population sample had a mean of 177 LCUs [27], which was higher than in our sample (133 LCUs) though that also included the adult respondents. One important undesirable event is sufficient to be part of the risk group. In the case of the first LEs combination exposed at Table 5 of breaking up with a boyfriend/girlfriend and failing a grade in school involves a sufficient impact to be part of a risk group. In addition, the combination of these two LEs seems to be usual. In fact, not only the LCUs sum score determine the effect, it's the undesirability of these LEs what makes respondents vulnerable after the experience. Further studies in larger samples would help to confirm or refute our results. Especially, in order to confirm the role of undesirable LEs and the tendency of change of gender differences. Moreover, it should be useful to have different measures of LEs in order to take into account also their effect in previous HRQoL. --- Conclusions The experience of LEs did impact the HRQoL of adolescents and youths in this sample, although the effect was moderate. Contrary to most previous studies, we didn't find that girls are more in risk than boys in the association between LEs and HRQoL. The occurrence of desirable life events did not produce a corresponding increase or decrease in HRQoL. In our sample, it is necessary to have lived LEs with at least a final sum score of 200 LCUs associated with undesirable events to have a moderate impact. This value may be useful as a cut-point to detect risk profiles in general population which may involve a considerable decrement in perceived health. --- List of abbreviations LEs: Life events; HRQoL: Health related quality of life; LCU: Life change unit Author details 1 CIBER en Epidemiolog<unk>a y Salud P<unk>blica (CIBERESP), Spain. 2 Health Services Research Unit, IMIM-Institut de Recerca Hospital del Mar, Parc de Salut Mar, Barcelona, Spain. 3 Agency for Health Information, Assessment and Quality, Barcelona, Spain. 4 NIHR School for Primary Care Research, Department of Primary Care, Division of Primary Care and Public Health, University of Oxford, UK. 5 Universitat Aut<unk>noma de Barcelona (UAB), Barcelona, Spain. Universitat Pompeu Fabra (UPF), Barcelona, Spain. Authors' contributions JMV, LR, and JA participated in the conception and design of the study. EVO, SRF, GV and JAPV analyzed the data. EVO, JMV, MH, MF, LR and JA participated in the drafting of the article. All authors contributed to a critical revision of the manuscript and made a substantial contribution to its content, and read and approved the final manuscript. --- Competing interests The authors declare that they have no competing interests.

Background: Most studies on the effect of life events (LEs) have been carried out in convenience samples which cannot be considered representative of the general population. In addition, recent studies have observed that gender differences in the health related quality of life (HRQoL) impact of LEs might be lower than believed. We assessed the relationship between LEs and HRQoL in a representative sample of Spanish adolescents/youths, focusing on gender differences. Methods: Participants (n = 840) completed the KIDSCREEN-27 to measure HRQoL at baseline and again after 3 years (n = 454). Follow-up assessment included the Coddington Life Events Scales (CLES) to measure LEs experiences in the previous 12 months. Respondents were categorized according to the amount of stress suffered. We calculated both the number of LEs and the Life Change Unit (LCU) score, a summary of the amount of stress inherent to the event and the time elapsed since occurrence. LEs were classified as desirable or undesirable, and family-related or extra-family. Effect sizes were calculated to evaluate changes in HRQoL. To assess the impact of LEs typologies, multiple linear regression models were constructed to evaluate their effect on HRQoL. Results: Girls reported a mean 5.7 LEs corresponding to 141 LCUs, and boys 5.3 and 129, respectively. The largest impact of LEs on HRQoL was observed in the group of boys that reported to have lived more stress (third tertil of LCUs distribution). The linear association between LEs and HRQoL tended to be stronger among boys than girls, but the difference was not statistically significant. The effect on HRQoL was deemed important when undesirable events had been experienced. To have an important impact on HRQoL, 200 LCUs due to undesirable events were necessary in boys. In girls, slightly higher scores were necessary for a similar impact. Conclusions: A moderate association was found between recent LEs and HRQoL, mainly among those who experienced several undesirable events that correspond to at least 200 LCUs. No gender differences were found in this association. Results may be useful for identifying adolescents with particular health risks, regardless of gender.

GETTING THE TERMS RIGHT: WHAT IS IMPACT ENTERTAINMENT The English-language literature uses the term "Social Impact Entertainment" to describe the use of cinema as a means of influencing public opinion and public awareness. Typically, such content covers a wide range of societal issues, such as the environment, cultural differences, racial and gender issues, and others. There is no similar term in Russian, but "impact entertainment" or "social impact cinema" can be used instead. The term also received a definition at the Vedomosti media business conference in 2012, albeit a narrowed-down one, which only referred to "drawing attention" -specifically, it was "an entertainment content production and distribution system that draws the audience attention to issues of social importance." 1 That is, the kind of content that does not only entertain the viewers, but also enables a discussion of important topics that would likely be left in obscurity if not shown on screen. It is important to note that social impact is a running theme in the entertainment industry, playing an important role. Figure 1 2 shows the relationship between impact entertainment and mainstream content production. The chart shows that social impact is an integral part of the entertainment industry without being a separate development sector or part of a development sector. Igor Mishin, CEO MTS Media, has the same view of the phenomenon, which he shared at the aforementioned conference: "Social impact content doesn't have to be something special. Viewers should love it and accept it normally, just like any other type of content. No need to take this topic into a reservation." 1 How do we go from declaring the importance of impact to measuring it and even guaranteeing that our goal is achieved? The global practice for assessing the impact of a motion picture is to use interdisciplinary tools aimed at measuring the social impact of content. In particular, the interest in assessing the social impact of a motion picture is related to the fact that such content can become a tool for addressing vital social and environmental issues by appealing to the public opinion. In this regard, social impact assessment tools are integrated into all stages in the development and production of film projects, as well as after their completion -just like with any project involving social change to be achieved. --- <unk> <unk>: <unk> <unk> <unk>-<unk> <unk>л<unk> о<unk>о<unk>на<unk>ени<unk> со<unk>иал<unk>но<unk>о во<unk>де<unk>стви<unk> кино в ан<unk>ло<unk>но<unk> литературе ис<unk>ол<unk>уетс<unk> термин «Social impact entertainment», котор<unk> трактуетс<unk> как ис<unk>ол<unk>ование кинемато<unk>ра<unk>а в ка<unk>естве средства во<unk>де<unk>стви<unk> на о<unk>ественное мнение и со<unk>нание. <unk>ак <unk>равило, тако<unk> контент о<unk>ват<unk>вает <unk>ироки<unk> кру<unk> о<unk>ественн<unk> <unk>ро<unk>лем, таки<unk> как <unk>коло<unk>и<unk>, кул<unk>турн<unk>е ра<unk>ли<unk>и<unk>, расов<unk>е и <unk>ендерн<unk>е во<unk>рос<unk> и дру<unk>ие. <unk> русском <unk>ке анало<unk>и<unk>но<unk>о термина не су<unk>ествует, но мо<unk>но ис<unk>ол<unk>оват<unk> «им<unk>акт-интерте<unk>нмент» или «со<unk>иал<unk>н<unk> им<unk>акт кино». <unk> 2012 <unk>оду на кон<unk>ерен<unk>ии «<unk>едомосте<unk>», <unk>осв<unk>енно<unk> медиа<unk>и<unk>несу, так<unk>е <unk>ло дано о<unk>ределение <unk>тому термину, но в сокра<unk>енном <unk>на<unk>ении, каса<unk>емс<unk> тол<unk>ко «<unk>ривле<unk>ени<unk> внимани<unk>» -<unk>то «система <unk>рои<unk>водства и дистри<unk>у<unk>ии ра<unk>влекател<unk>но<unk>о контента, <unk>ривлека<unk>а<unk> внимание аудитории к о<unk>ественно <unk>на<unk>им<unk>м <unk>ро<unk>лемам» 1. <unk>о ест<unk>, тако<unk> контент <unk>о<unk>вол<unk>ет не тол<unk>ко ра<unk>вле<unk> <unk>рителе<unk>, но и о<unk>судит<unk> ва<unk>н<unk>е тем<unk>, котор<unk>е не <unk>о<unk>али <unk> во внимание, если <unk> и<unk> не увидели на <unk>кране. <unk>а<unk>но отметит<unk>, <unk>то в индустрии ра<unk>вле<unk>ени<unk> со<unk>иал<unk>н<unk> им<unk>акт <unk>вл<unk>етс<unk> усто<unk>иво<unk> нормо<unk> и и<unk>рает ва<unk>ну<unk> рол<unk>. <unk>а с<unk>еме, <unk>редставленно<unk> на <unk>исунке 1 2, мо<unk>но увидет<unk> соотно<unk>ение ме<unk>ду им<unk>акт-интерте<unk>нментом и ме<unk>нстримн<unk>м <unk>рои<unk>вод ством контента. <unk>та диа<unk>рамма <unk>ока<unk>вает, <unk>то со<unk>иал<unk>н<unk> им<unk>акт <unk>вл<unk>етс<unk> неот<unk>емлемо<unk> <unk>аст<unk> ра<unk>влекател<unk>но<unk> индустрии, не <unk>вл<unk>с<unk> <unk>ри <unk>том отдел<unk>н<unk>м сектором или <unk>аст<unk> сектора ра<unk>вити<unk>. <unk>акое <unk>е <unk>редставление о <unk>влении в<unk>ска<unk>ано на в<unk>еу<unk>ом<unk>нуто<unk> кон<unk>ерен<unk>ии <unk>орем <unk>и<unk>ин<unk>м, <unk>ендиректором «<unk> <unk>едиа»: «Social impact контент не дол<unk>ен <unk>т<unk> <unk>ем-то осо<unk>енн<unk>м. <unk>рители дол<unk>н<unk> е<unk>о л<unk>ит<unk> и вос<unk>ринимат<unk> о<unk>но, как л<unk>о<unk> дру<unk>о<unk> вид контента. <unk>е надо <unk>ту тему уводит<unk> в ре<unk>ерва<unk>и<unk>» 1. <unk>ак <unk>ере<unk>ти от деклара<unk>ии о ва<unk>ности им<unk>акта к е<unk>о и<unk>мерени<unk> и да<unk>е <unk>арантии дости<unk>ени<unk> <unk>оставленно<unk> <unk>ели? <unk>л<unk> о<unk>енки им<unk>акта кино в мирово<unk> <unk>рактике ис<unk>ол<unk>у<unk>тс<unk> ме<unk>дис<unk>и<unk>линарн<unk>е инструмент<unk>, на<unk>равленн<unk>е на и<unk>мерение со<unk>иал<unk>но<unk>о во<unk>де<unk>стви<unk> контента. <unk> <unk>астности, интерес к о<unk>енке со<unk>иал<unk>но<unk>о им<unk>акта кино св<unk>ан с тем, <unk>то тако<unk> контент мо<unk>ет стат<unk> инструментом ре<unk>ени<unk> ва<unk>н<unk> со<unk>иал<unk>н<unk> и <unk>коло<unk>и<unk>ески<unk> <unk>ро<unk>лем <unk>осредством о<unk>ра<unk>ени<unk> к о<unk>ественному мнени<unk>. <unk> св<unk>и с <unk>тим инструмент<unk> о<unk>енки со<unk>иал<unk>но<unk>о во<unk>де<unk>стви<unk> кино внедр<unk>тс<unk> на все<unk> <unk>та<unk>а<unk> ра<unk>ра<unk>отки и <unk>рои<unk>водства кино<unk>роектов, а так<unk>е <unk>о и<unk> окон<unk>ании -<unk>о анало<unk>ии с л<unk>м <unk>роектом, <unk>ред<unk>ола<unk>а<unk>им дости<unk>ение со<unk>иал<unk>н<unk> и<unk>менени<unk>. --- SOCIAL SCIENCES: FILM AS A MULTIDIMENSIONAL SUBJECT Researchers assessing the social impact of motion pictures use various impact assessment tools (Kaufman & Simonton, 2013). The methods focus on different aspects of the social impact of the motion picture and study different subjects: the content of the film or the perception of the audience. One method is surveys, which involve collecting data from large groups of people about their attitudes, opinions, and behaviors related to film products. For example, a survey could be about how the film influenced viewers' opinions on a particular topic. Longitudinal studies track the same people or group over an extended period, to see how their experience with film changes and develops over time. Content analysis examines the content of films, identifies their topics and messages, and assesses their potential of influencing the audiences. Ethnographic research helps us understand how a community perceives film and how cultural context affects that perception. --- <unk> <unk>: <unk> <unk> <unk> <unk> <unk> <unk>сследовани<unk> со<unk>иал<unk>но<unk>о во<unk>де<unk>стви<unk> кино ис<unk>ол<unk>у<unk>т ра<unk>н<unk>е метод<unk> дл<unk> о<unk>енки е<unk>о вли<unk>ни<unk> (Kaufman & Simonton, 2013). <unk>етод<unk> <unk>окусиру<unk>тс<unk> на ра<unk>н<unk> ас<unk>екта<unk> со<unk>иал<unk>но<unk>о во<unk>де<unk>стви<unk> кино, име<unk> ра<unk>н<unk>е <unk>редмет<unk> исследовани<unk>: содер<unk>ание <unk>ил<unk>ма или вос<unk>ри<unk>ти<unk> <unk>рителе<unk>. <unk>ак, о<unk>рос<unk> <unk>одра<unk>умева<unk>т с<unk>ор данн<unk> от <unk>ол<unk>и<unk> <unk>ру<unk> л<unk>де<unk> о<unk> и<unk> отно<unk>ении, мнени<unk> и <unk>оведении, св<unk>анн<unk> с <unk>родуктами кинемато<unk>ра<unk>а. <unk>а<unk>ример, о<unk>рос мо<unk>ет <unk>т<unk> о том, как <unk>ил<unk>м <unk>овли<unk>л на мнение <unk>рителе<unk> <unk>о о<unk>ределенно<unk> теме. <unk>он<unk>ит<unk>дн<unk>е исследовани<unk> отсле<unk>ива<unk>т одни<unk> и те<unk> <unk>е л<unk>де<unk> или <unk>ру<unk>у в те<unk>ение о<unk>ределенно<unk>о <unk>ериода времени, <unk>то<unk> увидет<unk>, как мен<unk>етс<unk> и ра<unk>виваетс<unk> и<unk> о<unk>т о<unk>ени<unk> с кино. <unk>онтент-анали<unk> и<unk>у<unk>ает содер<unk>ание <unk>ил<unk>мов, в<unk>вл<unk>ет и<unk> тем<unk> и соо<unk>ени<unk>, а так<unk>е о<unk>енивает <unk>отен<unk>иал вли<unk>ни<unk> на аудитори<unk>. <unk>тно<unk>ра<unk>и<unk>еские исследовани<unk> <unk>омо<unk>а<unk>т <unk>он<unk>т<unk>, как соо<unk>ество вос<unk>ринимает кино и как кул<unk>турн<unk> контекст вли<unk>ет на <unk>то вос<unk>ри<unk>тие. --- <unk>а<unk>вле<unk>ение Entertainment-driven --- <unk> <unk> The social effects of cinema at the social attitudes level are not linear; they involve many other related variables, and vary depending on the assessment method and indicators selected. However, some tools based on the methods used in social sciences have long been used in filmmaking. University of Birmingham professor Finola Kerrigan, in her book on film marketing (Kerrigan, 2017), lists the following tools in this category: concept testing, positioning studies, test screenings, focus group testing, advertising testing, and exit polls: • Concept testing -this step assesses the potential for commercialization and success in the market, so as to determine the film budget. This can be viewed as part of the production process, because it can influence decisions such as script development and actor selection. • Positioning studies -this type of research analyzes elements that can act as marketing assets of a film. Marketers test the potential profitability of a film based on audience response to the genre, cast, and specific characters. • Test screenings. There are two types of test screenings in the film industry: B2B -for business partners, and B2C -for ordinary viewers. In the first case, film screenings are held for film distributors, as part of private screenings or film festivals. Such tests are used to learn the partners' opinions, as well as to attract them to the project. In the case of test screenings for viewers, the target audience members involved are invited to watch the film before the final version. It is here that changes can be made to the film, such as replacing the ending in Rambo or changing the drawing of characters in the Pikachu film. The respondents are selected by social and demographic qualities and are interviewed before and after watching the film. During the screening, researchers observe their reactions, and some of the invited viewers participate in a focus group after the screening. • Testing of marketing materials -posters, trailers, postcards and advertisements can be tested in focus groups to make sure they match the cultural codes of the target audience and are understood by them. As Ruth Vasey, author of a film research guide (Vasey, 2008), writes, Hollywood's peculiarity lies in the dichotomous nature of its "show business," which combines the financial imperatives of "business" with the ephemeral appeal of "show." There is an entire genre of articles in The Hollywood Reporter discussing how the results of <unk>о<unk>иал<unk>н<unk>е <unk>ект<unk> кино на уровне со<unk>иал<unk>н<unk> установок не <unk>вл<unk>тс<unk> лине<unk>н<unk>ми, вкл<unk>а<unk>т в се<unk> мно<unk>о дру<unk>и<unk> со<unk>утству<unk>и<unk> <unk>еременн<unk>, так<unk>е они вар<unk>иру<unk>тс<unk> от с<unk>осо<unk>а о<unk>енки и и<unk>ранн<unk> индикаторов. <unk>днако су<unk>еству<unk>т инструмент<unk>, основанн<unk>е на метода<unk> со<unk>иал<unk>н<unk> наук, котор<unk>е давно ис<unk>ол<unk>у<unk>тс<unk> в кино<unk>рои<unk>водстве. <unk>ро<unk>ессор <unk>ирмин<unk>емско<unk>о университета <unk>инола <unk>ерри<unk>ан в свое<unk> кни<unk>е <unk>о маркетин<unk>у <unk>ил<unk>ма (Kerrigan, 2017) (Vasey, 2008), осо<unk>енност<unk> <unk>олливуда состоит в ди<unk>отоми<unk>еско<unk> <unk>рироде е<unk>о «<unk>оу-<unk>и<unk>неса», котора<unk> со<unk>етает <unk>инансов<unk>е им<unk>ератив<unk> «<unk>и<unk>неса» и <unk>емерну<unk> <unk>ривлекател<unk>ност<unk> «<unk>оу». <unk>у<unk>ествует <unk>ел<unk> <unk>анр стате<unk> в <unk>урнале The Hollywood Reporter, <unk>осв<unk>енн<unk> о<unk>су<unk>дени<unk> то<unk>о, как ре<unk>ул<unk>тат<unk> тестов<unk> <unk>ока<unk>ов test screenings influenced marketing decisions for a particular film. From the same magazine you can learn 3 that test screenings were kept as part of the mandatory film preparation program even during the COVID-19 pandemic, because their results could lead to both additional filming and higher chance for success. Overall, we can conclude that cinema is an attractive and promising area for research. At the same time, representatives of different scientific disciplines use their own conceptual apparatus and formulate their own research questions related to this area. In particular, for a sociologist, cinema can be of interest in terms of representing social situations and phenomena, as described in Vadim Radayev's sociological essays (Radayev, 2023). There is even a special academic discipline, Cinema Studies (Bateman & Schmidt, 2013), which treats film as a separate semiotic space. The junction between the viewer's perception and the film plot can also be studied from the perspective of psychology (Petrenko & Dedyukina, 2019). From a linguistic point of view, there have been interesting studies on fictional languages from films, such as the Na<unk>vi language from the Avatar universe (Frommer, 2009). --- MARKETING: ASSESSMENT METHODOLOGY AND CULTURAL STRATEGY It is worth noting that neither social science nor psychology offers any ready-made recipes on how to incorporate their methods into the filmmaking process, nor do they aim to do so. The methodologies that include conducting research and considering its results as part of project activities are primarily developed by the actual stakeholders -the filmmaking community itself. In the United States an association for impact entertainment has been established, the Social Impact Entertainment Society. This association has a specially developed guide to impact assessment -the Four Impact Dynamics 4. Notably, these four elements mirror the social impact ladder and appeal to the logic of the theory of change. The theory of change (Gladkikh & Talitskaya, 2022) is the basic element of the social project evaluation model, connecting the links in the chain of a project's social outcomes. The social impact ladder describes a chain of social outcomes in terms of successive changes in the perceptions, attitudes and behaviors of target audiences, as well as changes in communities and society at large. <unk>овли<unk>ли на маркетин<unk>ов<unk>е ре<unk>ени<unk> в отно<unk>ении то<unk>о или ино<unk>о <unk>ил<unk>ма. <unk> <unk>то<unk>о <unk>е <unk>урнала мо<unk>но у<unk>нат<unk> 3, <unk>то да<unk>е во врем<unk> <unk>андемии тестов<unk>е <unk>ока<unk> не <unk>ли искл<unk>ен<unk> и<unk> о<unk>ател<unk>но<unk> <unk>ро<unk>рамм<unk> <unk>од<unk>отовки <unk>ил<unk>ма, <unk>оскол<unk>ку и<unk> ре<unk>ул<unk>тат<unk> мо<unk>ут <unk>ривести как к <unk>роведени<unk> до<unk>олнител<unk>н<unk> с<unk>емок, так и к <unk>ов<unk>ени<unk> о<unk>идани<unk> на ус<unk>е<unk>. <unk> <unk>елом, мо<unk>но <unk>акл<unk>ит<unk>, <unk>то кино <unk>вл<unk>етс<unk> <unk>ривлекател<unk>но<unk> и <unk>ерс<unk>ективно<unk> о<unk>ласт<unk> дл<unk> исследовани<unk>. <unk>ри <unk>том, <unk>редставители ра<unk>н<unk> нау<unk>н<unk> дис<unk>и<unk>лин ис<unk>ол<unk>у<unk>т сво<unk> со<unk>ственн<unk> кон<unk>е<unk>туал<unk>н<unk> а<unk>арат и <unk>ормулиру<unk>т свои исследовател<unk>ские во<unk>рос<unk>, св<unk>анн<unk>е с <unk>то<unk> с<unk>еро<unk>. <unk> <unk>астности, дл<unk> со<unk>иоло<unk>а кино мо<unk>ет <unk>редставл<unk>т<unk> интерес с то<unk>ки <unk>рени<unk> ре<unk>ре<unk>ента<unk>ии со<unk>иал<unk>н<unk> ситуа<unk>и<unk> и <unk>влени<unk>, как <unk>то о<unk>ис<unk>ваетс<unk> в со<unk>иоло<unk>и<unk>ески<unk> о<unk>ерка<unk> <unk>адима <unk>адаева (<unk>адаев, 2023). <unk>у<unk>ествует отдел<unk>на<unk> академи<unk>еска<unk> дис<unk>и<unk>лина Cinema Studies (Bateman & Schmidt, 2013), котора<unk> рассматривает кино как самосто<unk>тел<unk>ное семиоти<unk>еское <unk>ространство. <unk>о<unk>но и<unk>у<unk>ат<unk> ст<unk>к вос<unk>ри<unk>ти<unk> <unk>рител<unk> с с<unk>етом <unk>ил<unk>ма с <unk>о<unk>и<unk>ии <unk>си<unk>оло<unk>ии (<unk>етренко & <unk>ед<unk>кина, 2019). <unk> то<unk>ки <unk>рени<unk> лин<unk>вистики, ест<unk> <unk>анимател<unk>н<unk>е слу<unk>аи исследовани<unk> в<unk>думанн<unk> <unk>ков и<unk> <unk>ил<unk>мов -на<unk>ример, <unk>ка <unk>а<unk>ви и<unk> киновселенно<unk> «<unk>ватар» (Frommer, 2009). --- <unk>: <unk> <unk> <unk> <unk> <unk> <unk>тоит ска<unk>ат<unk>, <unk>то ни со<unk>иал<unk>н<unk>е науки, ни <unk>си<unk>оло<unk>и<unk> не <unk>редла<unk>а<unk>т (и не <unk>реследу<unk>т тако<unk> <unk>ели) никаки<unk> <unk>отов<unk> ре<unk>е<unk>тов о том, как встраиват<unk> свои метод<unk> в <unk>ро<unk>есс кино<unk>рои<unk>водства. <unk>етодоло<unk>ии, котор<unk>е <unk> вкл<unk>али в се<unk> <unk>роведение исследовани<unk> и у<unk>ет е<unk>о ре<unk>ул<unk>татов как <unk>аст<unk> <unk>роектно<unk> де<unk>тел<unk>ности, ра<unk>ра<unk>ат<unk>ва<unk>тс<unk> в <unk>ерву<unk> о<unk>еред<unk> самими интересантамисоо<unk>еством кинемато<unk>ра<unk>истов. <unk> <unk> су<unk>ествует ассо<unk>иа<unk>и<unk> дл<unk> им<unk>акт-интерте<unk>нмента -одноименна<unk> Social Impact Entertainment Society. <unk> <unk>то<unk> ассо<unk>иа<unk>ии ест<unk> с<unk>е<unk>иал<unk>но ра<unk>ра<unk>отанное руководство <unk>о о<unk>енке им<unk>акта «<unk>ет<unk>ре динамики во<unk>де<unk>стви<unk>» (Four Impact Dynamics) 4 To plan for impact, you must have a strategy lined up for the film, its mission, its image of social change, its target audiences, and the appropriate messaging. --- • Artistic vision -the inner vision of the film and of what it will be like. It is determined by the director and creative crew selected by the director. --- • The vision of the impact is the external vision of what the film will do. It is determined by the entire project team, including those in charge of public relations. History, narrative, and culture come together through theory of change (Borum Chattoo, 2021) into a cultural strategy, in which practitioners collaborate with the entertainment industry executives, writers and producers to shape positive images of socially excluded communities and raise awareness of social issues, produce and distribute their content. The basic principle behind the cultural strategy is that entertaining stories can influence public opinion and perception by fostering cultural dialog and public participation, which is used in social policies aimed at stimulating social justice. Cultural strategy refers to social results in media projects and is built on the principles of education-entertainment, an area at the junction of impact entertainment and educational projects. It is an established industry with clear methodologies and principles. A textbook on this direction of filmmaking (Singhal et al., 2003) cites a motion picture series made in South Africa as an example. An 18-month study was conducted to produce the series, which included in-depth interviews with social activists and government officials, case studies of women surviving violence, and audience research. To implement the study's findings, it was decided to focus the campaign on overcoming the structural barriers that prevented women from seeking help. The script was based on the materials from the interviews and group discussions. For example, character positioning was lined up based on focus group surveys and audience feedback during test readings of the script, and social issues were "sewn into" the fabric of the drama. термина<unk> <unk>оследовател<unk>н<unk> и<unk>менени<unk> на уровне <unk>редставлени<unk>, отно<unk>ени<unk> и <unk>оведени<unk> <unk>елев<unk> аудитори<unk>, а так<unk>е и<unk>менени<unk> в <unk>и<unk>ни соо<unk>еств и о<unk>ества в <unk>елом. <unk>о<unk>ласно у<unk>ом<unk>нутому руководству, 4 уровн<unk> им<unk>акта в кул<unk>турном <unk>рои<unk>водстве вкл<unk>а<unk>т в се<unk>: <unk>стори<unk>, нарратив и кул<unk>тура о<unk>един<unk>тс<unk> <unk>осредством теории и<unk>менени<unk> (Borum Chattoo, 2021) в кул<unk>турну<unk> страте<unk>и<unk>, <unk>ри которо<unk> <unk>рактики сотрудни<unk>а<unk>т с руководител<unk>ми, <unk>исател<unk>ми и <unk>род<unk>серами индустрии ра<unk>вле<unk>ени<unk>, <unk>то<unk> <unk>ормироват<unk> <unk>оло<unk>ител<unk>н<unk>е о<unk>ра<unk> со<unk>иал<unk>но искл<unk>енн<unk> соо<unk>еств и <unk>ов<unk>ат<unk> осведомленност<unk> в <unk>асти со<unk>иал<unk>н<unk> <unk>ро<unk>лем, <unk>рои<unk>водит<unk> и рас<unk>ростран<unk>т<unk> сво<unk> контент. <unk>сновно<unk> <unk>рин<unk>и<unk>, <unk>ало<unk>енн<unk> в кул<unk>турно<unk> страте<unk>ии, <unk>акл<unk>аетс<unk> в том, <unk>то ра<unk>влекател<unk>н<unk>е истории мо<unk>ут ока<unk>ват<unk> вли<unk>ние на о<unk>ественное мнение и вос<unk>ри<unk>тие, с<unk>осо<unk>ству<unk> кул<unk>турному диало<unk>у и о<unk>ественному у<unk>асти<unk>, <unk>то ис<unk>ол<unk>уетс<unk> в со<unk>иал<unk>н<unk> <unk>олитика<unk>, на<unk>равленн<unk> на стимулирование со<unk>иал<unk>но<unk> с<unk>раведливости. <unk>ул<unk>турна<unk> страте<unk>и<unk> относитс<unk> к со<unk>иал<unk>н<unk>м ре<unk>ул<unk>татам в медиа<unk>роекта<unk> и строитс<unk> на <unk>рин<unk>и<unk>а<unk> education-entertainmentо<unk>ласти на <unk>ересе<unk>ении им<unk>акт-интерте<unk>нмента и о<unk>ра<unk>овател<unk>н<unk> <unk>роектов. <unk> рамка<unk> нее су<unk>еству<unk>т <unk>еткие методики и <unk>рин<unk>и<unk>, <unk>оскол<unk>ку она <unk>редставл<unk>ет со<unk>о<unk> с<unk>ормировав<unk>у<unk>с<unk> индустри<unk>. <unk> <unk>рестомати<unk>но<unk> кни<unk>е о<unk> <unk>том на<unk>равлении кино<unk>рои<unk>водства (Singhal et al., 2003) <unk>риводитс<unk> <unk>ример сериала и<unk> <unk>. <unk>л<unk> <unk>рои<unk>водства <unk>то<unk>о сериала <unk>ло <unk>роведено 18-мес<unk>ное исследование, которое вкл<unk>ало в се<unk> <unk>лу<unk>инн<unk>е интерв<unk> с со<unk>иал<unk>н<unk>ми активистами и <unk>редставител<unk>ми <unk>осу<unk>ре<unk>дени<unk>, исследование слу<unk>аев <unk>ен<unk>ин, <unk>острадав<unk>и<unk> от насили<unk>, а так<unk>е исследование аудитории. <unk> ка<unk>естве им<unk>лемента<unk>ии на<unk>одок исследовани<unk> <unk>ло ре<unk>ение с<unk>окусироват<unk> кам<unk>ани<unk> на том, <unk>то<unk> <unk>реодолет<unk> структурн<unk>е <unk>ар<unk>ер<unk>, котор<unk>е ме<unk>али <unk>ен<unk>инам о<unk>ра<unk>ат<unk>с<unk> <unk>а <unk>омо<unk>. <unk>атериал<unk> интерв<unk> и <unk>ру<unk>ов<unk> о<unk>су<unk>дени<unk> As an example of the use of Four Impact Dynamics, we can refer to the crew experience of the US motion picture The Hunting Ground. This example stands out due to its comprehensive use of the assessment tools for the study of the social results achieved, as well as its integration of the media project and research. The team used theory of change to mobilize student interest and engage university leadership in new practices (Dick & Ziering, 2016). The crew held 70 screenings at 33 universities. They also worked with student leaders at conferences and national meetings where the film was used to influence campus life. The project team used the Four Impact Dynamics methodology to assess the social impact of the film. They worked with human rights activists and university associations to conduct a survey about sexual violence on campuses among the students on all campuses across the country. The following indicators were developed: student reactions to the screenings and follow-up activities, university staff and students' requests for screenings, adoption of violence prevention training programs by universities, and implementation of new policies recommended based on the results of the survey. The results of the team's work showed increased interest among students and university administrators in the issues raised in the film, and led to the introduction of new policies to combat sexual violence among the students. The applicability of the Four Impact Dynamics methodology to assess social impact of a motion picture is recognized by the media impact assessment community and used to assess the effects of impact entertainment. Its advantage is consistency with the steps of the impact ladder and the project logic of the Theory of Change. Therefore, it can be used as the basis for designing and building an impact assessment system for motion pictures. However, it is also worth noting the limitations of Four Impact Dynamics: estimating impact as the final stage ис<unk>ол<unk>овалис<unk> дл<unk> на<unk>исани<unk> с<unk>енари<unk>. <unk>а<unk>ример, <unk>о<unk>и<unk>ионирование <unk>ерсона<unk>е<unk> в<unk>страивалос<unk> на основании о<unk>росов <unk>елев<unk> <unk>ру<unk> и о<unk>ратно<unk> св<unk>и аудитории в <unk>оде тестов<unk> <unk>иток с<unk>енари<unk>, а со<unk>иал<unk>н<unk>е <unk>ро<unk>лем<unk> «<unk>а<unk>ивалис<unk>» в осо<unk>енности драм<unk>.

Кино -один из любимых видов искусства у людей разного возраста. Так же, как и музыка, театр, архитектура, литература или живопись, кино безусловно оказывает воздействие на человека. При этом, оно может не только привлекать внимание к социальным проблемам, но и обладает потенциалом менять установки и модели поведения людей, а значит и мир в целом. Данная статья представляет собой обзор подходов к проведению исследований социального воздействия кино и поднимает ряд важных вопросов, связанных с текущим состоянием отрасли измерений импакта в кино. Cinema is one of the most popular art forms among people of all ages. Just like music, theater, architecture, literature or painting, cinema certainly has an impact on people. At the same time, it has the capability to not just draw attention to social problems, but also to change people's attitudes and behaviors -thus changing the world as a whole. This article provides an overview of various approaches to researching social impact of motion pictures and raises a number of important issues related to the current state of the impact assessment industry in motion pictures.Оценка социального импакта кино и практики доказательности в кинопроизводстве

motion picture is recognized by the media impact assessment community and used to assess the effects of impact entertainment. Its advantage is consistency with the steps of the impact ladder and the project logic of the Theory of Change. Therefore, it can be used as the basis for designing and building an impact assessment system for motion pictures. However, it is also worth noting the limitations of Four Impact Dynamics: estimating impact as the final stage ис<unk>ол<unk>овалис<unk> дл<unk> на<unk>исани<unk> с<unk>енари<unk>. <unk>а<unk>ример, <unk>о<unk>и<unk>ионирование <unk>ерсона<unk>е<unk> в<unk>страивалос<unk> на основании о<unk>росов <unk>елев<unk> <unk>ру<unk> и о<unk>ратно<unk> св<unk>и аудитории в <unk>оде тестов<unk> <unk>иток с<unk>енари<unk>, а со<unk>иал<unk>н<unk>е <unk>ро<unk>лем<unk> «<unk>а<unk>ивалис<unk>» в осо<unk>енности драм<unk>. <unk> ка<unk>естве <unk>римера с <unk>рименением Four Impact Dynamics мо<unk>но <unk>ривести о<unk>т команд<unk>, ра<unk>отав<unk>е<unk> над американским <unk>ил<unk>мом «<unk>она о<unk>от<unk>» (The Hunting Ground). <unk>н в<unk>дел<unk>етс<unk> <unk>о сте<unk>ени <unk>олно<unk>енности о<unk>ено<unk>н<unk> средств дл<unk> исследовани<unk> дости<unk>нут<unk> со<unk>иал<unk>н<unk> ре<unk>ул<unk>татов, а так<unk>е <unk>о сте<unk>ени инте<unk>ра<unk>ии медиа<unk>роекта и исследовани<unk>. <unk>оманда ис<unk>ол<unk>овала теори<unk> и<unk>менени<unk>, <unk>то<unk> мо<unk>или<unk>оват<unk> интерес студентов и вовле<unk> руковод ство университетов в нов<unk>е <unk>рактики (Dick & Ziering, 2016). <unk>оманда ор<unk>ани<unk>овала 70 <unk>ока<unk>ов в 33 университета<unk>. <unk>ни так<unk>е ра<unk>отали с лидерами студен<unk>ески<unk> ор<unk>ани<unk>а<unk>и<unk> на кон<unk>ерен<unk>и<unk> и на<unk>ионал<unk>н<unk> встре<unk>а<unk>, <unk>де <unk>ил<unk>м ис<unk>ол<unk>овалс<unk> дл<unk> вли<unk>ни<unk> на <unk>и<unk>н<unk> кам<unk>усов. <unk>оманда ис<unk>ол<unk>овала методоло<unk>и<unk> Four Impact Dynamics дл<unk> о<unk>енки со<unk>иал<unk>но<unk>о им<unk>акта <unk>ил<unk>ма. <unk>ни ра<unk>отали с <unk>раво<unk>а<unk>итн<unk>ми и университетскими ассо<unk>иа<unk>и<unk>ми дл<unk> <unk>роведени<unk> о<unk>роса о сексуал<unk>ном насилии на кам<unk>уса<unk>, <unk>роведенно<unk>о студентами во все<unk> студен<unk>ески<unk> <unk>ородка<unk> стран<unk>. <unk>ли ра<unk>ра<unk>отан<unk> следу<unk>ие индикатор<unk>: реак<unk>ии студентов на <unk>ока<unk> и <unk>оследу<unk>ие меро<unk>ри<unk>ти<unk>, <unk>а<unk>рос<unk> университетов на <unk>ока<unk> дл<unk> сотрудников и студентов, <unk>рин<unk>тие о<unk>у<unk>а<unk>и<unk> <unk>ро<unk>рамм <unk>о <unk>редотвра<unk>ени<unk> насили<unk> университетами и внедрение нов<unk> <unk>олитик, рекомендованн<unk> на основании ре<unk>ул<unk>тата о<unk>роса. <unk>е<unk>ул<unk>тат<unk> ра<unk>от<unk> команд<unk> <unk>ока<unk>али <unk>ов<unk>ение интереса студентов и руководства университетов к <unk>ро<unk>лемам, <unk>атронут<unk>м в <unk>ил<unk>ме, а так<unk>е <unk>ривели к введени<unk> нов<unk> <unk>олитик <unk>о <unk>ор<unk>е с сексуал<unk>н<unk>м насилием в студен<unk>еско<unk> среде. <unk>рименимост<unk> методоло<unk>ии Four Impact Dynamics дл<unk> о<unk>енки со<unk>иал<unk>но<unk>о им<unk>акта кино <unk>вл<unk>етс<unk> <unk>ри<unk>нанно<unk> в соо<unk>естве о<unk>енки им<unk>акта в медиас<unk>ере и ис<unk>ол<unk>уетс<unk> дл<unk> о<unk>енки им<unk>акт-<unk>ил<unk>мов. <unk>е <unk>реиму<unk>еством <unk>вл<unk>етс<unk> со<unk>ласованност<unk> со сту<unk>ен<unk>ми лестни<unk> во<unk>де<unk>стви<unk> и <unk>роектно<unk> ло<unk>ико<unk> теории и<unk>менени<unk>. <unk>аким о<unk>ра<unk>ом, она мо<unk>ет <unk>т<unk> ис<unk>ол<unk>ована как --- <unk>то<unk> с<unk>ланироват<unk> во<unk>де<unk>ствие, нео<unk>одимо имет<unk> в<unk>строенну<unk> страте<unk>и<unk> в отно<unk>ении <unk>ил<unk>ма, е<unk>о миссии, о<unk>ра<unk>а со<unk>иал<unk>н<unk> и<unk>менени<unk>, <unk>елев<unk> аудитори<unk> и соответству<unk>и<unk> мессед<unk>е<unk>. To plan for impact, you must have a strategy lined up for the film, its mission, its image of social change, its target audiences, and the appropriate messaging. results needs an overly complicated model. The scope of subjects and film formats to which this methodology (and the general implementation of the principles of impact entertainment) is applicable, is also limited. --- RUSSIA'S EXPERIENCE IN EVALUATING THE SOCIAL IMPACT OF CINEMA When discussing the social impact of cinema, it is important to clarify that not all studies cover this social aspect of film production. Social impact is a subject of research that falls into the evaluation of social programs and policies, while it is possible to consider the plot, language, and audience perceptions of a film outside the context of social impact. One Russian study that addresses social impact but does not conduct or implement evaluation of a film, instead focusing on creating the conditions for mapping, is a study by the Center for Social Projects "Platforma", supported by the National Media Group. The aim of the study was to develop a social impact index and a map of films having social impact, taking into account both domestic and foreign experience. The study generated a list of 28 films released in 2021-2022 and identified by experts as having social impact, as well as a coverage rating and an impact map of the films, based on 11 parameters reflecting the film's impact on the audience. In order to create a map of social impact films, we used the results of a mass quantitative survey of Russian cities with more than 100,000 population, in which respondents chose up to five films they had seen, and rated each film by 11 parameters reflecting its impact on the audience. The study concluded that social impact films should consider the balance between popularity and social message, and not alienate the viewer by evoking negative emotions. One of the challenges of social impact films is to engage and retain the viewer. Cinema should be entertaining and emotional, as well as provide answers to the viewers' questions and support them in a challenging situation. It is hard to argue that these findings were unknown prior to the study. Notably, the authors use an integrated approach that includes both a common marketing framework for social impact assessment and a set of some social research methods. Speaking about the differences between this study and the testing practices in film production mentioned above, we can highlight the following points: • the Platforma study is done post-factum, after the films have been released, and is not a part <unk>роектна<unk> основа дл<unk> <unk>остроени<unk> систем<unk> о<unk>енки им<unk>акта кино. <unk>ем не менее, стоит отметит<unk> и о<unk>рани<unk>ени<unk> Four Impact Dynamics: дл<unk> о<unk>енки им<unk>акта как коне<unk>но<unk>о <unk>та<unk>а во<unk>де<unk>стви<unk> тре<unk>уетс<unk> сер<unk>е<unk>ное усло<unk>нение модели. <unk>ак<unk>е, о<unk>рани<unk>енно<unk> <unk>вл<unk>етс<unk> та с<unk>ера с<unk>етов и <unk>орматов <unk>ил<unk>мов, в которо<unk> во<unk>мо<unk>но <unk>рименение <unk>то<unk> методоло<unk>ии и, в <unk>елом, внедрение <unk>рин<unk>и<unk>ов им<unk>акт-интерте<unk>нмента. --- <unk> <unk> <unk> <unk> <unk> <unk> <unk>ри о<unk>су<unk>дении со<unk>иал<unk>но<unk>о им<unk>акта кино ва<unk>но уто<unk>нит<unk>, <unk>то не все исследовани<unk> о<unk>ват<unk>ва<unk>т данн<unk> со<unk>иал<unk>н<unk> ас<unk>ект кино<unk>родук<unk>ии. <unk>о<unk>иал<unk>н<unk> им<unk>акт <unk>редставл<unk>ет со<unk>о<unk> <unk>редмет исследовани<unk> и<unk> о<unk>ласти о<unk>енки со<unk>иал<unk>н<unk> <unk>ро<unk>рамм и <unk>олитик, в то врем<unk> как мо<unk>но рассматриват<unk> с<unk>ет, кино<unk>к и вос<unk>ри<unk>тие <unk>рителе<unk> вне контекста со<unk>иал<unk>но<unk>о во<unk>де<unk>стви<unk>. <unk>реди росси<unk>ски<unk> исследовани<unk>, котор<unk>е о<unk>ра<unk>а<unk>тс<unk> к со<unk>иал<unk>ному им<unk>акту, но <unk>ри <unk>том не <unk>ровод<unk>т и не внедр<unk>т о<unk>енку <unk>ил<unk>мов, но со<unk>да<unk>т услови<unk> дл<unk> картировани<unk>, стоит в<unk>делит<unk> исследование <unk> «<unk>лат<unk>орма» <unk>ри <unk>оддер<unk>ке <unk>а<unk>ионал<unk>но<unk> <unk>едиа <unk>ру<unk>. <unk>ел<unk> исследовани<unk> стала ра<unk>ра<unk>отка индекса со<unk>иал<unk>но<unk>о вли<unk>ни<unk> и карт<unk> <unk>ил<unk>мов с со<unk>иал<unk>н<unk>м <unk>ектом, у<unk>ит<unk>ва<unk> как оте<unk>ественн<unk>, так и <unk>ару<unk>е<unk>н<unk> о<unk>т. <unk> рамка<unk> исследовани<unk> <unk>ли с<unk>ормирован<unk> с<unk>исок и<unk> 28 кинора<unk>от, в<unk>ед<unk>и<unk> в 2021-2022 <unk>ода<unk> и о<unk>ределенн<unk>е <unk>кс<unk>ертами как име<unk>ие со<unk>иал<unk>н<unk> <unk>ект, а так<unk>е ре<unk>тин<unk>, состо<unk>и<unk> и<unk> о<unk>вата и индекса вли<unk>ни<unk> <unk>ил<unk>ма, расс<unk>итанно<unk>о на основе 11 <unk>араметров, отра<unk>а<unk>и<unk> вли<unk>ние <unk>ил<unk>ма на <unk>рител<unk>. <unk>л<unk> то<unk>о, <unk>то<unk> составит<unk> карту <unk>ил<unk>мов с со<unk>иал<unk>н<unk>м <unk>ектом, <unk>ли ис<unk>ол<unk>ован<unk> данн<unk>е массово<unk>о коли<unk>ественно<unk>о о<unk>роса <unk>ородов <unk> с населением <unk>олее 100 т<unk>с. <unk>еловек, в <unk>оде которо<unk>о рес<unk>ондент<unk> в<unk>ирали максимум 5 <unk>ил<unk>мов и<unk> <unk>исла те<unk>, котор<unk>е они видели, и о<unk>енивали ка<unk>д<unk> <unk>ил<unk>м <unk>о 11 <unk>араметрам, отра<unk>а<unk>им вли<unk>ние <unk>ил<unk>ма на <unk>рител<unk>. <unk> ре<unk>ул<unk>тате исследовани<unk> <unk>ли сделан<unk> в<unk>вод<unk>, <unk>то со<unk>иал<unk>ное кино дол<unk>но у<unk>ит<unk>ват<unk> <unk>аланс ме<unk>ду <unk>о<unk>ул<unk>рност<unk> и со<unk>иал<unk>н<unk>м <unk>осланием, а так<unk>е не отталкиват<unk> <unk>рител<unk>, в<unk>ва<unk> не<unk>ативн<unk>е <unk>мо<unk>ии. <unk>дним и<unk> в<unk>овов со<unk>иал<unk>но<unk>о кино <unk>вл<unk>етс<unk> вовле<unk>ение и удер<unk>ание <unk>рител<unk>. <unk>ино дол<unk>но <unk>т<unk> ра<unk>влекател<unk>н<unk>м и <unk>мо<unk>ионал<unk>н<unk>м, а так<unk>е дават<unk> ответ<unk> на во<unk>рос<unk> <unk>рител<unk> и <unk>оддер<unk>иват<unk> е<unk>о в трудно<unk> ситуа<unk>ии. <unk>рудно утвер<unk>дат<unk>, <unk>то <unk>ти в<unk>вод<unk> <unk>ли неи<unk>вестн<unk> до <unk>роведени<unk> исследовани<unk>. <unk>риме<unk>ател<unk>но, <unk>то автор<unk> ис<unk>ол<unk>у<unk>т инте<unk>рированн<unk> <unk>од<unk>од, котор<unk> вкл<unk>ает как о<unk>и<unk> <unk>ре<unk>мворк и<unk> маркетин<unk>а дл<unk> о<unk>енки со<unk>иал<unk>но<unk>о во<unk>де<unk>стви<unk>, так и на<unk>ор некотор<unk> методов со<unk>иал<unk>но<unk>о исследовани<unk>. <unk>овор<unk> <unk>ро отли<unk>и<unk> <unk>то<unk>о исследовани<unk> от у<unk>оминав<unk>и<unk>с<unk> ранее <unk>рактик тестировани<unk> в рамка<unk> кино<unk>рои<unk>водства, мо<unk>но в<unk>делит<unk> следу<unk>ие момент<unk>: of the filmmaking process that could have an effect on the final appearance of the film; • the study is conducted without focusing on specific films; that is, it does not focus on the media effects, narrative features, and special effects of a specific film whose crew is involved in the study design; • the study methods focus on the interpretation by an undetermined sample of viewers, so that there remains room for improvement of the approach by refining the sample of surveys; • the study does not include test screenings, an important element of applied research tools; • the only data source used is a survey, without the qualitative perspective of researching the plot of the film. However, the most important point worth noting in the Platforma study is that social impact is the final element in a complex chain of social effects, which cannot be proven by referring to post-factum social surveys of an undetermined number of individuals, without identifying the target groups, the social issues raised by the films, the effects of specific films on communities. The study still fails to map films by their social effect. As you read the report, the question arises: What is the social significance of the selected films? We can conclude that there is room for expanding the available set of tools for more effective research into the social significance of Russian cinema and incorporating them into the film production process. --- POSSIBLE APPLICATIONS OF THE ASSESSMENT OF THE SOCIAL IMPACT OF CINEMA Impact entertainment is the use of cinema to influence public opinion and awareness. In addition to entertaining viewers, social impact --- <unk> <unk> <unk> <unk> <unk> <unk>м<unk>акт-интерте<unk>нмент -<unk>то ис<unk>ол<unk>ование кинемато<unk>ра<unk>а дл<unk> во<unk>де<unk>стви<unk> на о<unk>ественное мнение и со<unk>нание. <unk>онтент с со<unk>иал<unk>н<unk>м во<unk>де<unk>ствием не тол<unk>ко ра<unk>влекает <unk>рителе<unk>, но <unk>м<unk>акт <unk>редставл<unk>ет со<unk>о<unk> <unk>редмет исследовани<unk> и<unk> о<unk>ласти о<unk>енки <unk>ро<unk>рамм и <unk>олитик, в то врем<unk> как мо<unk>но рассматриват<unk> с<unk>ет, кино<unk>к и вос<unk>ри<unk>тие <unk>рителе<unk> вне <unk>то<unk>о контекста. Impact is a subject of research that falls into the evaluation of programs and policies, while it is possible to consider the plot, language, and audience perceptions of a film outside this context. content also promotes a discussion of important topics, which can lead to increased awareness on various global issues and a change in social attitudes through the broadcasting of behavioral patterns. Impact entertainment is an integral part of the entertainment industry, existing within the framework of mainstream content production. The social impact of cinema is explored through numerous methods borrowed from the social sciences. Both qualitative and quantitative methods can be used to cover different aspects of the film's social impact. For example, surveys and longitudinal studies deal with viewers' attitudes and behaviors, content analysis deals with film content, and ethnographic studies address the cultural context of film perception. The perspectives of content analysis and perception can be combined using, for example, experimental psychosemantic methods. Different social research methods are used in film industry to focus on different aspects of the films' social impact. Applied research tools such as concept testing, positioning studies, focus group testing, test screenings, advertising testing, and exit polls are used during the film production. They help assess a film's potential in the marketplace and make decisions about script development and actor selection. Marketing tools such as project planning can be used to integrate social impact studies into the filmmaking process. The Russian experience of assessing social impact of a film is based on applying a combined framework of marketing and social science methods. The distinguishing feature of this study is that it is done after the films have been released, and is not a part of the production process that could have had an effect on the final appearance of the film. No test screenings were conducted as part of the study, and only an unspecified sample survey is used. The use of applied tools to determine the social impact of Russian films involves the integration of a research component into the film production process. Moreover, developing a cultural strategy that involves social impact assessment and design can contribute to greater social impact. This is how the image of evidencebased approach in the film industry emerges. The experience we have reviewed allows us to conclude that the field of social impact assessment in the film industry has a large number of growth points. One possibility is to develop an assessment model based on the project's theory of change. It involves describing the assessment tasks, indicators and methods at each stage of the development, creation and final evaluation of a film product. Accordingly, there are three types of evaluation: predictive evaluation during the preparation phase, monitoring after the launch of the film project, and impact assessment in the period of the anticipated achievement of social effects. As part of the evidence-based filmmaking strategy developed, research tools are adapted to different types of materials -script, intermediate materials (storyboards), finished content, informational materials, etc. и <unk>о<unk>вол<unk>ет о<unk>судит<unk> ва<unk>н<unk>е тем<unk>, котор<unk>е мо<unk>ут <unk>ривести к <unk>ов<unk>ени<unk> осведомленности л<unk>де<unk> о ра<unk>ли<unk>н<unk> <unk>ро<unk>лема<unk> в мире и к и<unk>менени<unk> о<unk>ественн<unk> установок <unk>ере<unk> трансл<unk>и<unk> <unk>оведен<unk>ески<unk> моделе<unk>. <unk>м<unk>акт-интерте<unk>нмент <unk>вл<unk>етс<unk> неот<unk>емлемо<unk> <unk>аст<unk> ра<unk>влекател<unk>но<unk> индустрии, су<unk>еству<unk> в рамка<unk> ме<unk>нстримно<unk>о <unk>рои<unk>водства контента. <unk>о<unk>иал<unk>н<unk> им<unk>акт кино исследуетс<unk> с <unk>омо<unk> мно<unk>о<unk>исленн<unk> методов и<unk> со<unk>иал<unk>н<unk> наук. <unk>ни мо<unk>ут <unk>т<unk> ка<unk>ественн<unk>ми или коли<unk>ественн<unk>ми, а так<unk>е ра<unk>отат<unk> с ра<unk>н<unk>ми ас<unk>ектами со<unk>иал<unk>но<unk>о во<unk>де<unk>стви<unk> кино. <unk>а<unk>ример, о<unk>рос<unk> и лон<unk>ит<unk>дн<unk>е исследовани<unk> ра<unk>ота<unk>т с мнени<unk>ми и <unk>оведением <unk>рителе<unk>, контент-анали<unk> -с содер<unk>анием <unk>ил<unk>ма, а <unk>тно<unk>ра<unk>и<unk>еские исследовани<unk> -с кул<unk>турн<unk>м контекстом вос <unk>ри<unk>ти<unk> <unk>ил<unk>мов. <unk>л<unk> о<unk>единени<unk> <unk>ерс<unk>ектив анали<unk>а контента и е<unk>о вос<unk>ри<unk>ти<unk> мо<unk>ут <unk>т<unk> ис<unk>ол<unk>ован<unk>, на<unk>ример, <unk>кс<unk>ериментал<unk>н<unk>е метод<unk> <unk>си<unk>осемантики. <unk>у<unk>еству<unk>т ра<unk>ли<unk>н<unk>е метод<unk> со<unk>иал<unk>н<unk> исследовани<unk> в кино, котор<unk>е <unk>окусиру<unk>тс<unk> на ра<unk>н<unk> ас<unk>екта<unk> со<unk>иал<unk>но<unk>о во<unk>де<unk>стви<unk> кино. <unk>о врем<unk> <unk>рои<unk>водства <unk>ил<unk>ма ис<unk>ол<unk>у<unk>тс<unk> <unk>рикладн<unk>е исследовател<unk>ские инструмент<unk>, такие как <unk>роверка кон<unk>е<unk>ии, исследование <unk>о<unk>и<unk>ионировани<unk>, тестирование <unk>окус-<unk>ру<unk>ами, тестов<unk>е <unk>ока<unk>, тестирование реклам<unk> и о<unk>рос<unk> на в<unk>оде. <unk>ни <unk>омо<unk>а<unk>т о<unk>енит<unk> <unk>отен<unk>иал <unk>ил<unk>ма на р<unk>нке и <unk>ринимат<unk> ре<unk>ени<unk> о ра<unk>ра<unk>отке с<unk>енари<unk> и в<unk>оре актеров. <unk>нструмент<unk> <unk>роектно<unk>о <unk>ланировани<unk> и<unk> маркетин<unk>а мо<unk>ут <unk>т<unk> ис<unk>ол<unk>ован<unk> дл<unk> встраивани<unk> исследовани<unk> со<unk>иал<unk>но<unk>о им<unk>акта в <unk>ро<unk>есс кино<unk>рои<unk>водства. <unk>ассмотренн<unk> росси<unk>ски<unk> о<unk>т о<unk>енки со<unk>иал<unk>но<unk>о во<unk>де<unk>стви<unk> кино основан на <unk>рименении <unk>ре<unk>мворка и<unk> маркетин<unk>а и методов и<unk> со<unk>иал<unk>н<unk> наук. <unk>тли<unk>ител<unk>на<unk> <unk>ерта <unk>то<unk>о исследовани<unk> <unk>акл<unk>аетс<unk> в том, <unk>то оно <unk>роводитс<unk> <unk>осле с<unk>емок <unk>ил<unk>мов и не <unk>вл<unk>етс<unk> <unk>аст<unk> кино<unk>рои<unk>водства, котора<unk> мо<unk>ла <unk> <unk>овли<unk>т<unk> на о<unk>лик <unk>ил<unk>ма. <unk> рамка<unk> исследовани<unk> не <unk>роводилос<unk> тестов<unk> <unk>росмотров, а ис<unk>ол<unk>уетс<unk> тол<unk>ко о<unk>рос не о<unk>ределенно<unk> в<unk>орки. <unk>с<unk>ол<unk>ование <unk>рикладн<unk> инструментов дл<unk> о<unk>ределени<unk> со<unk>иал<unk>но<unk>о вли<unk>ни<unk> оте<unk>ественно<unk>о кино <unk>ред<unk>ола<unk>ает внедрение исследовател<unk>ско<unk> составл<unk>е<unk> в кино<unk>рои<unk>водство. <unk>олее то<unk>о, ра<unk>ра<unk>отка кул<unk>турно<unk> страте<unk>ии с о<unk>енко<unk> и <unk>роектированием со<unk>иал<unk>но<unk>о во<unk>де<unk>стви<unk> мо<unk>ет с<unk>осо<unk>ствоват<unk> дости<unk>ени<unk> <unk>ол<unk>е<unk>о со<unk>иал<unk>но<unk>о <unk>екта. <unk>ак во<unk>никает о<unk>ра<unk> дока<unk>ател<unk>но<unk>о <unk>од<unk>ода в киноиндустрии. <unk>ассмотренн<unk> о<unk>т <unk>о<unk>вол<unk>ет <unk>акл<unk>ит<unk>, <unk>то о<unk>ласт<unk> о<unk>енки со<unk>иал<unk>но<unk>о им<unk>акта в кинемато<unk>ра<unk>е о<unk>ладает <unk>ол<unk>им <unk>ислом то<unk>ек роста. <unk>дно<unk> и<unk> во<unk>мо<unk>носте<unk> <unk>вл<unk>етс<unk> ра<unk>ра<unk>отка модели о<unk>енки на основании теории и<unk>менени<unk> <unk>роекта. <unk>на <unk>ред<unk>ола<unk>ает о<unk>исание <unk>ада<unk>, индикаторов и методов о<unk>енки на ка<unk>дом и<unk> <unk>та<unk>ов ра<unk>ра<unk>отки, со<unk>дани<unk> и <unk>инал<unk>но<unk> о<unk>енки <unk>родукта кинемато<unk>ра<unk>а. <unk>оответственно, в<unk>дел<unk>тс<unk> три вида о<unk>енки: <unk>ро<unk>ности<unk>еска<unk> о<unk>енка на <unk>та<unk>е <unk>од<unk>отовки, мониторин<unk> It should be noted that Russia currently lags behind in the field of evidence and has only a superficial level of discussion regarding the assessment of film products. The term "impact media" is far more popular than what is being done at the real scale to rigorously and reasonably assess film products. Self-assessment, expert review by the project team, and other approaches that are subjective and prone to wishful thinking are more common than conducting objective research and studying audience opinions and feedback. In this context, the phenomenon of impact washing is created, as players begin to add the word "impact" to the names of their companies or positions. For example, we now have "impact directors" instead of ordinary directors, which, however, does not lead to any significant changes in their work. We assume that once the market is saturated with experts capable of conducting research of sufficient complexity, it will become indecent to talk about creative products being immeasurable, or about the possibility of using the concept of "impact" in relation to simple social effects, or about drawing conclusions based on a sample of a few people. <unk>осле <unk>а<unk>уска кино<unk>роекта и о<unk>енка во<unk>де<unk>стви<unk> в <unk>ериод <unk>ред<unk>ола<unk>аемо<unk>о дости<unk>ени<unk> со<unk>иал<unk>н<unk> <unk>ектов. <unk> рамка<unk> в<unk>строенно<unk> страте<unk>ии дока<unk>ател<unk>но<unk>о кино<unk>рои<unk>водства ис<unk>ол<unk>у<unk>тс<unk> исследовател<unk>ские инструментарии, ада<unk>тированн<unk>е <unk>од ра<unk>ли<unk>н<unk>е ти<unk> материалов -с<unk>енари<unk>, <unk>роме<unk>уто<unk>н<unk>е материал<unk> (раскадровки), <unk>отов<unk> контент, ин<unk>орма<unk>ионн<unk>е материал<unk> и т. <unk>. <unk>ео<unk>одимо отметит<unk> нали<unk>ие в <unk>оссии ла<unk>а в о<unk>ласти дока<unk>ател<unk>но<unk> <unk>а<unk> и <unk>овер<unk>ностно<unk>о уровн<unk> дискуссии относител<unk>но о<unk>енки <unk>родуктов кинемато<unk>ра<unk>а. <unk>о<unk>ул<unk>рност<unk> термина «им<unk>актмедиа» <unk>на<unk>ител<unk>но <unk>рев<unk>ает в реал<unk>ном мас<unk>та<unk>е то, <unk>то делаетс<unk> дл<unk> стро<unk>о<unk> и о<unk>основанно<unk> о<unk>енки <unk>родуктов кинемато<unk>ра<unk>а. <unk>амоо<unk>енка, <unk>кс<unk>ерти<unk>а командо<unk> <unk>роекта и дру<unk>ие <unk>од<unk>од<unk>, св<unk>анн<unk>е с ни<unk>ко<unk> о<unk>ективност<unk> и в<unk>соко<unk> сте<unk>ен<unk> риска в<unk>дат<unk> <unk>елаемое <unk>а су<unk>еству<unk>ее, <unk>олее рас<unk>ространен<unk>, <unk>ем <unk>роведение о<unk>ективн<unk> исследовани<unk> и и<unk>у<unk>ение мнени<unk> и от<unk>вов аудитории. <unk> <unk>том контексте <unk>о<unk>вл<unk>етс<unk> <unk>еномен им<unk>акт-во<unk>ин<unk>а: и<unk>роки на<unk>али до<unk>авл<unk>т<unk> слово «им<unk>акт» к на<unk>вани<unk>м свои<unk> ком<unk>ани<unk> или дол<unk>носте<unk>. <unk>а<unk>ример, <unk>о<unk>вилис<unk> не <unk>росто ре<unk>исс

Кино -один из любимых видов искусства у людей разного возраста. Так же, как и музыка, театр, архитектура, литература или живопись, кино безусловно оказывает воздействие на человека. При этом, оно может не только привлекать внимание к социальным проблемам, но и обладает потенциалом менять установки и модели поведения людей, а значит и мир в целом. Данная статья представляет собой обзор подходов к проведению исследований социального воздействия кино и поднимает ряд важных вопросов, связанных с текущим состоянием отрасли измерений импакта в кино. Cinema is one of the most popular art forms among people of all ages. Just like music, theater, architecture, literature or painting, cinema certainly has an impact on people. At the same time, it has the capability to not just draw attention to social problems, but also to change people's attitudes and behaviors -thus changing the world as a whole. This article provides an overview of various approaches to researching social impact of motion pictures and raises a number of important issues related to the current state of the impact assessment industry in motion pictures.Оценка социального импакта кино и практики доказательности в кинопроизводстве

о<unk>ективност<unk> и в<unk>соко<unk> сте<unk>ен<unk> риска в<unk>дат<unk> <unk>елаемое <unk>а су<unk>еству<unk>ее, <unk>олее рас<unk>ространен<unk>, <unk>ем <unk>роведение о<unk>ективн<unk> исследовани<unk> и и<unk>у<unk>ение мнени<unk> и от<unk>вов аудитории. <unk> <unk>том контексте <unk>о<unk>вл<unk>етс<unk> <unk>еномен им<unk>акт-во<unk>ин<unk>а: и<unk>роки на<unk>али до<unk>авл<unk>т<unk> слово «им<unk>акт» к на<unk>вани<unk>м свои<unk> ком<unk>ани<unk> или дол<unk>носте<unk>. <unk>а<unk>ример, <unk>о<unk>вилис<unk> не <unk>росто ре<unk>иссер<unk>, а им<unk>актре<unk>иссер<unk>, <unk>то, однако, не <unk>риводит к су<unk>ественн<unk>м и<unk>менени<unk>м в и<unk> де<unk>тел<unk>ности. <unk>редставл<unk>етс<unk>, <unk>то с нас<unk>ением р<unk>нка <unk>кс<unk>ертами, с<unk>осо<unk>н<unk>ми <unk>роводит<unk> исследовани<unk> достато<unk>но<unk> сло<unk>ности, станет не<unk>рили<unk>н<unk>м <unk>оворит<unk> о неи<unk>меримости твор<unk>ески<unk> <unk>родуктов, о до<unk>устимости ис<unk>ол<unk>овани<unk> <unk>он<unk>ти<unk> «им<unk>акт» в отно<unk>ении <unk>рост<unk> со<unk>иал<unk>н<unk> <unk>ектов, а так<unk>е делат<unk> в<unk>вод<unk> на основании в<unk>орки и<unk> нескол<unk>ки<unk> <unk>еловек. --- <unk> <unk>

Кино -один из любимых видов искусства у людей разного возраста. Так же, как и музыка, театр, архитектура, литература или живопись, кино безусловно оказывает воздействие на человека. При этом, оно может не только привлекать внимание к социальным проблемам, но и обладает потенциалом менять установки и модели поведения людей, а значит и мир в целом. Данная статья представляет собой обзор подходов к проведению исследований социального воздействия кино и поднимает ряд важных вопросов, связанных с текущим состоянием отрасли измерений импакта в кино. Cinema is one of the most popular art forms among people of all ages. Just like music, theater, architecture, literature or painting, cinema certainly has an impact on people. At the same time, it has the capability to not just draw attention to social problems, but also to change people's attitudes and behaviors -thus changing the world as a whole. This article provides an overview of various approaches to researching social impact of motion pictures and raises a number of important issues related to the current state of the impact assessment industry in motion pictures.Оценка социального импакта кино и практики доказательности в кинопроизводстве

)1!+$!3>01/+$+$4!%73! &31/%+'$27+0!.3%533$!+$%3&93$%+'$2!'&!0'1+*+32!/$(!203*+;+*!73/1%7!')%*'632<unk>!M'5393&8!6'&3! ';!%7323!2%)(+32!/&3!$33(3(!%'!/((&322!73/1%7!+$3@)/1+%+32!()3!%'!%73!*'6013>!2:2%362!';! +$%3&/*%+'$2!'**)&&+$4!.3%533$!%73!2'*+/1!(3%3&6+$/$%2!';!73/1%7!/$(!%73!+$%3&93$%+'$2! 2%)(+3(!H<unk>GN/60'!3%!/1<unk>!PUUZV!=7'62'$!PUU<unk>J<unk>!!!!!! #$!%73'&:A(&+93$!39/1)/%+'$28!*'6013>+%:!+2!*/0%)&3(!.:!2%)(:+$4!7'5!%73!(+;;3&3$%! %73'&3%+*/1!31363$%2!%7/%!*'60'23!%736!/&3!+$%3&%5+$3(!HI%/63!PUU]J!/$(!.:!&3+$%34&/%+$4! %73!*'$%3>%!/2!/!?3:!31363$%!+$!%73!0&'()*%+'$!';!')%*'632<unk>!=73!%5'!6/+$!.&/$*732!';!%7+2! /00&'/*7!+$*1)(3!=73'&+32!';!N7/$43!/$(!b3/1+2%![9/1)/%+'$2<unk>!D'%7!/00&'/*732!7+471+47%!%73! +60'&%/$*3!';!)$(3&2%/$(+$4!*'$%3>%8!/2!+%!+2!+603&/%+93!%'!/%%&+.)%+$4!*/)23!/$(!*&+%+*+K3! PUU<unk>J<unk>!!!! =7+2!0/03&!3>01'&32!2'63!';!%73!*'6013>+%+32!2)&&')$(+$4!%73!1+$?!.3%533$!)&./$!&3$35/1! /$(!')%*'632!+$!73/1%7!/$(!73/1%7!+$3@)/1+%+32!)2+$4!/!%73'&:A(&+93$!/00&'/*7<unk>!='!('!%7+28!/! b3/1+2%!/00&'/*7!+2!6'&3!2)+%/.13!/2!53!/+6!%'!&3;+$3!3>+2%+$4!%73'&+32!/$(!%'!. f)/1+%/%+93!/$(!@)/$%+%/%+93!63%7'(2!53&3!)23(!%'!2%)(:!%73!3;;3*%2!';!%73!E3+47.')&7''(2! F/5!'$!%73!73/1%7!/$(!73/1%7!+$3@)/1+%:!';!$3+47.')&2<unk>!=7+2!*'$2+2%3(!';!%5'!0).1+273(! 2%)(+32!5+%7!%73!;+&2%!)2+$4!*'$*30%!6/00+$48!/!@)/1+%/%+93!63%7'(8!%'!3>01'&3!%73! 03&*30%+'$2!';!$3+47.')&2!+$!%5'!+$%3&93$3(!$3+47.')&7''(2!H'$3!+$!%73!*+%:!*3$%&3!/$(! /$'%73&!+$!%73!03&+073&+32J<unk>!!T/&%+*+0/$%2!53&3!/2?3(!%'!(32*&+.3!/$:!*7/$432!%7/%!7/(! '**)&&3(!+$!%73+&!$3+47.')&7''(2!+$!&3*3$%!:3/&2!/$(!573%73&!%7323!*7/$432!53&3!0'2+%+93! '&!$ +$%3&93$%+'$8!57+13!*'60/&+$4!%736!%'!/!4&')0!';!$'$A+$%3&93$3(!$3+47.')&7''(2!5+%7!/! 2+6+1/&!2'*+'A3*'$'6+*!2%/%)2!HB37(+0/$/7!3%!/1<unk>!PU"]J<unk>!<unk>)&!&32)1%2!+$(+*/%3!%7/%!.'%7! 5'63$!/$(!63$!1+9+$4!+$!+$%3&93$3(!$3+47.')&7 ''(2!7/(!+60&'93(!231! 0&'9+(3!39+(3$*3!'$!%73!0'2+%+93!3;;3*%2!'$!%73!73/1%7!/$(!73/1%7!3@)+%:!';!&32+(3$%2!1+9+$4! +$!+$%3&93$3(!$3+47.')&7''(2<unk>!X!%73'&:A(&+93$!/00&'/*7!*/$!.3!)23(!%'!*'6.+$3!.'%7!';! %7323!2%)(+32!/1'$4!5+%7!%73!3>+2%+$4!1+%3&/%)&3!+$!'&(3&!%'!(3931'0!%73!*/)2/1!0/%75/:2!';! 7'5!203*+;+*!)&./$!&3$35/1!0&'R3*%2!*/$!/;;3*%!%73!5311.3+$4!';!&32+(3$%2!/2!(32*&+.3(!+$! 1/%3&!23*%+'$2<unk>!!!! #19<unk>(L'(&3069<unk>15@I7M70?(1:9(I70N(D91F990(52D@0(2909F@I(@04(:9@I1:(( M3&3!53!)$*'93&!%73!*'6013>+%:!';!%73!*/)2/1!0/%75/:2!+$9'193(!.3%533$!%7323!0&'4&/62! /$(!73/1%7!+$3@)/1+%+32<unk>!=7323!/&3!0&323$%3(!+$!%73!;'&6!';!/!*'$*30%)/1!;&/635'&?!HO+4)&3! "J!&30&323$%+$4!/!./2+*!0&'4&/6!%73'&:!';!%73!2%302!$33(3(!;'&!/$!)&./$!&3$35/1! +$%3&93$%+'$!%'!+60/*%!73/1%7!/$(!73/1%7!+$3@)/1+%:!')%*'632<unk>!C3!%73$!;'*)2!')&!/%%3$%+'$! '$!%5'!?3:!0/%75/:28!/2!+$(+*/%3(!'$!%73!6/08!$3*322/&:!%'!/*7+393!%7'23!')%*'632<unk>!!! O+4)&3!"<unk>!X!*'$*30%)/1!;&/635'&?!';!)&./$!&3$35/1!/$(!+%2!3;;3*%2!'$!73/1%7!/$(!73/1%7! +$3@)/1+%:!!!!! "P =73!6'(31!2%/&%2!5+%7!)&./$!&3$35/1!+$%3&93$%+'$2!/((&322+$4!%73!/&3/2!';8!"4.$%/+,! %)56+$&63/&36(7!$#/%+,!%)&(*6+&%#)!/$(!(/#)#1%/!*6#8&4!Hb'.3&%2!PUUUV!B/*W&34'&!PU"UJ<unk>!#$! %7+2!6'(31!53!(36'$2%&/%3!)&./$!&3$35/1!0&'4&/62!/$(!*7/$432!+$!%73!$3+47.')&7''(!/2! /!57'13!%7/%!('!$'%!+$*1)(3!7')2+$4!(+63$2+'$28!57+*7!6/:!*'$2+2%!';!&31'*/%+$4!$3+47.')&2!.'%7!%360'&/&+1:!'&!03&6/$3$%1:<unk>!T&'R3*%2!+$!%7323!%7&33!/&3/2!)1%+6/%31:!13/(!%'! +60&'9363$%!+$!,%'(+-%,%&.8!57+*7!3$*'60/2232!*7/&/*%3&+2%+*2!%7/%!5')1(!6/?3!03'013! 5/$%!%'!1+93!+$!/!$3+47.')&7''(!+$*1)(+$4!3$9+&'$63$%/1!*'$(+%+'$28!2/;3%:!/$(!23*)&+%:8! %&/$20'&%/%+'$8!/32%73%+*28!&30)%/%+'$!';!%73!$3+47.')&7''(8!3*'$'6+*!'00'&%)$+%+32!/$(! 2'*+/1!*'732+'$!Hb:(+$!3%!/1<unk>!PU"PJ8!/11!0'%3$%+/11:!1+$?3(!%'!0'2+%+93!73/1%7!/$(!(3*&3/23(! 73/1%7!+$3@)/1+%+32<unk>!!! C7+13! 1+93/.+1+%:! +2! ';%3$! 6'&3! /22'*+/%3(! 5+%7! %73! 07:2+*/1! *'$(+%+'$2! ';! %73! $3+47.')&7''(8! %73! $#/%#0(/#)#1%/! 1+2(3"! +$*1)(32! ;/*%'&2! %7/%! +$;1)3$*3! %73! 2'*+/1! 3$9+&'$63$%! /%! 9/&+')2! 139312! +$*1)(+$4! 3601':63$%! 2%/%)28! 3()*/%+'$! 13931! /$(! 43$3&/1! 7')2+$4!/$(!1+9+$4!*'2%28!/11!';!57+*7!7/93!.33$!1+$?3(!%'!73/1%7!+$3@)/1+%+32!HD3&$/&(!3%!/1<unk>! PUUcJ<unk>!-&./$!&3$35/1!0&'4&/62!/&3!';%3$!(32+4$3(!%'!/((&322!2'63!';!%7323!+22)32!%7&')47! 0&'R3*%2!2)*7!/2!%73!32%/.1+2763$%!';!3601':63$%!*3$%&328!43$(3&!2)00'&%!0&'4&/628!/$(! 0'1+*+32!/+63(!/%!0&'%3*%+$4!0&'03&%:!9/1)32<unk>!I)*7!0&'R3*%2!&32)1%!+$!+$*&3/23(!'00'&%)$+%+32! %7/%!0&'9+(3!&32')&*32!$33(3(!.:!$3+47.')&2!/$(!%73+&!;/6+1+32!%'!13/&$8!5'&?!/$(!.3$3;+%! ;&'6! 2'*+/1! &31/%+'$27+028! /11! /;;3*%+$4! 73/1%7! ')%*'632! HD'&&311! 3%! /1<unk>! PU"QV! E'&%7&+(43! S! O&336/$!PU""J<unk>!!! M'5393&8!5+%7!%73!+60&'9363$%!';!1+93/.+1+%:8!'&!@)/1+%:!';!$3+47.')&7''(8!%73!2'*+'A 3*'$'6+*!6/?3)0!*/$!/1%3&!5+%7!$35*'63&2!.3+$4!/%%&/*%3(!%'!%73!$35!;3/%)&328!57+13! *'2%2!/22'*+/%3(!5+%7!7')2+$4!/$(!1+;32%:13!*/$!+$*&3/23!&32)1%+$4!+$!%73!43$%&+;+*/%+'$!';! +$(+9+()/12!HX&./*+!S!=/0/(/AD3&%31+!PU"PV!Y3/&$2!3%!/1<unk>!PUUZJ<unk>!T''&!73/1%7!')%*'632!7/93!.33$!/22'*+/%3(!%'!2'*+/1!$3%5'&?!(+2&)0%+'$2!()3!%'!;'&*3(!&31'*/%+'$!.3*/)23!';!+$*&3/23(! 7')2+$4!9/1)32!/$(!1+9+$4!3>03$232!()3!%'!+$*&3/2+$4!*'2%2!/22'*+/%3(!5+%7!&32')&*32!/$(! 23&9+*32!HM/4/$!3%!/1<unk>!"ZZ<unk>V!Y3/&$2!S!B/2'$!PU"QJ<unk>!L+201/*3(!&32+(3$%2!%3$(!%'!.3!';!1'53&! 2'*+'A3*'$'6+*!0'2+%+'$!/$(!%'!6'93!%'!6'&3!(30&+93(!$3+47.')&7''(2!&32)1%+$4!+$! 5'&23$+$4!*'$(+%+'$2!57+13!%73!/93&/43!73/1%7!2%/%)2!';!%73+&!0&39+')2!$3+47.')&7''(2!+2! 3>03*%3(!%'!+60&'93!5+%7!%73!6+4&/%+'$!';!7+473&!+$*'63!+$(+9+()/12!%7/%!/&3!/.13!%'!/;;'&(! $35!*'$(+%+'$2!HY3/&$2!3%!/1<unk>!PUUZJ<unk>!=73&3;'&38!0'0)1/%+'$!%)&$'93&!+2!/$!+60'&%/$%!9/&+/.13! %'!*'$%&'1!;'&!573$!1''?+$4!/%!73/1%7!/$(!73/1%7!+$3@)/1+%:!')%*'632!0'2%!/$!)&./$!&3$35/1! +$%3&93$%+'$<unk>!!! E'$3%7313228!+%!+2!+60'&%/$%!%'!$'%3!%7/%!1+93/.+1+%:!/$(!0'0)1/%+'$!%)&$'93&!('!$'%! &30&323$%!%73!0&'4&/6!+%231;!.)%!&/%73&!/!*7/+$!';!393$%2!5+%7+$!%73!0&'4&/6!%7/%!6/:! 6'(+;:!')%*'632!HT/52'$!S!I&+(7/&/$!PU"UJ<unk>!! O+$/11:8!%73!+60'&%/$*3!';!/#)&(9&!+$!%73!39/1)/%+'$!';!*'6013>!+$%3&93$%+'$2!7/2!.33$! 7+471+47%3(!.:!9/&+')2!/)%7'&2!HL'6+$4)3KAD3&R'$!3%!/1<unk>!PUU<unk>V!PU"]J<unk>!!#$!&3203*%2!%'!)&./$! &3$35/18!*'$%3>%!*/$!+$*1)(3!9/&+')2!31363$%2!2)*7!/2!43'4&/07+*!1'*/%+'$8!0'1+%+*/1!*1+6/%3! /$(!%73!3*'$'6+*!2+%)/%+'$!';!%73!*+%:8!*')$%&:!'&!&34+'$!/%!%73!%+63!%73!+$%3&93$%+'$!5/2! )$(3&%/?3$<unk>!#$!/((+%+'$8!%+63!1/4!;'&!*7/$432!+$!73/1%7!')%*'632!+2!/12'!*&+%+*/1!%'!*'$2+(3&!! "] 573&3!;'&!3>/60138!/%!/!'$3A:3/&!0'2%A+$%3&93$%+'$8!%73!%+63!1/4!6/:!$'%!.3!2);;+*+3$%!%'! 233!2+4$+;+*/$%!*7/$432!+$!73/1%7!')%*'632!1+?3!'.32+%:!'&!*/&(+'A9/2*)1/&!(+23/23<unk>!=73! *'$2+(3&/%+'$!';!%73!*'$%3>%!+$!57+*7!%73!+$%3&93$%+'$!5/2!+601363$%3(!+2!%73&3;'&3! 3223$%+/1!+$!'&(3&!%'!3>01/+$!0'%3$%+/1!)$+$%3$(3(!*'$23@)3$*32!';!%73!0&'4&/68!/2!53!5+11! (+2*)22!1/%3&!+$!%73!3>/60132!53!7/93!0&'9+(3(<unk>!!! O&'6!')&!*'$*30%)/1!;&/635'&?!/$(!')&!3>03&+3$*32!5+%7!%73!E3+47.')&7''(2!F/5!2%)(:8! 53!7+471+47%!H/2!233$!+$!O+4)&3!"J!%5'!0'%3$%+/1!0/%75/:2!)23(!%'!3>01/+$!7'5!)&./$! &3$35/1!0&'4&/62!/&3!*'$$3*%3(!%'!73/1%7!/$(!73/1%7!+$3@)/1+%+32<unk>!=73!;+&2%!'$3!1+32!.3%533$!%73!)&./$!&3$35/1!0&'R3*%!/&3/2!/$(!1+93/.+1+%:!/$(!(32*&+.32!7'5!36-+)!6()(8+,! %)/6(+$($!+//($$!+):!+'+%,+-%,%&.!#5!6($#36/($!+):!#""#6&3)%&%($!%)!)(%*4-#364##:$<unk>!=73! 23*'$(!'$3!1+32!.3%533$!1+93/.+1+%:!/$(!73/1%7!')%*'632!/$(!(32*&+.32!7'5!)(%*4-#36$! 13$&!3$(!+):!+:+"&!&4($(!/4+)*($!%)!#6:(6!&#!+/4%('(!#"&%1+,!4(+,&4!+):!4(+,&4!%)(;3+,%&.! #3&/#1($<unk>!!! =73!;+&2%!0/%75/:8!/2!0&39+')21:!63$%+'$3(8!+2!./23(!'$!%73!3>%3$2+93!1+%3&/%)&3!1+$?+$4! 2'*+/1!/$(!07:2+*/1!3$9+&'$63$%2!%'!73/1%7!')%*'632!/$(!%73!$'%+'$!%7/%!.:!+60&'9+$4! %7323!;/*%'&28!)&./$!&3$35/1!+60&'932!73/1%7!/$(!73/1%7!+$3@)/1+%:<unk>![9+(3$*3!%'!2)00'&%! %7+2!+$*1)(32!%73!1+$?2!.3%533$!';!+60&'93(!/**322!%'!&32')&*32!/$(!+60&'93(!5311.3+$4! HY3/&$2!3%!/1<unk>!PUUZV!E'&%7&+(43!S!O&336/$!PU""J<unk>!O'&!3>/60138!+$+%+/%+932!';!32%/.1+27+$4!! PQ O+4)&3!Q<unk>!N/)2/1!0/%75/:2!;'&!%73!&30/+&!';!2%&33%2!/$(!2+(35/1?2!! #$!')&!0&39+')2!2%)(:8!2%&33%!*'$(+%+'$2!53&3!/$!/&3/!';!*'$*3&$!;'&!&32+(3$%2!+$!%7+2! $3+47.')&7''(8!/$(!%73:!03&*3+93(!%7+2!+$%3&93$%+'$!/2!7/9+$4!93&:!0'2+%+93!/$(!+60'&%/$%! 3;;3*%2!'$!%73!5311.3+$4!';!&32+(3$%2!HB37(+0/$/7!3%!/1<unk>!PU"QJ<unk>!E3+47.')&2!;&'6!b'@)3%32! /4&33(!%7/%!'$3!';!%73!%'0!6'2%!+60'&%/$%!/$(!0'2+%+93!*7/$432!+$!%73+&!$3+47.')&7''(!+$!! P] %:!&32+(3$%2<unk>!<unk>1(6%/+)!F#36)+,!#5!@"%:(1%#,#*.8! "UZHPJ8!"<unk>APU]<unk>! ( D'&R/8!j<unk>!S!B)>+8!m<unk>!HPUU"J![20/+!Tg.1+*,!N+)%/%!#!N+)%/(/$+/<unk>!L+0)%/*+k!(3!D/&*31'$/8! D/&*31'$/<unk>! ( D'&&311!N8!L'6+$4)3KAD3&R'$!O8!T/2/&+$!B#8!O3&&/$('!j<unk>8!b'71;2!#<unk>!S!E3.'%!B<unk>!HPUUUJ!I'*+/1! +$3@)/1+%+32!+$!73/1%7!&31/%3(!.37/9+')&2!+$!D/&*31'$/<unk>!F!@"%:(1%#,!G#113)%&.!H(+,&4!PUUU8! _]H"J8!P]AQU<unk>!! D'&&311!N8!T'$2Aa+4)32!B<unk>!B'&&+2'$!j<unk>!S!L+3K![<unk>!HPU"QJ<unk>!O/*%'&2!/$(!0&'*32232!+$;1)3$*+$4! 73/1%7!+$3@)/1+%+32!+$!)&./$!/&3/2<unk>!F!@"%:(1%#,!G#113)%&.!H(+,&48!<unk>cH_J8!Q<unk>ZAZ"<unk>( ( N/938!D<unk>!S!N)&%+28!I<unk>!HPUU"J!L3931'0+$4!/!0&/*%+*/1!4)+(3!%'!/22322!%73!0'%3$%+/1!73/1%7! +60/*%!';!)&./$!&343$3&/%+'$!2*7363<unk>!I6#1#&%#)!J!@:3/+&%#)8!<unk>8!"PA"<unk>! ( N73$!M<unk>=<unk>!HPU"PJ!=73 +$!<unk>1(3&!X()1%2,!=73!&'13!';!03&*3+93(! 2/;3%:8!2'*+/1!*'732+'$8!/$(!5/1?/.+1+%:<unk>!F#36)+,!#5!A6-+)!H(+,&47!<unk>Z8!]UcA]"<unk>!! N'&:$8!N<unk>F<unk>I<unk>8!E'/?328!F<unk>!X<unk>8!C32%+$38!N<unk>L<unk>!S!I*7&'%3&8!L<unk>N<unk>!HPU""J!X!2:2%36/%+*!&39+35!';! %73'&:A(&+93$!39/1)/%+'$!0&/*%+*3!;&'6!"ZZU!%'!PUUZ<unk>!<unk>1(6%/+)!F#36)+,!#5!@'+,3+&%#)<unk>!QPHPJ,! "ZZAPP<unk>!!! N&/5;'&(8!L<unk>8!=+603&+'8!X<unk>8!W+132AN'&%+8!D<unk>8!D/118!Y<unk>8!M)638!N<unk>8!b'.3&%28!b<unk>8!3%!/1<unk>!HPUU<unk>J!L'! ;3/%)&32!';!0).1+*!'03$!20/*32!9/&:!/**'&(+$4!%'!$3+47.')&7''(!2'*+'A3*'$'6+*!2%/%)2p! H(+,&4!J!I,+/(8!"]H]J

%&'()*+%,#$-"*+./0.$*#.123.40*5-13--(*/&5.(*&-(-3.$6*2-.$12*.(7* 2-.$12*'(-8/.$'14*/0'()*.*12-%&497&':-(*.##&%.+2***** #$%&'(')!*+&,-.'('& '/01 /!2('!*'(3'

/1)/%+'$!0&/*%+*3!;&'6!"ZZU!%'!PUUZ<unk>!<unk>1(6%/+)!F#36)+,!#5!@'+,3+&%#)<unk>!QPHPJ,! "ZZAPP<unk>!!! N&/5;'&(8!L<unk>8!=+603&+'8!X<unk>8!W+132AN'&%+8!D<unk>8!D/118!Y<unk>8!M)638!N<unk>8!b'.3&%28!b<unk>8!3%!/1<unk>!HPUU<unk>J!L'! ;3/%)&32!';!0).1+*!'03$!20/*32!9/&:!/**'&(+$4!%'!$3+47.')&7''(!2'*+'A3*'$'6+*!2%/%)2p! H(+,&4!J!I,+/(8!"]H]J8!<unk>ZA<unk>ZQ<unk>!! L30/&%/63$%!(3!T'1l%+*/!=3&&+%'&+/1!+!<unk>.&32!Tg.1+@)32!HPUUZJ!F/!F13+!(3!D/&&+2,!-$/!/0'2%/! *'113*%+9/!03&!1/!*'732+k!2'*+/1<unk>!W3$3&/1+%/%!(3!N/%/1)$:/8!I0/+$<unk>!! L343$8!B<unk>!S!W/&*+/8!B<unk>!HPU"PJ!=73!=&/$2;'&6/%+'$!';!%73!qD/&*31'$/!B'(31G,!X$!X$/1:2+2!';! N)1%)&/18!-&./$!b343$3&/%+'$!/$(!W'93&$/$*3<unk>!D)&(6)+&%#)+,!F#36)+,!#5!A6-+)!+):!B(*%#)+,! C&3:%($8!Q<unk>H_J8!"UPPA"UQ<unk>!!! PZ!! L'6+$4)3KAD3&R'$8!O<unk>8!D'&&3118!N<unk>8!b'(&+4)3KAI/$K8!B<unk>8!B/&lAL311G<unk>16'8!B<unk>8![2$/'1/8!I<unk>8! T&+3%'AI/1*3(8!B<unk>L<unk>8!L)@)38!#<unk>8!b'(&+4'8!B<unk>T<unk>8!W&)0'!(3!L3%3&6+$/$ -&./$!&343$3&/%+'$!/2!/!0).1+*!73/1%7!+$%3&93$%+'$<unk>!F#36)+,!#5!C#/%+,! D)&(6'()&%#)T!V4(#6.!+):!I6+/&%/(8!"ZHQJ8!Q<unk>A_"<unk>!! B/%732'$!X<unk>8!L&35!Y<unk>!S!N)66+$4!j<unk>!HPUUZJ!N'6013>+%:8!39/1)/%+'$!/$(!%73!3;;3*%+93$322!';! *'66)$+%:A./23(!+$%3&93$%+'$2!%'!&3()*3!73/1%7!+$3@)/1+%+32<unk>!H(+,&4!I6#1#&%#)!F#36)+,!#5! <unk>3$&6+,%+8!PU8!PP"APP<unk>!! B37(+0/$/78!b<unk>8!B/16)2+8!L<unk>8!B)$%/$3&8!N<unk>8!S!D'&&3118!N<unk>!HPU"QJ!X$!39/1)/%+'$!';!/$!)&./$! &3$ ';!3>0'2)&3!%'!$/%)&/1!3$9+&'$63$%!'$!73/1%7! +$3@)/1+%+32,!/$!'.23&9/%+'$/1!0'0)1/%+'$!2%)(:<unk>!W+)/(&8!QcP8!"<unk>__A"<unk>U<unk>!! B'&&+2'$8!L<unk>8!T3%%+*&358!B<unk>!S!=7'62'$8!M<unk>!HPUUQJ!C7/%!/&3!%73!6'2%!3;;3*%+93!5/:2!';! +60&'9+$4!0'0)1/%+'$!73/1%7!%7&')47!%&/$20'&%!+$%3&93$%+'$2p![9+(3$*3!;&'6!2:2%36/%+*! &39+352<unk>!!F#36)+,!#5!@"%:(1%#,#*.!+):!G#113)%&.!H(+,&4!_c,!QPcAQQQ<unk>!! E31d1'8!<unk>!HPU"UJ!=73!*7/113$432!';!)&./$!&3$35/1<unk>!=3$!1322'$2!;&'6!%73!N/%/1/$!3>03&+3$*3<unk>! <unk>)+,%$(!C#/%+,8!"]H"ZcJ8<unk>_Ac"_!! E'&%7&+(438!B<unk>![<unk>8!S!O&336/$8!F<unk>8!HPU""J<unk>!-&./$!T1/$$+$4!/$(!M3/1%7![@)+%:<unk>!F#36)+,!#5! A6-+)!H(+,&4!<unk>8!_<unk>PA_Zc<unk>!! <unk>GN/

%&'()*+%,#$-"*+./0.$*#.123.40*5-13--(*/&5.(*&-(-3.$6*2-.$12*.(7* 2-.$12*'(-8/.$'14*/0'()*.*12-%&497&':-(*.##&%.+2***** #$%&'(')!*+&,-.'('& '/01 /!2('!*'(3'

Introduction Online Social Networks (OSN) thrive on getting users to share information about themselves. One of the ways humans build relationships is through sharing of personal information to build trust, therefore it is unsurprising that they use social networks to do so. OSNs encourage sharing personal information by adding prompts like "Wish Pat a happy birthday today!" or "Tell your friends about your new shoe purchase". Even LinkedIn, which is an OSN focused on professional networking, has prompts about birthdays. Historically, date of birth (DOB) was considered as private information; this is why it has been used widely in authentication. Even today, some organizations such as phone companies and banks still use DOB as one of several authentication questions when users phone in (Kaur et al. 2020;Lee et al. 2020;Robinson 2019). The treatment of DOB as private data can also be seen in the European GDPR regulation where DOB is legally considered to be personal data (Commission 2020) and is also regularly reported in data breach reports to the public as important personal information that may or may not have been lost during the breach (Jackson, Vanteeva, and Fearon 2019). Still, the use of DOB in authentication these days is significantly lower compared to a couple of decades ago before the spread of social media. Early research by Rabkin (Rabkin 2008) surveyed the password recovery mechanisms of 20 banks with the aim of showing how vulnerable they were. They found that DOB is used in the process of password recovery of some of these banks, and highlighted that this information can be inferred using public data found in OSNs, which opens the accounts to automatic attacks. This early study was the first to highlight that the trend of how DOB is seen is changing in the era of social media, shifting from a fact that is shared only with close friends and family to a fact that is publicly shared with complete strangers online. In this paper, we characterize the disclosure of birthdays/DOBs on the Twitter platform and reactions of the users to such celebrations. We explore the tension between birthdays being open celebrations and birth dates as private information and investigate behavior by measuring the disclosure of birthday wishes on Twitter and users' reactions to them; and attitudes through a user survey asking Twitter users about their thoughts on the topic. Our main research question is: Do social media users see their birthdays/DOB as private information anymore? More precisely we investigate the following sub-research questions: RQ1 How many birthday wishes on Twitter are posted publicly? What percentage of those indicate the mentioned person's exact date of birth? RQ2 Do protected accounts, who are theoretically more privacy concerned, see less disclosure of birth days/dates than public accounts? RQ3 How do Twitter users react to these public wishes? Is there a difference in the reactions of public and protected accounts? RQ4 How aware/comfortable are Twitter users with having their birth day/date disclosed online? To answer our research questions, we collected over 18 million tweets/retweets mentioning "happy birthday" over 45 days. Of those, 2.8 million tweets directly mention one non-verified user account. The number of tweets shows just how many birthdays are being disclosed over a single OSN1. Interestingly, we found that over 66K of these tweets likely disclosed the age of almost 50K unique users (e.g. "Happy 16th birthday @user"), which makes easy to directly infer the exact date of birth (DOB) of the user. Some of the mentioned accounts were "protected", where users have explicitly indicated that their tweets should be kept private and only visible to an approved list of people; yet these users still had their birthday, and sometimes DOB, publicly disclosed by their followers. While public account holders' ages are tweeted more often than the age of the users with protected accounts, still over 5K protected account holders' DOB were likely exposed within the 45 days of our collection period. Finally, our user survey measured Twitter users' opinion/awareness on birth day/date exposures through celebration on the platform. 48% of the participants were comfortable with others tweeting publicly about their birthdays including their ages. Our findings indicate that indeed Twitter users are publicly expressing birthday wishes, sometimes also exposing the full DOB, even for protected accounts. The majority of the users are reacting positively to having their birthday and DOB disclosed publicly. These findings show that the view of social media platform designers is the closest to the reality; a large number of users do not think that birthday and DOB are sensitive information anymore. This finding should be taken into account by the organizations that still use this piece of information in their authentication process. --- Background People use OSNs to share their experiences, interact with each other, as well as, to gain social capital (Ellison et al. 2011). Some people may use OSNs in professional contexts (Mahrt, Weller, and Peters 2014) and to build reputation (Syn and Oh 2015), while others may use them to seek social support from others (Yin et al. 2016). Users maintain weak ties by interaction on social media (Vitak 2014) and celebrating birthdays is one of the popular ways to do that (Viswanath et al. 2009). However, these interactions can leak information if they are public. Users can actively utilize privacy settings provided by the platforms to control the flow of their information (Keküllüoglu, Vaniea, and Magdy 2022). However, individuals' privacy is connected to their networks in OSNs (boyd 2012; Amon et al. 2020), which can lead to unintentional disclosures (Keküllüoglu, Magdy, and Vaniea 2020). Most of the time, users cannot directly control these disclosures, and they might not even realize the reach of it (Bernstein et al. 2013). While there is research on utilizing these privacy leaks to get information on individual users (Jurgens, Tsvetkov, and Jurafsky 2017), to our knowledge, there is no research focused on the characterising birthday and DOB disclosures by networks or the users' privacy concerns regarding this situation. --- Awareness of Data Sharing OSN users want to both share information and control its reach. However, given that users underestimate audience size (Bernstein et al. 2013), do not fully understand the visibility to third parties (King, Lampinen, and Smolen 2011), and have difficulty understanding that information shared online can result in other types of information being inferred (Acquisti and Gross 2006), an argument could be made that truly controlling information flow is quite challenging for an OSN user. OSN users' privacy awareness has increased from the early days of Facebook (Tsay-Vogel, Shanahan, and Signorielli 2018), particularly in regards to their understanding of the visibility of their data to the public and the visibility to their connections. However, even with these improvements, people still struggle to understand how broad the reach of their posts are. Bernstein et al. (Bernstein et al. 2013) looked at the true audience reach of 220,000 Facebook users as well as surveyed users about their perceived audience. They found that the imagined size of the audience was only 27% of the true size. --- Sharing Other People's Data The term "networked privacy" is pitched by boyd (boyd 2012) to reflect on the collective aspects of privacy in social media. Even when individuals protect their private information, their networked relations may disclose them. An example would be a photo which is uploaded by a user onto an OSN and then tagged with the people in it. Each member of this collective (photo taker, photo subjects, event space owner) has some privacy stake in the photo and therefore its management is a collective issue. However, most OSNs give the right to manage the privacy settings of a content only to the uploader. While this enables users to control their self-disclosures, they have no say over what others share about them. Trusting privacy protection to the users' network might not be sufficient (Pu and Grossklags 2017) and priming the network might even backfire, leading them to share more (Amon et al. 2020). Some research has looked into how to handle this type of situations automatically on behalf of the user (Keküllüoglu, Kökciyan, and Yolum 2018;Kökciyan, Yaglikci, and Yolum 2017), but it is still in the proof-of-concept stage. Unlike offline interactions, there is a certain permanence to online posts and interactions. Hence, users might want to edit or delete some of their posts (Y<unk>lmaz et al. 2021). However, some parts of the interactions can stay in the platform and leak information (Keküllüoglu, Magdy, and Vaniea 2020). For example, Twitter keeps the replies to a protected/deleted tweet visible in the platform. Some social media users choose to pause or stop using their accounts for various reasons including being "in tune with" themselves (Baumer, Sun, and Schaedler 2018), productivity (Grandhi, Plotnick, and Hiltz 2019), religious practices (Schoenebeck 2014), as well as privacy protection (Grandhi, Plotnick, and Hiltz 2019;Lampe, Vitak, and Ellison 2013). However, users' friends can still share posts that lead to privacy violations (Amon et al. 2020;Lampe, Vitak, and Ellison 2013). Even without sharing any posts, networks of non-users could still disclose enough information to create shadow profiles (Garcia et al. 2018). --- Finding and Inferring Personal Data Several studies have looked at the types of private information shared on OSNs as well as how to use that data to infer information which has not been shared. Mao et al. (Mao, Shuai, and Kapadia 2011) looked at the information shared deliberately on Twitter; they find that events such as vacation, illness, and drinking are shared. One type of information can also be used to infer more information. For example, burglars can use the above types of tweets to know that a user's house is vacant (e.g. PleaseRobMe.com). Insurance companies could also increase their premiums according to the severity of illnesses shared in the tweets (Heussner 2009). Jain et al. (Jain, Jain, and Kumaraguru 2013) studied phone numbers posted publicly on Twitter and Facebook in India. They found that most of the phone numbers were intentionally posted by their owners. However, they were also able to use the phone numbers to find the name of the owner, voter ID, family details, age, home address, and father's name. By adding the numbers to WhatsApp they were able to get further information such as their US numbers, relationship status, and so on. A user's connections on OSNs can also be used to learn quite a bit about the user, even if that user has "locked down" their account using settings. Jurgens et al. (Jurgens, Tsvetkov, and Jurafsky 2017) showed that an analysis of tweets mentioning a user is enough to determine their gender, age, religion, diet and personality traits. Kekulluoglu et al. (Keküllüoglu, Magdy, and Vaniea 2020) found that life events such as marriage, graduation, surgery recovery of a user can be inferred by only looking at the replies sent to them. Analysis of OSN friend networks has shown that knowing information about a user's friends is sufficient to accurately infer attributes such as age, gender, location, political orientation, and sexual orientation (Al Zamal, Liu, and Ruths 2012;Zheleva and Getoor 2009;Jernigan and Mistree 2009;Jurgens 2013;Aldayel and Magdy 2019). Magdy et al. (Magdy et al. 2017) inferred users' gender and age with 92% accuracy from their network interaction and comments, which allowed them to spot "fake" accounts that might be used for catfishing on adult social networks. Similarly, Garcia et al. (Garcia et al. 2018) found that using networks of people on Twitter, allows detecting the physical location of a user with median error of 68.7km, and identify the city the user lives in with 32% accuracy. --- Birth Dates in the Authentication Process Best practices advise against using knowledge-based questions in the authentication process (Grassi, Garcia, and Fenton 2017) which includes asking for the birth dates as security questions. Usage of DOB should especially be avoided since it is considered easily discoverable information (OPC 2016), especially with the spread of social media (Rabkin 2008;Irani et al. 2011). Even with these warnings, some organizations such as banks (Kaur et al. 2020;Murdoch and Anderson 2010;Smyth 2010), wireless carriers (Lee et al. 2020), and email service providers (Li, Mehta, and Yang 2017;Al Maqbali and Mitchell 2018) still use DOB while authenticating users. Against the best security practices (ENISA 2020), birth dates are also commonly used by people while constructing passwords (Brown et al. 2004;Bonneau, Preibusch, and Anderson 2012;Wang et al. 2019). People also use DOB in their PINs which make them easier to be predicted. According to Bonneau et.al. (Bonneau, Preibusch, and Anderson 2012), lost or stolen wallets will lead thieves to correctly guess PINs up to 8.9% of the time and the primary reason for that is the identification cards with DOB found in the wallets. --- Data Collection and Analysis Methodology We collected tweets containing the words "happy" and "birthday" and then analyzed them in regards to the amount of disclosure, type of account (public, protected), age disclosure, and engagement by the mentioned person. In the following, we describe our data collection and annotation methodology that enables our initial quantitative analysis. --- Collecting Tweets We used the Twitter streaming API (Twitter 2019) to collect public tweets in real time. We filtered for English language tweets that contained both the words "happy" and "birthday", resulting in only tweets containing both those words, but not necessarily in consecutive order. We collected tweets for 45 days between January and March 2019 resulting in nearly 18 million tweets and retweets. We filtered out the 11 million retweets as we are only interested in the initial birthday mention. In addition, we filtered out 2.3 million tweets that had no mentioned account along with 630K tweets mentioning multiple users where it was unclear whose birthday was disclosed. For some accounts, Twitter will verify the identity of the account holder and add a blue tick beside their user name. These tend to be owned by public figures rather than average users. Hence, we also removed 1 million tweets that mentioned verified accounts as well as 3K tweets where the user mentioned themselves. After this cleaning process, we ended up with a set of around 2.8 million tweets that use the words "happy" and "birthday", as well as mention only one non-verified account. We refer to this dataset as "BD" tweets dataset. Two days after the last tweet was collected, we batch processed all BD tweets by: 1) identifying any mentioned accounts, 2) checking if the mentioned accounts are public or protected. We excluded 44K tweets where the mentioned account could not be reached (e.g. deleted or suspended) at the time of processing. We then labeled each tweet in the collection with two labels in terms of: 1. account status of the mentioned accounts: either mentioning a public account (mPublic) or mentioning a protected one (mProtected). 2. tweet conversation type: either a reply or, directed to a user. A reply is in response to an existing parent tweet, such as when a user tweets about their own birthday and a follower replies. A tweet directed to a user is a new tweet without a parent, mentioning the user (e.g. "@username Happy 21st birthday"). These are likely to be wishes by friends of the mentioned user who already know their birthday. Its worth mentioning that protected account tweets are visible to their approved followers only and cannot be retweeted or quoted by other users. However, if a public account replies to a protected account's tweet, the reply can be seen publicly. This is also the case for any tweet mentioning a protected account. Protected accounts can also be mentioned by non-followers. --- Gathering Reactions on BD Tweets We also measure the reaction of the mentioned user accounts to birthday tweets. For measuring the reactions, we collected the engagement of the mentioned account with these tweets either by replying to or liking the tweet. Inspecting all these tweets individually to check interaction with them was impractical, due to Twitter API limitations. Thus, we randomly sampled a set of 10,000 tweets (5,000 mPublic, 5,000 mProtected) from the BD dataset. To avoid bias, we took samples equally from each day. We refer to this sample of our dataset as BD-react. Twenty days after the last tweet in our main data set was collected, we measured the amount of engagement tweets in BD-react had experienced. The average time to reply to a tweet in our set was 3.5 hours, so we are fairly confident that the majority of engagement will have happened within our 20+ day time period. For protected accounts, it is not possible to see if the account has interacted with a tweet via API. However, how many protected accounts have retweeted or liked a tweet is visible via Twitter's user interface (UI). Thus, for mProtected tweets, we scraped if they have been liked or retweeted by a protected account. Note that we can only understand whether a protected account interacted with the tweet but we cannot get the usernames of those users to check whether the interaction was by the mentioned protected account. To determine if any of the mentioned accounts had replied to the tweet, we collected the tweets of the mentioned account and searched for replies to our recorded tweet. Doing so was necessary because the Twitter API does not have a method to collect replies to a particular tweet. After this process, each tweet in our BDreact was labeled as being liked, retweeted, and/or replied to (in the case of mPublic) by the mentioned user. --- Gathering Birthdays on Profiles While our main focus is on birthday disclosure by others, users themselves might be self-disclosing the information publicly on their profile pages. In this case, the user may be fine with birthday exposure and others might feel encouraged to tweet about their publicly visible birthday. To see whether users shared their birthday or date information in their profiles, we collected the public birthday information from each account. This information could be gathered for both public and protected accounts. We collected the selfdisclosed birthday information for all accounts in our BDreact collection. We have applied this process a few months after our initial collection, which led to losing access to some of the accounts due to deletion, deactivation, or suspension; resulting in getting the information of only 4364 public and 4159 protected unique accounts. --- Tweets Disclosing User's Age Some tweets explicitly mention the age of the person. An example from our BD dataset (username anonymized): "Happy 40th Birthday to @username. Have a great day and night". If the age is combined with the date the tweet was posted, it becomes trivially possible to reconstruct the full birth date. To understand the scope of this disclosure, we further analyzed the tweets to extract those containing a twodigit number between 10 and 99. We looked for all instances of two digits on their own or in combination with an ordinal indicator (i.e. "st", "nd", "rd", "th"). We selected the 10-99 range, because numbers below 10 might mean something other than the age, and technically Twitter does not allow users younger than 13 years old. Similarly, few people live to over 99, so the number of errors in this numeric range is expected to be large compared to the number of true ages. The percentage of tweets that contain two digits with our criteria are shown in Table 1. To verify if the tweets containing two-digit numbers are referring to the user's age, we manually labeled a random sample of of 4000 tweets (2000 mPublic, 2000 mProtected) from the tweets that had two-digit numbers. We took samples equally from each day as we did with BD-react. We refer to this sample of tweets as BD-age. For the annotation, we used the online crowdsourcing platform Appen2 participants were asked: "Can we tell that this person: @username has their age disclosed in the tweet?" where @username was replaced with the mentioned person's account from the actual tweet. Each tweet was judged by three trusted workers and we used majority voting to label tweets. A test-set of 64 pre-labeled tweets, that we manually annotated, was provided for quality control of the annotation. If a worker got more than 20% of the pre-labeled tweets incorrect, their annotations were discarded. The final inter-annotator agreement rate was 93%. --- Ethical Considerations This work aims to measure the prevalence and practice of birthday celebrations on Twitter, a question that is most practically answered by sampling public tweets from Twitter itself. However, such sampling comes with some ethical conundrums since we are using data for research purposes that was initially shared presumably to connect with others. We are also aware that some users publicly mention protected accounts, potentially sharing information, like DOB, that the protected user would prefer to keep private. To limit potential negative impacts of our work we take several steps. We only collect publicly available tweets. In the paper we report on aggregate information and refrain from singling out individuals in quotes, links, or anything else identifying. When using quotes, we carefully select those that are generic and represent common tweet content (i.e. "Happy sweet 16th birthday!"). We also do not collect our survey participants' Twitter data or link it to their answers. We received ethical approval from our institution for this work, including the Twitter data collection and the following user survey. The Share of Birthday Wishes on Twitter --- BD Dataset Statistics Table 1 shows the overview of our BD dataset, which contains 2.8m original tweets, broken out by the type of the mentioned account -mProtected, mPublic. These 2.8m tweets were directed to 724K unique accounts. The majority (56%) of these accounts received only one birthday celebration tweet. 99% of them received 30 or less birthday wishes. Protected accounts tended to get less birthday wishes on average with 99% of them receiving 15 or less tweets. One public user in our collection received 3934 birthday wish tweets, while the most popular protected user received 394 birthday wishes. We also check the percentage of tweets that were replies to other tweets or written to a user directly. We see that tweets were more likely to be directed to a user without replying to an existing tweet. Only 42.7% of the tweets were a reply to an existing tweet, whereas 57.3% of them were directed to a user without the user tweeting about their birthday. This difference is even higher for mProtected, where 69.5% were tweets directed to them and only 30.5% were replies (Table 1). Finally, looking at the tweets containing two-digit numbers (including ordinal indicators) mProtected tweets had higher percentage (3.2%) than the mPublic tweets (2.7%). These numbers can be an indication of the mentioned person age, which can lead to easily inferring the person's exact DOB. We provide further analysis for the meaning of these numbers later. --- Twitter Users' Reactions to Birthday Wishes We carried out another analysis on BD-react in which the tweets were processed in more depth to analyze the reaction of the mentioned users to the BD tweets mentioning them (Table 2). At the time of processing, 10.5% of these were no longer accessible due to various reasons such as the deletion of the tweet, the author protecting their account, and so on. mPublic tweets received high interaction from the mentioned accounts. We observed that 56.1% of the mPublic tweets had likes from the mentioned account. For mProtected tweets, we only know that a protected account interacted with the tweet, not which account. However, hidden interactions can give us an idea. 51.6% of mProtected tweets had hidden likes while 13.8% of them had hidden retweets. 66.6% of the public mentioned accounts interacted with the tweets mentioning them while 54.1% of the mProtected tweets had hidden interactions. This result shows that people frequently interact with the tweets that wish them a happy birthday in a positive way such as liking and retweeting, regardless of those people's accounts being public or protected. In addition, the large number of interactions can indicate that the tweets are seen by other people who might not be necessarily following the birthday person. --- Sharing Birthdays on Profile By checking the birthday information on the profiles of the 8522 reachable accounts, we found 7077 (83%) shared no birthday information, 1250 (14.7%) shared the birthday (BD), 73 (0.9%) shared only the birth year (BY), and 123 (1.4%) shared their full DOB. Public accounts were more likely to share information on their birthday (890, 20.4%) than protected accounts (556, 13.4%). In total only 196 (2.3%) of the users disclosed their birth year. We report the birthday sharing behavior on profiles broken out by the account type in Table 3. Users who shared their birthday information reacted similarly to the birthday tweets with those who did not. This was also the case for protected users. --- DOB Leakage on Twitter Regarding the BD-age tweets, 82.9% of the tweets with twodigit numbers refer to the mentioned person's age, according to Appen annotators. The percentage is slightly higher for the mentioned protected accounts (84.4%) than public ones (81.3%). We noticed that this percentage becomes higher (95.3%) if the two-digit number is followed by ordinal indicator (st, nd, rd, th). Using these rates, we extrapolated to the whole data set, taking into account the total number of tweets containing two-digit numbers, the results are shown in Table 4. We look at the unique accounts mentioned in the BD dataset to understand the potential DOB disclosure for birthday people. There were 56K (8%) accounts in total that received at least one birthday tweet that contained a two-digit number, of those 33K received at least one tweet accompanied by an ordinal indicator. 51K of them were public accounts, while 5.5K of them were protected accounts. Based on the results of the annotation, we can estimate that the actual age of the person is exposed for over 49K accounts which when combined with the date of the tweet, likely exposed the full birth date and year. This is 6.8% percent of the accounts that were mentioned in the tweets we collected. The mean of the two-digit numbers we found is 25 with median 21. The most celebrated ages were 18 and 21, followed by ages at multiples of ten (Figure 1). From our collection, we see that over 1K accounts receive birthday wishes that exposes their DOB every day, where 10% of those are protected accounts. While these users are mostly young adults, there are also users who are teenagers and elderly. Public accounts got more age exposing tweets than the protected accounts which suggests that people treat accounts differently depending on their type. Interestingly, accounts that shared no birthday info got more birthday messages with two-digits. Combining these results with the reaction of those users on the tweets, it becomes necessary to understand how Twit-ter users see this phenomena and if they perceive the DOB as private information. --- Measuring Users' Opinions and Awareness We conducted a survey to better understand how Twitter users think about the public sharing of birthday wishes on Twitter (RQ4), as well as their understanding of tweet visibility settings. We advertised the survey on Prolific Academic (PA) (Prolific 2022) as "Wishing a Happy Birthday on Social Media". The advertisement limited participants to Twitter users from the United States or United Kingdom to ensure similar culture and English label proficiency. We followed our University's ethics protocol in the design and running of the survey. Participants were compensated £0.5 (£8.34 per hour). --- Survey Instrument The survey started with informed consent followed by a screening question about if they had a Twitter account and if they used it more or less than once a month. Those without a Twitter account were screened out. We then asked if their primary Twitter account, was public, protected, or sometimes protected where they change the settings, followed by if they associated their Twitter account with their "real identity", and if they had their birthday publicly visible on any social media account. To gauge understanding of Twitter setting impacts, we asked what would happen in two scenarios where public and protected accounts interact. We also asked if they can tell that a poster's account is public or protected when replying to a tweet, and if they look to see if the account is protected when engaging with tweets (reply, mention, retweet). To understand their comfort with public birthday and date disclosure we asked them how comfortable they would be with friends and family publicly tweeting about their birthday with and without age information. We also asked how they might engage with such a tweet (like, retweet, reply, direct message (DM), ask to remove). We then asked them a similar question around the participant tweeting about a friend or family member's birthday with and without age. Finally, to gauge participants' understanding of the positive and negatives of public birthday wishing we asked them two free-text questions: "Give at least one example of a good thing that could happen if someone knew your birthday and age." and the same question with "bad thing". The survey ended with an optional comment box. As PA provides common participant demographics to researchers, we did not directly ask for any demographics. --- Survey Results Participant demographics The survey had 151 participants, 118 from the United Kingdom (UK) and 33 from the United States (US). Their average age was 31.2 years (<unk> = 10.6) with a median of 28.5. Respondents were primarily female (n=103, 68.2%) vs. male (46, 30.5%) with 2 preferring to not respond. For context, Twitter users from US have median age of 40 and half of them are female (Wojcik and Hughes 2019). Only 44% of the UK Twitter users are female (Statista 2021b) with more than half of the users older than 35 (Statista 2021a). Our participants are generally younger than the general Twitter population and have a higher percentage of female representation. Account types Most participants had public accounts (92, 61%) with the rest having protected accounts (44, 29%) or switching between public and protected (15, 10%). Figure 2 illustrates sex and age distributions of the survey participants broken out by their account types. 84 (55.6%) participants associated their Twitter accounts with their real identity. Of those, 27 (32%) had a protected account or switched between public and protected (9, 11%). In other words, roughly half of the people whose accounts were linked with a real identity were also protected. Publicly listing a birthday on at least one social media account was common, with 96 (63.6%) publicly listing a birthday, 46 (30.5%) not listing, and 9 (6%) not sure. 65% of the participants with public accounts shared their birthday on at least one social media account, whereas 57% of the protected accounts shared it. 75% of the users who switch between public and protected shared their birthday on social media. 13% of the adult US Twitter users have protected accounts (Wojcik and Hughes 2019), which is much lower than the share of protected accounts in our survey participants. --- Visibility of protected accounts Participants were asked about two scenarios: imagine that "Alice (Public) retweeted one of Bob's (Protected) tweets using the Twitter website?" and "Alice (Public) tweeted at Bob (Protected) using his handle (@bob) in the tweet?" We then asked them if Twitter would: allow Alice, warn Alice, allow Alice but restrict visibility to Bob's friends, allow Alice with no restriction, or they didn't know. For retweets, the Twitter website does not allow public accounts to retweet protected accounts. Only 45 (29.8%) of participants gave this answer. The most common answer was that Alice could retweet but only Bob's followers could see it (70, 46.4%), which is incorrect. For tweeting at protected accounts, the Twitter website allows public accounts to mention protected ones with no visibility restrictions. Participants were split on this question with 58 (38.4%) thinking that the tweet would be visible (correct), and 58 (38.4%) thinking that it would be visible only to Bob's followers (incorrect). The confusion over how Twitter protected accounts work is further highlighted by comments from participants about why they switch their accounts between protected and public. "If I post things I don't mind everyone seeing I changed it to public but if [I] post photo that I only want my followers to see I make it private." The comment highlights a potential misconception that the protections are per-post instead of per-account. When an account is protected, a padlock appears beside the username. However, when asked, only 33.8% of the participants agreed that they can easily see whether the Twitter account they are replying to is protected or not. Even less check the type of account they are replying to (24.5%). Birthday tweet opinions Looking at participants' comfort with friends and family tweeting about their birthday, the majority of participants were comfortable with their birthday being tweeted (103, 68.2%). Most gave the same answer for birthday and birthday with age (90, 59.6%), indicating that the addition of age had no impact on their comfort. A further 58 (38.4%) indicated that they would be less comfortable with birthday tweets containing an age. Looking at participants' likeliness of tweeting about a friend or family member's birthday, the majority of participants indicated they would be unlikely to do so (79, 52.3%), vs likely to do so (53, 35.1%). The majority of participants (80, 53%) gave the same answer for both birthday and the birthday with age, again indicating that the addition of age had no impact on tweeting likelihood. The rest (69, 45.7%) were less likely to post a birthday tweet containing age. Regarding their reactions to birthday wishes online, 90 (60%) participant said they would reply with a thank you, or like the tweet (84, 56%). Replying via direct message was less common (18, 12%). And it was rare to ask the person to remove the tweet (7, 4.6%) or retweet it (7, 4.6%). Those with public and protected accounts indicated similar reactions to tweets. These findings are similar to Table 2 where likes and replies were the main reactions to birthday wishes. Good & bad impacts of sharing We asked our participants to list some of the good and bad things that could happen if someone knew their birthday and age. Two researchers read through all the free-text answers and jointly grouped them into themes, discussing throughout to reach agreement. The most common good things mentioned were getting birthday wishes (45%) and gifts (31%). For the bad things, the most prominent worry was identity theft (37%), and the next most mentioned was being harassed, ridiculed or harmed (19%). 8.6% of the participants said nothing bad would happen. The responses of participants on our survey showed that users are aware of the role birthday wishing can have in connecting them with others as well as some of the dangers of birthday and DOB disclosure can cause. Nevertheless, nearly

Date of birth (DOB) has historically been considered as private information and safe to use for authentication, but recent years have seen a shift towards wide public sharing. In this work we characterize how modern social media users are approaching the sharing of birthday wishes publicly online. Over 45 days, we collected over 2.8M tweets wishing happy birthday to 724K Twitter accounts. For 50K accounts, their age was likely mentioned revealing their DOB, and 10% were protected accounts. Our findings show that the majority of both public and protected accounts seem to be accepting of their birthdays and DOB being revealed online by their friends even when they do not have it listed on their profiles. We further complemented our findings through a survey to measure awareness of DOB disclosure issues and how people think about sharing different types of birthday-related information. Our analysis shows that giving birthday wishes to others online is considered a celebration and many users are quite comfortable with it. This view matches the trend also seen in security where the use of DOB in authentication process is no longer considered best practice.

, or like the tweet (84, 56%). Replying via direct message was less common (18, 12%). And it was rare to ask the person to remove the tweet (7, 4.6%) or retweet it (7, 4.6%). Those with public and protected accounts indicated similar reactions to tweets. These findings are similar to Table 2 where likes and replies were the main reactions to birthday wishes. Good & bad impacts of sharing We asked our participants to list some of the good and bad things that could happen if someone knew their birthday and age. Two researchers read through all the free-text answers and jointly grouped them into themes, discussing throughout to reach agreement. The most common good things mentioned were getting birthday wishes (45%) and gifts (31%). For the bad things, the most prominent worry was identity theft (37%), and the next most mentioned was being harassed, ridiculed or harmed (19%). 8.6% of the participants said nothing bad would happen. The responses of participants on our survey showed that users are aware of the role birthday wishing can have in connecting them with others as well as some of the dangers of birthday and DOB disclosure can cause. Nevertheless, nearly half of them are still comfortable with their friends sharing their age along with birthday tweets online. --- Discussion --- Summary of Findings Our first research question concerns the number of birthday wishes posted on Twitter and how many include information revealing the DOB of the mentioned user. We collected 2.8 million happy-birthday tweets mentioning 724K unique users over 45 days, which shows that a large number of happy-birthday tweets can be easily linked to a specific account. Further, we found that around 7% of those accounts received at least one tweet that disclosed their age, allowing for easy combination of posting date and age to compute DOB. Considering only around 2% of the users in our dataset share their birth years on their profiles, DOB is being actively exposed for users who did not proactively share it. Our second research question concerns how birthday wishes differ towards protected and public accounts. 88% of all the birthday wishes we collected were towards public accounts. These accounts also received more birthday wishes per account on average than protected ones. We found that public accounts are also slightly more likely to receive birthday tweets that reveal their DOB than protected accounts, 8% vs. 6.2% respectively (Table 4). The result indicates that having a protected account does not prevent online disclosure of birthday or date information. Finally, we measured users' reaction behavior (tweet interaction -RQ3) as well as their attitudes (survey -RQ4). We found that 66.6% of tweets mentioning public accounts were reacted to by the mentioned account and 54.1% of tweets mentioning a protected account were likely reacted to, though we have limited visibility of protected accounts. Both Twitter and survey data show that liking and replying are popular ways to react to birthday tweets. 56% of the mPublic tweets in BD-react received a like from the mentioned person. Similarly, 56% of the survey participants said they would like a birthday tweet they receive on Twitter. 60% of the survey participants would reply to the birthday tweet, while 44% of the mPublic tweets in BD-react received replies from the mentioned user. Our Twitter data collection (20%) and user survey (5%) differ on the cases of retweets. 5% of our survey participants selected that they would ask the person to remove the tweet if they received a birthday celebration over Twitter. While it is not possible to measure this reaction from the Twitter data directly, we recorded 428 (9%) cases where the tweet or the user was deleted while collecting the replies to the tweets in BD-react. However, there is no way to differentiate from the API response whether it was the tweet that was deleted or the user's account. There is also no way to make sure that the birthday person requested the tweet deletion in any case. These findings suggest that users are aware of happy birthday tweets and react to them positively. Our survey verifying our tweet analysis findings show that Twitter users are comfortable with public celebrations of birthdays, both with and without explicit mention of their ages. This result is evident both in the scale of current birthday wishing on Twitter as well as the attitudes of survey respondents. However, most of the respondents were less likely to publicly celebrate birthdays of their friends and family with tweets containing their ages. Our survey also showed that Twitter users might not be fully understanding who can see their tweets when they mention protected accounts and sometimes not aware of account types of users they interact with. We further discuss these findings in the implications section below. --- Limitations Our research analysis was limited to tweets on Twitter which explicitly mentioned both the words "happy" and "birthday". While the term "happy birthday" is culturally quite common in English speaking countries, there are many other ways to express the sentiment. Additionally, we did not look for common misspellings or abbreviations such as "hbd", "happy bday". Therefore, the numbers presented in this work should be seen as a lower bound, or what an opportunistic data gatherer might be able to locate easily. Another limitation is that we only look at the two-digit numbers that have a leading space and no trailing alphanumeric characters other than the ordinal indicators. Because of this, we are missing some age exposing tweets. Some tweets may also have the age spelled instead of writing with numbers like "twenty first", or "sixteen". We ran a test on BD data set (Table 1), to understand if people spelled out ages. We looked at a commonly celebrated birthday "twenty one" or "twenty first" and compared the occurrences with the numeric "21" and "21st". We found that <unk> 0.01% of our BD tweets spell out 21, likely due to the character limit pressures imposed by Twitter. Hence, we expect the effects of excluding spelled out numbers to be minimal. --- Implications Popularity of birthday wishing Twitter is clearly considered a suitable platform to wish someone else a "happy birthday", which is further reinforced by verified account reactions as well as Twitter itself (Castro 2015). In our initial data collection we observed about a million tweets wishing a verified account a happy birthday. While we excluded these from analysis, the size of the dataset speaks to the public reinforcement that birthday wishing is a normal public Twitter activity. Twitter itself also encourages birthday celebrations by displaying a balloon animation on their birthday when others visit it. Other OSNs, such as Facebook, also actively encourage people to wish others a "happy birthday", which can impact the number of birthday wishes (Fleishman 2018). Birthday wishing was also common among non-verified accounts, accounting for nearly three quarters of the wishes. As an OSN, one of Twitter's roles is in helping people maintain weak and strong ties through sharing information (Vitak 2014) which can have benefits on their mental health and sense of belonging (Dym and Fiesler 2018). These ties are also useful at helping people gain access to prospects such as jobs and opportunities (Hoyle et al. 2017), so maintaining them has value. Viswanath et al. (Viswanath et al. 2009) found that users who do not interact on social media frequently, mostly only exchange birthday messages. Hence, birthday wishes can support and encourage users to maintain their social ties. Disclosure control Even if a user is inclined towards not sharing their DOB on Twitter, they have limited control over their network. Most OSNs only provide control to the poster, not the data subject. Hence, networked privacy means that the control over who can see what data is not solely in the hands of the person whose data it is. While some OSNs allow subjects to remove their tags from a post, they do not allow complete removal of the post. Consequently, if a user would prefer their birth date not be known, they would need to ask each poster to remove their post. Such a request may be socially challenging, especially if the majority of users feel that wishing someone a happy birthday is a good thing and nothing to be concerned about. Such a request might also risk being labelled as "paranoid" (Gaw, Felten, and Fernandez-Kelly 2006). Controlling information disclosure is also dependent on Twitter users and their networks' accurate understanding of tweet visibility. Previous research on social media with granular privacy options show that users find it hard to comprehend and configure these settings (Madejski, Johnson, and Bellovin 2011), while also underestimating the audience size (Bernstein et al. 2013). Twitter has a relatively simplistic access-control approach for a modern OSN. An account is either public or protected. If public, anyone on the internet can see the posts, if protected, then only a selected set of users can see them. Yet even this simplistic model confused our participants. In our study we found that 38.4% of participants thought that posts in reply to a protected account would also be protected. Findings suggest that users' mental models of Twitter protections might be inaccurate which may lead them to disclose information about protected accounts while honestly believing that the posts are not publicly visible. --- Authentication question vs. celebration There are surprisingly few questions that work well for authenticating identity. A good authentication question should: apply to nearly everyone, have a large number of possible answers, have equal distribution of answers, be easy to remember, and should not change. Based on those requirements, it is obvious why facts like birthdays were regularly used as a part of authentication. However, with birthdays celebrated publicly on OSNs and reaching more people than before, organizations justifiably shifted their use of DOB. DOB is also used as one of the attack vectors in social engineering and re-identification methods (Sweeney 2000;Krombholz et al. 2015). People often use their DOB when constructing passwords (Brown et al. 2004). Bonneau et.al. (Bonneau, Preibusch, and Anderson 2012) showed that up to 8.9% of the time, lost or stolen wallets will lead to thieves to correctly guess PINs. A primary reason is that DOB can be obtained using identification cards found in the wallet. However, as we can also see from our Twitter data and the following user survey, birthdays are not seen as secret by the general public. Similar with our findings, Markos et al. (Markos, Milne, and Peltier 2017) also found that DOB was considered as a low-privacy segment data along with e-mails by their participants compared to mother's maiden name, home addresses, and phone numbers. This attitude by public leads to a tension between data privacy and the reality of cultural sharing. One of the main recommendations that we can learn from our study is that organizations, such as banks (Kaur et al. 2020;Murdoch and Anderson 2010), email service providers (Li, Mehta, and Yang 2017;Al Maqbali and Mitchell 2018), wireless carriers (Lee et al. 2020), that still consider DOB sensitive information should withdraw from using it as a part of their authentication system. Users should also not incorporate their birthdays into constructed passwords. Twitter design implications Birthday sharing is widespread on Twitter and users are comfortable with it. Encouraging this behavior might help users to feel valued and appreciated, as well as maintain friendships (Vitak 2014). In order to encourage birthday wishing, Twitter could notify followers of a user on their birthday if the user wants to receive such messages. Displaying a small indicator (e.g. balloon, cake) next to the username of a person having their birthday on their tweets may also act as a reminder for the birthday and encourage a message. Providing users personalized messages and collating the birthday tweets in one thread will help the birthday person to feel special and also allow easy access for replying to those messages. On the other hand, some people might not want their birthdays to be celebrated publicly. However, Twitter accounts cannot manage the tweets that mention/quote/retweet them. Some users solve this by asking other users explicitly not to interact. For example, some users state that they do not want other users to quote them in their profiles or usernames. Information about protected accounts can also be revealed through public replies to their tweets. Twitter recently introduced a feature to select user groups who could reply to specific tweets. While this is a positive step, these tweets are still publicly visible. Another recently added feature is the option to hide replies, however, it is easy to access these tweets with an extra click. They are also reachable by search. In addition, users might be socially uncomfortable to hide celebratory messages that leak information about them. Hence, users should be given control on the visibility of the tweets that mention them. Another point is the confusion over the visibility of the posts when public and protected accounts interact. Our survey showed that users' mental models of tweet privacy might be quite different from the actual Twitter functionality. Our participants expected these tweets to be visible only to the followers of a protected account when a public follower interacts with the protected account. Hence, it is essential to disambiguate the interactions between them. This can be achieved by having an indication when interacting with protected accounts and let users know who can see the tweet if posted. The account types of the users mentioned in a tweet should be easy to check. As of now, when replying to a protected tweet, there is a padlock near the username indicating the status. However, when drafting a new tweet, the public or protected status is obscured. --- Conclusion In this study, we investigated the sharing of birthday wishes on Twitter, how they can reveal the date of birth of some users, and privacy concerns of the Twitter users regarding the DOB disclosure. Our objective was to provide an indepth analysis of how social media users see their DOB, as a private personal information, or as a happy event to be celebrated publicly. Our aim was to assist designers in the security and social media field to get a clear answer of how to treat this information about users when designing their systems. We both conducted an analysis of 2.8 millions tweets sharing birthday wishes, and a survey of Twitter users to understand their opinions around public celebration of birthdays on the platform. We found that birthday celebrations are common over Twitter and over 1K tweets disclose the DOB of the mentioned account daily, where 10% of those are protected. While the majority of those accounts do not share their birthday publicly, they still seem to be comfortable with others sharing birthday tweets publicly regardless of their account type, even when it discloses their DOB. We show that birthdays and DOB are not considered as sensitive information by users; they are celebrated publicly. Our findings should move any organization that is still using DOB as a part of their authentication process to phase its use out.

Date of birth (DOB) has historically been considered as private information and safe to use for authentication, but recent years have seen a shift towards wide public sharing. In this work we characterize how modern social media users are approaching the sharing of birthday wishes publicly online. Over 45 days, we collected over 2.8M tweets wishing happy birthday to 724K Twitter accounts. For 50K accounts, their age was likely mentioned revealing their DOB, and 10% were protected accounts. Our findings show that the majority of both public and protected accounts seem to be accepting of their birthdays and DOB being revealed online by their friends even when they do not have it listed on their profiles. We further complemented our findings through a survey to measure awareness of DOB disclosure issues and how people think about sharing different types of birthday-related information. Our analysis shows that giving birthday wishes to others online is considered a celebration and many users are quite comfortable with it. This view matches the trend also seen in security where the use of DOB in authentication process is no longer considered best practice.

Background Community health workers (CHWs), including promotores de salud, are trusted members of the community who have an intimate understanding of the population they serve [1][2][3][4][5][6]. Promotores de salud are a subset of CHWs who primarily serve Latinx communities and are grounded in a social, rather than a clinical model of care [7,8]. Their community membership, language, and cultural relationships allow CHWs and promotores de salud to bridge health care and social services divides within the current population health service system, facilitating the delivery of health promotion and culturally informed interventions [1][2][3][4][5][6]. CHWs build individual and community capacity by increasing health knowledge and self-sufficiency through conducting outreach, providing education, connecting communities to social support services, and advocacy [9]. CHWs may be employed by a clinic, hospital, health department, or community-based organization, or as volunteers [10,11]. In most states, CHWs and promotores de salud work outside of any system of formal certification. Nationally, there are no unified training standards for CHWs [12]. However, there are nationally recognized CHW skills and competencies [13]. Approximately 20 states have developed certification standards [14]; these are intended to enhance the credibility of qualifying professionals, increase service quality, and assure those served of CHW competency [15][16][17][18]. Statewide CHW certification involves developing a standardized process for documenting the proficiency of individuals across the state in the core skills and roles of a CHW [17]. Training alone does not lead to certification. Certification processes identify requirements related to CHW training or experience, formal education, language, and criminal background and are usually overseen by a certifying agency (i.e., state, educational institution, or private entity) that administers a competency-based examination to certify individuals with the skills and knowledge necessary to perform relevant tasks [17,19,20]. Certification differs from an educational certification of completion and also credentialing or licensing, a legislative directive where an individual must obtain a credential or license to practice or work under that job title [19]. A key policy question is whether and how a state should introduce certification for CHWs [21,22]. The CHW workforce stands at a crossroads: to advocate for certification or oppose it. One path leads to a certified workforce integrated into health care systems via formalized training and qualifications [23]. The other path retains the current emphasis on CHWs as part of the communities where they work, valued for their community relationships [23]. As there is very limited evidence about how certification impacts the workforce [24], we cannot say with certainty that certification standardizes the CHW workforce or maintains CHWs' community connection. Importantly, CHWs historically have not led their own workforce development, possibly because they are often from marginalized communities and are economically vulnerable [10,19,25]. This is now changing. CHWs have been involved in developing certification in some states (e.g., Arizona and Massachusetts) and have led national CHW efforts (e.g., National Association of Community Health Workers) [26][27][28][29]. This study explores the debate through data from California, presenting diverse CHW stakeholder perspectives on CHW certification. As in many states, California's CHW workforce includes CHWs with diverse roles and training, who represent distinct ideologies of care [7,8]. Clinical CHWs are most often employed by clinics and hospitals. Community-based CHWs work for health departments and community-based organizations. Promotores de salud are typically volunteers and provide services to primarily Latinx communities through a social rather than a clinical model of care [7]. Most recently, California's managed care organizations have started employing CHWs to provide health promotion services [20,30]. This study is relevant for other states considering certification to address their CHW workforce and poses a certification option that may preserve CHW identity while also protecting diversity within the workforce. --- Methods We aimed to explore perceptions among diverse stakeholders about possible benefits and consequences to implementing CHW certification in California. The lead author employed a purposive sampling approach to recruit three categories of stakeholders (all from California): 1) CHWs (part-time, full-time, and volunteer), who could be clinical CHWs, community-based CHWs, or promotores de salud, 2) managers of programs that employ CHWs, and 3) system-level participants, such maintain CHW identity, and protect diversity within the workforce. Positioning CHWs as decision makers will be critical when designing state certification processes. Keywords: Community health workers, Promotores de salud, Community health, Certification, Workforce development, Qualitative methods as employers, coalition leaders, academic researchers, foundations, policymakers, and health officers. They represented various organization types (e.g., hospital, community-based organization, government, clinics, managed care organizations, advocacy organizations, academia), and communities served. We also recruited experts from the federal government, national coalitions, and organizations with technical expertise in CHWs and certification. The lead author identified experts if they had provided technical assistance to states implementing certification, had publications related to CHW workforce development, including certification, or led CHW workforce initiatives (e.g., taskforces). All participants were recruited by phone and email and provided verbal consent to participate in the study. Participants were offered up to $100 compensation (for up to three-hour participation in focus group or interview, travel time, and childcare expenses), which was not always accepted. The study was approved by the University of California Berkeley Committee for Protection of Human Subjects. We used focus groups, interviews, and observation of public forums to gather diverse CHW stakeholder perspectives on CHW certification in California. The lead author conducted focus groups primarily with CHWs to allow for discussion among workforce members. The lead author conducted interviews with program managers and system-level participants to obtain in-depth contextual information on California's CHW workforce in a manner that was feasible with their time constraints. Focus groups and interviews were semi-structured with a single guide that sought input and recommendations on CHW certification using examples of certification programs (one voluntary, one required) from two other states (i.e., Texas and Massachusetts). During focus groups and interviews, participants received an overview of certification and how it differs from credentialing or licensing. Focus groups lasted up to three hours, and interviews lasted approximately one hour and were audio recorded and transcribed. Focus groups and interviews were conducted in either English or Spanish. From October 2018 -November 2019, the lead author conducted 44 focus groups, group interviews, and individual interviews, with 66 CHWs, 11 program managers, and 31 system-level participants. Focus groups ranged from three to nine participants. The lead author conducted predominantly individual (i.e. one-on-one) interviews with program managers and system-level participants, but in some cases, program managers and system-level participants requested group interviews (two to three people) when both worked at the same organization. Focus groups were conducted with CHWs to facilitate discussion across workforce peers. Most of the focus group and interviews were conducted in English, while six focus groups were conducted in Spanish. All focus groups and interviews were conducted at the organization where stakeholders worked. Nearly all interviews with system-level participants were conducted in person, although some were conducted by phone when in-person interviews were not feasible. In addition, we conducted ten observations in public forums (e.g., conferences, taskforce meetings, coalition meetings) where CHW workforce issues, including certification, were discussed. Observations offered context to how stakeholders leveraged CHW work and their plans to apply CHW certification in real-world settings that may not have been discussed in individual interviews or focus groups. The lead author took detailed field notes to document how CHW workforce issues were discussed, the context in which certification was described. We used a thematic analysis approach to identify, analyze, and report patterns or themes within the focus group, interview, and observation data [31]. All focus group and interview recordings were transcribed in their original languages. Spanish-language focus groups were transcribed into Spanish and then professionally translated and checked by the Spanish-speaking authors for accuracy. We reviewed the focus group and interview transcripts and observation field notes to identify preliminary analytic categories. The two Spanish-speaking authors developed the codebook through inductive coding and analyzed the data line-by-line through focused coding. We developed codes and code definitions in English, based on both English-and Spanish-language transcripts and observation notes. The two authors coded initial data separately, then compared, discussed, and reached an agreement if codes or emerging patterns matched or did not match [32]. After code testing and consensus coding, we applied codes to the data using qualitative analysis software, Dedoose (SocioCultural Research Consultants, LLC, Los Angeles, CA). We wrote memos to document patterns in the data and to comment on our methodological decisions. We re-sorted codes into themes and refined those themes by checking how well the coded extracts illustrated the themes. --- Results As shown in Table 1, CHWs, program managers, and system-level participants represented government and non-governmental organizations, health care providers, health plans, academia, and foundations, that served urban areas or a combination of urban and surrounding rural areas. All stakeholders described the opportunities posed by plans for CHW certification in California but also warned of its unintended consequences. Four themes emerged among the three stakeholder groups (Table 2): 1) certification may enhance recognition of CHWs from health care providers and the community; 2) certification may offer more upward mobility and professional growth for CHWs; 3) certification may threaten CHW identity by pushing the CHW role into a clinical one and creating unintentional hierarchies (perhaps unintended) among certified and non-certified CHWs; and 4) certification may exclude some current or potential CHWs due to entry requirements. --- Certification may enhance recognition of CHWs Two themes emerged about CHW recognition. First, given that California's CHWs lack a common training background, all types of participants believed that certification could validate CHW work and provide recognition from health care providers. CHWs described how Table 1 Characteristics of study participants a Non-governmental organizations (NGOs) included community-based organizations, coalitions, and advocacy and policy organizations b Government included Local, state, and federal government c Health care providers included individuals or organizations providing clinical services (e.g., federally qualified health centers, clinics, hospitals) d Two focus groups with 14 total participants included a mixture of NGO and health care provider organizations but data were not collected to identify the type of organization for each individual participant e Two focus groups with 14 total participants included a mixture of full-time, part-time, and volunteer CHWs but data were not collected to identify the employment status for each individual participant f When both English and Spanish languages were spoken during an interview or focus group, the primary language of all participants was labeled "bilingual. " g Stakeholder service area was labeled "not defined" when the stakeholders or stakeholder organization did not provide direct services to a specific geographic area --- CHWs N (%) Program Managers N (%) --- Systems-level N (%) Number Urban and rural 22 g Not defined 0 (0) 0 (0) 20 (64) --- Geographic Region Northern California 1 (1) 1 ( 9 CHWs expressed that their community experience, gained in their vital and intimate work with families in the community, is not understood or valued by many of the organizations they work for, especially in clinical settings. CHWs also shared their community experience positions them to identify community needs, communicate with families, and address social determinants of health better than most other health care providers. While all stakeholders commented on their desire for increased recognition of CHW skills, some commented that recognition from health care providers should not be dependent on certification. A system-level participant shared, "It's a little sad that [certification] is how you see it's necessary to get the respect of other people. " All types of participants posed that educating health care providers about CHW contributions to care could be more impactful than certification. A few systemlevel participants believed that institutional racism and discrimination within the health care system discounts the CHW role and creates poor working conditions. "If you think about who the CHWs are, " a system-level participant shared, "they tend to be women... [and] women of color. " The participant described how the tendency to discount CHWs as "real health care providers" "justifies that they get paid less; they don't get regular hours, shoddy training. It reinforces gender discrimination and racial discrimination. " System-level participants argued that certification could validate CHW work which may mitigate health care provider discrimination experienced by CHWs. Racism and discrimination towards CHWs impact the communities they serve because CHWs may be discouraged from facing racism in healthcare environments. Second, CHWs speculated that certification could help them more effectively deliver information within their community by strengthening their credibility. A few CHWs recalled that some families they serve ask why they are qualified to provide education. A CHW shared, "Creating a certification gives the person who is receiving the information the assurance that you are saying things correctly. " CHWs shared that certification could give them the confidence to demonstrate their skills and training. CHWs believed that a certifying agency would have more weight with the families they serve and provide confidence in the services provided. A CHW shared, "We don't have that support of saying, 'I have this certification, I know what I do, please pay attention to me.'We still don't have that support of being trained. " CHWs suggested these questions about credibility undermines their training and expertise to deliver health education in their communities. --- Certification may offer more upward mobility and professional growth for CHWs Three themes emerged about upward mobility and professional growth. First, all types of participants overwhelmingly agreed that unified training and consensus about core CHW skills and knowledge could create the foundation for the professionalization of the CHW workforce in the state. "[CHWs are] not all the same depending on who trained them and what they've got, " commented one system-level participant. This uneven training, continued the participant, could open the workforce to outside criticism, as employers may speculate: "'You're missing some basic skills, but yet you've been a CHW for 10 years. '" CHWs suggested that this continued failure to unify training and skills through certification could have negative effects for the families with whom they work, as some CHWs receive structured training programs and mentorship while other CHWs receive piecemeal trainings. Second, all types of participants believed that CHW certification may result in career mobility. CHWs welcomed the opportunity that certification would afford them to grow in their careers. Certification could set boundaries for a workforce that other health professionals may be unfamiliar with and distinguish their work from health care providers with areas of overlapping expertise, such as social workers. Further, certification could establish a career ladder or series of positions that would enable them to advance professionally with increasing experience. System-level participants commented that, in California, CHW job transitions are mostly lateral. A system-level participant shared, "CHWs will get hired and trained for one specific job, and then that job ends. Then they have to start from scratch and they just have whatever job they can find. " Participants believed that certification could generate a stronger field of positions when CHWs sought work transitions. Professionalizing the workforce could enable CHWs to leverage better pay, better positions, and career mobility through recognition of their skills. In observations, some stakeholders showed concern that a CHW career ladder would threaten the roles of health care providers. Some health care provider advocacy organizations feared that CHW certification would lead to "scope creep, " with CHWs potentially infringing on their professional scope of practice, diminishing their current clinical roles and clinical support for their licensed professions. Conversely, some system-level participants noted the hypocrisy embedded in this reaction to building CHW career mobility: "Other professions forget that established professions objected to them on the way up. Nurses had to struggle for...professionalization. Midwives did. Health educators did. MSWs [social workers] did. But they're really quick to turn around and say, 'But I don't know about you. '" Last, all participants were united in their belief that CHW certification could lead to higher compensation and an established pay rate. CHWs felt that the potential for increased pay from certification could result from better recognition of their value, which is modestly remunerated or, in the case of promotores de salud, not compensated. Uncompensated promotores de salud clarified that, though pay was not the motivation for their work, certification may enable them to earn money: "It's not that money is important, I have 15 years of being a promotora and, believe me that all promotoras, yes, it is the love for the work, but we would be at a more recognized level [if we earned money.]" Program managers and system-level participants believed that certification may facilitate the payment of CHW services by insurers, allowing payers clarity of the CHW role in care. A system-level participant shared that payers' confusion about CHW roles was a barrier to supporting their role in care: "They don't want to pay for something that they're not clear about. " Certification could enable sustainable financing models from federal payers, such as Medicaid, that include strict regulations on spending categories and service providers. Without certification, noted a systemlevel participant, "I find it hard to imagine that CHWs' time is going to be reimbursed in the model and structure of the health care setting that we have now. These are federal dollars, there's rules. " Yet others cautioned that certification alone will not guarantee Medicaid financing. Another system-level participant suggested that Medicaid financing for CHW services relies more on the advocacy of CHW champions, state-specific legislation, and payer systems rather than singly on certification. --- Certification may threaten CHW identity Three themes related to CHW identity emerged. First, all participants expressed fears that certification may dissolve the identity of CHWs and promotores de salud and push them into a more clinical role, further away from the historical community-centered role. Some CHWs and program managers were concerned that certification could attract people without a commitment to the community-centered CHW role, but instead sought to use certification as a "stepping stone" into other health care positions, such as social work or nursing. Some systemlevel participants commented on evidence from states that have implemented CHW certification that validated these fears. A system-level participant who worked on certification outside California shared, "Institutions were under some pressure to get people into their program, and they were recruiting some people who might have been inappropriate for the work. They were certified, but they weren't really CHWs. They were not from the community. They didn't have anything in common with the community, but they had the training, and so they were entitled to call themselves certified CHWs, even though the community would probably look at them and say, 'You're not a CHW. You got a piece of paper, but you're not a CHW. You're not from here. '" A system-level participant reflected that instructors in CHW certification programs could distinguish who in their class possessed "real CHW" qualities, and others who may eventually leave the profession, believing that individuals were motivated by the certification rather than serving their community. As California CHWs represent distinct ideologies of care, program managers and system-level participants were wary that certification could push more CHWs into clinical roles. Because clinical organizations place more emphasis on certifications and degrees, participants anticipated all CHWs employed by these organizations would pursue certification. Since certifications are not required by community-based organizations, participants believed that community-based CHWs and promotores de salud may not pursue certification. All types of participants worried that certification could produce a disparity between clinical and community-based CHWs or promotores de salud due to the demands of their organizations. Program managers and system-level participants also feared that the clinical orientation of certification would transform CHWs to fit the health care system and undermine promotores de salud historic emphasis on impacting the social determinants of health through community-based work. Second, system-level participants expressed concern that the tension between clinically focused and community-based roles could exacerbate identity differences between CHWs and promotores de salud in a state with no universal definition of CHWs. A system-level participant described how funding sources already separated promotores de salud and CHWs. Promotores de salud have historically served Latinx communities, working at the intersection of health and social justice, predominantly as volunteers. CHWs were more recently employed by health departments, clinics, and hospitals, driven by single source funding opportunities. (Other stakeholders described CHWs and promotores de salud as interchangeable titles but serving different populations.) Yet promotores de salud participants were less likely to consider certification relevant for their work, citing the intrinsic motivation for their work. A promotor de salud shared, "We have never thought of a certification because we never thought of receiving money for our service. It's always volunteer, we always do everything from the heart. " Promotores de salud, describing their work as service rather than a professional career, asserted that certification was not relevant to their motivation to serve their communities and should not dictate who may work as a CHW. For community-based CHWs, promotores de salud, and program managers, certification does not challenge the identity of the CHW role. Their identity goes beyond a certification process because their intrinsic motivation to serve their community defines their CHW role instead of skills identified by a certifying authority. A CHW shared, "If you don't have that certificate, that doesn't mean you're not a CHW. " Third, CHWs and program managers anticipated that certification may create a hierarchy between certified and non-certified CHWs, in which certified CHWs may have an unfair advantage for employment and feel superior to non-certified peers because of the certification, exacerbating other existing inequalities between CHWs in employment type (e.g., paid or volunteer) and work setting (e.g., community or clinic). A program manager shared, "We have the entitlement piece where it's like, 'I'm certified and you're not so I'm better at my job.'That's going to be a barrier, unfortunately. " Still, system-level participants from states that have implemented CHW certification shared they have not directly observed a hierarchy among CHWs that was feared for California. --- Certification may exclude some current or potential CHWs Four themes emerged about exclusion. First, nearly all participants were concerned that instituting certification could exclude CHWs from working in their communities-whether they were existing CHWs or those considering the career path. "When you're certifying people, " one CHW shared, "you're limiting other groups of people getting the job done. You think it's best for them, but when you get the certification you have to be literate, able to learn, be multi-tasking. It requires a little bit more steps that other people are not willing to do. " The very qualities that make CHWs so effective in providing culturally and linguistically appropriate care can also be vulnerabilities that certifying authorities may not value or recognize. A CHW shared their fear of excluding people with crucial shared experiences but little formal education, commenting, "Moms who cannot read or write will not have the chance to be promotores. " All participants believed that certification, as a complex process overseen by the state, posed multiple barriers to a workforce largely comprised of women of color from marginalized, multi-lingual communities. Potential barriers to obtaining certification mentioned were numerous, such as literacy level, training costs, availability of certification in languages other than English, education requirements, and legal residence status. Further, participants believed the many CHWs who lived in communities with geographic and technological barriers to accessing training sites and materials risked being shut out of certification entirely. Second, all participants were concerned about the implications of certification for undocumented CHWs in California since most certification processes require legal residence documentation. System-level participants worried that certification could prevent undocumented CHWs from continuing to work in immigrant communities. "They're already facing racism, xenophobia, ICE [Immigration and Customs Enforcement] raids, and they don't want to see this profession move along without remembering the unique contributions that they make in our state, " explained one participant. "And they're worried particularly that certification will leave them behind. " Third, some CHWs who experienced prior incarceration worked with individuals to navigate the psychosocial and structural struggles after "coming home" from prison. All types of participants feared that certification requirements for a criminal background check could exclude these CHWs with prior felonies. "Society [holds] stigma against people who are incarcerated, " noted one system-level participant, "There's some really great people there, who have had this experience, who can turn it around and really help others and, that's what makes them so successful. " Excluding CHWs with felony convictions could eliminate a vital point of support for an extremely vulnerable population. Last, many feared that certification could exclude those CHWs who are most effective in their communities by placing more emphasis on state requirements than the vital social skills of connecting with communities. Participants questioned the extent to which certification could assess whether CHWs are equipped to do their work and cautioned that relationship-building skills and lived experience are difficult to evaluate through a certification system. A system-level participant shared that such training requirements "would actually knock a lot of people out of the workforce who actually demonstrated that they were great at connecting with [high-risk and high-needs patients]. " Participants believed that state certification requirements could overshadow what CHWs do best: developing trusting relationships within their communities and advocating for positive change. A system-level participant illustrated this concept with a promotores de salud motto: "Don't change who we are but give us a chance, " emphasizing that workforce standardization could inevitably alter the qualities that make CHWs and promotores de salud effective. --- Discussion We set out to understand diverse stakeholder perspectives on CHW certification as California considers certification's capacity to advance or potentially exclude members of this critical workforce. This study was strengthened by CHW perspectives about certification and provided crucial insights about the diverse membership of CHWs. Peer-reviewed CHW certification literature is limited, and existing scholarship largely omits CHW perspectives on setting their own workforce standards [19], despite their expertise and deep knowledge of the organizational and systemic levers necessary to implement community programs. Echoing literature on certification [17,22,[33][34][35][36][37], we found conflicting perspectives within and between all stakeholders groups on certification. Our findings confirm some findings from a 2021 study that explored perspectives on certification from seven states that have implemented or considered implementing certification [22] (e.g., CHW participation in decision making, concerns about certification requirements excluding CHWs). In this study, some participants affirmed the financial and career opportunities offered by certification, while others feared it may inadvertently exclude vulnerable CHWs and the communities they serve and downplay advocacy at the heart of CHW identity. Participants also highlighted how certification does not address all challenges related to integrating CHWs into health care systems. Our findings underline how California CHWs desire opportunities to grow. We confirmed that CHWs view certification as an opportunity to increase compensation and to build career opportunities; not just through expanding the capacity of organizations to work with CHWs, but also by building a career ladder [38]. Participants agreed certification may bring health care provider and community recognition and increase the demand for CHW services. Our findings confirmed literature asserting that health care providers feel more confident when CHWs are certified because they can ensure a standard of care [39]. At the same time, certification provoked fears across many participants that CHWs will transform into a more clinical role, potentially diminishing the tradition of advocacy, social justice, and community connection, echoing existing literature [23,33,40,41]. CHW participants emphasized their skill in addressing the social determinants of health and noted that this essential contribution might be undermined if CHW work were restricted within a medical model of health. By "professionalizing" the CHW workforce, certification could threaten the qualities that make CHWs effective, such as gaining community trust [38]. Promotores de salud stakeholders fear that certification washes away the "essence" of promotores by attracting people without "the heart" for the work [42]. A broader group of participants doubted whether certification could cultivate or measure the relational skills that help constitute this "heart. " Existing research suggests these fears may be warranted. A Zambian study found that focusing on skills or career incentives, rather than community service, displaced CHWs with desirable social connections and lessened the quality of services they provided [43]. Our findings also demonstrate CHWs' skepticism that individuals will use certification as a "stepping stone" to other health professions, potentially resulting in high turnover of the CHW workforce. A national survey of CHW certification programs confirms this prediction-finding that many CHWs, after attending community college certification training programs, later advanced to nursing and social work professions [24]. Participants in this study also stressed the limits of certification as a tool to advance the CHW workforce. Our findings echo existing literature asserting that certification does not guarantee employment or Medicaid financing for CHWs [38,39,44,45] but continues to be a motivation for financial sustainability for the workforce [22]. This is a timely finding as California's Department of Health Care Services (the agency responsible for administering Medicaid in California) has taken federal actions to finance CHW-delivered services for Medi-Cal members and California's managed care organizations have begun integrating CHWs into their standard of care [20,30,46,47]. While some participants appreciated the increase in recognition that certification may offer, other studies show that funding streams and return on investment have been identified by employers as the most important factors in whether to hire CHWs [48]. While our participants hoped that certification would affirm the quality of care delivered by CHWs in their communities and to their employers, there is not yet conclusive evidence CHW training and certification programs ensure quality of services delivered, as there have been limited evaluations in states to date [24]. One of the key contributions of CHWs and promotores de salud in this study was that they identified their uncertainties that certification will engender unintentional hierarchies: both between clinical and community-based CHWs, and between CHWs and promotores de salud. Over time, employers may prefer certified CHWs since they could ensure a standard skill set and knowledge base [38], unofficially making the certification "required" [49]. Since community-based CHWs and promotores de salud have little need for certification in their communities, their work-which already receives less funding-may become further marginalized. We fear certification may result in fewer community-based CHWs and promotores de salud working in marginalized communities, ultimately diminishing access to services delivered by culturally and linguistically appropriate professionals who embody community trust. Our study highlights how racism and discrimination within the health care system may have obstructed the advancement of the CHW workforce into health care systems, a theme unacknowledged in previous CHW research. This insight adds evidence of documented racism towards paraprofessionals, particularly among structurally similar positions like certified nursing assistants who experience institutional racism, cultural insensitivity, and discrimination from supervisors and coworkers [50][51][52][53][54]. While public health and health care institutions have supported the concept that CHWs are key to diversifying the health care and public health workforces [55], their recommendations have not acknowledged that racism within the health care system remains a critical barrier to this diversification. To the best of our knowledge, this is the first study to describe perspectives of CHW certification from CHW, promotores de salud, program manager, and system-level stakeholders using existing certification models. However, this study is not without limitations. The CHW participants were almost exclusively from Southern California. CHWs from Northern California may have had different perspectives on certification. There were two focus groups for which employment status and organization type data were not collected. We do not have basic demographic information for participants (e.g., age, gender, race/ethnicity). One type of CHW, known as a Community Health Representative (CHR), works with Native American clients as trusted members of their tribal communities and connects them to health care and social services [56]. Scheduling logistics limited our ability to reach CHR informants. Patient perspectives were also not included in this study. All these issues limit the conclusions we can draw about an entire state. To address the concerns identified by participants, we suggest a certification process where CHWs may choose whether they want to become certified and consists of multiple, accessible pathways to certification: a training pathway (where new CHWs learn required skills and knowledge) and a work experience pathway (existing CHWs may certify based on amount and type of experience) [19,57].To date, nearly every state that has certification offers existing CHWs an opportunity to certify based on their work experience [58]. Approximately 15 states have implemented certification based on training [44,46,58]. Echoing a recent peer-reviewed study of certification, this certification approach would limit requirements of formal education, English-language proficiency, history of incarceration, and legal residence status [19,22,57,59,60] to continue to foster the diversity that is a key feature of an effective CHW workforce. Most importantly, this approach respects promotores de salud and other CHWs that may not need certification for their work by not mandating certification and acknowledging their contributions [19]. When designing CHW certification, stakeholders must position CHWs as key decision makers. Our study echoes existing literature identifying CHWs as essential and best equipped to lead workforce discussions because they can better anticipate the needs and aspirations of their workforce and should make up at least half of the decision makers [22,61]. We anticipate if CHWs are not included in the decision making, certification could reflect a process that may unintentionally exclude workforce members and limit services to communities in need. --- Conclusion Considering the high stakes posed by certification, we appraised its potential through evaluating the perspectives of diverse stakeholders who represented various CHW types and organizational roles. Certification may increase CHWs' legitimacy within clinical organizations and communities by expanding recognition of their contributions and increasing their professional security. Yet certification poses genuine challenges to delivering care by CHWs that value relationships, prioritize cultivating the social determinants of health outside clinical settings, and favor connection with vulnerable communities largely excluded from formal systems of education. Our findings affirm neither certification nor its absence, but instead a nuanced path ahead that mixes opportunities with continued reflections on how the CHW workforce can be supported for what it does best. Above all, CHW participation in designing certification processes is essential to protect "the heart" and diversity of the workforce. previous versions of the manuscript. All authors read and approved the final manuscript. --- Availability of data and materials Participants did not give consent for transcripts to be published in public repository. Codebook available upon request. --- Abbreviations --- Declarations Ethics approval and consent to participate This research was approved by the UC Berkeley Committee for the Protection of Human Subjects (Protocol ID #2018-02-10721) and determined exempt. All participants provided verbal consent, including their acknowledgement of the purpose, benefits, and potential risks of participation, and that their participation would have no bearing on employment within their organization or future collaboration with the researcher, as well as the option for non-participation. --- Consent for publication All participants agreed that their words and points may be presented for publication but that their name or other identifiable details would not be included. --- Competing interests The authors have no conflicts of interest to declare. • fast, convenient online submission • thorough peer review by experienced researchers in your field --- • rapid publication on acceptance • support for research data, including large and complex data types •

Background: For community health workers (CHWs) and promotores de salud (CHWs who primarily serve Latinx communities and are grounded in a social, rather than a clinical model of care), the process of certification highlights the tension between developing a certified workforce with formal requirements (i.e., certified CHWs) and valuing CHWs, without formal requirements, based on their roles, knowledge, and being part of the communities where they live and work (i.e., non-certified CHWs). California serves as an ideal case study to examine how these two paths can coexist. California's CHW workforce represents distinct ideologies of care (e.g., clinical CHWs, community-based CHWs, and promotores de salud) and California stakeholders have debated certification for nearly twenty years but have not implemented such processes.We employed purposive sampling to interview 108 stakeholders (i.e., 66 CHWs, 11 program managers, and 31 system-level participants) to understand their perspectives on the opportunities and risks that certification may raise for CHWs and the communities they serve. We conducted focus groups with CHWs, interviews with program managers and system-level participants, and observations of public forums that discussed CHW workforce issues. We used a thematic analysis approach to identify, analyze, and report themes. Results: Some CHW participants supported inclusive certification training opportunities while others feared that certification might erode their identity and undermine their work in communities. Some program managers and system-level participants acknowledged the opportunities of certification but also expressed concerns that certification may distance CHWs from their communities. Program managers and system-level participants also highlighted that certification may not address all challenges related to integrating CHWs into health care systems. CHWs, program managers, and system-level participants agreed that CHWs should be involved in certification discussions and decision making.To address participant concerns, our findings recommend California stakeholders build a voluntary certification process structured with multiple pathways to overcome entry barriers of traditional certification processes,

all, CHW participation in designing certification processes is essential to protect "the heart" and diversity of the workforce. previous versions of the manuscript. All authors read and approved the final manuscript. --- Availability of data and materials Participants did not give consent for transcripts to be published in public repository. Codebook available upon request. --- Abbreviations --- Declarations Ethics approval and consent to participate This research was approved by the UC Berkeley Committee for the Protection of Human Subjects (Protocol ID #2018-02-10721) and determined exempt. All participants provided verbal consent, including their acknowledgement of the purpose, benefits, and potential risks of participation, and that their participation would have no bearing on employment within their organization or future collaboration with the researcher, as well as the option for non-participation. --- Consent for publication All participants agreed that their words and points may be presented for publication but that their name or other identifiable details would not be included. --- Competing interests The authors have no conflicts of interest to declare. • fast, convenient online submission • thorough peer review by experienced researchers in your field --- • rapid publication on acceptance • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year --- • At BMC, research is always in progress. --- Learn more biomedcentral.com/submissions Ready to submit your research Ready to submit your research? Choose BMC and benefit from:? Choose BMC and benefit from: --- Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Background: For community health workers (CHWs) and promotores de salud (CHWs who primarily serve Latinx communities and are grounded in a social, rather than a clinical model of care), the process of certification highlights the tension between developing a certified workforce with formal requirements (i.e., certified CHWs) and valuing CHWs, without formal requirements, based on their roles, knowledge, and being part of the communities where they live and work (i.e., non-certified CHWs). California serves as an ideal case study to examine how these two paths can coexist. California's CHW workforce represents distinct ideologies of care (e.g., clinical CHWs, community-based CHWs, and promotores de salud) and California stakeholders have debated certification for nearly twenty years but have not implemented such processes.We employed purposive sampling to interview 108 stakeholders (i.e., 66 CHWs, 11 program managers, and 31 system-level participants) to understand their perspectives on the opportunities and risks that certification may raise for CHWs and the communities they serve. We conducted focus groups with CHWs, interviews with program managers and system-level participants, and observations of public forums that discussed CHW workforce issues. We used a thematic analysis approach to identify, analyze, and report themes. Results: Some CHW participants supported inclusive certification training opportunities while others feared that certification might erode their identity and undermine their work in communities. Some program managers and system-level participants acknowledged the opportunities of certification but also expressed concerns that certification may distance CHWs from their communities. Program managers and system-level participants also highlighted that certification may not address all challenges related to integrating CHWs into health care systems. CHWs, program managers, and system-level participants agreed that CHWs should be involved in certification discussions and decision making.To address participant concerns, our findings recommend California stakeholders build a voluntary certification process structured with multiple pathways to overcome entry barriers of traditional certification processes,

Introduction When creating and presenting social media content, we are faced with many challenges: How should we phrase the title of our submission? Should the title be long or short, specific or generic, unique or derivative, flowery or plain? Which community should we submit our content to? At what time? How should the title change when targeting a different community? As content creators, our end-goal in asking such questions is achieving high popularity for the content we submit. Therefore, answering these questions involves analyzing the effect of each of these aspects on the popularity of the submissions. Insights stemming from such analysis are extremely valuable as it is much more appealing to a content creator to increase the chances of the content's success by changing the title, posting time, and community, rather than changing the content itself. Na<unk>vely, given enough content and enough submissions we might train classifiers that predict the success of social media content directly from its title, community and posting time. While this is a reasonable strategy for predicting the overall success of the content, it does not effectively capture the confounding effects arising from the complex interplay between these factors. For example, we might learn that occurrence of certain words in the title increase the chances of a submission's success (Brank and Leskovec 2003). While this kind of insight is interesting, it might not accurately tease apart the effect of the content and the title on the submission's popularity. To illustrate, consider the case of kitten images. Such images are accompanied by titles such as 'cute kittens', and while such titles are correlated with high popularity, it would not be appropriate to conclude that words such as 'kittens' contribute to a 'good' title in general. Rather, such submissions are popular because of the images themselves, and not because there is anything particularly effective about the title. It is exactly this interplay between the content, title, community, and posting time that makes it difficult to study factors that determine the success of social media content (Artzi, Pantel, and Gamon 2012;Yang and Leskovec 2011;Yano and Smith 2010). Our goal in this paper is to directly study this interplay by observing how well the same content performs when posted with multiple different titles to multiple communities at different times. To do this, we consider submissions to the website reddit.com, each of which is an image, uploaded with a particular title to a particular community at a particular time, and rated by the community. The success of a Reddit submission depends on many factors other than the quality of the content. When the same content is submitted to Reddit multiple times, the popularity of each submission is not independent: an image is far less likely to be popular the twentieth time it is submitted regardless of how well the other factors are accounted for. A particularly unique and important property of our dataset is that every image we consider has been submitted multiple times, with multiple titles to multiple communities. This means that our dataset represents a natural large-scale experiment, which allows us to tease apart the quality inherent in the content itself, and directly assess the extent to which factors such as the title, community and posting time influence the popularity of the content. We develop a statistical model which accounts for the effect of four factors all of which play an important role in influencing the popularity of online social media content: (1) the content of the submission, (2) the submission title, (3) the community where the submission is posted, and (4) the time when it is posted. Our approach consists of two main components. First is a community model that accounts for factors such as the number of times an image has been submitted previously, the time of day it is submitted, and the choice of communities that it has been submitted to. The second is the language model that accounts for the quality of the title. A good title is then considered to be one that further improves the success of that submission. Our community model accounts for all the factors that influence a submission's success other than the title itself. The language model then uncovers properties of good titles. For instance, we discover that the vocabulary used in the titles should account for the preferences of the targeted community, yet the titles should be novel compared to the previous submissions of the same content within the community. Such findings have applications when targeting existing content (e.g. movies, books, cars) to new markets, though our findings are sufficiently general that they can be applied even when brand new content is submitted. A Motivating Example. An example of the type of data we collect is shown in Figure 1. Here, the same image (of a bear riding Abraham Lincoln) is submitted to Reddit 25 times, with several different titles, in several communities ('subreddits'). The top plot shows the popularity of each resubmission (the number of upvotes minus the number of downvotes), together with our community model's prediction about how well each submission ought to do without knowing what title was used. From this picture we begin to see the complex interplay between content, community, and title. The choice of title and community has a significant impact on whether a submission becomes popular. The title 'Merica!' (or some variation) is used four times, but is successful only the second time it is used. Thus its success must be due to factors such as the time of day the image was submitted or the community it was submitted to. The purpose of our community model is to account for such factors. We make use of features such as the number of times an image has been previously submitted (resubmissions of the same content are naturally less likely to be popular than original submissions); the community (subreddit) the image was submitted to (submissions to more active subreddits have the potential to become more popular, but also face more competition); and the time of day of the submission (submissions posted during certain hours of the day are more popular). In this example (Fig. 1), the submission is first successful the fifth time it submitted. However, our community model predicts a similar level of success (due to factors such as the subreddit and time of day etc.), thus this initial success was not due to the title. The submission first achieves major success the tenth time it is submitted. Indeed, our community model predicted that the tenth submission would be popular, though it was even more popular than expected. Later, the submission achieves renewed popularity when a novel title is proposed ('God bless whoever makes these'). It is precisely these effects (Fig. 1, bottom) that we use to determine!!! " # " $ % &'('! &)*+, -*!!!!. /' 0 0 &!!! -1 '!!!!!2 0 /' '!!! 3!!!,!!! $4)5# 6!!!# 6 $ % *! $()5# % $ + /!!! 7 % ) )' '!!! # % 7 / % # % 7 / % 8 0 9 7 /,!!!, 0! )1 0!!!. $ '! "!!! /! : : ; <unk> = " 8 + > Figure 1: Submissions of the same image (a bear riding Abraham Lincoln, top right) to Reddit, with different titles. The top plot shows how successful each submission is, along with our community model's prediction of its popularity without knowing the title. Our community model accounts for factors such as the number of previous submissions of the same content, the community (i.e.,'subreddit') the image is submitted to, and the time of day of the submission. which titles and communities are good, having factored out the inherent popularity of the content itself. --- Contribution and Findings Our main contribution is to study the effect that titles, submission times, and choices of community have on the success of social media content. To do so, we introduce a novel dataset of <unk>132K Reddit submissions with a unique property: every piece of content has been submitted multiple times. This number is made up of <unk>16.7K original submissions, each of which is resubmitted <unk>7 times on average. We use this data to disentangle how much of a submission's popularity is due to its inherent quality, and how much is due to the choice of commutniy, submission time, and title. The former part is handled by our community model, while the latter part is handled by our language model. These two models are the main contribution of our paper. The models we develop help us to uncover several novel results. Indeed, we confirm our intuition that good content'speaks for itself', and can achieve popularity regardless of what title is used. Choosing a good title has a secondarythough still important-effect. We find that features such as length, descriptiveness, and even sentence structure can be predictive of whether a title will be successful. We find that the choice of community also plays a major role. --- Related Work Predicting the popularity of social media content is an important problem, and has been approached from many angles. One approach is to use measurements of an item's early popularity, such as view counts on youtube and digg.com (Szabo and Huberman 2010) to predict its future success (Lee, Moon, and Salamatian 2010;Tatar et al. 2011), though others have tried to forecast popularity before an item is submitted (Tsagkias, Weerkamp, and de Rijke 2009; Bandari, Asur, and Huberman 2012). Others have even used such media to predict the outcomes of external events, such as using tweets to predict box-office revenues (Asur and Huberman 2010). Works that predict the future success of social media content include (Bandari, Asur, and Huberman 2012;Tsagkias, Weerkamp, and de Rijke 2009;Yano and Smith 2010). Typically, the goal of such works is to predict the popularity, or the number of comments that will be generated by an article based on its content. Although this is a similar goal to our own, such approaches differ from ours in that they typically rely on the content of the article rather than its title. For instance, in (Yano and Smith 2010) the authors learn that words like 'Obama' are predictive of comment volume: while a valuable insight, this is precisely what we do not want to learn-such content is successful because it discusses Obama-not because it has a clever title or was posted to a particularly appropriate community. Many authors have used Twitter to study the relationship between language and social engagement (Boyd, Golder, and Lotan 2010;Danescu-Niculescu-Mizil, Gamon, and Dumais 2011;Hong, Dan, and Davison 2011;Petrovic, Osborne, and Lavrenko 2011;Suh et al. 2010). For example, in (Artzi, Pantel, and Gamon 2012), the authors use language models (in addition to social network data) to predict which tweets will be successful. This approach is related to our own, in that the authors consider a similar predictive task using lexical features, however the data is critically differentin the case of short messages such as those on Twitter, there is no meaningful distinction between title and content. Another related study is that of (Danescu-Niculescu-Mizil et al. 2012), which considers how phrasing effects the memorability of a quotation. Among other findings they discover that a memorable phrase should use less-common word choices, but should be syntactically familiar. Although their problem setting is quite different, this is similar to our own finding that a successful title is one that employs novel words, yet at the same time conforms to the linguistic norms of the community to which it is submitted. To our knowledge, few studies use Reddit as a source of data (Wang, Ye, and Huberman 2012;Gilbert 2013). How-ever, other works have considered similar'social news' sites (i.e., sites where users rate each other's submissions), such as digg.com (Hogg and Lerman 2010;Lerman and Galstyan 2008;Lerman and Hogg 2010). These works study features such as the relationship between 'visibility' and 'interestingness' (which is somewhat analogous to our study of the relationship between title and content), the effect of 'friendships' on social voting, and aspects of the website design to predict the popularity of social news content (Hogg and Lerman 2010;Lerman and Galstyan 2008;Lerman and Hogg 2010). --- Proposed Method We consider submissions to reddit.com, a community news site where users create, comment on, and evaluate content, which essentially consists of a title and a url. Feedback comes in the form of positive and negative ratings ('upvotes' and 'downvotes'). These ratings are then used to promote content, so that highly rated submissions are more visible (closer to the top of the page). Content can be posted to one of hundreds of communities, or'subreddits'; posting to a large community means that the submission faces more competition, but also that it will be more visible if it becomes successful. --- Dataset Description Rather than studying all submissions, we focus on submissions of images, which consist of a title and an image url. For each submission we obtain metadata including the time of the submission, the user who submitted the image, the community they submitted it to, the number of upvotes and downvotes, and the comment thread associated with the image. Naturally, we could use other types of web content other than images, though using images has several advantages over other types of web data. For example, webpage data may not be static, meaning that multiple submissions of the same url may not actually refer to the same content, while for an image submission we can more reliably detect duplicates. We note that images are the dominant form of content in many of the most active communities on Reddit. To identify resubmissions of the same image, we use a reverse image search tool specifically designed for Reddit.1 This allows us to discover resubmissions of the same image even if the submission urls are not the same. This tool maintains a directory of resubmitted content, from which we collect resubmissions of content dating back to 2008. In total, we collect 132,307 images, 16,736 of which are unique. In other words, each of the images we obtain has been submitted 7.9 times on average. This data consists of roughly 5 million comments, 250 million ratings (56% upvotes, 44% downvotes), from 63 thousand users to 867 communities ('subreddits'). Our dataset is made available for public use. 1 <unk> h i,n <unk>(c h,i = c h,n )<unk> c h,i penalty from communities of previous submissions (Fig. 3, rows) + penalty for submitting to the same community twice (Fig. 3, diagonal) <unk>(c h,i = c h,n )<unk> c h,i A h,i --- Problem Setup and Evaluation Our dataset consists of images, each submitted with a certain title to a certain community. Since we are studying images that are submitted multiple times, we use h to donate a specific image, and V h,n to denote the success of that image the n th time it was submitted. Initially we will assume that this quantity refers to the rating (upvotes -downvotes) of the image, but later we will show that it can straightforwardly be replaced by other measures, such as the number of comments a submission receives. The notation we use throughout this paper is briefly summarized in Table 1. We first develop the community model by considering those aspects of a submission's success that are not related to its title. A significant factor is the choice of community ('subreddit') and the time of day of the submission. Submissions can potentially receive more attention if they are submitted to more active communities, and at busier times of day (though they will also face more competition). Figure 2 (left) shows the average popularity of submissions to some of the most active communities (and one less active one) at different times of day. From this we observe that there are vast differences between the most active communities compared to smaller communities (such as GifSound). There is also an apparent periodicity, though interestingly the peak does not occur when the website is most active (around 8pm UTC). We compute the average popularity of submissions to a community c at time t (in one-hour intervals), avg c,t. This is the quantity depicted in Figure 2 (left). Niche communi-ties, where fewer than 50 submissions are available in any one-hour interval, are combined to use a single background parameter. We use this quantity to normalize our output variable V h,n. Instead of predicting V h,n directly, we predict A h,n = V h,n avg c,t,(2) i.e., how much more popular was this submission compared to others in the same community, at the same time. Initially we performed training to predict V h,n directly, but found that this skewed our evaluation to favor submissions to popular communities. By normalizing the submission in this way, we attempt to treat each submission equally, regardless of whether it was submitted to a popular or a niche community. Our goal shall be to propose a model whose predictions <unk>h,n are similar to the observed values A h,n. We shall measure performance using the coefficient of determination (the R 2 statistic) between A and <unk>, using appropriate train/test splits, though we defer further explanation until our experimental section. --- Community Model Having accounted for the overall activity of each community at each time of day through normalization by avg c,t, we move on to model more subtle community effects that influence a submission's popularity. Our final community model, whose parameters we shall now explain, is shown in Equation 1. Firstly, <unk> h is intended to capture the inherent popularity of the image h, while all other parameters capture how the popularity changes as it is submitted multiple times in different communities. In Figure 2 (center), we observe that on average, content becomes less popular each time it is submitted. We model this using an exponential decay function with importance <unk> h, and decay parameters <unk> and <unk> to be described shortly. We find that the exponential decay effect diminishes with the increase in the time interval (in days) between successive submissions, as shown in Figure 2 (right). To account for this, we use <unk> h i,n to denote the time difference (in days) between the i th and the n th submission of the image h. The presence of 1/<unk> h i,n in the exponent means that the resubmission penalty gradually disappears if there is enough time between submissions. Finally, we want to model the interaction effects between communities, i.e., the cost of resubmitting an image to the same community versus the cost of submitting it to a different community. For instance, if a submission first becomes successful in a high-visibility community (such as the subreddit 'pics'), then it is unlikely to be successful if it is later posted to a smaller community (since users have already seen it). However, this is not symmetric, since a successful submission from a low-visibility community still has a potential audience in a larger community. Figure 3 shows the probability that the n th submission of an image is less popular (A h,n <unk> 1) given that the previous submission of the same image was successful in the community it was posted (A h,n-1 > 1). There are two main observations: Firstly, a submission following a previously successful submission of the same content within the same community is unlikely to be popular, as evidenced by high values along the main diagonal. Secondly, a submission is unlikely to be popular if it has previously been successful in a high-visibility community, as evidenced by high values in rows corresponding to popular communities. In short, content submitted to a popular community will not be successful again, whereas content submitted to a niche community may yet become popular in a different community. Critically, since this interaction matrix is essentially a low-rank matrix plus a diagonal term, it is sufficient to model it using two first-order parameters. The decrease in popularity due to have seen the same content previously in a different community c is modeled by <unk> c, and the penalty for having seen it in the same community is modeled by <unk> c. We fit all of the above parameters by minimizing the leastsquares criterion A-<unk> 2 2 using L-BFGS, a standard quasi-Newton procedure to optimize smooth functions of many variables (Nocedal 1980). --- Language Model Having accounted for the effects coming from the choice of the community and submisison time, we shall now move on to model linguistic aspects of a submission's title. At this point, our goal is to model the impact that a title has on a submission's success. After factoring out the effects captured by the community model, we attribute the residual popularity to the title of the submission: y h,n = A h,n -<unk>h,n. (3) This is depicted in our introduction: Figure 1 (top) depicts our community model's prediction of <unk>h,n, while Figure 1 Figure 3: Community effect on successive submissions of the same content. Rows and columns are Reddit communities; an entry in row x, column y indicates the likelihood that a submission to community y will be (un)popular if the same content was previously submitted to community x and achieved a high popularity; a value of 0.9 indicates that the re-submission will be less popular 90% of the time. (bottom) depicts the residual y h,n of this prediction, i.e., the difference between the actual and the predicted value. Below we propose several features and predictors associated with the titles of submissions that can be used to model the output variable y h,n. Later, we combine these elements in a simple way by training a linear regressor per community with weights associated with each of their predictions. Modeling good and bad words. Each community may appreciate certain choices of words in titles and dislike others. We should be able to aggregate the effect of all the words in the title and associate it with the residual popularity y h,n. In order to achieve this, we extend the supervised LDA framework (Blei and McAuliffe 2007) (a topic of 'bad' words) or a neutral way to the title's popularity in the community. Each word in a given title falls into one of these topics. In order to aggregate the effect of the different words in the title, we model the residual popularity as a function of the proportion of the words in the title belonging to each of the three topics. Further, this function incorporates the notion of accounting for higher (lower) popularity when there are more words in the title which are received favorably (unfavorably) by the community. More formally, our topic model is defined as follows: each title is associated with a topic distribution (a stochastic vector) from which the topic of each word is sampled. A linking parameter <unk> c (per community) relates the affinity scores of each topic to our response variable y h,n so that for each topic we learn which words are likely to generate positive, negative, or neutral values of y h,n. We learn such topics per community, so that each community's choices of words are accounted for appropriately. Our generative process is shown in Table 2. The inference procedure for estimating the various latent parameters in this model is discussed later. Modeling community and content specific words. An important aspect when studying submission titles is'specificity'. Three different kinds of words come into play when phrasing a title-words which are specific to the image h, words which are specific to the community c, and 'generic' syntactic words which tie together the content and community specific words. Understanding how much content or community specificity is necessary for a good title is crucial. Further, this might be different for different communities. In order to quantify the content and community specificity of a given title and associate it with the residual popularity y h,n, we propose a topic model which again extends the supervised LDA framework of (Blei and McAuliffe 2007). Our model associates with each word in the title a latent variable spec p,i which indicates if the word is a generic word (spec p,i = 0), a community specific word (spec p,i = 1) or a content specific word (spec p,i = 2). Further, we associate the residual popularity with the specificity by modeling y h,n as a function of the proportion of the words in the title belonging to each of these three categories. Our model can be formally explained as follows: for each title, we sample a stochastic vector of length three corresponding to the proportion of generic, community specific, and content specific words used in the title. For each word in the title, we sample from this stochastic vector to determine which word distribution to use. In this way we learn topics for each item h, each community c, as well as a back-1. For each title p (image h, sub. n, comm. c) a. Sample <unk> p <unk> Dirichlet(<unk>) b. for each word position i in p i. spec p,i <unk> Multinomial (<unk> p ) if (spec p,i = 0) w p,i <unk> Multinomial (<unk> generic ) else if (spec p,i = 1) w p,i <unk> Multinomial (<unk> community c ) else w p,i <unk> Multinomial (<unk> content h ) c. y p <unk> Normal (<unk> T c <unk> p, <unk> 2 ) Table 3: Generative Process for specificity. ground topic containing generic words. These topic distributions are tied to our output variable by modeling y h,n as a normal variable whose mean depends upon the proportion of content specific, community specific and generic words. and a linking parameter <unk> T c (per community). The complete generative process is shown in Table 3. Topic model inference. The models described above are motivated by the supervised topic models framework (Blei and McAuliffe 2007). We employ a collapsed Gibbs sampling approach (Griffiths and Steyvers 2004) to estimate the latent topics and topic distributions of titles. The hyperparameters <unk> and <unk> (note that <unk> is a prior for <unk>) are set to the values of 0.01 and 0.1/K (where K is the number of topics) respectively. The parameters <unk> and <unk> 2 are estimated by maximizing the likelihood after the latent variables are sampled. The newly estimated parameters are in turn used to sample the latent variables, and this process is repeated until the log-likelihood converges. --- Other Linguistic Features Part of speech tags. Certain communities may favor highly descriptive, adjective laden titles, while others may prefer simpler titles consisting of a few nouns. We use binary features to indicate the presence or absence of each part-of-speech, in order to measure how much they influence the success of a title in each community. Specifically, we consider determiners, pronouns, nouns, adjectives, adverbs, interjections, and prepositions. Sentiment. We found that 'positive' sentiment contributes to a title's popularity in certain communities. In order to capture the sentiment of a title, we employed a hierarchical classification approach (Pang and Lee 2004) which first identifies if a given piece of text is subjective or not and then categorizes the subjective text into positive and negative sentiment classes. This enables us to classify each title into positive, negative and neutral sentiment classes. In this work, we used an off-the-shelf implementation of this algorithm. 3Length of the title. We found that the length of a title does not significantly impact the popularity of a submission's success, unless the title is either extremely long or extremely short. Therefore we use a simple binary feature that indicates whether a title is very short (fewer than four words), or very long (more than sixteen words). Weighted Jaccard similarity. Finally, we consider the Jaccard similarity of the current title compared to previous submissions of the same content. Much as we did in (eq. 1), we learn two parameters <unk>, and <unk> to measure the effect of this feature, depending on whether the post was submitted to the same, or to a different community. Specifically, the parameters <unk> and <unk> measure these effects according to n-1 i=0 <unk> similarity compared to submissions in different communities <unk>(c h,i = c h,n ) J(p h,n, p h,i ) <unk> h i,n + <unk> <unk>(c h,i = c h,n ) J(p h,n, p h,i ) <unk> h i,n similarity compared to submissions in the same community, (4 ) where J(p, q) is the Jaccard similarity between the titles p and q. Our final language model is a linear combination of all the above factors, including predictions made by our supervised topic models. Again, we fit the parameters of our language model by minimizing the least-squares criterion y -<unk> 2 2. --- Quantitative Evaluation We use our models to predict three measures of a submission's success: the rating (upvotes -downvotes) that the image receives, the attention (upvotes + downvotes), and the engagement (total number of comments). In the previous sections we assumed that the output variable V h,n referred to the rating (and our analysis shall focus on this variable), though our methods are easily adapted to other variables. Training proceeds as described in the previous section: we first normalize the output variables V h,n to account for the overall popularity of submissions in a certain community at a certain time, producing A h,n. Our community model produces predictions <unk>h,n, and finally we use our language model to fit the residuals y h,n. We evaluate our models by computing the coefficient of determination R 2 (x, x) = 1 -i (x i -xi ) 2 i (x i -x) 2, (5 ) which measures how accurately our regression fits the data (a value of 1.0 indicates a perfect fit). Since our language model fits the residuals of our community model, we report R 2 (A, <unk>) when evaluating our community model, and R 2 (A, <unk> + <unk>) when evaluating our language model. To ensure that we are not overfitting to our dataset, we also evaluate our model when trained using two train/test splits: in the first split, we train our model for each image h using all but the two most recent submissions of the image h, and then evaluate it on the held-out submissions. Although testing on the most recent submissions most accurately captures the performance we would expect if we were to employ our algorithm 'in the wild' today, it may be biased in the sense that the most recent submissions are unlikely to be the most popular (since they are by definition content that has been submitted many times). For this reason, we also evaluate our model on a randomly selected test set of the same size. In both cases we report R 2 on the test set. Baselines. In addition to our community and language model, we consider the following four baselines: • Simple decay model: Models the success of the n th resubmission using a simple exponential functin <unk> h e -<unk>n. • Per-community decay: Same as the above model, but adds a per-community decay rate. The success of the n th resubmission is modeled as <unk> h exp<unk>-n-1 i=1 <unk> c h,i <unk> • +forgetfulness: Same as the above model, but adds the forgetfulness parameter <unk> h i,n. The success of the n th resubmission is modeled as <unk> h exp<unk>- n-1 i=1 <unk>c h,i <unk> h i,n <unk> • Language-only model: Uses our language model to predict A h,n directly, rather than using it to predict residuals. The language-only model is intended to demonstrate a claim we made in our introduction, namely that trying to train a model to directly predict the quality of a title is not effective, but rather we need to explicitly control for the effect of content and language separately. Results for each of these baselines (as well as our final models) are shown in Table 4. We see that each of our baselines gradually builds towards the performance of our final community model, indicating that each of the components in (eq. 1) is a meaningful contribution to our final prediction. A simple decay model (our first baseline) is ineffective, achieving an average R 2 score across all tasks and testing conditions of only 0.071. Adding per-community decay terms improves this by nearly fourfold, to 0.352. Adding a forgetfulness parameter that gradually reduces the resubmission penalty further improves the performance by 40%, to 0.494. Adding the remaining terms from (eq. 1) further improves the performance by 12%, to 0.556. Combining our language model with our community model further improves the performance by 15%, to 0.639. Thus, while our community model performs well, there is still significant variation that can be explained by modeling the linguistic features of the title. This confirms that a user's choice of title is impactful on the success of that submission. On the other hand, trying to predict the success of our title using our language model alone (i.e., without combining it with our community model) leads to extremely poor performance, with an average R 2 score of only 0.139. Thus we confirm a claim made in our introduction: namely that when trying to model what makes a title successful, it is necessary to control for the quality of the content itself. We see similar performance across all three prediction tasks we consider (rating, attention, engagement). We also see similar performance for different train/test splits, indicating that our model is not overfitting, and ought to generalize well to novel data. In Figure 4 we show performance on the same tasks as those in in order to fit the model, whose performance we evaluate on the entire dataset. We find that around seven submissions is enough to train an accurate model, after which performance does not improve significantly. --- Qualitative Evaluation We now move on to discuss our findings of what makes a good title. Earlier, we suggested that a title should be targeted to a community, in the sense that it should be linguistically similar to other titles used within that community, yet at the same time it should be novel in order to capture readers' attention (Danescu-Niculescu-Mizil et al. 2012). We capture this effect in Figure 5, left. Here, the x-axis shows the fraction of the words in the title of the current submission which are community specific. This fraction is obtained from one of our supervised topic models (generative process shown in Table 3). The y

Creating, placing, and presenting social media content is a difficult problem. In addition to the quality of the content itself, several factors such as the way the content is presented (the title), the community it is posted to, whether it has been seen before, and the time it is posted determine its success. There are also interesting interactions between these factors. For example, the language of the title should be targeted to the community where the content is submitted, yet it should also highlight the distinctive nature of the content. In this paper, we examine how these factors interact to determine the popularity of social media content. We do so by studying resubmissions, i.e., content that has been submitted multiple times, with multiple titles, to multiple different communities. Such data allows us to 'tease apart' the extent to which each factor influences the success of that content. The models we develop help us understand how to better target social media content: by using the right title, for the right community, at the right time.

see similar performance across all three prediction tasks we consider (rating, attention, engagement). We also see similar performance for different train/test splits, indicating that our model is not overfitting, and ought to generalize well to novel data. In Figure 4 we show performance on the same tasks as those in in order to fit the model, whose performance we evaluate on the entire dataset. We find that around seven submissions is enough to train an accurate model, after which performance does not improve significantly. --- Qualitative Evaluation We now move on to discuss our findings of what makes a good title. Earlier, we suggested that a title should be targeted to a community, in the sense that it should be linguistically similar to other titles used within that community, yet at the same time it should be novel in order to capture readers' attention (Danescu-Niculescu-Mizil et al. 2012). We capture this effect in Figure 5, left. Here, the x-axis shows the fraction of the words in the title of the current submission which are community specific. This fraction is obtained from one of our supervised topic models (generative process shown in Table 3). The y-axis shows the probability of the submission being successful. For most communities, we find a 'peak' in terms of how closely a submission should be targeted to that community. On either side of the peak, a submission is either poorly matched to its community (too few community specific words), or is too similar to content the community has already seen (too many community specific words). For a few communities (e.g. 'atheism', 'gaming'), there is no such dip: we might argue that such communities prefer content that is extremely closely targeted to their accepted standards. Figure 5 (center) gives further evidence of this phenomenon. Here we compare a submission title to previous titles used for the same image h. Although using an original title (low Jaccard sim.) is itself not indicative of good per-formance, using a very unoriginal title (high Jaccard sim.) is a strong indication that the submission will do poorly. Finally, Figure 5 (right) shows how individual parts-ofspeech affect the success or failure of a particular title. To compute the probabilities we merely consider the presence or absence of each part of speech in a particular title. Here we see some general trends, e.g. nouns and adjectives impact the success of a title more than verbs and adverbs. However we also see that these effects differ between communities, for instance pronouns, adverbs, and determiners are highly negative in the atheism community. In Figure 6 we visualize the outputs of one of our supervised topic models (Table 2). This topic model finds topics that are correlated with the output variable y h,n A na<unk>ve classifier for the same problem (that directly predicts A h,n from the title) is not successful, and leads to predictors that are composed largely of proper nouns which are specific to popular content. This is no longer an issue in our model: many of the words that our model finds to be effective are general words that could be applied to any title. In Table 5 we look more closely at some of the titles used for a particular submission (the same one from Figure 1), and our model's explanation of why that title was good or bad. The main message from this figure is that the model correctly predicts that the most successful titles have positive attributes, while the least successful titles do not. For instance the title 'MURICA' (short for America) is penalized for not using language that is common for that community, but is rewarded for its originality (at least the first time it is submitted). In practice, we believe that such findings could be reported using an interface where a user proposes a title for their submission and our model outputs its expected success, along with its failings and qualities. In Situ Evaluation. In order to evaluate our models on Reddit directly, we carried out an experiment in which we pooled a sample of 85 images from our dataset and assigned two titles to each of them-one which our language model gave a high score (a 'good' title), and another which was given a low score ('bad' title). We posted the same image with these two titles on Reddit at approximately the same time in two different communities in order to determine whether our notions of good and bad submissions as determined by our models match those of the real world. After one day, we collected rating, attention and engagement metrics for these submissions. The R 2 values for each of these metrics with our community + language model were Figure 5: Three linguistic components that affect a submission's chance of being successful (i.e., more popular than average, or A h,n > 1). Submission titles should be novel, but should still use language that is familiar to the community (left). A submission is more likely to be successful if its title is original compared to previous submissions of the same content (center). Different types of words should be used in different communities (right). 0.565, 0.588, and 0.647 respectively. Overall, the total rating of the 'good' submissions (which is the sum of the ratings of all the 'good' submissions) was about three times higher than that of the 'bad' ones (10,959 vs. 3,438). Moreover, two of our 'good' submissions reached the Reddit front page and three others were on the front pages of their respective communities. 4 --- Discussion Although we have described how our dataset can be used to develop models of whether a title is well-matched to its content and community, we have yet to describe how this type of data might be used in the absence of such a submission history. Firstly, many of our findings are quite general, for instance the fact that a title should match a community's expectations-yet not be too dissimilar to other titles within that community-is a finding that can easily be applied without observing multiple submissions of the same content. Secondly, we believe that there exist real-world applications where one does have access to data such as ours. For example, certain products, such as movies, books, and cars, are given different names when distributed in different markets. With a model such as ours, we could learn how to select a good name for a particular market, based on the success of previous names of that product in different markets. We have concerned ourselves with classifying whether titles are good, though a natural next step is to automatically generate good titles. We believe that these goals are not orthogonal. Current work on title generation is generally concerned with whether generated titles are accurate, in that they meaningfully summarize the content they refer to (Tseng et al. 2006;Jin and Hauptmann 2002;2001). Our methods could be combined with such approaches, to identify titles that are accurate, yet are also interesting. A dataset such as ours could also be used to study other problems where content is a confounding variable. For example, how does the title of a submission influence the tone of a discussion about its content? Are popular users popular 4 In Reddit terminology, a 'front page' refers to the page hosting the most popular content because they submit better content, or is their content more successful merely because they are popular? --- Conclusion When movies, books, or social media submissions become successful, it is difficult to assess how much of that success is due to the quality of the content, and how much is due to having used a good title, at the right time, in the right community. We have developed models that disentangle the interplay between these factors. We proposed a novel dataset of images posted to reddit.com, each of which has been submitted multiple times, with multiple titles. We developed community models to account for effects other than the title, such as the choice of community the image is submitted to and the number of times it has been submitted. We also developed language models that account for how the success of each submission was influenced by its title, and how well that title was targeted to the community. These models allowed us to study features of good title, and to understand when, where, and how a submission should be targeted. --- Atheism Pics Gaming Good words --- Bad words Figure 6: 'Good' and 'bad' words discovered by our supervised topic model, i.e., words likely to lead to a high or a low residual y h,n (respectively). Results are shown for three communities. Word size is proportional to likelihood. 5: Some titles and our model's explanations for their success. Titles are selected from Figure 1. The first two values (SW, SB) indicate whether the title received a high rating, and whether it received a better rating than our community model predicted (i.e., whether it is a 'good' title). The remaining values are four of our language model's predictors. Each predictor's score is summarized using a percentile: top 10% (++); top 25% (+); bottom 25% (-) bottom 10% (--). Danescu-Niculescu-Mizil, C.; Cheng, J.; Kleinberg, J.; and Lee, L. 2012. You had me at hello: How phrasing affects memorability. In ACL.

Creating, placing, and presenting social media content is a difficult problem. In addition to the quality of the content itself, several factors such as the way the content is presented (the title), the community it is posted to, whether it has been seen before, and the time it is posted determine its success. There are also interesting interactions between these factors. For example, the language of the title should be targeted to the community where the content is submitted, yet it should also highlight the distinctive nature of the content. In this paper, we examine how these factors interact to determine the popularity of social media content. We do so by studying resubmissions, i.e., content that has been submitted multiple times, with multiple titles, to multiple different communities. Such data allows us to 'tease apart' the extent to which each factor influences the success of that content. The models we develop help us understand how to better target social media content: by using the right title, for the right community, at the right time.

Introduction The current SARS-CoV-2 pandemic has raised concerns for the homeless in both social service agencies and medical communities, as the homeless are more likely to be both victims and vectors of disease spread. Homeless individuals historically have greater exposure to viral epidemics (influenza as a prime example), as well as greater vulnerability for more severe viral illness, due in part to comorbidities such as pre-existing respiratory and cardiac disease, drug use and diabetes. [1][2][3][4][5][6] Contributory environmental risks include inadequate sanitation, lack of access to healthcare services, crowded living conditions (e.g., shelters, hotels and rooming houses), interactions with legal and penal systems and general lack of personal protective equipment (PPE). 7 Baggett et al. states that "the high number of asymptomatic SARS-CoV-2 infections and the potential for rapid spread in congregate settings support the need for proactive, universal COVID-19 testing strategies" in homeless populations. 3 Testing for both active and recent SARS-CoV-2 infection, via reverse transcription-polymerase chain reaction or antibody detection, are key components of general community prevention. Testing in 4 major U.S. cities between late March and mid-April 2020 found that up to 66% of San Francisco homeless shelter residents and 16% of homeless shelter staff were positive for SARS-CoV-2. 8 Community incidence outside of homeless shelters was highest in Boston, at 14.4%, as a comparison. After a rapid outbreak among 3 homeless service sites in King County, Washington State, the Centers for Disease Control stated that rapid interventions, including testing and isolation to identify cases and minimize transmission, were necessary. 9 A pilot study of COVID-19 testing and support to reduce outbreak risk in Hamilton, Ontario (Canada) used strategies of shelter facility restructuring, daily symptom screening and rapid testing via nasopharyngeal swab (NPS) for those endorsing viral symptoms. These strategies resulted in isolation precautions, but no denominator was given as to whether any residents refused testing or isolation, or what they understood about the virus and how to manage symptoms. 10 Kumar Kar et al. note that homeless mentally ill people have a lack of health awareness and poor help-seeking behaviors due to marginalization. 11 Homeless populations receive preventive screening or test results less often, including for HIV and hepatitis (as well as for cervical and breast cancer in homeless women), despite greater risk. 3,6,10,[12][13][14] Public health officials and homeless service providers emphasize that homeless persons need to follow virus precautions while maintaining usual ongoing health care due to their high risk for medical and psychiatric decompensation. 7 To prevent viral outbreaks in homeless communities, work has included: 1) enhanced communication strategies, 2) infection control actions including isolation quarantine methods, 3) resource allocation, to mitigate against illness exposure and environmental immune stresses (such as lack of sleep and nutrition) and 4) planning for future outbreaks. 15,16 Spirituality is an additional important factor to consider, as a study by Tsai and Rosenheck concluded that religious faith in chronically homeless adults may influence clinical and psychosocial outcomes. 17 The homeless must have basic knowledge and ideally connections with case managers, outreach staff and faith leaders to partner together in mitigation, infection control and ongoing health maintenance. It is, therefore, critical to know the attitudes of homeless individuals about COVID-19 testing and their rates of accepting testing, so that the homeless and their communities may be safely triaged for illness, quarantined if exposed or tested and if positive, isolated, protected and appropriately treated. To better understand the testing attitudes, perceptions, and motivations in the Duval County, Jacksonville, FL homeless population, Sulzbacher Center, a Federally Qualified Healthcare Center, undertook a quality improvement (QI) project. The QI project was performed simultaneously to a clinical testing outreach and a collaborative research study with a university medical center to learn more about current health status and the comorbid health conditions in the Duval County homeless population. The QI project consisted of a questionnaire given before and after giving and reviewing a Centers for Disease Control (CDC) fact sheet on COVID-19 and/ or NPS testing. The information learned from this project was to immediately inform feasible ways to increase testing and compliance with treatment and infection control strategies in predicted future waves or new pandemics. --- Methods The The Bell Homeless Perceptions of COVID-19 Checklist (Figure 1) was given to the shelters and drop-in centers ahead of the testing dates. Those aged 18 years or older, who were homeless residents of the shelters, walk-in homeless or homeless center staff members were asked to participate. Staff members gave a paper form of the Bell Checklist to those who were interested in being tested for COVID-19 before the NPS testing. The form also included non-identifiable demographic information and reasons why they may be agreeing to testing. Clinical staff then reviewed key points of the CDC education form about the COVID virus (see Appendix A) or gave out the CDC education form about the COVID virus. Those who were not considered a staff member were provided with a $10 gift card at the end of the data collection. Testing and survey collections were performed by staff from Sulzbacher Center, Changing Homelessness, UF Health (including paid, previously furloughed, physical therapists) and Or- The top reasons all participants as a group wanted to get the NPS testing (beyond believing it would help them or others, which was endorsed when stating a desire to be tested) (Table 2) were: 1) "I'm curious about the test and/or the results" (47.8%), 2) "The shelter encouraged me to go" (37.6%), 3) "My faith will protect me from the virus" (31.3%), 4) "I'll get a gift card" (24.2%), 5) "My family and/or friends have said I should" (14.1%), and 6) "My healthcare team said I should" (8.1%). Reasons to not be tested included: 1) "I am already isolated from society" (20.3%), 2) "I am afraid it may hurt" (14.8%), 3) "I'm worried about what will be done with my information" (3.3%), 4) "I don't really care -there are other things that worry me more" (2.9%), and 5) "I don't really want to know" (0.4% For those who wanted testing, the most endorsed attitudes were: 1) "I'm curious about the test and/or the results"; 2) "My faith will protect me from the virus"; and 3) "The shelter encouraged me to go." (Table 3) For those who didn't want testing, the top attitudes were: 1) "The shelter encouraged me to go"; 2) "I'm curious about the test and/or the results"; and 3) "I'll get a gift card." All the attitude respons- es differed between those who wanted NPS testing compared to those who did not want the test, based on Pearson Chi-Square Tests (chi-square 130.467, df 11, sig. 0.000; Table 3). Further subanalysis teased apart individual differences in attitudes between the two groups. First, whether or not participants wanted the gift card did not influence their desire to get testing (chi-square 0.45, df 1, sig 0.5). Second, people who wanted testing were more likely to believe that faith would protect them from the virus (chi-square 28.558, df 1, sig. 0.000). People who were not faith-influenced were less likely to want testing. --- Discussion This quality improvement project was unique in that it examined perceptions, interest, and motivations of homeless people to seek or receive active viral testing during a pandemic. The Sulzbacher Center undertook this project with the aim not only to test clinically for SARS-CoV-2, but to understand the beliefs and perceptions about why homeless individuals agreed to be tested, with a client-centered, collaborative, quality improvement focus. Too often the homeless population is ignored or directed to comply with edicts. If their attitudes are considered, organizations that serve the homeless will be better informed to approach and partner with homeless individuals on major health crises, such as a pandemic, with true buy-in and informed consent. The quality improvement study was conducted in Jacksonville, Florida, a state that did not expand Medicaid. It is not surprising therefore that the majority of individuals (66.9%) in this study lacked insurance. Black/African American was the most common race identified in this homeless population, which nationally makes up 40 percent of the homeless population despite representing only 13 percent of the general population, due to factors such as poverty, rental housing discrimination, incarceration and lack of mental health treatment. 19 The percentage of males and females in this study (66.0% and 27.1%, respectively) were in keeping with the annual point-in-time count data for Florida in 2019 (64.8% male and 34.9% female). 19 Although the top reason for the total number of homeless persons to get the nasopharyngeal swab testing was curiosity, for those not wanting testing, the most common attitude was being encouraged by the shelter to go. Some attitudes varied by site, with some sites taking a more mandatory approach to testing (CRM McDuff, Trinity/Clara White and CRM State), --- Limitations The number of homeless tested at the Sulzbacher Village was lower than expected, as well as least robust in data collection. Also of note, the mobile medical bus and internet were experiencing technical difficulties that day. Other limitations of this project include realworld behaviors. Different shelter sites had variable approaches toward encouraging testing by staff; this may reflect the reality that some staff and sites have more paternalistic or authoritative attitudes rather than encouraging self-actualizing behavior, or may themselves may be fearful of COVID and its complications for themselves and their clients and understandably want their clients tested. Therefore, bias in attitudes of those completing testing may be present. The current study did not a priori characterize or analyze homeless shelter sites by the subgroups of populations served. Those in a rest stop may be considerably different than those in an overnight shelter. Those in the Sulzbacher Village were by definition more likely to be women. Additionally, the project did not include homeless adolescents. They are another important subgroup whose attitudes often differ from adults. The survey itself was brief and not psychometrically designed or tested for validity or reliability. It also lacked a full range of attitudes to inquire about, including more positive variables, potentially creating a bias in valence of response. There is little literature with which to compare this project to. The majority of publications about beliefs, attitudes, and perceptions are toward the homeless, rather than of the homeless. The closest comparison is with an influenza mobile outreach immunization program for vulnerable populations in Melbourne, Australia. Although over half the group had at least one risk factor for severe influenza, 60% had not received an influenza vaccine the prior year, with most reporting that they were 'not worried about influenza' as their reason. Those who were pregnant listed a health care provider's recommendation as the most frequently given reason for why they desired immunization. 21 These results for a potentially similar respiratory viral outbreak indicate that paternalism and cultural and educational variables affect health care literacy and promotion and alter perceptions of preventive care in vulnerable populations. Findings from this quality improvement study support considering a variety of strategies to encourage participation in testing events with large numbers of homeless individuals. These include: education, gift cards, shelter staff encouragement (especially for retesting), local faith leader involvement and increased support from the community, to encourage self-isolation, when necessary, and to help the homeless meet self-care and basic needs. The best strategies may be specific to the shelter and their staff and clients, as some between-group differences are noted for site of participation. Future projects can ideally allow adequate planning time so that focus groups including homeless persons, shelter staff and health care workers from a variety of settings may together discuss survey items and how to best test strategies to promote compliance with pandemic recommendations for screening, prevention and treatment. --- Conflicts of Interest Dr. Gracious reports personal fees from Novo Nordisk, outside the submitted work. Drs. Bell and Bosi declare they have no conflicts of interest. The authors are employees of Orange Park Medical Center, a hospital affiliated with the journal's publisher. This research was supported (in whole or in part) by HCA Healthcare and/or an HCA Healthcare affiliated entity. The views expressed in this publication represent those of the author(s) and do not necessarily represent the official views of HCA Healthcare or any of its affiliated entities. --- Author Affiliations

Attitudes of those in the homeless population toward testing, particularly during a pandemic, are critical to understand, so that they and their communities may be safely triaged and protected. Homeless persons are more likely to be exposed during viral epidemics, and have greater vulnerability for more severe viral illness, due to greater medical comorbidities. The literature reflects a dearth of published papers describing the perceptions, interest, and motivations of homeless people to seek or receive viral testing, despite their status as a high-risk population.A quality improvement project consisting of a cross-sectional survey took place at 8 SARS-CoV-2 infection testing sites (local shelters and drop-in sites) within Duval County, Jacksonville, FL.The vast majority of homeless individuals approached for testing completed demographic data and a checklist of beliefs and attitudes about testing (N=764) and underwent COVID-19 nasopharyngeal swab testing (n=679). Mean age was 48; 66% were male, and the predominant race was Black (51%) with 89% of non-Hispanic ethnicity. Of the total participants, 59.2% wanted testing and 4.6% declined testing. Attitudes toward testing varied by site and by wanting vs. not wanting to be tested. Top reasons in those wanting testing included curiosity; a belief that faith would protect them from the virus; and having shelter encouragement to be tested. Top attitudes among those not wanting to be tested were: shelter encouraged me; curiosity; and receiving a gift card. For the total group of subjects, being offered a $10 gift card did not affect their desire to be tested. Those who were not faith-influenced were less likely to want testing.Findings from selected literature and this quality improvement study support the use of a variety of strategies to encourage participation in testing events with large numbers of homeless individuals, including education, gift cards, shelter staff encouragement, involvement of local faith leaders and more broad support by the community. An additional qualitative study would complement these findings, as populations appear to differ in beliefs and attitudes depending on their location and other demographics. Motivational strategies to influence testing rates can be fine-tuned if beliefs, perceptions and attitudes are better understood.

Background There is convincing scientific evidence that exposure to air pollution, in particular ambient particulate matter (PM), is related to both short and long-term health effects. Increased mortality and hospitalizations for cardiopulmonary causes have been noted in several studies evaluating the short-term effects of PM 10 or PM 2.5 [1]. Some longitudinal studies conducted in the US and in Europe did find a consistent association between long term exposure to air pollution and natural mortality, especially for cardiovascular diseases [2][3][4][5][6][7]. In urban areas a relevant source of air pollution is vehicular traffic as the importance of other sources (e.g. industry, power plants) is declining. In some of the studies evaluating health effects, proxy measures of exposure have been used, such as distance from busy roads, distance-weighted traffic density or the length of main street segments within a buffer from home [8][9][10][11][12]. These proxy measures tend to be correlated with other more objective measures (for example, NO 2 concentrations estimated by dispersion modelling and/or land use regression models) and are relatively easy to obtain using Geographic Information System (GIS) [13]. Some studies have directly assessed the relationship between living along busy roads and mortality or morbidity. For example, Hoek and colleagues found, in a cohort of adults aged 55-69 years, that cardiopulmonary mortality was associated with living near a major road [5]. In the SAPALDIA (Swiss Cohort Study on Air Pollution and Lung Diseases in Adults) study an increased risk of regular phlegm and wheezing in non smokers living within 20 m of a main street was found [8]. All the literature has been recently reviewed by a scientific panel from the Health Effects Institute (HEI panel on the Health Effects of Traffic-Related Air Pollution) in the US [14]. The Panel concluded that the evidence was "sufficient" to infer a causal relationship between exposure to traffic and exacerbation of asthma and "suggestive but not sufficient" to infer a causal relationship with onset of childhood asthma, non-asthma respiratory symptoms, impaired lung function, and total and cardiovascular mortality. When evaluating the relationship between living along busy roads and mortality/morbidity, the potential role of socioeconomic status should be considered. The relationship between socioeconomic position and exposure to air pollution has been extensively reviewed and "environmental justice" concerns have been raised [15][16][17][18][19][20][21]. In fact, low socioeconomic groups of the population seem to suffer from the worst environmental conditions, including poor air quality [18][19][20][21][22][23][24], and tend to be more vulnerable to air pollution [19,[25][26][27]. However, the extent to which low socioeconomic position and proximity to busy roads are related is not well-defined. Few studies have characterised subjects living near high traffic roads in urban areas to better understand the exposure characteristics and the susceptibility factors of air pollution. The lack of studies is somewhat surprising given the potential public heath impact of traffic exposure. We established a large cohort of people who were resident in Rome in 2001 with information on socio-demographic characteristics and baseline health status to study long-term effects of traffic-related air pollution. We wished to use diverse indices of socioeconomic position (both at the individual and neighbourhood level) and different geographic information system (GIS) measures to characterize the residents who live close to busy roads in Rome. In the present paper we describe the main baseline characteristics of the cohort and examine the association of exposure to traffic with socio-demographic characteristics. --- Methods --- Setting Rome is the largest Italian city with a population of about 2.6 millions inhabitants on a surface of 1290 km 2. It is a radiant city, and the most important roads are still the ancient roman roads that starting from the centre, the Roman Forum, connect the city with the rest of the country in all the directions. During the last century, the urban development in Rome took place gradually from the centre to the suburbs, with a higher population density in the centre compared to the periphery [28]. --- Study Population The cohort was defined from Rome Municipal Register's data. We enrolled all residents of Rome on the 21 st October 2001; data were available on gender, age, and residential history. Using a variety of record-linkages procedures, under strict control to protect individual privacy, we collected additional information for each study member. In particular, individual data from the 2001 Census were used for indices of socioeconomic position at the baseline. The 2001 residential addresses were used to estimate environmental exposures and traffic-related air pollution indices. Individual hospital admissions from public and private hospitals in Italy, during the period 1996 to 2001, were available to provide the morbidity history of the subjects. --- Area-based and individual information on the socioeconomic position A composite area-based index of socioeconomic position (SEP) by census block was built using the 2001 Census of Rome. Briefly, we used 4,888 census blocks with at least 50 inhabitants (average population: 500 subjects) as the units of observation. We considered census information that represented various socioeconomic parameters (occupation, education, housing tenure, family composition, and foreign status (yes or no)) and each census block was characterized. We performed a factor analysis to create a composite indicator, and we used the quintiles of its distribution in census blocks to obtain a 5level area-based index [29]. To obtain the index for all census blocks of Rome, we assigned a SEP level to census blocks with fewer than 50 inhabitants (0.4% of the population) according to the levels of contiguous blocks. The area based SEP has been validated with individual census data [29] (for example in the highest category of area based SEP there was 29% of people with a university degree vs. 5% in the lowest category of SEP), the index is highly correlated with a small area income index based on 1998 Tax Register data [30], and it has been associated with overall and cause-specific mortality and incidence of specific diseases such as stroke [31,32]. Figure 1 shows the map of the city by SEP. From the 2001 Census we obtained individual data on educational level (grouped into four categories: University, High school, Secondary school, Primary school), employment status (Employed, Looking for first employment, Unemployed, Student, Housewife, Retired, Military or civil service, Unable to work, Other), occupation (Non-manual: Managers, Highly-skilled, Medium-skilled, Unskilled; Manual: Services, Farmer, Highly-skilled, Medium-skilled, Unskilled; Military forces). --- Environmental characteristics at the residential address Geographical information system (GIS) indices were developed for each individual. We geocoded each subject's residence as of 21 st October 2001 using the interpolation method within road segments. To locate the address on the map we used the Italian road network (Tele Atlas, Italy). The City Council of Rome provided the traffic data for all major roads in Rome as of 2005, i.e. 6,585 road segments which represented the 26% of all roads, and included the totality of roads with more than 10,000 vehicles per day (2,228 segments). We defined as high traffic (HTR) roads all road segments where at least 10,000 vehicles travelled per day. Figure 2 shows the map of Rome with the HTRs. We defined different GIS indicators for each residential address: the distance from the residence to the nearest HTR, the total length of the HTR segments within a 150 m buffer zone, the daily average traffic counts from the closest HTR within 150 m, and the traffic density within 150 m. The latter was defined as the sum of the products of each HTR segment length by the estimated annual average daily traffic count of the HTR segment (within the 150 meter buffer zone around the residence address) [10], divided by the area of the buffer: We also defined a categorical variable of traffic density within 150 m as the quartiles of the distribution of the continuous variable. Similar to the SAPALDIA study [8], we applied buffers of different radii (50, 100, 150, and 250 m) to the residences and intersected the buffers with the list of high traffic roads to create a categorical variable indicating distance to HTR (high traffic road more than 250 m, between 150 and 250 m, between 100 and 150 m, between 50 and 100 m, and less than 50 m away). We calculated a five category variable of total length of high traffic road segments within a 150 m buffer zone as the quartiles of the sum of segments' lengths within the 150 meters buffer (none, low <unk>166 m, medium 166-266 m, high 266-323 m, very high 323-1445 m). We collected and stored all geographical variables using ArcGis 9.1 (ESRI, Redlands, California, USA). We used the Word Geodetic System of 1984 with the Universal Transverse Mercator 33N as the coordinate system and map projection. --- Baseline health status: individual morbidity history In Italy there is a National Health Service that provides medical care to all the population. The morbidity history of the study population was based on data from the Health Information System of the Lazio region, where Rome is located. The regional Health Information System collects individual discharge records from all hospitals, both public and private. All the records are linkable using a unique identifier, but privacy protection is assured from strict management rules. Discharge records are routinely collected and contain: patient demographic data, admission and discharge dates, up to six discharge diagnoses (International Classification of Disease, 9 th revision, Clinical Modification [ICD-9-CM]), medical procedures or surgical interventions (up to six), and status at discharge (alive, dead, transferred to other hospital). In order to describe the baseline health characteristics of the cohort, we used hospital discharges from 1996 to 2001 to identify those individuals who had at least one hospitalisation. We considered hospitalisations for all causes excluding accidents, those with a principal diagnosis of cardiovascular diseases (ICD-9-CM: 390-459), with principal or secondary diagnoses of hypertension (ICD-9-CM: 401-405), with principal diagnosis of ischemic heart disease (ICD-9-CM: 410-414), congestive heart failure (ICD- 557.9, V43.4), arrhythmia (ICD-9-CM: 426.0, 426.13, 426.7, 426.9, 426.10, 426.12, 427.0-427.4, 427.6-427.9, 785.0, 996.01, 996.04, V45.0, V53.3), cerebrovascular disease (ICD-9-CM: 430-438); for cancer (ICD-9-CM: 140-239); diabetes (ICD-9-CM: 250), and chronic obstructive pulmonary disease (ICD-9-CM: 490-492, 494, 496). --- Statistical analysis We studied the association of traffic variables with age and socioeconomic position. We considered two SEP indices (area-based SEP and individual level of education) and three traffic measures (a binary variable to identify who lives within 50 meters of an HTR, the distance from HTR, and the traffic density within 150 m). We used logistic regression to evaluate the associations with living close (50 m) to an HTR (Odds ratios, OR, with 95 percent confidence intervals were calculated). We log-transformed both the distance from an HTR and the traffic density to obtain two normally distributed variables, and we used them as dependent variables in a multivariate linear regression analysis. For both logistic and linear regression models, the independent variables were age (0-17, 18-34, 35-64, 65-74, 75+ years), education, and area-based SEP. In the presentation of the results of the linear regression models, we calculated the exponential function of the regression coefficients in order to estimate the ratio of the dependent variable in the specific subgroup compared to the reference group (geometric mean ratio, GMR). We calculated 95 percent confidence intervals of the GMR. To take into account the clustering of the subjects within census blocks, we performed all multiple regression analyses with robust variance estimate. As a final step, we performed a stratified analysis dividing the entire population by area of residence (inside and outside the central railway ring) to better understand the relationship between age, small area SEP, and education with traffic exposure. --- Results The total population of Rome in October 2001 was 2.56 million (2001 Census) people. For 2,118,670 residents (84%) we had information on the address of residence and on individual socioeconomic position. We selected only those individuals whose address was matched and geocoded by the Tele Atlas system with optimal quality, resulting in 1,898,898 subjects that were considered in this study. There were no significant differences between subjects included and excluded from the analysis for age, and gender. Table 1 shows the socio-demographic characteristics of the study population in five age groups. Fifteen percent of the study population was under 18 years of age, while 19% was aged 65 or more. The population by age group was not evenly distributed across the area-based socioeconomic (SEP) index: the youngest groups were more likely to live in low SEP areas compared to the oldest group. Also level of education increased with age. Descriptive data of exposure to traffic by age are reported in table 2. The subjects' homes were an average of 272 meters (standard deviation, SD, 459) from a high traffic road. A total of 34% of the population lived further than 250 meters from an HTR, while 45% had an HTR within 150 meters, and 17% lived closer than 50 m from an HTR. The daily average traffic count from the closest HTR (within 150 m from home) was 21,533 vehicles (SD 15,094). The majority of subjects with an HTR within 150 meters had traffic counts ranging from 10,000 to 15,000 while a minority had more than 30,000. For residents living closer than 150 meters from an HTR, the mean traffic density was 107 vehicle meters of HTR/m 2 (SD 88). Table 3 shows the percentages of subjects who were hospitalised for specific diseases during the period 1996-2001, before the enrolment in the study. Overall hospitalisations include births but not accidents. Twenty-nine percent of the study population had been admitted to hospital in the five years before enrolment (5% for cardiovascular causes). The percentage with a previous hospitalisation increases with age, reaching 16.5% for hypertension in the 75+ age group. Table 4 shows the association (odds ratios and geometric mean ratios) of age, area-based socioeconomic position, and level of education with the three measures of traffic exposure (living 50 meters from an HTR, distance from HTR and traffic density within 150 m from home). The table reports odds ratios or GMR adjusted only for sex and age group and adjusted for all the factors in the table. We observed that older people were more likely to live in high-traffic areas and the association was confirmed for the three traffic indicators when adjustment was made for the SEP variables. Twenty-one percent of subjects living in medium SEP census blocks had a high traffic road within 50 meters of home, versus 18% of high area-based SEP, and 12% of low-area based SEP. The results of the multivariate analyses indicated that those belonging to the medium area-based SEP index level were the most likely to have heavy traffic within 50 meters. In addition, those living in the lowest two area-based SEP categories were less likely to live close to an HTR compared to others even when age group and educational level were taken into account. The results were similar when we used the continuous variables indicating distance from HTR and traffic density as the exposure measures. All three indicators of traffic exposure showed that less educated subjects were less exposed to traffic than more educated residents; these associations were attenuated when area-based SEP was taken into account. Table 5 shows the association between SEP and traffic exposure by area of residence (inside and outside the central railway ring, corresponding to 298,326 and 1,600,572 cohort members, respectively). The central part of Rome, delimited by the railway ring and corresponding to the historical centre, is characterized by a resident population older (65+ years: 24.3% versus 18.5%) and with a higher socioeconomic position than the rest of the city (both at individual and area level; e.g. for high area-based SEP: 35.9% versus 16.4%; for low area-based SEP: 3% versus 23.7%). In both the areas there is evidence of association between older age and living in proximity of an HTR, however, the association between traffic exposure and SEP had a different sign in the two areas. In the city centre where traffic is higher and 25% of the residents lives close to HTR, less affluent and less educated people tend live closer to HTR than more affluent and highly educated people, the opposite is seen in the rest of the city (where traffic is lower and only 15% lives close to HTR) in agreement with the overall results. --- Discussion The study clearly shows that a large fraction of the Rome population is exposed to traffic at home. This is not surprising when one considers that out of a population of 2.6 million inhabitants, almost 2.3 million vehicles were circulating in the city in 2001, including moped/motorcycles, and commercial vehicles. The public health problem of traffic is amplified as one considers that in 2001 a large proportion of the vehicles were in the high emissions categories (EURO 0 and EURO 1) and that 17.4% of them were diesel-powered vehicles. Exposure to traffic is not evenly distributed in the city: we found that all the measures of traffic exposure were higher in the elderly population than in younger groups. In general, traffic was not higher in the poorest sector of the population, on the contrary, low SEP was associated with greater distance from HTRs and with lower traffic density within 150 m from residence. However, when we restricted the analysis to the city centre, higher exposures to traffic were found among subjects in the low socioeconomic groups. In any case it appeared that area level more than individual level drives the association between SEP and traffic exposure. The explanation for the traffic exposure and age and SEP relationships is clearly related to the urbanization history in Rome with the consequence that people of medium-high social class are located in the more central and prestigious areas of the city where there are more high traffic roads. The prices of dwellings in the city centre are much higher than in the suburbs and the poor air quality is rewarded by prestigious buildings, good quality of public transportation, and access to several facilities. This clearly explains why elderly are more exposed to traffic air pollution: in the past it was still affordable for young couples to buy a flat in the city centre, but now the value of dwellings is so high compared to the suburbs that in the last decades young new families have tended to live in the periphery. The same reasons explain the different socioeconomic distribution of population by area of the city, new residents in the city centre tend to be very well-off whereas less advantaged people move from the centre to the suburbs for economic reasons. It is interesting, however, that restricting the analysis to the city centre, the relationship between SEP and traffic exposure changes, with low exposures in residents better educated or who live in a high SEP area, suggesting that environmental inequalities do exist in the central area. As indicated before, the large number of people residing close to busy roads in Rome underlines the need for a public health attention to the issue. Unfortunately, there are few comparable measures from other cities in the world to evaluate whether other places have similar distribution values to what we found in Rome. In the Netherland cohort study, almost 5 percent of the study members lived within 50 meters of a busy road, whereas it has been reported that the proportion of the population living within 50 meters from a major road in Los Angeles and Toronto is approximately 32% [6,14]. A meta-analysis of the available evidence indicates that for most pollutants there is a distant decay gradient in the range of 100-400 meters from the source and therefore the population with direct exposure to traffic-related pollution is even higher [33]. Our results on the association between SEP and traffic exposure were similar to those from a study in the Netherlands [5], but in contrast with the results from studies on environmental equity from other countries, especially the US [15,[34][35][36]. However a study by Buzzelli and Jerrett found that in Toronto racial minority groups were less exposed to air pollution than other groups, and that dwelling values predicted total suspended particulates levels in Canada [37]. Most of the early evidence that people of lower socioeconomic status are more exposed to air pollution comes from studies that have considered point sources, mainly industrial settlements [14]. On the other hand, the recent literature available on exposure to traffic is more limited. Comparability is also difficult given the use of diverse measures of exposure to air pollution, and different measures of socioeconomic position. In the studies conducted in the US, it is taken for granted that disadvantaged groups of the population live in the worst environmental conditions including high exposure to outdoor air pollution [17]. It should be recognized that the literature on the issue is still limited and additional observations are needed from different parts of the world. Our finding in Rome is not new. In a previous investigation on short-term effects of PM 10 in the city, we pointed out that emissions of PM, CO, NOx, and benzene were higher in areas of higher socioeconomic status [25]. In addition, in a large cross-sectional study of adults, GIS indices of exposure to traffic and estimated NO 2 level at the residential address from a land use regression model were directly linked with socioeconomic position [13]. Despite the socioeconomic gradient of exposure, the short-term mortality effects of PM 10 in Rome were larger in the low socioeconomic groups perhaps as a result of a different susceptibility level of poor people rather than to different proximity to traffic [25]. On the other hand, the fact that higher socioeconomic groups live in high traffic areas does not necessarily mean that they are more exposed than underprivileged groups are. More advantaged people work in air-filtered office buildings than outside, they use private transportation more than public transportation, they are more likely to live closer to their work places, they are more likely to spend the week-ends outside the city than low socioeconomic groups. Clearly, investigations on personal exposure to air pollutants are needed to better qualify the issue. Important new evidence has been accumulating on the effects of long-term exposure to traffic-related air pollutants. Although the relationship between traffic-related air pollutants and cardiovascular mortality and morbidity has not been clearly defined yet, it is clear that particulate matter has an effect on the cardiorespiratory system [14,38]. An important study on long-term effects of PM 2.5 on the cardiovascular system was conducted in the US [39]. The follow-up included more than 65,000 postmenopausal women (Women Health Initiative) without previous cardiovascular disease in 36 US metropolitan areas from 1994 to 1998. Each 10 <unk>g/m 3 increase in PM 2.5 was associated with a 24% increase in the risk of a cardiovascular event and a 76% increase in the risk of death from cardiovascular disease. A report from Oslo has linked long-term exposure to traffic-related pollutants to cause-specific mortality [40]. A consistent effect on all causes of death was found for both sexes and age groups; the effects were particularly strong for COPD. The study shows that people with COPD disease and the elderly seem to be affected by air pollution at lower levels than the general population. Important results are available for myocardial infarction as its occurrence [9,41] and survival have a strong links with particulate matter exposure [42]. This study has some limitations. First, we used proxy measure of exposure at residence rather individual measures of exposure to traffic related pollutants. Second, we used the interpolation method for geocoding instead of a building/parcel method, and this could result in an inaccurate exposure characterization, especially in the periphery of Rome where street segments are longer than in the centre. Finally, when we geocoded the addresses of participants living in a multi-storey building, we assigned the distance to the front door to all residents in the building, irrespective to their real position. The results of our investigation with respect to the history of previous hospitalization should be interpreted for descriptive purposes only. They are useful to evaluate the specific vulnerability of the population in relation to the environmental exposure and they indicate that from a public health point of view there is a subpopulation that is more susceptible that needs to be protected in a more intensive way. --- Conclusions We have shown that living nearby traffic is a widespread phenomenon in Rome, and that people who live close to traffic are older, better educated, and reside in higher socioeconomic areas than those who live far. The study population will be followed to better understand the potential health effects of traffic-related air pollution exposure. --- Abbreviations HTR: high traffic road; SEP: socioeconomic position; GIS: geographic information system; PM: ambient particulate matter; SD: standard deviation; OR: odds ratio; GMR: geometric mean ratio; CI: confidence interval; CVD: cardiovascular diseases; COPD: chronic obstructive pulmonary disease; CO: carbon monoxide; NOx: nitrogen oxide. Authors' contributions GC participated in the design of the study, performed the analysis, and drafted the manuscript. CB performed all GIS analysis. VR provided data on traffic and helped to discuss the results. ED provided GIS data, and helped in interpreting the results. CAP participated in coordination and interpretation. All authors read and approved the manuscript. FF participated in the design and coordination of the study, and helped to draft the manuscript. --- Competing interests The authors declare that they have no competing interests. and emissions data to predict the spatial distribution of traffic-related air pollution in Rome.

Background: Subjects living close to high traffic roads (HTR) are more likely to suffer from air-pollution related morbidity and mortality. The issue has large public health consequences but few studies have described the main socio-demographic characteristics of people exposed to traffic. Objectives: To characterise a large cohort of residents in Rome according to different measures of traffic exposure, socioeconomic position (SEP), and baseline health status. Methods: Residents of Rome in October 2001 were selected. Individual and area-based SEP indices were available. GIS was used to obtain traffic indicators at residential addresses: distance from HTR (> = 10,000 vehicles/day), length of HTR, average daily traffic count, and traffic density within 150 meters of home. Hospitalisations in the 5-year period before enrolment were used to characterise health status. Logistic and linear regression analyses estimated the association between traffic exposure and socio-demographic characteristics. Results: We selected 1,898,898 subjects with complete SEP information and GIS traffic indicators. A total of 320,913 individuals (17%) lived within 50 meters of an HTR, and 14% lived between 50 and 100 meters. These proportions were higher among 75+ year-old subjects. Overall, all traffic indicators were directly associated with SEP, with people living in high or medium SEP areas or with a university degree more likely to be exposed to traffic than people living in low SEP areas or with a low level of education. However, an effect modification by area of residence within the city was seen and the association between traffic and SEP was reversed in the city centre. Conclusions: A large section of the population is exposed to traffic in Rome. Elderly people and those living in areas of high and medium SEP tend to be more exposed. These findings are related to the historical stratification of the population within the city according to age and socioeconomic status.

used to denote significant differences between racial and ethnic groups and differences between groups by selected characteristics. This study was reviewed by CDC and was conducted consistent with applicable federal law and CDC policy. <unk> <unk> Nearly 10% of adults aged <unk>45 years (9.6%) reported experiencing SCD in the past 12 months (Table 1). SCD prevalence increased among successive age groups and was lower among adults with health insurance (9.5%) than among those without health insurance (11.6%) and decreased with increasing formal educational attainment. The highest overall percentages of SCD occurred among adults who did not have a high school diploma (16.4%), those who had been married, but were not currently married (13.6%), and those aged <unk>75 years (13.3%). Age-adjusted SCD was higher among AI/AN (16.7%) and Hispanic adults (11.4%) than among White adults (9.3%) and was lower among A/PI adults (5.0%); prevalence among Black adults (10.1%) was similar to that among White adults (9.3%). This pattern of racial and ethnic differences was observed across most demographic subcategories examined. Unadjusted overall prevalence of SCD was highest in Alabama (14.3%), Oklahoma (14.1%), Florida (13.6%), Louisiana (13.6%), West Virginia (13.6%), Tennessee (12.9%), and New Mexico (12.8%), and lowest in Illinois 6.1% (Table 2). Because of small sample sizes, estimates for A/PI and AI/AN adults were unstable in most states. * Except for age-specific and overall unadjusted estimates, estimates were age-standardized to the 2000 projected U.S. Census Bureau population aged <unk>45 years using three age groups: 45-64, 65-74, and <unk>75 years. https://www.cdc.gov/nchs/data/statnt/statnt20.pdf <unk> Defined as the self-reported experience of worsening confusion or memory loss in the preceding year. <unk> Estimates are aggregated to include the most recent survey year for each jurisdiction. These estimates include data collected from jurisdictions in 2020 (Alaska, Arizona, Arkansas, California, Delaware, District of Columbia, Hawaii, Idaho, Illinois, Kentucky, Maine, Michigan, Nevada, New Hampshire, New York, North Carolina, Ohio, Oregon, Utah, Vermont, Washington, Wyoming, and Puerto Rico); 2019 (Alabama, Connecticut, Florida, Georgia, Indiana, Iowa, Kansas, Louisiana, Maryland, Minnesota, Mississippi, Missouri, Nebraska, New Mexico, North Dakota, Oklahoma, Pennsylvania, Rhode Island, South Carolina, South Dakota, Tennessee, Texas, Virginia, West Virginia, and Wisconsin); 2018 (New Jersey); 2016 (Massachusetts and Montana); and 2015 (Colorado). <unk> Persons of Hispanic or Latino (Hispanic) origin might be of any race but are categorized as Hispanic; all racial groups are non-Hispanic. ** Total study population includes respondents in the five racial and ethnic groups (206,025) and an additional 9,381 adults reporting multiple races and other race categories. <unk> <unk> The estimate for this racial and ethnic group differs statistically from that of White adults (p<unk>0.05). <unk> <unk> Reference group for comparisons for groups defined by a characteristic within the overall population and within each racial and ethnic group. <unk> <unk> The estimate for this group differs statistically from the reference group defined by this characteristic (p<unk>0.05). *** Dashes indicate estimate is not reported because relative SE >30% or sample size <unk>50. Among 21,299 respondents with SCD, 47.3% reported talking with a health care professional about confusion or memory loss; women (50.7%) were more likely than men (43.3%) to do so (Table 3). Overall and within racial and ethnic groups, adults with SCD symptoms who were less likely to talk to a health care professional about their symptoms were aged <unk>75 years, had less education, did not have health insurance, did not have a personal doctor, and had not visited a doctor in the past year. --- Discussion Prevalence of SCD varied across adults by demographic characteristics and race and ethnicity. Among racial and ethnic groups, SCD was lowest among A/PI adults and highest among AI/AN adults. Prevalence of SCD was higher among persons with less formal education than among college graduates across all racial and ethnic groups. This finding is consistent with other studies that suggest that persons with more years of formal education have a lower risk for dementia than do those with fewer years of formal education (5,6). Low prevalence of SCD among adults with higher education suggests that education might be protective against SCD. More research is needed to better understand the roles that education and related systemic factors play in sustaining cognitive health, particularly across diverse racial and ethnic populations. For example, modifiable risk factors for ADRD are less prevalent among adults with higher education, differ among racial and ethnic groups, and are associated with high prevalence of SCD (2). The findings of this study can help health care providers identify groups of patients who would benefit from risk reduction behaviors and further cognitive assessment. Persons who talked with a health care professional about SCD were more likely to be women, had at least some college education, were aged <unk>75 years, had a personal doctor, had a doctor visit within the past year, and had health insurance. Public health strategies are needed to support access to health care for persons who lack access to routine health care or to have a designated preventive health care professional. For example, programs such as Welcome to Medicare <unk> <unk> and Medicare Annual Wellness Visit for adults <unk> <unk> aged >65 provide coverage for preventive care screenings including cognitive assessments. Health care providers could consider asking patients as young as age 45 years about experiences of worsening memory loss or confusion during visits to initiate discussions about early signs of dementia and strategies to reduce risk and sustain cognitive health. The findings in this report are subject to at least three limitations. First, sample sizes for some racial and ethnic groups were too small to detect statistical differences, particularly at the state level. Second, BRFSS represents the noninstitutionalized adult population only and therefore cannot be generalized to institutionalized adults. Finally, BRFSS data rely on self-report rather than medical examination records; survey questions about cognitive decline may be subject to recall and social desirability biases and responses may reflect cultural differences. However, the self-perception of cognitive decline has been shown to discriminate preclinical ADRD from normal aging (7). Early detection and diagnosis are important to rule out conditions other than ADRD that might be treatable, and to establish a care plan to manage co-morbid conditions and avoid unnecessary hospitalizations. These potentially avoidable hospitalizations can be both costly and detrimental to quality of life, especially given that some persons with SCD <unk> <unk> https://www.medicare.gov/coverage/welcome-to-medicare-preventive-visit --- ** https://www.cdc.gov/brfss/annual_data/2020/pdf/2020-response-ratestable-508.pdf; https://www.cdc.gov/brfss/annual_data/2019/pdf/2019response-rates-table-508.pdf; https://www.cdc.gov/brfss/annual_data/2018/ pdf/2018-response-rates-table-508.pdf; https://www.cdc.gov/brfss/annual_ data/2017/pdf/2017-response-rates-table-508.pdf; https://www.cdc.gov/brfss/ annual_data/2016/pdf/2016moduleanalysis.pdf; https://www.cdc.gov/brfss/ annual_data/2015/pdf/2015moduleanalysis.pdf * Except for age-specific and overall unadjusted estimates, estimates were age-standardized to the 2000 projected U.S. Census Bureau population aged <unk>45 years using three age groups: 45-64, 65-74, and <unk>75 years. https://www.cdc.gov/nchs/data/statnt/statnt20.pdf <unk> Defined as the self-reported experience of worsening confusion or memory loss in the preceding year. <unk> Estimates are aggregated to include the most recent survey year for each jurisdiction. These estimates include data collected from jurisdictions in 2020 (Alaska, --- Summary What is already known about this topic? Subjective cognitive decline (SCD), self-reported memory loss or confusion that is occurring more frequently, might be a symptom of early-stage dementia. What is added by this report? During 2015-2020, approximately 10% of adults aged <unk>45 years reported SCD, with the highest prevalence among American Indian or Alaska Native adults (16.7%); prevalence declined with increasing formal educational attainment. Fewer than one half of persons with SCD had discussed their concerns about SCD with a health care professional. What are the implications for public health? Discussing changes in cognition with a physician is important to identifying potentially treatable conditions, obtaining early diagnosis and detection, developing support systems for caregivers, and establishing a treatment or care plan to help adults remain healthy and independent for as long as possible.

Subjective cognitive decline (SCD), the self-reported experience of worsening or more frequent memory loss or confusion, might be a symptom of early-stage dementia or future serious cognitive decline such as Alzheimer disease* or a related dementia (ADRD) (1). Established modifiable risk factors for ADRD include high blood pressure, inadequate physical activity, obesity, diabetes, depression, current cigarette smoking, and hearing loss ( 2). An estimated 6.5 million persons aged ≥65 years in the United States live with Alzheimer disease, the most common dementia (1). This number is projected to double by 2060, with the largest increase among non-Hispanic Black or African American (Black), and Hispanic or Latino (Hispanic) adults (1,3). Using data from the Behavioral Risk Factor Surveillance System (BRFSS), CDC assessed racial and ethnic, select demographic, and geographical differences in SCD prevalence, and prevalence of health care professional conversations among those reporting SCD. The age-adjusted prevalence of SCD during 2015-2020 was 9.6% among adults aged ≥45 years (5.0% of Asian or Pacific Islander [A/PI] adults, 9.3% of non-Hispanic White [White] adults, 10.1% of Black adults, 11.4% of Hispanic adults, and 16.7% of non-Hispanic American Indian or Alaska Native [AI/AN] adults). College education was associated with a lower prevalence of SCD among all racial and ethnic groups. Only 47.3% of adults with SCD reported that they had discussed confusion or memory loss with a health care professional. Discussing changes in cognition with a physician can allow for the identification of potentially treatable conditions, early detection of dementia, promotion of dementia risk reduction behaviors, and establishing a treatment or care plan to help adults remain healthy and independent for as long as possible. * Although the term "Alzheimer's disease" is frequently used, this report uses "Alzheimer disease" in accordance with the American Medical Association Manual of Style, 11th Edition, and MMWR style. BRFSS is a random-digit-dialed annual landline and cellular telephone cross-sectional survey of noninstitutionalized U.S. adults aged ≥18 years. † The BRFSS six-question cognitive decline optional module (4) was administered to adults aged ≥45 years by all 50 states, Puerto Rico, and the District of Columbia at least once from 2015 to 2020; for states that implemented the module in multiple years, the most recent year of data was used. § To maximize sample sizes for each † https://www.cdc.gov/BRFSS/index.html § Most recent data collection year for each jurisdiction: 2020 (Alaska,

The socio-economically disadvantaged people living in the cities of the UK, northern Europe and North America have relatively poor health, and I propose that an important and potentially correctable reason is a relatively low exposure to sunlight and thereby insufficiency of vitamin D. The inequalities of health within society were obvious in the 19th and early 20th centuries. One of the objectives of the NHS, introduced in the UK in 1948, was to reduce the inequalities but by 1974 this had not been achieved and Townsend reported that the mortality gap was widening. 1 In 2010, the UK Office for National Statistics (ONS) reported that although the health of all social groups had improved, that of the socio-economically advantaged had improved at the greatest rate. In other words, the social gap had widened still further. 2 The USA has a similar social gradient of disease and premature death, but this is compounded by an ethnic factor that is so obvious in the USA. The poor have worse health and earlier death than the affluent, especially if both poor and black. 3 However, a recent study showed that bad outcome after heart surgery is more the result of socio-economic disadvantage than ethnicity. 4 Successive UK governments have set objectives to close the social gap, but how to achieve this has been a challenge. The simple explanation is that! The Author 2011. Published by Oxford University Press on behalf of the Association of Physicians. All rights reserved. For Permissions, please email: [email protected] the socio-economically underprivileged have poor health because they have little money, but how money converts into good health in the UK is far from clear. The possible importance of exposure to the sun and thereby vitamin D synthesis has not been officially acknowledged. To demonstrate the potential importance of vitamin D, I will apply the causal criteria of Sir Austin Bradford Hill, namely: strength of association, consistency of association, temporality, biological gradient, plausibility, coherence, experimentation and analogy. 5 It is also necessary to examine paradoxes. --- The social-health problem Being relatively poor in the UK results in a surprisingly wide range of diseases and premature mortality, described well by Professor Sir Michael Marmot of the London School of Hygiene and Tropical Medicine. Marmot led the Whitehall study of 17 530 civil servants and noted that the 10-year mortality rate was three times higher in those with the lowest grading compared with the administrators who had the highest grading. 6 He and his co-authors concluded: ''For every cause of death except genitourinary disease.... the two lower grades have higher mortality risks than the two higher grades. For nearly every cause there is a step-wise relation between grade and mortality.'' The specific diseases noted in this study included coronary heart disease (CHD), lung cancer, other cancers, stroke, bronchitis, accidents and suicide. On re-investigation 10 years later, there was found to be an overall improvement but increasing divergence. 7 The same phenomenon was shown in Sweden, with a mortality risk in the unskilled workers twice that in the professional/executives group. 8 In Sweden and in England and Wales the prevalence of reported long-term illness was higher in the socio-economically disadvantaged, with a 3-fold gradient across social groups. 9 The plight of the long-term unemployed was not considered in these studies, probably because of very low unemployment rates at that time. The UK Office of National Statistics (ONS) tells us that in 2005 a man living in a prosperous location would be expected to live to an average age of 77.4 years and would have a healthy life expectancy of 66.2 years, whereas a man in one of the least prosperous locations of the UK would be expected to live on average to 71.4 years and have a healthy life expectancy of just 49.4 years. 10 Historically, rickets and tuberculosis were more common in the poor, and emerged in particular in the workers of the new industrial cities. Glasgow was perhaps the world's most important industrial city at the end of the 19th century. Its population developed perhaps the highest incidence of rickets and tuberculosis, and a century later the highest incidence of coronary heart disease. 11 It was first thought that CHD was most likely to occur in the business executive, leading to the philosophy of work blaming. It became clear later that coronary events and mortality (including very early mortality) were more common in the socioeconomically disadvantaged. 12,13 This has generally been interpreted as victim blaming, that the poor are responsible for their own poor health because of their behaviour, especially cigarette smoking and dietary characteristics. 14 Heart failure was also found to be more common in the socioeconomically disadvantaged (7.2 vs. 6.4 per 1000 population) with a reduced survival (mean 2.8 vs. 3.5 years). 15 Lung cancer is more common in the socioeconomically disadvantaged, 6 but this is only in part the result of more cigarettes smoked. A study of 4913 men in Denmark identified that for a given number of cigarettes smoked over a long period of time, the risk of dying from lung cancer was three times higher in the most socio-economically disadvantaged group compared to those who were least disadvantaged. 16 Full adjustments were made for factors such as age and inhalation habit. Socioeconomic disadvantage must be acting through a biological factor to produce such a dramatic difference in mortality rate. This factor has not yet been defined. Death rates from several other cancers are greater in the socio-economically disadvantaged. 6 Not only are cancers more common in this group but survival is worse. This has been demonstrated in the UK for all cancers, there being a gradient of survival from the most affluent to the most deprived. 17 The 5-year survival of malignant melanoma is 10% less in the socio-economically disadvantaged, 17 with better survival known to be in those with higher blood levels of vitamin D. 18 Diabetes also has a social distribution in favour of the affluent. A study from Liverpool investigated the variation of diabetes prevalence in the various localities within the city. It demonstrated a gradient, the prevalence of diabetes increasing with socioeconomic disadvantage. 19 Peptic ulcer mortality has also been shown to be two or three times greater in the socio-economically disadvantaged, both in the USA and in England and Wales. 20 It has been pointed out from Canada that the 'poverty is a greater predictor of heart disease than risk factors such as smoking, obesity, stress, or blood cholesterol concentrations'. 21 It appears therefore that being socio-economically disadvantaged does not lead to a specific disease but that it makes individuals susceptible to a number of diseases, and it leads to a worse prognosis. It is as though body defence mechanisms are somehow impaired. --- The poverty paradoxes In northern countries poverty is associated with poor health whereas in sunny southern countries this does not appear to be the case. For example, Albania has the lowest per capita income in Europe and a high infant mortality rate (greater than 40 deaths per 1000 live births compared to less than 10 in the UK). But Albania has one of the lowest mortality rates for CHD in Europe, similar to Greece, Italy and Portugal. 22 The low cardiovascular mortality in these countries has been attributed to a Mediterranean diet but when this is seen in context with other geographical data it seems likely that the Mediterranean climate itself may have benefits. A poverty paradox was also observed in studies comparing Scotland and India in the early years of the 20th century. Dr HS Hutchison together with his assistant SJ Shah, undertook a particularly detailed and important study in the Nasik district of the Bombay Presidency. 23 A house-to-house survey identified rickets to be present in 13.7% of children in Class 1 Hindu families and 10.2% of Class 1 Moslem families. This compared with 0.3% of children in the poor Class 2 Hindu families and 2.7% of Class 2 Moslem families. They also noted tuberculosis to be rampant in the wealthy families and rare in the poor families. Their interpretation was that the better health of the poor families was the result of outdoor activity and fresh air, whereas the women in particular in wealthy families were able to practice purdah and stay indoors. Dr Hutchinson had undertaken his studies in response to detailed studies of rickets and tuberculosis in Glasgow undertaken by Dr Margaret Ferguson, who was working for the Medical Research Council. 24 Dr Ferguson had found that rickets and tuberculosis were particularly common in the poor in Glasgow and frequently co-existed. Whereas in India lack of outdoor activity in clean air was considered to be the cause of the serious health problems in the wealthy, in Glasgow this was a characteristic of the poor. Further evidence for the poverty paradox comes from a study conducted in India almost a century later. In this case the subject was diabetes, and it was found to much more common in those who were working, educated, prosperous and living in an urban environment, but rare in the illiterate and poor. 25 This is the opposite of what is found in the UK. 19 We can see therefore that whereas in the UK and other northern latitude countries (Sweden, Canada, USA) the socio-economically disadvantaged have the worst health, in Albania they do not have a health disadvantage and in India they have in some respects the best health. In other words, whereas poverty is clearly associated with increased ill health in northern latitudes, it does not have this association in sunny countries closer to the equator. The wide range of diseases found to excess among the socio-economically disadvantaged in northern countries suggests that it might be futile to look for a common cause. It might be more productive if we were to search for a common susceptibility factor that would lead to the suppression of body defence mechanisms. Three analogies will help us to understand the high incidence of disease in the socio-economic disadvantaged in northern Europe and North America. --- Evidence from geography It is known that latitude of residence is associated with health risk. Within the UK there is a gradient of risk of death from CHD with minimum mortality rate in the south-east and maximum in north-west of England, in Scotland and in Northern Ireland. 26 This cannot be explained on the basis of diet or other conventional factors. 27 Within France there is a similar latitude effect for deaths from myocardial infarction (MI), with the age-adjusted mortality rate being higher in Lille (21 per 1000) compared with Toulouse (11 per 1000), while Strasbourg (18 per 1000) takes an intermediate position in both latitude and mortality rate. 28 Within Europe there is a mortality gradient from the south to the north-west, with maximum mortality in Scotland and Finland. 29 The variation cannot be explained on the basis of diet. The most detailed study compared middle-aged men in Belfast with those of Toulouse. The health differences were striking, with much higher mortality rates from all causes, CHD and cancer in Belfast. 30 Apart from a higher rate of cigarette smoking among middle-aged men in Toulouse, there were no significant dietary or social patterns. 28 Simply living in Toulouse rather than Belfast appeared to deliver a health advantage. And it is not red wine: evidence from the Albanian paradox tells us that the country with the lowest wine consumption Europe is among those with the lowest CHD mortality rate. An obvious factor explaining the influence of latitude on health is the sunlight energy reaching ground level, and this of course determines the synthesis of vitamin D in the skin. It is interesting to note that in the USA there is an inverse relationship between altitude of residence and cardiovascular disease. 31,32 UVB which promotes synthesis of vitamin D in the skin is absorbed by the atmosphere and so increases in strength with altitude. This evidence supports the view that sunlight, acting via vitamin D, promotes cardiovascular health. Cancers also have an incidence related to latitude, first identified in North America in 1941. As latitude of residence increased from 10-30 to 50-60 degrees north, the mortality rate due to cancer increased by a factor of 2.5. 33 Death rates from breast and colon cancer show a gradient increase from the south of Europe to the north. 34 Lifetime risk of developing adenocarcinoma of the oesophagus is maximal in Scotland, reducing by 60% in France and the USA, with England and the Netherlands in intermediate positions. 35 Survival of men from cancer is better in the south of England than in the north, 36 also specifically from colon cancer. 17 Within the UK there is an additive effect of social class and geography, as identified in the 2001 Report of the Chief Medical Officer of the Department of Health. 37 The age-standardized death rate in the north-west compared with the south-east for a person in social Class I was only minimally higher. But the gap widened with progression along the social scale. In social Class V the age-standardized death rate was twice as high in the north-west compared with the south-east, 1080 vs. 540. Mortality in Scotland is higher in all social classes than it is in England, and a number of diseases, notably multiple sclerosis, are more common in Scotland than in England. 38 The analogy of geography strengthens the proposal that relative deficiency of vitamin D is responsible for a high incidence of several diseases and that it may account for much ill health of the socio-economically disadvantaged. --- Evidence from seasonality The second important physical factor is the variation of illness, disease and death throughout the year, and there are many examples of this. Excess deaths in the winter months are particularly well recorded, mainly because death is more accurately measured than illness. 39 It is recorded in many temperate countries in both hemispheres. 40 Specific diseases identified as having a higher incidence or mortality rate in the winter include respiratory illnesses and CHD. 41 Tuberculosis has its peak incidence in the early spring, when the minimum levels of vitamin D allow its activation. 42 The spring peaks of tuberculosis notifications in India are more obvious in northern states, where there is an overall higher incidence, 43 again suggesting a vitamin D effect. In the first half of the 20th century in western Scotland there was an increased incidence of perforated peptic ulcer during the winter. 44 There are some lesser known and perhaps surprising influences of season on health. An autumn or early winter birth following late gestation during the summer gives the offspring a significantly reduced risk of developing multiple sclerosis 45 and diabetes 46 in later life, as well as a later age of menopause. 47 The analogy of season again suggests an association between vitamin D insufficiency and ill health supporting the proposal that relative deficiency of vitamin D is responsible for much ill health of the socio-economic disadvantaged in high latitude countries. --- Evidence from ethnicity Ethnicity is an important factor determining life expectancy and the incidence of a wide range of diseases, including CHD, diabetes, chronic renal failure, tuberculosis, rickets and osteomalacia. The major ethnic groups at risk in North America and Europe are South Asians, African Blacks and Caribbean Blacks. In the UK the incidence of tuberculosis is much higher in all these ethnic groups than the national average, in recent years it has been especially high in African Blacks as a result of AIDS. 48 Rickets is still seen in both ethnic Black and South Asian children. 49,50 They share a high incidence of diabetes 51 and also chronic renal failure independent of diabetes, 52 which in both ethnic groups is linked to low blood levels of vitamin D. 53 The high incidence of death from CHD in both groups is a serious problem. In ethnic South Asians in the UK it is a greater problem in the second generation immigrants. 54 Skin pigmentation is an obvious common factor between these ethnic groups, and the reduced efficiency of vitamin D synthesis by pigmented skin 55 suggests that vitamin D deficiency might be critical. There is also an additive effect of ethnicity and geography. Within Great Britain it is men living in Scotland who have the highest CHD risk, but among these it is the international migrants (mainly from South Asia) within Scotland who have the highest risk of all, 50% higher than the Scottish average and five times higher than those immigrants who settled in the south of England. 56 The analogy of ethnicity strengthens the proposal that relative deficiency of vitamin D is responsible for poor health of the socio-economically disadvantaged. --- Vitamin D studies Blood levels of vitamin D are no longer just a research tool but part of clinical practice, leading to an increasing awareness of the extent and significance of vitamin D deficiency. The results of important recent studies have received little attention. In a USA study, the all-cause standardized mortality risk of the lowest vitamin D quartile was 80% higher than the highest quartile. 57 The lowest quartile had a blood level of vitamin D as 25(OH)D <unk>40 nmol/l (17.8 ng/ml) and the highest quartile >70 nmol/l (32.1 ng/ml). The overall range is higher than in UK studies, reflecting the fact that the vitamin D levels are lower in the UK. This may be expected because the UK is much further north, New York being on the same latitude as Madrid. A UK study also used quartiles and provided a great deal of social information. 58 The social classification was the usual I-V but in addition identified the unemployed, known to have the worst health. Profound vitamin D deficiency was defined, as in clinical practice, as 25(OH)D <unk>25 nmol/l (<unk>10 ng/ml). In the winter it was found in 25% of the unemployed, twice that of social Classes I and II. It is remarkable and of great concern that one quarter of the unemployed had such profound deficiency of vitamin D during the winter months, and about 8% during the summer. It is at this level of vitamin D that osteomalacia can be expected to occur. The amount of time spent out of doors is important in determining adequacy of vitamin D levels. The mean summer level of vitamin D in those spending <unk>30 min per day out of doors in the UK was found to be 50 nmol/l (20 ng/ml) compared to 65 nmol/l (28 ng/ml) in those spending >4 h per day out of doors with a gradient between these two extremes. 58 The low vitamin D levels of poor people can be explained by the observation that many of them spend less time outdoors: 'The poorest spend around 11 hours per week more than the richest watching television.... A cultural underclass might be developing as those on low incomes are restricted to the cheaper entertainment of television while those on higher incomes take advantage of other leisure activities'. 56 Television viewing is also associated with low levels of vitamin D, most obvious in the summer months, and once again there is a gradient effect. 59 A north/south gradient in vitamin D deficiency is found in the British Isles. Profound vitamin D deficiency was found in 23.5% of the population sample in Scotland in the winter, compared with 12.5% in the south of England. In the summer the corresponding results were 8.3% and 2.7%. 58 The observation of a health gradient within France 28 and comparison between Toulouse and Belfast 30 did not include vitamin D measurements or assessments. This was an unfortunate omission but vitamin D measurements within France have been performed subsequently. The result is as expected, that vitamin D levels correspond to latitude, with the highest levels in the south of France, the lowest in the north and a gradient between the two extremes. 60 The highest morbidity and mortality rates are found where the blood levels of vitamin D are lowest. The social and physical geography of vitamin D levels generally match observations of health data. The seasonal effect on vitamin D levels also matches the seasonal variation in illness and mortality. The UK peak mean vitamin D in September was found to be 68 nmol/l (29 ng/ml), falling to 32 nmol/l (13 ng/ ml) minimum in February. 58 Finally the ethnic dimension. There are many reports of the re-emergence of rickets in the UK 49 and the USA, 50 usually in South Asian or Black ethnic children. Vitamin D deficiency is well established in these groups, a significant proportion of whom are socio-economically disadvantaged. Studies of vitamin D status have shown their blood levels to be particularly low. For example in a US study, 86% of white-skinned people had a blood level >70 nmol/l (>30 ng/ml) compared to only 4% in Black and 3% in Mexican Americans. 57 --- Behavioural factors The analogies between the geographic, social and ethnic patterns of illness cannot be ignored, and are suggestive of a common factor-vitamin D deficiency. There are obviously overlaps between ethnicity and socio-economic disadvantage, both in Europe and North America. Poor survival following cardiac surgery in the USA suggests that social factors are much more important than skin colour. 4 If vitamin D is the key factor then this observation is entirely plausible. There is however some controversy as to the efficiency of vitamin D synthesis in pigmented and non-pigmented skin. Recent carefully controlled research suggests that pigmented skin might have an equal ability to synthesize vitamin D as white skin. 61 In other words it is possible that vitamin D deficiency in South Asians and Blacks in the UK and North America is due not just to skin pigmentation but also to behavioural factors of deliberate or cultural sun avoidance, as appears to be the case with the white socio-economically disadvantaged. Perhaps maximizing sun exposure to the skin is genetically determined behaviour in white-skinned people. It is obvious however that dark-skinned people have no incentive to sit or lie in the sun to develop a tan. They are culturally adapted to cover the skin so as to protect themselves from the equatorial sun, and Moslem women have religious/cultural pressures to cover their skin almost completely. Hindus are usually vegetarian and obtain little or no dietary vitamin D, and this has been considered to be the reason for the excess risk of tuberculosis in Hindus compared to Moslems in London. 62 It is the enforced behaviour of the socioeconomically disadvantaged in the UK and other northern latitude countries that leads to little sun exposure and therefore vitamin D deficiency. Residence is likely to be in inner city areas with more atmospheric pollution. Housing is likely to be without a garden (having a garden is associated with higher vitamin D levels 63 ) and there is unlikely to be the opportunity for outdoor leisure or a usable local park. The families particularly of the long-term unemployed will be unlikely to have access to the countryside or appropriate clothing to blend into this social environment. Leisure is likely to be indoors, focussed on the television. There is unlikely to be sufficient money for holidays in the sun, even in the present era of unprecedented opportunities for travel. Relative deficiency of vitamin D is thus inevitable. --- Hill's criteria So far Hill's criteria have been met. I have demonstrated the strength of associations between socioeconomic deprivation, poor health with premature death and vitamin D deficiency. I have demonstrated the consistency of association and a biological gradient of effect. I have provided analogies of the geography and seasonality of vitamin D deficiency and a variety of illness linked to it. I have also demonstrated the important ethnic dimension of interactions between poor health and vitamin D deficiency, which are similar to those of socioeconomic disadvantage. Temporality is incomplete at the present time as widespread blood testing for vitamin D is only just becoming available. We have seen that blood levels of vitamin D are associated with life expectancy. 57 We know that the lowest blood levels of vitamin D are associated with the highest risk of subsequent multiple sclerosis, 64 and similarly associated with the subsequent development of colorectal cancer. 65 There are other measures of temporality. We know from previous observational studies that a low dietary intake of vitamin D is associated with a subsequent increased incidence of disease, for example multiple sclerosis. 66 We also know that vitamin D supplement in only minimal dose provides future protection against low birth weight, 67 diabetes, 68,69 multiple sclerosis, 66 prostate cancer (ultraviolet radiation exposure), 70 colorectal cancer, 71 and it improves outcome from lung cancer following surgery, with an additional beneficial effect from diagnosis and treatment during the summer. 72 What is not yet known is whether vitamin D supplements given to the socio-economically disadvantaged, in a dose adequate to optimize blood levels, will close the social gap. This is the potential for future research. The synthesis of 7-dehydrocholesterol (7-DHC) by primitive organisms such as plankton became a fundamental step in evolution. By a physical process a single bond of this molecule is split by the action of ultraviolet radiation from the sun to form vitamin D. Its wide-ranging effects are only just being appreciated but they include not just bone development but immune competence, 73 control of malignant process 74 and modulation of gene expression. 73,75 The model presented is coherent and plausible. I do not suggest that vitamin D deficiency causes directly any disease other than rickets and osteomalacia, but it does lead to a form of immunodeficiency syndrome, by no means as severe as that caused by the human immunodeficiency virus (HIV), but much more common. This leads to susceptibility to a number of disease processes-acute infections, chronic infections, inflammatory diseases and several cancers. If we wish to improve resistance so as to minimize disease then attention to vitamin D deficiency should be a priority. This would be particularly important if we wish to close the social health gap, something that has failed spectacularly in the past 60 years. Hill's final criterion is experimentation. This is generally considered to be the randomized controlled trial, but this will only take place if there is a plausible basis for it, effectively meaning that the other criteria have been met. It is only this that will ultimately determine if the social gap can be closed by giving attention to vitamin D, starting during gestation. The difficulty would be placebo control with informed consent, knowing that vitamin D is a natural vitamin component of food and not a drug, and that low blood levels are associated with significant health disadvantage. Clinical trials of drug treatment are well established, usually after the introduction and use of the pharmaceutical agent, but disease prevention by a natural food component presents a much greater organizational challenge. 76 A safe and effective dose of vitamin D would be 2 000iu per day, 77 but a trial should ideally involve adjusting the vitamin D supplement given to an individual so as to achieve a target blood level of 100-150 nmol/l (40-60 ng/ml). Nature has provided us with an experimental model-vitamin D receptor (VDR) polymorphisms. Vitamin D in its fully activated form calcitriol 1,25(OH) 2 D is a hormone that combines with VDRs, complex intracellular proteins that are genetically encoded, creating a heterodimer that activates vitamin D-responsive elements (VDREs) concerned with gene expression. This is particularly important within T-lymphocytes leading to a rapid defence process in response to infection. 73 If defective VDR polymorphisms are inherited, then although the blood level of vitamin D might be ideal the effects of vitamin D deficiency will be seen. The most obvious and well recognized are vitamin D resistant rickets and osteomalacia. However several other conditions are associated with VDR polymorphisms and these include tuberculosis, 78,79 advanced prostate cancer, 80 occult hepatitis B virus infection, 81 Crohn's disease, 82 colon cancer in ulcerative colitis. 83 Such experiments of nature, by virtue of determination at the moment of conception, provide temporality to complete Hill's criteria. --- Conclusion Conventional thinking provides no answer to the poor health of the socio-economically disadvantaged. Marmot, who has done so much to identify the problem, suggests that the reason is lack of perceived status within peer group and within society. 84,85 Although this sounds superficially attractive, it is based on psychological models which themselves are very doubtful. The link between 'lack of status', disturbed physiology and pathology lacks plausibility-there is no obvious mechanism based on existing scientific principles, and Hill's criteria are far from fulfilled. It is an untestable hypothesis, pseudoscience. 86 There have been in the past attempts to identify a common cause for the so-called diseases of western civilization, including CHD, stroke, diabetes and cancers. The principle was 'association is a clue to causation', 87 and the suggested culprit was diet, in particular the industrial refining of carbohydrate. This idea was superficially attractive and had an effect on eating patterns. However it lacked the power to explain so much and it had to compete with the diet-cholesterol-heart hypothesis, despite the many paradoxes and shortcomings of the latter. Associations can be more than coincidence. For example the increased risks of various cancers with pre-existing diabetes, and their worse outcome, are not easily understood, but in a recent review the social dimension and the possible role of vitamin D were not considered. 88 It must be remembered that in science there is no such thing as absolute truth and Hill's criteria are not absolute. Science works by paradigms, models of understanding that are the most coherent and plausible at the time, models that explain most observed phenomena, and which have the fewest paradoxes. The social divergence of health and premature death is at present in the pre-paradigm era: there is no understanding and there has been a total failure of helpful intervention. I propose on the evidence presented that relative vitamin D deficiency is the major reason for the widening social gap of health in the UK. There are of course other behavioural factors such as drug-related deaths and accidents that have other explanations, but these are not so readily recognized or reversible. 89 This is the challenge for the future, but today it is possible to identify and correct vitamin D deficiency. This could be met by population-based vitamin D supplement, but ideally it should be targeted and corrected on an individualized basis, as with diabetes and hypertension. --- Conflict of interest: None declared.

The strong association between socio-economic disadvantage and poor health in northern Europe and North America causes great public health concern but as the mechanism is not understood, government actions have been completely unsuccessful. The social gap continues to widen. Aim: I propose that an important mechanism involved is relative deficiency of vitamin D in the socio-economically disadvantaged people. Design: The evidence is presented by explaining apparent paradoxes, together with the analogies of geography, season and ethnicity. Vitamin D studies are also demonstrated. Method: The picture is put together according to Hill's criteria of causation: strength of association, consistency of association, temporality, biological gradient, plausibility, coherence, experimentation and analogy. Results: Hill's criteria of causation are met, acknowledging that they are pragmatic and not absolute. The associations between vitamin D deficiency and socio-economic deprivation are supported by analogies of geography, seasonality and ethnicity. There is also biological plausibility. The paradoxes of Albania and India show that health disadvantage and socio-economic disadvantage are not necessarily connected, if relative deficiency of vitamin D is the mediating biological factor.The poor health and average early death of the socio-economically disadvantaged in the UK, together with the widening social gap of health, can be explained by relative deficiency of vitamin D. At present there is no explanation and attempts by government to close the social gap have failed. The understanding presented gives an important opportunity to improve the health disadvantages of the socio-economically disadvantaged.

Introduction This paper focuses on what mightbeunderstood as social and psychological resistanceto changing gender relations and transforming masculinities. Concerned primarily with work on or with Black masculinities in South Africa, the argumento ffered here is, however, informed by the international and nationalresearchliterature and activism that has focused on challenging, deconstructing and transforming dominantf orms of masculinity by engaging boys and men towards progressive masculinities and gender justice (e.g., Connell1 995; Kimmel, Hearn, and Connell 2005;Morrell 2001;Morrell, Jewkes, and Lindegger 2012;Segal 1993;Shefer, Stevens, and Clowes 2010). Overt he lastt wo decades, concern for masculinities has permeated South African social and health sciencesaswell as the humanities (e.g., Clowes, Ratele, and Shefer2013;Morrell1 998;Reid and Walker 2005;Shefere ta l. 2007;Vetten and Ratele 2013). This focus on men and boys has been evident in programmaticinterventions as well as activist work, just as it has been globally (e.g., Greig and Edstro <unk>m2 012; Minerson et al. 2011;Peacock2 012;Stern,P eacock, and Alexander 2009;World Health Organization 2007). The turntomen and boys has not gone without resistance, contestation and contradictions, however. --- Why resistance? The term resistanceisused in various ways in the health and social sciences, and the range of practices described by the term is broad.Resistance can be against oppressive structures (such as hegemonic hetero-sexist masculinity), but it can also be against practices that subvert oppression (for instance, against alternative, non-heteronormative practices)(see Govender 2011). In addition, resistancec an be collective or individual in character. Drawing upon insights from psychoanalytic theory, resistancehere denotes that which opposesw hat we seek to overcome ( Rycroft 1968( Rycroft /1972)). In psychoanalysis, resistance refers to that which we have to work through -o ritrepeats itself in various guises (Freud 2001(Freud /1958)). The intriguing aspect of resistancet hus conceived is that it highlights the fact that even if structuresand practices of masculinities may lead to physical,social and psychological injury to complicit men themselves (besidesthe violenceagainst women and children), men and boys might still cling to violent or hetero-patriarchal masculinitiesjust as some people hold on to their illness (Freud 2001(Freud /1958)). Even thought he psychoanalytic notion of workingt hrough resistance( rather than dismissing it) looks potentially like adevice for engaging men and boys towards gender equality, in this case, resistancei ss till apprehendedi ns ocial terms (without neglecting the psychological). Hence, insteado fd efining'resistance as the process and act of challenging one's subordinated position in agivensocial system ', as Haslamand Reicher (2012, 155) have done, resistance here is takena sareactionary social and psychological opposition to challenges to existing gender relations and hegemonic forms of masculinity. Resistance is worth examining because it is, as Foucault(1990) observed, beckoned by power. In workingw ith boys and men against the subordination of girls and women, gender and sexualviolence, and towards healthier forms of masculinities, we are likely to encounter resistance. Yet, whilst the effectso fr esistancea re commonly observable in work with men and boys, the phenomenon has not attracted enough gender-critical attention. And so it is proposed that we ought to learn to anticipatethat our engagements with boys and men will be characterised by dynamism, contradictionsa nd contestations whose final goal is to resist transformation, frustrate change and conserve existing relations of gender power. Above all, and at different points, we should expectand plan to work through resistanceinengaging men and boys toward gender justice and progressive, pro-feminist masculinities. --- Reactionary resistance at different points Resistance is evident at different points in work to transform masculinities. The obvious case of resistance is that 'from men still committed to gender privilege,orfrom men who accept gender equality in principle but do little about it in practice' (Connell2 003, 9). Othersh ave remarked on the 'conflicts, contradictionsa nd resistance' to change with respect to government policy and practice (Shefer et al. 2008, 160). Resistance to gender equality policies and programmes may alsoresult from material benefits men accrue from gender inequality and ideological justification of male supremacy (Connell 2003). An interesting case is that of:... resistance on the part of some women to involving men in gender and development work -d riven by fear about the dilution of the feminist agenda, and by anxieties over the diversion of limited resources away from women's empowerment initiatives and back into the hands of men. (Esplen 2006, 1) Jewkes, Flood and Lang (2014) have observed that somef emaleg ender activists have resisted the 'prioritisation of, and resource allocation for, work with men on violence K. Ratele S146 prevention' (1). Such resistancecouldemanatefrom the fact that 'gender analysis is often confined to women and girls' (Dover 2014, 92). In relation to South Africa, ac loser eading of policy documents and discussion papers reveals that although the ruling party has produced progressive, well-intentioned documentso ng ender, therei saclear conflationo fg ender with women and girls. For example, even though the 2012 African National Congressg ender policy discussion document starts offb yo bserving that 'gender is as ocially constructed understandingo f what it is to be am an and what it is to be aw oman' (2), and there is reference to men and relatively more mention of boys in the document,w hat the document is basically aimed at is women'sa nd girls' empowerment. Women's and girls' empowerment is significant in working towardg ender justice. However, the empowerment of women and girls is not exhaustive of gender relations and practices. Boys and men are gendered too. Contestations and hierarchies within masculinities matter. Violence against men and boys should be of concern in interventions, policy discussions and policymaking on health, gender and sexuality. Broadlys peaking, then, there appears to be al ack of conviction about men, and to al esser extent about boys, as gendered subjects-ad isinclination that possiblyc ontributes to the neglect of men as vulnerable to physical and sexualv iolencea nd health problemsa rising from gender regimes (Clowes 2013). Resistance is alsoevident at the point of the individual men's behaviour,for example regarding anxiety-associated identity problemsa bout gender change and emotional antipathy against gender-based programmes (Connell 2003). Resistance at the level of individual men and boys is often encountered in face-to-face interactions, when working in small groups with boys and men or in an interview with ar esearch participant. We should therefore not neglect psychological resistance, just as we should not neglect forms of resistance characterising institutions, cultures and structures. And while much of our attention may need to be dedicated to resistancew ithin structures like the state, the wider cultureand intermediate institutions like families, we mustbeprovoked to imagine what makes aboy desire gender equality or not; by what prompts aschool-boy in arecent focus group discussion to say: Roughly translated: 'Men are heads of households. Women must respect them [men]. If a woman wants to be aman, it shows disrespect for her husband' Thea bove expression of reactionary resistanceb yaboy emergedi nastudyo f 14 -16-year-old school-going boys from seven schools in the WesternCape (Ratele et al. 2007). What is interesting in the boy's account is the way it conveysh ow cultural traditions are af undamental yet neglected issue in workingw ith boysa nd men toward gender justice and progressive masculinities. Nowhere in that exchange was cultureo r tradition explicitly mentioned, although they would become moree vident in the focusgroup discussion. Also, nowhere are gender or sexuality mentioned. But accounts of tradition and gender intersect to produce the resistance conveyed in the extract. Of particular importance is that the effect of the boy's account is ultimatelyt or esist efforts for gender and sexuality change and transforming masculinities. In engaging boys and men, we ought to be aware that we are engaging the cultural traditions in which their practices are embedded. At the same time, in working through the resistancethat we are likely to encounter in trying to change gendered traditions, we have to be criticalo fo ur own biases too. Uncovering howateenager acquires such a' languageo ft radition' is perhaps the first step in overcoming someo ft he difficulties of getting boys and men to change and support gender justice. Culture, Health &S exuality S147 Three possiblec onsequences follow from underscoring the reactionaryr esistance encountered in working with boys and men. First, we observe that there could be an association between resistance to change and the fact that despite increasing work on gender and sexualviolence, levels of violence remain high in South Africa (Jewkesetal. 2012;South African Police Service 2012). Second, we start to notice that men and boys are disproportionatelyrepresented in such violence, as perpetrators and victims. And after nearly two decades of work with boys and men, as well as laudable gender equality policies and programmes,success in preventing violence, particularly against men and boys, has been minimal (Seedat et al. 2009). Third, we recognise that reactionary resistancet oc hanging men and boys is often touched on, perhaps often too lightly, in several South African qualitative and ethnographic studies (e.g., Bhana and Pattman 2011;Langa and Kiguwa2013;Naidu and Ngqila2013). As such, even though it has contributed to literature and practice on gender and sexualities, the body of studies and practical interventions that focus on boys and men is confronted with encumbrances that are hard to overcome. Given that the ultimateaim of working with boys and men is changing their subjective and social identities, relations with otherm en, relations with girls and women and the social order,t he body of worko nb oys and men in South Africa may do better by strategising aroundand working through reactionary resistance. Abetter understandingof why and how as ignificant proportion of men resist gender transformation, even where hegemonicmasculinity injures them,could contribute towardmoreeffective programming and politics in engaging boys and men towardhealthy,caring, pro-feminist masculinities and gender relations. --- Working through resistance in working boys and men An array of forces continues to frustrateprospectsfor comprehensive, swift and concrete changeso ng ender justice and healthy masculinities. These forces include poverty, unemployment, inequality, race and racism, cultural tradition and gender itself. This is not an exhaustive list, of course, but highlights some of the factors that need closer consideration if worktoe ngage boys and men is to be more effective. --- Working throughpoverty, unemployment and income inequality in engagingboys and men towardsg ender equality On 10 August, 2012, rock drillers at the Lonmin platinum mine in Marikana went on a wildcat strike.They demanded araise in their basicsalary to R12,500 (approximately US $1,130 in 2014 terms, where 1S outh African Rand is approximately 0.090 US$). They never got whatt hey wanted. The Marikana strike resulted in the deaths of 44 people, mostly mineworkers, killed by the police on 16 August. The tragedy or massacre, as it is variously referred to by different observers, is said to be the single most lethal use of force by the police against civilians in post-apartheidS outh Africa. In 2014, the mineworkers again wento ns trike, possibly indicative of continuing struggles aroundi ncome-related issues such as economic inequality, relative and absolutepoverty, decent employment and aliving wage. Commenting in an open editorial on the mineworkers' strike, Botha and Ratele (2014) noted the absence of analyses and politics that link of the mineworkers' strike to wider contestations aroundmasculinities. Maintaining that'money is central to the definition of what makes men' powerful or subordinate, they claimed 'thestruggle of the black men at the mines is therefore astruggle for them to be lookedupon as worthy men as any others' K. Ratele S148 (16). Furthermore, they stated:'If we continue to delink the struggles of blackmen for a better salary and ignore the economic inequalities that characterise society, and the low regard by capitalism for anything but capital, we increaset he risk of failing to convince black men of the value of gender equality' (16). Literature from aroundt he world has supportedt he contention that economic issues such employment, decent work, poverty and income inequality are central to as ense of manhood. Theg eneration of monetary income has been shown to be important in the subject position takenb yA ngolan male war veterans particularly vis-a <unk>-vis their wives (Spall 2014). Financialindependence and gainful employment is said to central to asense of manhood in Africa (Barker and Ricardo 2005). The ability to earn aliving is among the primary elements regarding what it means to be ag rownm an in mostc ontemporary societies (Correia and Bannon 2007). Silberschmidt (2011) has contended that 'poor, frustrated men with no access to income-generating activities' (99) are unlikely to have any enthusiasm for engaging in gender equality. Her main argumenti st hat while it is argued that gender equality creates development, it is not gender inequality that hampers sustainablep overtyr eduction ands ocioeconomic development,b ut povertya nd socioeconomic under-development that are obstacles to gender equality. In sum,being(un)employed, having adecentincome or having moneyvia other means matters both to men and masculinities. Aconcern with issues of poverty, unemployment and income inequality affects, or shoulda ffect,h ow we approach men and boys in our work. But how employment or lack thereof,poverty and income inequality matter needs closes tudy andw ith deeper contextualisation. Broadly, though,t ransforming masculinities and having men supportg ender justice appears to be related to their opportunities to have decent work, earn moneyorhave other income. Being poororrich, unemployed or employed and the economic inequality that increasingly characterises the world, cannot be ignored in our engagement with men. With respect to boys in particular, as an example, poverty, unemployment and income inequality comes into play through their families, cultures and peers. In interacting with their peers, family members and otherc ultural agents, they learn the role of economic inequalities in gender relations.O nce again, it is worth noting that how money-related issues comet om atter to how boys think of gender power is not straightforward. In the study referred to earlier by Shefer et al. (2007), boyshad an interesting exchange around lobola (bridal gifts, historically cattle and now money) aw ould-be groom sent to the bride's family: Roughly translated: Respondents: She must to do all the chores as Ihave paid lobola for her. Respondents: But the person you will hire might seduce you. Respondents: Helping each other has nothing to do with lobola, it only shows love. Respondents: So if you make lobola the priority, it means you don't love you partner. Respondents:I do love her but she must work. Respondents: That is totally not asign of love but abuse. (also see Shefer et al. 2007,...,e mphasis added) Themain topic of the above interchange was housework. It ends up beingabout sex, lobola,love, gender relations and aman beingable to pay for help.Gender is imbricated with othersocial and personal issues. And poverty, unemployment and income inequality Culture, Health &S exuality S149 are one set of those issues. Getting boys and young men to listentowhy gender justice is neededm ight mean workingt hrough issues of poverty, unemployment and income inequality. Ultimately, it is important to devote moret hought to realities around poverty, (un) employmenta nd income inequality as imperatives in workingw ith boys and men. Connecting to, and working with, unemployed and precariouslyemployedmen as well as labour unions and government structures on poverty and inequality issues will inform how to better engage men and boys aroundg ender matters. --- Working throughr acism andracial inequality in engaging boys andmen towards gender equality Work with boys and men may also wanttorecognise, develop more sophisticated tools to examine, and challenge the effectsofracialised hegemonies and racism in their lives. The main reason is that silenceo n racial inequality can cast doubt on work on gender inequality. Above all, though, both race and racism are pro-feminist issues. Ihave indicated that the subjectsofprimary interest here are Black heterosexual men and boys and the gender ideologies that interpellate them. Theh etero-sexualisation and racialisation of the men and boys (Shefer and Ratele 2011) is as ignal to the facts of coloniality and contemporary sexualised racism.C oloniality and racism are entangled with constructions and representations of masculinities, turning somemen and boys into objects (as opposedt ot reating them as fully human) (Fanon 1967(Fanon /1952)).T his suggests that engagingboys and men to change hetero-patriarchal and violent forms of masculinity and supportg ender transformation can never be simply about workingt owardg ender justice. Something more is needed -afocus on the potential or reality of objecthood of some men. When workingwith Black men on masculinities and gender relations, questions of the interwovenness of race and racism are likely to materialise, certainly in countries like South Africa. 'The masculine-feminine dynamic also has racial form', Gordon (2000, 125) has contended.T his dynamic holds for countries that have the experience of racialised imperial and colonialstructuring. However, the subtle or gross costs and benefitsofrace and racism applytoall boys and men in different ways because race and racism position women and men differentiallyw ith respect to the ruling patriarchal masculinity in their particularcontext. All the same, race and racism have bodily resonance for non-Caucasian men in countries such as the Brazil, France, the UK, the USA and South Africa, countries where race has cometoplay akey role in social and gender relations. At the same time, the issue of migration of differently racialised subjects from the global South to the global North has meantthat racial discrimination and racism have become resurgent concerns for many othercountries. Although it is theoretically possible to separate gender from race, in everyday reality, gender relations and identities cannot easily be disentangled from race relations and identities in the lives of Black subjects, similar to whatM orrell (2001) has indicated for issues of class and gender among settlers in colonial Natal.Aquestion we might be asked in workingwith Black boys and men,then,iswhetheragender-equal world means one in which white men are equal to Black women, and Black men are equal to White women, andn ot onlyi ntra-racialg endere quality. Health-enhancing progressive genderdeconstructive projects therefore need to consider issues around liberating Black masculinities from colonialist, racialised regimes of power and knowledgea st hey rub against hetero-patriarchy( Ratele 2013). --- K. Ratele S150 Underscoring the racial aspectofmasculinities does not imply that women, girls and subordinate men do not experience oppression by hegemonic Black masculinities. It does not meanthat male-targeted gender equality interventions focus only on Black boysand men or shouldn ot focus on these groups.A nd it does noti ndicate that workingw ith raciallyunmarked ruling masculinities is unproblematic (Morrell 2001). The task is how to retain ac ommitment to struggles aroundg ender equality with boys and men while recognising the difference that racism and race inequality introduces (Phoenix 2006); in other words, how to analyse the fact of race poweri ns ome boys and men's lives. This suggests, as masculinities scholarsh ave said, the ever-present need to recognise the construction, multiplicity and hierarchy of masculinities within and betweens ocieties (Connell 1995;Connell and Messerschmidt 2005;Hearn 2006). Neglecting racism and racialised poweri ne ngaging gender and sexuality is not the best way to go about developing healthy black boyhood and manhood. Health-related effects of racism and racial inequality may undermine the workf or sexual and gender equality. In the face of racism and racial inequality,i ti sn ot an uncommon to hear that struggles for gender equality are aW hite feminist agenda (Ratele, Shefer, and Botha 2011). Whereas gender-critical work on masculinities shows awareness of the problems boys and men create and experience as agender and in their sexualrelations, we need to better appreciate the struggles of someb oys and men aroundr acial marginalisation. Understanding the experience and structuresofinjusticearound race and racism is just as important as understanding structures and experiences around gendered blackness (Clowes 2013). Work with Black boys and men needs to show that gender and sexuale quality and racial equality are not antagonistic. Just as racial equality is incomplete without gender and sexualequality, workingwith masculinities and men's sexualhealth without placing them in the appropriate context of racialised power potentially undermines ourefforts. We do not have to completely reinvent ways of doing the kind work that is sensitive to gender and race. Black feminist scholarship has already provided approaches to workingi nt he interstices between gender,sexuality and race (Collins 1989;Crenshaw 1991). --- Working throught radition and culture in engaging boys and men towards gender equality Tradition and cultureo ccupy as ignificant yet largely misrecognisedr ole in the lives of boys and men, as well as women and girls (e.g., Jackson, Ho, and Na 2013;Moolman 2013). Tradition or culturesignal self-reflexive symbolic resources through which subjects act as members of in-or out-groups. Accounts of cultureo rt radition reference how experience, in the context of life with others and shared pasts, is authorised, contested, transmitted, inherited,i nterpreted and reinvented (Ratele 2013). Such accounts are often encountered in gender work with and on boys and men.Thisist he case,especially, in a world where cultureand tradition have cometooccupy asignificant place in struggles for land, nationalf reedom,n ationhood and citizenship. Tradition and culturea re often embraced as aresource by men and women, and boys and girls who may not have access to other vehicles of power. But in mucho fm asculinities studies, tradition, for instance, usually does not make an appearance on its own strengthbut is often incarnated in the term 'traditionalm asculinity'. In otherw ords, there is limited direct examination of men's gender and men's tradition. Working towards gender transformation seems to invite those who work with men and masculinities to enteri ntoo ngoing discussions aroundc ultures and traditions per se (Mama 2006). --- Culture, Health &S exuality S151 Traditions and cultures are doubtless not fixed, beingreinvented over time (Hobsbawm 1983). They are like masculinities and sexualities in that way -c ontinually being performed, remade,contested and changed. Fictional or not, though, we need to concern ourselvesw ith tradition and culturei np ro-feminist work because, like gender and sexualities, as discursive and symbolic resources, they are 'embraced by actors as aframe within which they may choose to act or against which they react' (Osborne 2008, 284). Spiegel (1989) has arguedt hat: The notion of tradition as ar esource from which ideas for dealing with contemporary situations can be drawn is crucial in that it directs our attention towards analysis of those situations where the process of cultural transmission is actively pursued, i.e. in situations where it may be useful to those who transmit and receive the ideas comprising tradition. ( 65) In work to engage men and boys, talk of tradition or culture has to be grasped as signalling their experiences in the context of life with other men and boys, women and girls. In speaking of cultureortradition, men and boys ought to be understood as trying to speak with and of authority, to contest power, to speak of what they may have inherited, how they arrive at their interpretations of the world around themselves. Accountst hat reference tradition and cultures hould alert us to the fact that subjects self-consciously position themselves in relation to amoreorlesstacitly understood set of beliefs or past. Hence, paying reparative yet critical thought to accounts of and contestations around cultureand tradition may be useful in workingwith men and boys, especially in contexts where tradition and culture are historically colonised or rapidly changing. --- Working throughg ender in engaging boys and men Engaging boys and men is engaging gender.B oysa nd men, like girls and women, are hailed into gender structures from conception to death. Working with boys and men as boys and men requires ac onsideration how they comprehend, construct and perform gender in their subjective lives; how they position themselves on the terrain of gender; how masculinities are contested or embraced; and the hierarchies that exist within groups of boys and men. Yet, as Kimmel (2005) noted, 'though we now know that gender is a central axis aroundw hich social life revolves, mostmen do not know they are gendered beings' (106). Barker and Ricardo (2005) have maintained that 'a gender analysis of young men musttake into account the plurality of masculinities in sub-SaharanAfrica.... There is no typical young man in sub-SaharanAfrica and no single African version of manhood' (v). Masculinities do notcomeinaone-size-fits-all model, and there cannot be aone-sizefits-all intervention when engaging boys and men towards gender justice and transformation of masculinities (Jewkes, Flood, and Lang 2014). In any setting or group, several waysofexpressing of gender are observable. While this may be commonplace to social constructionism-informed gender interventions (Magnusson and Marecek 2012, 36), spectres of essentialisms continue to haunt societal perceptions of and work with Black men and boys. Sometimese ssentialisms slip through in terms such as traditional gender roles, traditional sexualscripts and traditional masculinities. Although masculinity may not be viewed as given with maleness, there is somee vidence of an undertow of essentialism whereby gender is viewed as inherent in male bodies and minds. Those who work with men and boys to different degrees support the thesis that 'work with men and boys is necessary, can be effective, and can have apositive, transformative impact for the lives of women and girls, but also for the lives of men and boys' (Minerson et al. 2011, 2). Several interventions from around the world, such as thoseb yt he South African organisation Sonke Gender Justice and the Brazilian Promundo, have shown that K. Ratele S152 boys and men can be engaged to workt owards gender equality and equitable decisionmaking (e.g., Peacock 2012;Ringheim and Feldman-Jacobs 2009;World Health Organization 2007). However, convincing the majorityo fw omen and men that that pro-feminist work with boys and men is worktowards gender equality still requires effort. There are indications in government policies, civil society projects, politicala nd social activism,a sw ell as in the researchl iterature, that varying and often conflicting understandings exist in workingwith boys and men precisely entails. The ANC's Gender Policy Discussion Document has been referred to already.Policymakers, researchers and activistsw ho workw ith boys and men may need to do more of the kind of persuasion suggested by Minersone ta l. ( 2011) and other authors. Firstt hough, we may need to'recognize the reasons for resistance to gender equality among men and boys' as Connell (2003, 10) has argued. And if thisisnot such an issue in some parts of the world, in many countries in Africa we may need to do moreadvocacy, lobbying and research to show and convince governments and civil societyt hat'men are embedded in gendered systems to the same degree as women' (Lorentzen 2011, 111). It was noted earlier that we need to keep engaging with gender in workingwith boys and men because, although there is much to be found in South African government policies and programmes on gender,boysand men still appeartobeconsidered as less or not as gendered as girls and women. Men tend not to be atarget of gender transformation work and initiatives led by government and big business (but boys do fares omewhat better). In South Africa, somenon-governmental organisations, like Sonke Gender Justice, are highlighting the need to focus on men (Stern, Peacock, and Alexander 2009). But, generally, gender keeps being fused with being aw oman and girl, and boys and men are not considered as so vulnerable to violencea nd gender-related health problems as women and girls. Characterised by atendency to gloss contestations among masculinities, the resistancea ti ssue here is encountered in documents from government and its agencies, politicalparties, non-governmental organisations, as well as from some gender researchers. The construction, multiplicity and hierarchy of masculinities are not adequately incorporated into government policies and programmes. Given the goal of changing gender powerrelations,thoseworkingwith boys and men cannot but engage with discourses on gender that leave out boys and men.They may also want to do moret oi nsert within gender projects the under-recognised impact of heteropatriarchal ideologies and violenceonboys and men. They need to findwaystoovercome the resistance around the gender of boys and men by providing am ore economically, politically, culturally-informed understanding of men's and boys' lives in relation to women'sand girls' (Shefer, Stevens, and Clowes 2010). And it may be that men who work in this area need to learn from and form alliances with women activists, women's organisations and feminist women interested in boys and men as to how effectively to engage the state, policymakers and other women. --- Conclusion This paper has sought to show that boys and men live within aset of dynamic and usually contradictory social structures, identity networks, interpersonal relationships and material forces that include but exceedgender. These dynamics and contradictions are frequently intensified by processeso fm arginalisation and subordination. Some of these structures, networks, relationshipsand forces challenge, complicateand resist efforts to change men and boys. Of special relevance here are income-related issues such as employment, decent work, income poverty and inequality; racism and racial hierarchies and injustices; and Culture, Health &S exuality S153 cultural tradition as they intertwine with sexuala nd gender inequalities. These realities separately, jointly and complexly subvert men's and boys' health, just as they undermine women'sand girls' health. Transforming boys and men suggests workingtoward health in the thick sense of gender,s exual, racial, economic and cultural wellbeing. In this paper,Ihave focused on some of the reactionaryr esistancea sw ell as the dynamics, contradictions and contestations from individualsand structures that need to be considered and overcome in the worktoengage men and boys toward gender justice and egalitarian masculinities. It suggestedt hat those of us who work with boys and men take into account such psychological, institutional and ideological resistance if we hope to liberate masculinities and achieve gender justice. Transforming masculinities and persuading men and boys to workf or gender equality requiresu st or ecognisea nd overcome, and help men and boys recognisea nd overcome, resistancet og ender justice efforts. As Connell (2003) once wrote: the task for gender equality policy is to recognize the reasons for resistance to gender equality among men and boys, to find answers to the arguments advanced by opponents, and to find better solutions to the underlying social concerns that find expression through resistance to gender equality. (10) --- Disclosures tatement No potential conflict of interest was reported by the author. --- Re <unk>sume <unk>es vingt dernie <unk>res anne <unk>es, en Afrique du Sud, les sciences sociales, de la sante <unk>et humaines ont e <unk>te <unk>mpre <unk>gne <unk>es du focus sur la remise en question et la transformation des formes dominantes de la masculinite <unk>et sur la mobilisation des garc ons et des hommes pour l'e <unk>galite <unk>des genres et des masculinite <unk>ssaines. Ce focus sur les hommes et les garc ons s'est e <unk>galement re <unk>ve <unk>le <unk>e <unk>vident dans les actions lance <unk>es par les programmes et les activistes. Ce virage ne s'est pas produit sans re <unk>sistances, contestations et contradictions. Une se <unk>rie de re <unk>alite <unk>sl ocales et mondiales ac ontrarie <unk>une grande part de l'enthousiasme pour des changements rapides, radicaux et concrets concernant l'e <unk>quite <unk>de genre et la de <unk>termination de masculinite <unk>sp rogressistes. Les forces discursives et mate <unk>rielles qui s'opposent aux actions incitant les garc ons et

Over the last two decades, af ocus on challenging and transforming dominant forms of masculinity and engaging boys and men towards gender equality and healthy masculinities has permeated South African social and health sciences and the humanities. This focus on men and boys has also been evident in intervention and activist work. However, the turn to boys, men and masculinities has not gone without resistance, contestation and contradictions. Arange of localised and global realities has frustrated much of the enthusiasm for rapid, sweeping and concrete changes regarding gender justice and the making of progressive masculinities. Among the discursive and material forces that oppose work that engages boys and men are those to do with income-related issues, race and racism, cultural traditions and gender itself. Because of this, it is contended that engagement with boys and men needs to consider not only gender but also economic inequality, poverty and unemployment, divisions created by race, and struggles around tradition. This paper discusses these forces that undermine and counteract work with boys and men and how we might work through resistance in engaging with men and boys.

ons et des hommes pour l'e <unk>galite <unk>des genres et des masculinite <unk>ssaines. Ce focus sur les hommes et les garc ons s'est e <unk>galement re <unk>ve <unk>le <unk>e <unk>vident dans les actions lance <unk>es par les programmes et les activistes. Ce virage ne s'est pas produit sans re <unk>sistances, contestations et contradictions. Une se <unk>rie de re <unk>alite <unk>sl ocales et mondiales ac ontrarie <unk>une grande part de l'enthousiasme pour des changements rapides, radicaux et concrets concernant l'e <unk>quite <unk>de genre et la de <unk>termination de masculinite <unk>sp rogressistes. Les forces discursives et mate <unk>rielles qui s'opposent aux actions incitant les garc ons et les hommes a <unk>s'engager sont celles qui sont en rapport avec la question des revenus, la race et le racisme, les traditions culturelles et le genre luime <unk>me. C'est pourquoi nous soutenons que l'implication des garc ons et des hommes doit e <unk>tre envisage <unk>e, non seulement par rapport aux questions de genre, mais aussi par rapport aux ine <unk>galite <unk>s e <unk>conomiques, a <unk>la pauvrete <unk>et au cho <unk>mage, aux divisions ayant les appartenances ethniques pour fondement, et aux conflits lie <unk>saux traditions. Cet article traite des forces qui s'opposent aux actions mene <unk>es aupre <unk>sdes garc ons et des hommes, et de la manie <unk>re selon laquelle nous pourrions faire face a <unk>ces re <unk>sistances. --- Resumen En las dos u <unk>ltimas de <unk>cadas, la importancia de cuestionar yt ransformar las formas dominantes de masculinidad ylaparticipacio <unk>ndechicos yhombres hacia la igualdad de sexos ylas masculinidades saludables es un planteamiento que ha impregnado las ciencias sociales ys anitarias yl as humanidades en Suda <unk>frica. Este planteamiento en hombres ychicos tambie <unk>nhasido evidente en el trabajo de intervenciones yactivistas. Sin embargo, la reorientacio <unk>nhacia los chicos, los hombres y las masculinidades no ha ocurrido sin resistencia, confrontaciones yc ontradicciones. Una serie de realidades localizadas yg lobales han frustrado gran parte del entusiasmo por los cambios ra <unk>pidos, radicales yconcretos con respecto alajusticia de los ge <unk>neros yeldesarrollo de las masculinidades progresivas. Entre las fuerzas discursivas ym ateriales que se oponen al trabajo donde participan chicos yhombres existen aquellas que tienen que ver con cuestiones relacionadas con los ingresos, la raza yelracismo, as<unk> <unk>como las culturas tradicionales yelge <unk>nero. Debido aesto, se sostiene que la Culture, Health &S exuality S157 participacio <unk>ndechicos yhombres debe considerarse no solamente teniendo en cuenta el ge <unk>nero sino tambie <unk>nlas desigualdades econo <unk>micas, la pobreza yeldesempleo, divisiones que surgen por la raza ylas luchas que giran en torno alatradicio <unk>n. En este art<unk> <unk>culo se debaten las fuerzas que contrarrestan el trabajo con chicos yhombres yseanaliza co <unk>mo podr<unk> <unk>amos conseguir la participacio <unk>ndehombres ychicos atrave <unk>sd elar esistencia. K. Ratele S158

Over the last two decades, af ocus on challenging and transforming dominant forms of masculinity and engaging boys and men towards gender equality and healthy masculinities has permeated South African social and health sciences and the humanities. This focus on men and boys has also been evident in intervention and activist work. However, the turn to boys, men and masculinities has not gone without resistance, contestation and contradictions. Arange of localised and global realities has frustrated much of the enthusiasm for rapid, sweeping and concrete changes regarding gender justice and the making of progressive masculinities. Among the discursive and material forces that oppose work that engages boys and men are those to do with income-related issues, race and racism, cultural traditions and gender itself. Because of this, it is contended that engagement with boys and men needs to consider not only gender but also economic inequality, poverty and unemployment, divisions created by race, and struggles around tradition. This paper discusses these forces that undermine and counteract work with boys and men and how we might work through resistance in engaging with men and boys.

Introduction The organizational structure of startups is different from traditional organizations. Startups have flatter hierarchies, less bureaucracy, more flexible structures, a team mentality, and cross-functional collaboration (Lopes Fo. et al., 2019). These characteristics bring professionals together and foster the promise of equality in terms of career opportunities (Mickey, 2019;Spender et al., 2017). However, due to their strong technological base, high degree of replicability, scalability, and competitiveness (Lopes Fo. et al., 2019;Kuester et al., 2018), women are not immune to discrimination in startups. Systemic and institutional gender biases reinforce male stereotypes that are entrenched in the technology field (Sperber & Linder, 2023;Ughetto et al., 2020). Official data presented by the Startup Heatmap Europe shows that in Europe, 15.5% of founders or co-founders of startups are women. They also receive 38% less funding than their male counterparts when starting up under the same conditions (Startup Heatmap Europe, 2020). In the United States, in 2018, for every dollar raised by a startup with a female founder, exclusively male startups raised $8.33 (West & Sundaramurthy, 2019). In Brazil, a country strongly characterized by a patriarchal, sexist, and macho culture (Carrieri et al., 2013), 29.5% of startups have women as partners and only 4.4% were founded exclusively by women (Female Founders Report, 2021). This echoes recent data published by Serviço Brasileiro de Apoio às Micro e Pequenas Empresas (2023), which indicates that even though Brazil has 10.3 million women business owners, the largest contingent of female entrepreneurs in its history, women's participation in the information technology (IT) services sector remains minimal. These data highlight that in countries where progress has been made in terms of gender equality (e.g. Europe and the United States), startups are more likely to be male organizations, and even more so in countries with gender inequality (e.g. Brazil). In addition to being characterized as a male professional context because of their close ties to the technological field, startups are innovative businesses that depend on the entrepreneurial potential of their founders to develop. Previous empirical studies point to entrepreneurship as a male context, characterized by gender discrimination (e.g. Welsh et al., 2023). Thus, women founders of startups face double gender barriers (Sperber & Linder, 2023) as they struggle simultaneously with two male professional contexts: technology and entrepreneurship. According to Derks et al. (2016) and Faniko et al. (2021), traditionally male organizational contexts are conducive to the emergence of the queen bee phenomenon (QBP). The QBP suggests that some women leaders who experience gender discrimination in traditionally male environments may engage in behaviors that are more harmful than helpful to other women (Derks et al., 2016). Thus, considering the influence of the male professional context on the prominence of queen bee (QB) attitudes among women leaders, we propose as the objective of this study to analyze the adherence of women founders and co-founders of startups to QBP attitudes. Confronting the QBP premise that the male work context is more susceptible to the emergence of QB attitudes with the fact that women founders of startups face double gender barriers (see Sperber & Linder, 2023), we make the assumption that women founders and co-founders of startups exhibit strong QBP adherence. To fulfill the proposed objective, we conducted a qualitative methodological study. The female startup founders interviewed worked and lived in the Brazilian Northeast, a region where patriarchalism is even more accentuated (Nicholus, 2019). We believe that applying a qualitative methodological approach, which is rarely used for studies on the QBP, and highlighting a sociocultural group with pronounced gender inequalities could help develop theoretical aspects of the QBP. In general, gender issues are still rarely studied in Brazilian startups (e.g. Gomes No. et al., 2020). The international gender studies in startups highlight the investment barriers faced by women (e.g. Balachandra et al., 2019). However, no study on the QBP in startups was found in the Brazilian or international literature. Research adopting the QBP has been conducted specifically in universities (e.g. Faniko et al., 2021), police forces (e.g. Derks et al., 2011), and the health sector (e.g. Sengul et al., 2019). Thus, this study is justified in advancing the debate on women's performance in the technological field. It is necessary to go beyond the examination of barriers related to access to capital and explore sociocultural factors of the environment in which women are embedded, as well as explore the cognitive processes underlying the relationships established between women in this environment (Wheadon & Duval-Couetil, 2019). This study sheds light on the influence of social and organizational factors on cooperative and competitive relationships among women in a Brazilian region historically characterized by patriarchalism (Carrieri et al., 2013;Nicholus, 2019). At the same time, startups are organizations that attract young founders, directors, and employees. Women Founders of Startups: an Examination Through the Prism of the Queen Bee Phenomenon They belong to a generation that has been more exposed to gender movements (e.g. #metoo) and are therefore more sensitive to gender diversity. Therefore, it is important to investigate how the antagonistic forces of patriarchalism and gender diversity movements interact and impact the relationships among women. Finally, our study can inform the actions of governments and development agencies that seek to contribute to the advancement of women's careers in startups. Besides this introduction, this paper is divided into five other sections. The theoretical framework is presented in two sections. In the first one we present the QBP, the psychological processes behind the perpetuation of gender discrimination by senior women, empirical results of previous studies, and some criticisms of the term. In the second one, we present a brief overview of women and work in Brazil, adding some information about Brazilian startups founded by women. Our methodological decisions are presented and justified in the method section. We then describe and discuss our results in two different sections. In the conclusion section, we address the main limitations of our study, its theoretical and practical implications, as well as a future research agenda. --- Psychological processes that explain the queen bee phenomenon Women who perceive a threat to their gender identity in the workplace face the dilemma of whether to adopt individual or collective strategies in the face of the threat. For example, if women perceive themselves as disadvantaged due to their gender, they can either work collectively for women's development or focus on developing their personal opportunities. Actions at the collective level emphasize the needs of the group as a whole. Individual-level actions, on the other hand, emphasize personal outcomes by distancing oneself from the disadvantaged group and adopting traits typical of the advantaged group (Derks et al., 2016). Strategies at the collective level can have two different ramifications. The first involves stereotypically female qualities that are positive. The second concerns actions aimed at social change, such that women collectively protest against career decisions that illegitimately harm the female collective (Derks et al., 2016). Solidarity, cooperation, and collective behavior presuppose that women support each other on the basis of their gender identity, especially at higher levels where they act as role models and mentors (see Mavin, 2006). Previous empirical findings point to networking and mentoring processes as mechanisms for women to advance the careers of other female professionals (e.g. McAdam et al., 2019). Some leaders are seen as symbols that increase opportunities, help to chart a possible path to professional success, and serve as role models and inspiration for their female colleagues (O'Neil et al., 2011). The more participative and collaborative leadership style provides greater emotional, social, and professional support for other women (O'Neil et al., 2008). Hurst et al. (2017) point out that women in positions at the bottom of the organizational hierarchy have higher expectations for emotional understanding and support from their managers. At the same time, they expect their managers to have a more holistic view and better understand their complexities. Thus, for some women, supportive behavior and building mutual alliances among women emerges as a strategy to reduce gender inequality in the workplace and create new opportunities for professional advancement (Abalkhail, 2020). The presence of women in leadership positions, as well as their participation in women's collectives, has positive outcomes in reducing gender discrimination and improving the career prospects of other women. Building collaborative networks, whether formal or informal, promotes opportunities for women's professional development and allows them to challenge the status quo (McAdam et al., 2019), reducing stereotypes that position men as natural and more successful entrepreneurs than women. On the other hand, it is worth noting that these collaborative experiences are situated in and influenced by personal, organizational, and social contexts (Hurst et al., 2017). Thus, O'Neil et al. (2011) reinforce that in a system and culture that does not support gender diversity, strong collaborative networks among women are not sufficient to address gender inequality. Even when pursuing legitimate causes, women who adopt collective strategies to overcome inequalities are often associated with the image of troublemakers. This negative image can discourage other women from adopting collective strategies (Ellemers & Barreto, 2015). The individual strategy emerges as a more discrete mechanism, in which the woman seeks to be accepted by the favored group by distancing herself from the disadvantaged group. Thus, some women in leadership positions may behave according to the norms of the dominant group because they expect some personal gain from this behavior (Ellemers & Barreto, 2015). This type of strategy, in which women deal with gender discrimination by adapting to the system, promotes the emergence of attitudes characteristic of the QBP. The pioneering work on queen bee focused on competitive behaviors among women (Staines et al., 1974). Thus, the diffusion of this metaphor damaged the image of female leaders and blamed them for the low representation of women in high hierarchical positions (Mavin, 2006). More recent studies on the QBP clarify the psychological process behind women's competitive behavior. Some of the psychological processes are also found in other disadvantaged groups, such as self-group distancing in some disadvantaged ethnic groups (Derks et al., 2016). Therefore, unlike the pioneering studies, the recent ones attempt to erase the image of hereditary rivalry among women and account for queen bee attitudes toward male organizations and gender discrimination faced throughout their careers (Grangeiro et al., 2023a). Therefore, QBP attitudes have been organized into three dimensions in the scientific literature, as we present below. The first dimension is called male identification. It refers to stereotypically masculine characteristics that provide higher status in organizations (Wood & Eagly, 2012). Women in leadership positions exhibit more agency traits (e.g. analytical, assertive, competitive, ambitious, dominant, and self-reliant) and resemble men in the way they dress, speak, and relate to subordinates (Faniko et al., 2021). Women report being more committed and ambitious than their early career peers (Faniko et al., 2016;Grangeiro et al., 2023a). They adhere to masculine stereotypical behaviors because characteristics associated with male leadership styles confer more status and power in organizations (Derks et al., 2011). In the technological field, women reinforce the importance of becoming invisible, becoming "one of the guys", and avoiding drawing attention to their feminine characteristics (Harvey & Tremblay, 2020). The second dimension refers to self-group distancing. This dimension suggests that women in prestigious positions report having low identification with different subgroups of women. They also report making more sacrifices (Ellemers, 2014) for the sake of their careers, resulting in low identification with female co-workers in lower positions in the organizational hierarchy who have not achieved professional success. Women in higher positions may feel compelled to distance themselves from female colleagues in order to escape the negative expectations and low status associated with their gender (Ellemers, 2014). Faniko et al. (2016) emphasize that this phenomenon does not refer to a generalized physical and psychological distancing directed at all women, as queen bee women approach and support colleagues who are at the same hierarchical level and have similar trajectories. Finally, the third dimension is called gender hierarchy legitimation. It highlights that QB women reinforce the status quo i) by denying gender discrimination (Derks et al., 2011). Previous research suggests that academic and police women do not realize that some of the barriers they face are gender barriers (Gomes No. et al., 2022;Derks et al., 2011); ii) by defending the meritocratic system even in the face of clear evidence of inequality (Webber & Giuffre, 2019); and iii) by being hostile toward affirmative policies that seek to promote gender equality in positions of high responsibility, such as women who oppose gender quotas, since such policies somehow neglect the efforts they have made to advance in their careers (Faniko et al., 2017;Derks et al., 2011). The image conveyed by the queen bee metaphor suggests the presence of hereditary female competitive behaviors because it uses an image from the biological sciences. Mavin (2006) adds that the term is sexist and tarnishes the image of women who have reached high hierarchical organizational levels. Mavin (2006) warns that the use of the term may perpetuate a caricatured image of women leaders as ambitious and as more masculine than men. Nevertheless, the scientific research on the QBP from the last decade does not point to women themselves as being responsible for the emergence of the phenomenon, but highlights the role of the male organizational context as critical to the emergence of the phenomenon (Faniko, Ellemers, Derks, & Lorenzi-Cioldi, 2017). Similarly, it is believed that women who adhere to these attitudes do so because of the numerous difficulties they have experienced throughout their careers (Faniko et al., 2017). Despite researchers' efforts to portray QB as a consequence of gender discrimination faced by women in male organizations, criticism of the phenomenon persists, leading the most influential researcher on the topic to propose replacing the QBP with self-group distancing (Faniko et al., 2021). --- A brief overview of Brazilian women in the workplace Brazil is a country that has historically been characterized by the sexual division of labor (Nicholus, 2019). Women Founders of Startups: an Examination Through the Prism of the Queen Bee Phenomenon The shared belief that women were biologically more adapted to housework and caregiving isolated them in the private domain. According to this mindset, men are more competitive, ambitious, and self-reliant due to their biological traits and are therefore considered more capable of pursuing privileged professions of higher social status (Hirata & Kergoat, 2007). Thus, gender inequalities remain due to the persistence of gender roles deeply rooted in the culture of a macho Brazilian society (Coelho Jr. et al., 2022). Since the 1970s, there has been a greater increase of women in the labor market (Ara<unk>jo, 2021). Entering into paid work is the first step towards achieving gender equality (Cole, 2020). In the 1980s, we can observe the achievement of legal equality through the 1988 Federal Constitution (Lopes, 2006). Data from 2021 allow us to confirm the presence of gender parity in the labor market in general, since 42.5% of those in the Brazilian formal labor market are women (Instituto Brasileiro de Geografia e Estat<unk>stica, 2022). However, traces of the sexual division of labor persist, and we can cite three pieces of evidence for this persistence. First, even highly qualified women who hold high managerial positions (Carvalho No. et al., 2010) and who are prominent in their university careers (Censon et al., 2022) report having more responsibility for domestic and caregiving activities than their partners. Second, the gender pay gap persists, with women being paid approximately 30% less than men (Instituto Brasileiro de Geografia e Estat<unk>stica, 2022). The third indication relates to horizontal and vertical segregation. Horizontal segregation refers to the confinement of women to so-called female professions and the difficulty of entering and remaining in historically male professions, such as science, technology, engineering, and mathematics (STEM). Vertical segregation refers to the low representation of women in the highest hierarchical positions in organizations (Gomes No. et al., 2020). The development of women in the labor market has been accompanied by an increase in the number of enterprises founded by women in Brazil (Santos et al., 2022). Nevertheless, as in the labor market, Brazilian entrepreneurship in general is characterized by sexual division. Female businesses are more often based on gender stereotypes, i.e. related to care, beauty, fashion, or food. Also, female enterprises are smaller than male ones and have less access to financing (Female Founders Report, 2021; Serviço Brasileiro de Apoio às Micro e Pequenas Empresas, 2023). Even if we can list some advantages for women's careers in startups, such as organizations that are more open to diversity, offer greater opportunities to balance personal and private life, and promote actions to support women and women's collectives (Gomes No. et al., 2020), gender inequalities persist. The Female Founders Report (2021) indicates that 4.7% of Brazilian startups were founded exclusively by women. The same report shows other data that prove the gender disadvantage in startups: in addition to less relevant funding for women's ventures, more than 70% of female founders said they had experienced moral harassment. There was little diversity among the women running startups: 76.5% identified as white and 87.5% as heterosexual (Female Founders Report, 2021). Still, in a study on the profile of the founders of the first Brazilian startups to become unicorns, Sousa (2021) identified 18 startups and 44 founders, only one of whom is a woman. Pavan et al. (2021) show some reasons for the low number of startups founded by women. The social and historical context in which women are embedded is one of the reasons, since in cultures where women are seen as homemakers and childcarers, clients, stakeholders, and investors tend not to associate their image with entrepreneurship, and women themselves have difficulty feeling legitimate as entrepreneurs. Second, women have lower expectations for the high growth of their ventures (Ruiz Arroyo et al., 2016). Even though women's ventures are as risky as men's ventures, female businesses are less likely to succeed. The gender of the founder impacts the survival of the venture (Bertolami et al., 2018), so women have to improve compensatory mechanisms to increase the chances of success of their business. Women need greater investment in human and social capital and management practices for their ventures to have the same chances of survival as male ventures (Bertolami et al., 2018). --- Method To achieve the proposed objective, we adopted a qualitative research approach. As a data collection strategy, we used semi-structured interviews, carried out between February 2020 and February 2021. We interviewed 30 women founders of startups in the Brazilian Northeast, specifically 11 from the city of Fortaleza (CE), 9 from Recife (PE), and 10 from Salvador (BA). These cities occupy the top three positions in the ranking of Northeastern cities in terms of number of startups (Associaç<unk>o Brasileira de Startups, 2021). Regarding the sample size, we considered two criteria. --- Rebeca da Rocha Grangeiro / Manoel Bastos Gomes Neto The first was information redundancy, and the second was the balance between the number of participants from each city. --- Data collection procedures The research instrument was divided into five blocks of questions. First, personal and professional questions were asked, followed by specific questions about the startups they manage, women's overall participation in the startup ecosystem, factors that positively influence women's access and permanence, and finally, factors that hinder women's performance (see Supplementary Data 1interview script). To access women working in startups, we first mapped Brazilian startups (Associaç<unk>o Brasileira de Startups, 2021) and searched their websites and social networks for information about their teams. In the next phase, the participants were contacted through social networks such as WhatsApp and Instagram. We also asked the women interviewed for contacts of other professionals involved in startups. In this way, participants were approached according to convenience and through snowball sampling, both non-probabilistic methods. After conducting the first three interviews, the authors met to discuss the appropriateness of the interview script. We did not find the need for major changes, but we did notice that some questions made sense to some female founders but not to others, and that this depended on the stage of the startup (e.g. early stage, growth stage, or late stage). So we designed a large script and adapted the questions according to the startup characteristics. The first three pilot interviews were kept and the participants are part of this study. We conducted the first interviews in person ( 5), but due to the COVID-19 pandemic, the remainder of the interviews had to be conducted remotely (25), using Google Meet. Before starting the interview recordings, the objective was presented and the participants were asked to sign the free and informed consent form, guaranteeing their voluntary interest in participating in the study, the ethical criteria, confidentiality, and anonymity of the information provided. The interviews were conducted in Portuguese, the official language of the researchers and the participants. They averaged about 50 minutes of audio each and were transcribed without software assistance (see Supplementary Data 2 -transcription of the interviews). The participants' quotes presented in this study were translated by the authors and back-translated by a bilingual (English-Portuguese speaking) researcher not involved in this study. To ensure the participants' anonymity, information that could identify them was removed and their names were replaced by a code consisting of a letter (F for founder) followed by a number indicating the order in which the interviews were conducted (from 1 to 30). --- Participants The study participants are women between the ages of 21 and 56, with a high level of education, 54% have a college degree, and approximately 43% have a post-graduate degree. Their experience in the startup ecosystem varied, for the most part, from 1 to 5 years. We tried to diversify the startup market sector, as can be seen in Table 1. --- Data analysis procedures The coding process used is characterized as abductive, as it combines deductive and inductive methods (Halpin & Richard, 2021). First, we relied on pre-established theoretical categories in the literature on the QBP to analyze the interviewees' statements. Then, the dimensions that make up the QBP were considered to start the coding process. We had no software assistance in the coding process (see Supplementary Data 3 -codebook and codes). Thus, for the male identification dimension, we sought to identify language that addressed agency traits, strong commitment to the career, and life choices that favor the career. For the self-group distancing dimension, we looked for statements in which the participants declared that they did not identify and/or compete with other women. For the gender hierarchy legitimation dimension, we sought to identify adherence to the meritocratic discourse, hostility to policies that favor gender equality, and denial of discrimination, recognized as such when the participant claims not to have experienced discrimination but reports situations characterized as sexual and sexist violence (Jaspard, 2011). After the first stage of coding the transcripts, we noticed low frequencies of codes referring to self-group distancing (N = 01) and gender hierarchy legitimation (N = 03). Due to the low frequencies of the second and third dimensions of the QB phenomenon, we chose to exclude these theoretical categories. Then, we performed additional free readings of the text corpus to identify segments that referred to the relationship between women and how they signified the discriminatory processes they experienced. Thus, through the inductive method, codes related to support for other women and identification among women resulted in the category of cooperation. Women Founders of Startups: an Examination Through the Prism of the Queen Bee Phenomenon Also, codes related to the recognition of gender discrimination as such, non-adherence to the meritocratic discourse, and support for gender equality policies gave rise to the category of fighting the status quo (Figure 1). --- Figure 1. Interview analysis procedure Once new second-order themes and categories were defined, the researchers performed another reading of the text corpus to count the frequency of the categories in each of the participants' statements. Each researcher carried out this procedure individually and independently. Thus, each interview was coded twice. In order to analyze the reliability of the categorization process carried out, we calculated the percentage of agreement between researchers and the kappa coefficient for each of the study categories (see Table 2). The percentages of agreement ranged from 90.0 to 100. All kappa coefficients were significant at p <unk> 0.001 and ranged between 0.757 and 1. These values are adequate according to Vanbelle (2016). After evaluating the levels of agreement among the researchers, we met to settle disagreements. At this point, for each category, we took the transcripts of the interviews where there was disagreement, and after reading and discussing the text corpus, we reached a consensus on the categories where there was disagreement. --- Results --- Male identification In line with the underlying QBP behaviors noted in the literature, the statements of female founders of startups highlight the need to adopt masculine characteristics: 83.3% cite the adoption of agency traits, 80% cite the need for greater commitment, and 33.3% cite the need to make life choices that favor the career. These three second-order behaviors, their frequencies, and percentages are shown in Table 3. --- Agency traits Gender stereotypes suggest that male characteristics as more suited to technology and entrepreneurship, making male overrepresentation the norm and disfavoring female careers. The interviewees confirm this statement when they state that: "[...] in education in general, women aren't encouraged to be leaders, [...] Since it is a mostly male environment that values stereotypically masculine traits, the women interviewed affirm the need to adopt a firmer posture in order to occupy more space in technology, in leadership positions, and in strategic decision making: "[...] at times it requires you to have a more dominant voice, a firmer grip, to know how to position yourself" (F24). Thus, 83.3% of the participants (n=25) reported having agency traits. The statements of participants F05 and F06 exemplify such traits. F06 describes herself as an active, dynamic person and states the need to "[...] be really incisive, really objective [...]. So, I think that women shouldn't be afraid, you know, to expose themselves, to be leaders, to be firm when they need to be." F05 points out that she adopted a more masculine stance in the process of obtaining resources to found her startup, due to the fact that it was an exclusively male environment: Do you know what my nickname was? Peireir<unk>o. They called me Peireir<unk>o. I took it as a joke, and even today I still do [...] But it has weight, right? Because there were only men. I was the only woman there, so I had to be a Peireir<unk>o. --- Strong career commitment Because startups present a less hierarchical and more dynamic business model, a faster pace of work, and a culture of greater belonging, some interviewees cited the need for greater commitment as a general obligation: In the startup environment, not everything is ready, not everything is thought out. So you need to make yourself very available to develop, [...] make suggestions for improvements, be a professional who has initiative and is aligned with the business purpose, you know? (F15). Even though a high level of commitment is required of all employees, 80% of the women (n=24) highlight the need to commit and sacrifice more than their male colleagues: When there's a business in which a woman is growing and being valued, we have to applaud, because you can be sure that the effort she made was much greater than the effort of a man or a team of men. (F09). Interviewee F18 emphasizes that "[...] some women need to dedicate themselves more, work harder, study more, and end up having this pressure to spend more time at the company, to show that they are performing, while some men don't have this concern." Thus, "[...] a woman has to work 2 or 3 times harder to... at least achieve the same result as a man," confirms F12. F04 reinforces this by stating that "I think women work more hours, I think women care more, they are much more dedicated. Women have a much higher level of self-demand!". The women also highlight the need to work harder to obtain professional credibility: Everything becomes more difficult, you have to be twice as assertive. You have to study the subject twice, you have to be and prove to others that you are twice as good, so that you'll be respected and be heard. (F09). --- Cooperation Going against queen bee attitudes, 80% of the participants did not show self-group distancing and reported cooperative behaviors and strategies for support among women. Only one participant (3.3%) stated that in startups there are competitive behaviors among women, as shown in Table 4. Given the recognition of gender inequalities, the underrepresentation of women, and the need to make greater sacrifices to integrate into the startup ecosystem, we observed a movement of mutual support among women, especially at the higher levels where they act as role models and mentors. Interviewee F17 points out that "I contribute like this, I've been a mentor, I've led seminars, I've given talks... because they have... they always do hackathons, things like that". F26 points out that she develops actions focused on women: "[...] with programs for women, grants for the acceleration of startups led by women, seminars, we bring many female mentors and entrepreneurs, [...] who are angel investors". --- Table 4 Number of occurrences of the second-order themes of cooperation --- Category Second-order themes Occurrences Frequency --- Cooperation Female competition 1 3.3% --- Female cooperation 24 80% Rebeca da Rocha Grangeiro / Manoel Bastos Gomes Neto These cooperative behaviors are also directed toward women at the beginning of their careers. F01 states that she usually gets involved in "[...] training, conversations, mentoring to stimulate this point [...] of women's entry into the job market, especially in companies where they don't think they are suited to work in that position, right?". The leaders state that they also fight to ensure that female stereotypes are perceived as positive. They fight for structural changes and challenge beliefs that point to a lack of support among them. F10 says that she notices a [...] very great mutual help, very great. So, I don't see it as... 'ah, because you're a woman you're going to take my place.' On the contrary, 'ah, because you're a woman, I'm going to help you so that we can go further together'. (F10). F14 reinforces this by stating, "I see the effort that we're all making to break down our prejudices, our barriers, and to come together, and create and fight for space" (F14). Thus, F30 states that "[...] when there are more women, women always start with this view of 'let's hire women, right?'". In addition to mentoring and combatting gender stereotypes, the participants mention formal and informal collaborative networks that promote opportunities for women's professional development. Regarding formal meetings, the interviewees describe their participation as both organizers and attendees of events: "I've already organized the women's Startup Weekend, I've already participated in women's hackathons. The digital port here in Recife has a branch that is Minas,Women,and Innovation" (F16). They confirm the existence of Informal collective actions are also described by the interviewees as important spaces for exchanges and mutual support among women. F07 provides another example of the construction of informal support: We have a group for the women who work here. We had a meeting with the women who are here on a daily basis, [...] we had a debate and so on, and then this group was created and we still have it today, we meet every two months. And then we have [...] a WhatsApp group, that when something happens, we send it there too. And I think that helped a lot, right? (F07). --- Fighting the status quo The results of our study also go against gender hierarchy legitimation. Contrary to what the QB attitudes suggest, 83.3% of the female founders of startups acknowledge experiences of discrimination in the organizational environment and 66.7% stated that they support gender equality policies, as shown in Table 5. --- Recognizing discrimination According to the academic literature on the QBP, gender hierarchy legitimation occurs through the denial of discrimination, adherence to the meritocratic discourse, and hostility to policies that aim to promote gender equality. As for the denial of the existence of discrimination, only one participant (F16) states that it is difficult to see gender discrimination: Women Founders of Startups: an Examination Through the Prism of the Queen Bee Phenomenon People tend to say that women are discriminated against in terms of qualification. You see, I have a hard time thinking that this is... to assume that this is true, okay? I was never mistreated for being a woman, I was never discriminated against for being a woman, and if I ever was, I didn't even notice, because I'm not the kind of person who asks for permission to do things, I do them! (F16). When asked about their relationships with co-workers, clients, and investors, 25 women reported situations in which they experienced gender discrimination. Interviewee F11 affirms that "[...] you always have to be prepared for a confrontation and to show that there is inequality". In everyday life, in moments with her peers, F12 points out that "[...] there are embarrassing situations, there are still sexist situations. [...] For example, sometimes we sit at a meeting table, there is only me and another woman, and sometimes our skills are put to the test". This situation is emphasized in F08's statement when she says that her male colleagues make sexist remarks and do not realize it: "I also identify [...] remarks that are a sexist, but without the intention of being sexist, but that men don't recognize that it is". Situations of discrimination and lack of respect are also reported in relation to customers. F15 describes a situation she experienced: [...] arriving at the customer's office, he asked, 'yes, where's the person in charge of IT? 'It's me.' Then he looks at you and he gets suspicious, then you spend an hour for the person to give you credibility, you know? Other situations of questioning and mistrust are also reported: "Sometimes there are clients who [...] 'ah, let me test her because she's a woman and I'll try to do it this way,' but then you have to keep your posture, right?" (F28). Contact with investors is also characterized

Purpose -To analyze the adherence to queen bee phenomenon (QBP) attitudes among women founders and co-founders of startups. Theoretical framework -We used the three dimensions of the QBP to understand how female founders of startups relate to other women in the work environment and how they deal with the male professional culture in which they are embedded. We also used the gender career literature to understand cooperative relationships between women. Design/methodology/approach -Our research approach is characterized as qualitative. We employed semi-structured interviews with 30 women founders of startups. We used the abductive analytical procedure. The authors carried out the coding process independently and we obtained adequate reliability values in the coding procedures. Findings -The participants stated that they assimilated male traits. However, they did not distance themselves from their gender identity group. Nor did they legitimize the gender hierarchy. Thus, we rejected our assumption that women founders of startups exhibit strong QBP adherence. Practical & social implications of research -This study contributes to the theoretical advancement of the QBP by analyzing a male context where QB attitudes are not salient. At the same time, we suggest gender bias awareness as a variable that impacts QB attitudes. Although cooperation among women reduces QB attitudes, it does not reduce gender inequality in the workplace. Thus, in terms of social implications, we reinforce the need for structural changes to achieve gender equality. Originality/value -The relationship between women has been scarcely studied in the startup context, and we found no studies that analyzed female founders of startups through the prism of the QBP.

For example, sometimes we sit at a meeting table, there is only me and another woman, and sometimes our skills are put to the test". This situation is emphasized in F08's statement when she says that her male colleagues make sexist remarks and do not realize it: "I also identify [...] remarks that are a sexist, but without the intention of being sexist, but that men don't recognize that it is". Situations of discrimination and lack of respect are also reported in relation to customers. F15 describes a situation she experienced: [...] arriving at the customer's office, he asked, 'yes, where's the person in charge of IT? 'It's me.' Then he looks at you and he gets suspicious, then you spend an hour for the person to give you credibility, you know? Other situations of questioning and mistrust are also reported: "Sometimes there are clients who [...] 'ah, let me test her because she's a woman and I'll try to do it this way,' but then you have to keep your posture, right?" (F28). Contact with investors is also characterized by uncomfortable situations, with interviewees reporting situations of harassment and invisibility. During a visit to an investor, F05 reported that the investor said to her male co-founder: "It must be very hard for you to carry the startup on your shoulders alone. And me by his side. Then I thought: Really? I carry the company on my shoulders together with him, he's not alone." F02 reported participating in an event where investors avoided contact with her: "They didn't want to keep in touch with me. They wanted to keep in touch with my male partner. And they'd always bypass contact with me and go straight to him." F04 reports that: Two investors said, "I'll give you... easily that 1 million you need, but you have to change your niche, because yes...women, it'll take a long time for them to realize that the business needs to grow, women aren't greedy. If you place your product for men, you'll capitalize very fast and then the startup will grow much faster". (F04). The participants' statements indicate that they are aware of their experiences of discrimination and that women have fewer opportunities than their male counterparts: "In several award events, we noticed that startups led by men or that had one man on the team were prioritized and considered more credible than startups led by women". (F21). --- No adherence to the meritocratic discourse Consistent with the recognition of the existence of gender discrimination in the startup ecosystem, only two of the 30 women interviewed expressed adherence to meritocratic principles. The other women interviewed perceived the need to work harder for their own careers and the organization than their male colleagues. The participants perceived the quantitative superiority of men and the difficulties in reaching strategic positions: "Women have difficulty reaching higher positions, right? So, there's no point in having 80% women on your team if the 20% that are the leadership, the CEOs, are all men, right?" (F01). Interviewee F26 reinforces this by stating that "[...] in larger startups, the leadership roles are all held by men. This is because of male benefits, so just the fact of being a man is already a great privilege". F09's statement recognizes the startup ecosystem as a limiting environment for female performance and that strategic decisions are made by men: "[...] women don't have the opportunity to show that they work as well as men, that they're equal in the situation". In addition, the interviewees emphasize that they are not in favor of a war of the sexes, but that they seek greater gender equality, as can be seen in the report of F04: "[...] women should receive enough support to achieve gender equity". --- Support for gender equality policies The women affirm that they are aware of policies aimed exclusively at female entrepreneurship in startups, and reaffirm their admiration for organizations that have this type of action, as can be seen in the words of F06: "I think it's interesting, I think it's valid, it's really cool who can do this type of action". --- Rebeca da Rocha Grangeiro / Manoel Bastos Gomes Neto The participants also report efforts to hire other women and mention public funding programs that seek to reduce gender inequality. Interviewee F23 explains that her "[...] female partner herself has already been a mentor, including in a hackathon that was only for women in technology, which was to develop solutions and the inclusion of women in the technology market". The selection process focused on recruiting women in technology was also presented by one of the interviewees: "Similar curricula, we choose the woman, because we'd like to have more representation, more women on the team. We understand the importance of diversity, of representation" (F06). Participant F10 points out that although there is no set policy at her startup, she has been involved in hiring other women. F10 recognizes that this management practice aims to reduce inequalities, while acknowledging the extra effort needed to integrate women into the labor market: "I have a very strong philosophy in the company, that the hiring priority, if I have two very good people for a function, I will always prioritize hiring a woman". Besides these initiatives in startups, the participants cite examples of incentives and report that they participate in actions developed by private and public institutions that aim to integrate women in the entrepreneurship and technology ecosystem, as F22 reports: I participate in SEBRAE Delas, which is focused on the development of women in companies and business." F20 also highlights the actions of other companies: "Ita<unk> Bank has Ita<unk> Women Entrepreneurs, Bradesco Bank [...] There are many women's initiatives coming from the private sector. --- Discussion The purpose of this study was to analyze the adherence of women founders and co-founders of startups to QBP attitudes. From the textual corpus analyzed, we observed that, like women leaders in other professional contexts, women founders and co-founders of startups conform to male gender stereotypes (Derks et al., 2011, Faniko et al., 2017, 2021). They reported high levels of commitment, with a significant portion of their life devoted to work. They also reported having a number of agency traits, which they said was necessary to gain professional legitimacy in a predominantly male context. These agency traits are: assertiveness, not being afraid, and not feeling intimidated. Previous empirical studies analyzing the effects of gender and position in the organizational hierarchy have shown that holding a leadership position has a greater statistical effect on male identification than the gender of the respondent (e.g. Faniko et al., 2016). Sperber and Linder (2023) mention that female entrepreneurs working in science, technology, engineering, and mathematics fields face the barrier between "being a woman" and "being an entrepreneur," which requires an adaptive approach to doing and undoing gender. In the role of entrepreneur, women are pressured to display agency traits (e.g. assertive, competitive, independent, and dominant) that are stereotypically attributed to men. And, at the same time, they need to exhibit behaviors that are desirable for women and associated with communal traits (e.g. patient, kind, nice, and compliant). Previous empirical studies have shown that women are not disadvantaged in the financial fundraising process because of their gender, but rather that stereotypically feminine behaviors undermine the likelihood of obtaining the desired investment (e.g. Balachandra et al., 2019). Furthermore, Sperber and Linder (2023) state that the male dominance in the field is not related to women's external conditions, such as low education, lack of experience, or low quality of their startups, but rather because they are perceived as female and do not conform to male self-references. Similarly, we believe that adherence to stereotypically masculine traits is a requirement of the professional context in which the participants are embedded, rather than a genuine QB attitude. Contrary to the assumptions of the QBP, the women interviewed in this study do not distance themselves from the self-group. Contrary to what is suggested by QB syndrome, which informs us that some women who have achieved professional and personal success would behave in ways that would create a barrier to the professional advancement of other women (Staines et al., 1974), we observed high levels of cooperation among women founders of startups. The participants reported participating in official and unofficial women's groups aimed at sharing professional support among women who share the same gender identity and have similar professional experiences. The competitive and anti-feminist behaviors that some scholars (e.g. Staines et al., 1974) have identified as typical behaviors of the queen bee syndrome were refuted in this study with women in the startup ecosystem. The interviewees reported more cooperative than competitive behaviors among women. Women Founders of Startups: an Examination Through the Prism of the Queen Bee Phenomenon Furthermore, the participants identified as feminists and/or showed that they shared the ideals of the feminist cause, such as being sympathetic to the idea of gender equality, promoting organizational practices that aim to establish gender equality, and promoting recruitment processes geared towards women when they notice a quantitative gender imbalance in some position or sector of the organization. Solidarity reinforces the feminist identity and strengthens women's demand for equality, as found by Cheng and Silva Jr. (2022). The feminist identity then looks to reduce self-group distancing and increase cooperative behavior among women. Regarding the third dimension of the QBP, the findings of this study allow us to propose that the women interviewed do not legitimize the gender hierarchy. Similarly to what was observed for the second dimension of the QBP, the participants' statements do not correspond to the assumptions of the QBP. Thus, contrary to the assumptions of the third dimension, the interviewees support and implement policies that favor gender equality. They are aware that women need to work harder to obtain the result that a man would obtain with little effort, they feel penalized for this and do not adhere to the meritocratic discourse. The awareness of gender bias therefore seems to reduce gender hierarchy legitimation. In view of the particularities of Brazilian culture (Coelho Jr. et al., 2022), the persistence of the markers of sexual division of labor (Hirata & Kergoat, 2007), and gender barriers in startups (Pavan et al., 2021), we assumed that women founders and co-founders of startups exhibit strong QBP adherence. Our study does not support this assumption. On the contrary, the participants were shrewd in denaturalizing sexist behaviors typical of the macho culture of the Brazilian Northeast and presenting sexual and sexist violence as experiences of gender discrimination. Our study corroborates previous research conducted with Brazilian women in public service (Arvate et al., 2018) and university (Grangeiro et al., 2023b), which suggested, respectively, the inexistence and low levels of QB attitudes in the professional contexts examined. Other non-Western studies on the QBP present similar results. A study conducted in private Turkish enterprises did not find queen bee attitudes among women leaders. Female employees of the private enterprises analyzed stated that their female managers were empathetic and supported them in their career development (Kobal, 2021). The findings of Xiong et al. (2022) also refute the existence of queen bee attitudes among Chinese female managers. Finally, the female founders of startups who participated in this study did not display all of the traits and attitudes that characterize the QBP. Although they reported strong professional commitment and adherence to agency traits, we did not find a direct relationship between the presence of these attitudes and the participants' characterization as queen bees. We suggest that adherence to male identification is strongly impacted by the professional context in which the participants are embedded. In general, the participants reported attitudes contrary to the second and third dimensions of the QBP, except for one participant who reported distancing herself from other women and three participants who legitimized the gender hierarchy by denying discrimination, adhering to meritocratic discourse, or opposing equality policies. Even though these attitudes were rare among our participants, it is important to highlight them, since one queen bee leader can negatively impact the professional experiences of many junior women or women at lower levels of the organizational hierarchy (Abalkhail, 2020). --- Conclusions Examining the reliability of the interview coding process between researchers can be considered a strength of our study. The calculation of Cohen's kappa index and the subsequent harmonization of the coding process by two researchers were essential steps to increase the reliability and robustness of the analyses presented in this manuscript. Another positive aspect of our research is the use of qualitative methods. Usually, QBP studies apply quantitative approaches, and the fact that we conducted a qualitative study allowed us to access data that are inaccessible through surveys. By giving voice to our participants, we were able to go beyond findings that confirm or refute the existence of queen bee attitudes among female founders of startups. Moreover, we observed women who are sensitive to gender issues and aware of historical and social processes that disadvantage women. In order to respond to the sexist culture in which they are embedded, these women ally with colleagues of the same gender to promote women's professional development. The fact of having participants from different economic sectors (e.g. finance, social economy, agriculture, education), some male-dominated and others feminized, ensured the heterogeneity of the professional experiences lived by the participants, which we consider a strength of our study. --- Rebeca da Rocha Grangeiro / Manoel Bastos Gomes Neto On a theoretical level, this study contributes to the advancement of the QBP by examining a male professional context where QBP attitudes are not salient. Previous studies have indicated that the male organizational context is conducive to the QBP (Faniko et al., 2021). This research highlights the opposite. Our results suggest that women's cooperation is facilitated by adherence to gender equality values. Based on our findings, we observe that adherence to gender equality values is related to adherence to feminist ideology. Thus, we believe that adherence to feminist ideology or feminist identity promotes the development of collective strategies to confront the gender threat faced by women in leadership positions. The cooperative and collective strategies strengthen women's connection to their self-group. We therefore suggest that adherence to feminist identity reduces the dimension of the QBP called self-group distancing. We also suggest that awareness of historical female subjugation leads women to fight the status quo. Awareness of gender bias enables women to identify the gender barriers they face and to question male privilege, not to adhere to meritocratic discourse, and to support gender quotas. Thus, women who are aware of the gender bias tend to oppose the gender hierarchy legitimation dimension of the QBP. Therefore, the findings of this study allow us to contribute theoretically to the advancement of the QBP, as it enables us to review and include variables not yet retrieved in the QB literature that mitigate QB attitudes, such as adherence to feminist identity and gender bias awareness. Also regarding theoretical contributions, the results of this study reinforce that women's collaboration may not be seen as the key to achieving gender equality at work. Women, individually or collectively, cannot be blamed for a social problem that has deep historical roots (Mavin, 2006). Collaboration among women favors women's career advancement but does not guarantee gender equality (O'Neil et al., 2011). This requires a willingness on the part of men to share positions of power and deeper changes in organizational culture. In terms of practical implications, this study contributes to practitioners, diversity managers, and policy makers. Given the relationship we found between gender bias awareness and the reduction of QB attitudes, namely the legitimization of gender hierarchy, we propose the first practical contribution of our study. We suggest the implementation of practices that raise awareness of gender issues by exposing historical sexist processes that disadvantage women but also create masculinities that harm men. In addition, informal recruitment practices established by women to achieve gender balance among employees should be formalized as human resource policies in startups. The non-probabilistic sampling method should be considered as a limitation of this study, since it can lead to biases regarding the participants accessed. Our second limitation relates to the snowball sampling strategy, which may have biased our sample. This could explain the large number of participants who are sensitive to gender issues and aware of gender inequalities. Even if generalization from a sample to a large population is not a concern for qualitative research, since generalizability differs from qualitative to quantitative research (Osbeck & Antczak, 2021), we acknowledge that 30 participants poorly represent female founders of startups in the Brazilian Northeast. Thus, the number of participants may be considered as a limitation of our study. In this research, we focused on female founders from the Brazilian Northeast, which contributes to gender studies since it is a region strongly characterized by sexism (Nicholus, 2019). Nevertheless, we suggest that future research expand to other Brazilian regions and compare the results of regions that are less and more characterized by sexism. Future research could also consider analyzing the impact of gender composition and of organizational context (male, mixed, female) on the male identification dimension of the QBP. In addition, we suggest analyzing, by means of quantitative research, the impact of feminist identity and gender bias awareness on the QBP dimensions, more specifically, the impact of adherence to movements favoring gender equality (feminist identity) on self-group distancing and the impact of gender bias awareness on gender hierarchy legitimation. --- SUPPLEMENTARY MATERIAL Supplementary material accompanies this paper. Supplementary Data 1 -interview script Supplementary Data 2 -transcription of the interviews Supplementary Data 3 -codebook and codes This material is available as part of the online article from https://doi.org/10.7910/DVN/PVNHCE. --- Financial support: There are no funding agencies to report. Open Science: Grangeiro, Rebeca;Gomes Neto, Manoel Bastos, 2023, "Supplementary Data -Women founders of startups: an examination through the prism of the Queen Bee Phenomenon", https://doi.org/10.7910/DVN/PVNHCE, Harvard Dataverse, V1. --- Conflicts of interest: The authors have no conflict of interest to declare. --- Copyrights: RBGN owns the copyrights of this published content. --- Plagiarism analysis: RBGN performs plagiarism analysis on all its articles at the time of submission and after approval of the manuscript using the iThenticate tool. --- Authors: 1. Rebeca da Rocha Grangeiro, PhD. in Psychology, Universidade Federal do Cariri, Juazeiro do Norte, Brasil. E-mail: [email protected] 2. Manoel Bastos Gomes Neto, Undergraduate Degree in Business Administration, Pontif<unk>cia Universidade Católica de Minas Gerais, Belo Horizonte, Brasil. E-mail: [email protected] Authors' contributions: 1 st author: Definition of research problem; development of hypotheses or research questions; definition of methodological procedures; data collection; literature review; analysis and interpretation of data; critical revision of the manuscript; manuscript writing. 2 nd author: Definition of methodological procedures; data collection; literature review; analysis and interpretation of data; critical revision of the manuscript; manuscript writing.

Purpose -To analyze the adherence to queen bee phenomenon (QBP) attitudes among women founders and co-founders of startups. Theoretical framework -We used the three dimensions of the QBP to understand how female founders of startups relate to other women in the work environment and how they deal with the male professional culture in which they are embedded. We also used the gender career literature to understand cooperative relationships between women. Design/methodology/approach -Our research approach is characterized as qualitative. We employed semi-structured interviews with 30 women founders of startups. We used the abductive analytical procedure. The authors carried out the coding process independently and we obtained adequate reliability values in the coding procedures. Findings -The participants stated that they assimilated male traits. However, they did not distance themselves from their gender identity group. Nor did they legitimize the gender hierarchy. Thus, we rejected our assumption that women founders of startups exhibit strong QBP adherence. Practical & social implications of research -This study contributes to the theoretical advancement of the QBP by analyzing a male context where QB attitudes are not salient. At the same time, we suggest gender bias awareness as a variable that impacts QB attitudes. Although cooperation among women reduces QB attitudes, it does not reduce gender inequality in the workplace. Thus, in terms of social implications, we reinforce the need for structural changes to achieve gender equality. Originality/value -The relationship between women has been scarcely studied in the startup context, and we found no studies that analyzed female founders of startups through the prism of the QBP.

Introduction: Social determinants of health include both specific characteristics of and pathways by which societal (including cultural) conditions impact health. Examples include income, education, social capital, occupation, community structure, social support, availability of health services etc 1.Persons can be categorized into subcategories by the method of social stratification. They are classified on the basis of variables which are regarded as significant by the society like income, gender, education etc. Individuals with better parameters are categorized in the higher social status 1. Health disparities denote a less proportionate burden or risk of death, disease etc. Studies have reported that females visit a dentist more often than males. But very less females get oral care during gestational period which is a crucial period in their life 1. Gestation is a natural process during which changes in hormonal status occurs. These variations enhance the risk of dental infections like gingivitis during gestation 2. Periodontal diseases are related to health problems like heart disease, diabetes, preterm birth etc. Some cultural beliefs negatively affect proper nutrition and the ability of females during gestational period to attain better S-26 dental health 2. Additionally some research indicates that being pregnant increases the risk of dental caries 3. During different stages of life social and demographic characteristics affect heath status and even lead to subcategories of riskier groups, during which prenatal care is important. Females during the period of gestation is a target group for giving educational strategies which identifies gestational period as a better stage for establishing healthier practices. Expecting mothers are psychologically more accepting of new information which helps them to advocate good life style habits which can also help the remaining members in the family 4. Adoption of better life style practices by females during the gestational period depends on socioeconomic variables like literacy, number of kids & age 4. Dental health behaviour during the gestational period like visit to a dental surgeon, dental hygiene, eating sweets, have a vital impact on their dental health and on their kid's future dental health. Females during their gestational period are advised to brush regularly and floss, and to avoid having high levels of sugar intake. Females during the gestational period do not know about the impacts of dental health on the foetus and its effects on pregnancy. Researches indicate that females during the gestational period had poor attitude towards dental health care during gestation 5. Hence, the present study was undertaken to assess Social and demographic factors affecting the oral health status and behaviour of pregnant women in Kerala. --- Methodology: The present cross sectional study was conducted in a group of expecting mothers in two private hospitals in Trichur district of Kerala. The Protocol for study was presented before the ethical committee and Ethical approval was obtained from the University Ethics committee for Human Trials of MS Ramaiah University of applied Sciences. Those who consented to participate in the study were included and those who were 35 years of age and above were excluded from the study. 128 participants were included in the study. The study was carried out from November 2021 -March 2022. Before that a review of literature [8][9][10][11][12][13][14][15][16][17][18][19][20][21] was conducted to derive the social factors affecting the dental health status and behaviours of expecting mothers. Questions were made based on these derived social factors. A questionnaire was designed to collect demographic details and information on social factors affecting the oral health status and behaviour of the study subjects. The independent variables were questions related to social and demographic factors. Questions related to social factors were used to gather information on job status, work stress, getting adequate leisure time in a day, financial independence, hours of work, whether a BPL card holder, enrolment in government programs, preference of availing dental health care services during pregnancy, free /paid treatment, adequate access to transportation, family encouragement in obtaining Oral Health Care regularly etc. Questions related to behaviour like frequency of brushing, use of tooth paste and also assessment of oral health status by assessing dental caries, periodontal status and also oral mucosal lesions were considered as dependent variables. The content of the questionnaire was validated by a total of 6 experts including 2 gynaecologists and 4 public health dentists. All the questions got a CVR value of 0.66 or above, hence all questions were retained. The questionnaire was translated and back translated by a Malayalam expert. After which, suitable modifications were made. Oral health status was assessed using DMFT index and CPI Index, Oral mucosal lesions scoring and coding adopted from WHO oral health assessment proforma 1997.Statistical analysis was done using SPSS version 21.Chi square test and fischer's exact test was performed to test association between categorical variables. Fischer's exact test was done in case where value is less than 5 for more than 20 percent of the expected frequencies. --- Results: The Mean age of the study subjects is 25 There was a significant difference between financial independence and total missing teeth where, 74.1 % of study subjects who reported having financial independence did not have any missing teeth compared to 20.8% of the study subjects who had no financial independence. (p value<unk>0.001). There was a significant difference between whether there were BPL card holder and frequency of tooth brushing, only 7.9 % of the study subjects who were BPL card holders had a habit of brushing twice daily compared to 66.2% of study subjects having no BPL Card. ( p value=0.000). In the present study, 61.7 percent reported that they had no miscarriage which is comparable to the study done by Barbieri W et al 4 on pregnant women attending the primary health care unit in Sao Paulo Brazil, 77.9 percent reported that they had no miscarriages. --- Sociodemographic factors affecting the frequency of percent had upto higher educational level regarding education. In the present study, 22.4% of study subjects having miscarriage had a CPI score of 4 compared to 10 % of study subjects having no miscarriage. This is in comparison to the study done by Barbieri W et al 4 which reported that miscarriage and abortion are factors impacting the oral health of pregnant women. In the present study socioeconomic condition/status and educational status negatively influence the CPI & LOA score and DMFT status. This is similar to the study done by Nota A et al 20 where the level of education is a vital component of socioeconomic condition and educational level of mothers was an important factor on the kid's general and oral health. In the current study lack of adequate access to transportation, the greater the distance from dental clinic to house/residence was negatively associated with CPI, LOA score and DMFT status. Also, many of them were not sure whether their health insurance covered dental treatment needs. This was similar to the study done by Patrick DL et al 6, where in Patients with limited socioeconomic resources assess the amount of time & money taken to visiting dental clinic once, in visiting the dentist. The eligibility for Medicaid insurance does not necessarily result in patients' enrollment, and enrollment does not ensure the availability, accessibility, and obtainment of needed dental care. In the present study, none of the pregnant women visited the dentist during pregnancy. This is in comparison to the study done by Saddki N et al 22 where poor socio-economic conditions denoted by low education level, lack of a job and small household income, were significant factors correlated with decreased probability of not going for a dental visit during the gestational period. In the present study, only 7 percent of the study subjects were caries free. This is in comparison to the study done by El-Mahdi Ibrahim HM et al 2 in Sudan where in only 24.5 % of the study subjects were free from dental caries. This is in comparison to the study done by El-Mahdi Ibrahim HM et al 2 where,10.7 percent were illiterate, 36.9 percent had educational level of primary school, 30.2 percent upto high school, 22.2 --- S-30 In the study done by El-Mahdi Ibrahim HM et al 2, 22.6 percent reported that their present dental health was poor. El-Mahdi Ibrahim HM et al 2, reported only 9.5 percent used other dental hygiene methods like dental floss. In the present study, 41.4 percent of the study subjects reported that they had no health insurance covering dental needs, whereas, 35.9 percent of the study subjects reported that there might be dental insurance, but they were not sure. 22.7 percent of the subjects reported that they did not know whether they had health insurance covering dental needs. In the studies done by Al Habashneh R, Guthmiller JM et al 23,Vergnes JN, Pastor-Harper D et al 24, Amin M, ElSalhy M et al 25 health insurance coverage was positively correlated with visits to dental surgeon during gestation. Only 9.4 percent of the study subjects in the present study reported that they frequently visited a dentist. There was no significant association between income and visits to the dentist. This is in contrast to the study done by Sun et al 26 2014, where low income women who were pregnant had a regular oral care. This is also in contrast to the studies done by Boggess KA et al 27 and Amin and Elsalhy et al 25 where income was associated with seeking oral health care during pregnancy.The finding of this study is similar to studies done by Al Habashneh R et al 23, Saddki et al 22 which showed no association between income and dental care use during gestation period. --- Recommendations: Further researches have to be carried out in a longitudinal fashion to assess the effect of socio demographic factors identified from this study in detail. --- Conclusion: Sociodemographic factors like SES, job status, financial independence, BPL status, enrolment in government programs, preference regarding availing oral health care services during pregnancy, choice of free / paid treatment, access to transportation, family encouragement in obtaining regular oral health care etc. acts as a strong barrier resulting in less use of dental health care services, which has to be effectively overcome by proper dental health education.

Aim: To assess Social and demographic factors affecting the oral health status and behaviour of pregnant women in Kerala.The present cross-sectional study was conducted in a group of expecting mothers at 2 private hospitals in Trichur district, Kerala, India. A questionnaire was used to assess the social and demographic factors affecting the oral health status and behaviour of pregnant women. Results: Study subjects having a professional education status did not having any periodontal pocket (CPI score 3 and 4). 85.7 % of the Study subjects having professional education status brushed twice daily whereas only 7.7 % of the study subjects having intermediate education or diploma brushed twice daily. There was a statistically significant difference between BPL status and frequency of brushing, as well as financial independence and number of teeth missing. CPI & LOA score was associated with sociodemographic factors. Conclusion: Social factors like SES etc act as a strong hindrance which lead to less use of dental health care services. Dental health education has to be given to overcome this.

The COVID-19 pandemic has caused long-lasting changes in many people's employment, social relationships, and mental and physical well-being. Caregivers of youth, children and seniors, representing 7.8 million Canadians (25% of the total Canadian population), tend to suffer in all aspects of life (finance, social, physical, emotional and mental) owing to their unique role in society during the pandemic. [1][2][3] Caregivers are people who tend to someone else's needs for a period of time, and the context of this study refers to caregivers of children and youth. 4 They act as a social support system, assist with daily tasks and ensure basic needs are met. 4 Caregivers often face challenges and difficulties, including poor time management, worsening emotional and physical stress, lack of privacy, financial strain and sleep deprivation. [5][6][7][8][9] Caregivers can struggle to balance this role with other responsibilities in their lives, and these strains can affect the youth they care for emotionally and mentally. 10 Since the COVID-19 pandemic began, many challenges typically faced by caregivers have been exacerbated, leading to physical and mental deterioration of these caregivers. [10][11][12] Children may have been at home for online schooling, routines may have been disrupted owing to limited access to childcare and social support systems, and many children and youth with mental health and/or addiction (MHA) concerns may have had diminished access to mental healthcare. 10 The increased stress associated with the pandemic also led to increased alcohol use in youth during the pandemic. 13,14 This has led to increased stress in youth and caregivers. Recent studies have shown that COVID-19 has negatively affected caregivers of children and youth and has been associated with substantial changes in caregiving responsibilities. [15][16][17][18][19][20] For example, when family caregivers were compared with noncaregivers in Pittsburgh, caregivers reported higher levels of anxiety and depression, increased food insecurity and financial worries, and lower social participation. 15 Furthermore, those with more COVID-19-related disruptions had worsening mental health and physical outcomes, owing to the increased difficulties added to their responsibilities, and thus had more negative perceptions of their responsibilities. 15 In Germany, the care situation of caregivers has worsened during the pandemic; 25.5% of caregivers that were unable to receive professional help reported that their responsibilities worsened during the pandemic, including overabundant demands, loss of social support and problems implementing COVID-19 measures. 16 Caregivers had diminished abilities to complete their responsibilities owing to pandemic-induced changes in their responsibilities, and this had a substantial negative impact on their physical and mental health. 16 In Japan, caregivers of children aged 3 to 14 years had an increase in mental distress compared with how they felt before the COVID-19 pandemic, with 24.1% reporting moderate mental distress and 29.3% reporting severe mental distress. 20 It was also found that such mental distress could stem from changes in daily routine (increase in screen time, decrease in time spent outside, closure of schools) and increases in child health issues. 20 Furthermore, caregivers' stress during the COVID-19 pandemic has been shown to be internalised by children, causing psychological and behavioural issues including yelling/screaming, name-calling and blaming others. [20][21][22] Thus, there is a need to continue to understand how the pandemic has affected the perception of caregiving responsibilities for children/ youth over time in other jurisdictions, including Canada. As the pandemic has progressed, it has affected caregiving in an everchanging way, as children and youth have returned to school and caregivers are going back to work and broader social contact. This current study aimed to investigate the sociodemographic and caregiver and child/youth mental health factors associated with perceived changes in caregiving responsibilities among caregivers of children and youth aged 0-25 years in Ontario, Canada, during the later stages of the COVID-19 pandemic. --- Method Study design and participants The current study used wave 3 data from the COVID-19 Mental Health & Addictions Service Impacts & Care Needs Study, which aimed to examine the effects of the COVID-19 pandemic on mental health and substance use among adults residing in Ontario, Canada. Data reported in this study were collected from January to March 2022. See the study protocol for details regarding participant inclusion/exclusion criteria, recruitment and sampling. 23 The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. All procedures involving human subjects/patients were approved by Sunnybrook Health Sciences Centre Research Ethics Board (ref. 1931). Written informed consent was obtained from all participants. Participants (n = 5000) were adults over the age of 18 living in Ontario, Canada. Data from 1381 participants who identified caregivers of children/youth aged 0-25 years were used for the purposes of this analysis. --- Measures Variables related to the impact of the COVID-19 pandemic on caregiving responsibilities in relation to participants' own and their child/youth's MHA concerns and sociodemographic characteristics were assessed through a series of measures. Complete details of the following measures can be found in the study protocol, including the sociodemographic characteristics of the overall study sample (n = 5000). 23 Caregiver strain during the pandemic was assessed using the Peabody Treatment Progress Battery 2010 and the Caregiver Strain Questionnaire-Short Form 7 (Adult Caregiver) questionnaire (both Cronbach's <unk> = 0.89). 24 The total strain was further categorised into tertiles: low (score between 2 and 3.83), medium (3.83 to 7.0) and high (7.0 to 10). 24 Depression and anxiety were assessed using the DSM-5 Self-Rated Level 1 Cross-Cutting Symptom Assessment for Adults (Cronbach's <unk> = 0.81). 25 Alcohol use was assessed using the Alcohol, Smoking and Substance Involvement Screening Test version 3.0 (Cronbach's <unk> = 0.85). 26,27 Six items about changes in caregiving responsibilities were created for the purpose of this study. Participants were asked if they had received any of a list of MHA services or supports in the past 3-4 months. Analysis was performed with a dichotomised response of yes (had access to at least one service/support) or no (received no service/support). Satisfaction with the availability of various types of social support (friends, family, romantic partners, community, co-workers, pets) since the COVID-19 pandemic was also assessed using a three-point Likert scale. 23 Analysis was performed with a dichotomised response of yes (satisfied with the availability of social support) or no (neither satisfied nor dissatisfied or not satisfied with the availability of social support). Caregivers were also asked to estimate the degree to which their child/ youth's mental well-being had changed from before the pandemic using a three-point Likert scale. 23 Analysis was performed with a dichotomised response in at least one child/youth per caregiver of either a positive or no change as distinct from a negative change. Caregivers were asked whether or not at least one child/youth per caregiver had used at least one substance in an unhealthy/excessive manner since the pandemic. 23 Analysis was performed with a dichotomised response of yes (used at least one substance in an excessive/unhealthy manner in at least one child/youth per caregiver) or no (had no excessive/unhealthy substance use in any child/youth cared for by caregiver). Access to MHA services since the pandemic commenced was also assessed through a dichotomised response of yes/no (yes: a service was accessed by at least one child/youth per caregiver; or no: no service was accessed by any youth). Participants were asked how their caregiving responsibilities had affected their well-being since the pandemic commenced. 23 Analysis was performed with a dichotomised response of a positive or no impact as distinct from a negative impact. --- Statistical analysis Data analysis was conducted in SPSS version 28.0 (IBM Corporation) with statistical significance set at P <unk> 0.05 (twotailed). Logistic regressions were performed with independent variables, namely sociodemographic characteristics, caregivers' and children/youth's MHA concerns, and the dependent variables, namely, change in caregiving responsibilities. There was no evidence of multicollinearity among the independent variables (all variance inflation factor values <unk> 2.5). --- Results --- General characteristics Demographics for the sample are presented in Table 1. The age range for this sample of caregivers (n = 1381) was 18 to 91 years, with an average mean (<unk>s.d.) of 43.60 years (<unk>10.47) (Table 1). In this study, 649 (46.0%) identified as male, and 719 (52.1%) identified as female. The majority (480, 34.8%) lived in Central Ontario, were Caucasian (893, 64.7%), had at least some post-secondary education (1271, 92.0%) and were married/common-law partners (1185, 85.8%) (Table 1). These caregivers cared for a total of 2423 children/youth between the ages of 0 and 25 years with a mean (<unk>s.d.) of 1.8 (<unk>0.9) children/youth per caregiver (Table 1). The mean age of children/youth being cared for was 10.3 years (<unk>7.0) (Table 2). Of the children and youth being cared for, there were 1284 (53.0%) males, 1129 (46.6%) females and ten (0.4%) who were reported to be non-binary (Table 2). --- Impact of COVID-19 pandemic on caregivers --- Caregiver strain Caregivers scored a mean (<unk>s.d.) of 2.22 <unk> 1.02 (range 1-5) on objective strain and a mean of 2.46 <unk> 1.12 (range 1-5) on subjective strain, and 4.68 <unk> 2.01 (range 2-10) on total caregiver strain (Table 3). Five hundred and forty-seven (39.6%) caregivers were in the low-strain group, 652 (47.2%) in the medium-strain group and 182 (13.2%) in the high-strain group (Table 3). Since the pandemic was declared, almost half of the caregivers indicated that their mental health had worsened (n = 629, 45.5%) (Table 3). It was found that 211 (15.3%) were at no risk, 880 (63.7%) were at low risk, 219 (15.9%) were at moderate risk and 71 (5.1%) were at high risk (Table 3) for alcohol misuse. Mental health and/or addictions services or support Over three-quarters of caregivers did not receive MHA services or support (n = 1050, 76%). By contrast, 1048 (76.6%) were satisfied with their level of social support since the pandemic was declared (Table 3). --- Caregiving responsibilities Four hundred and forty-eight (32.4%) caregivers indicated that their caregiving responsibilities had had a negative impact on their wellbeing since the pandemic was declared (Table 3). --- Impact of COVID-19 pandemic on children/youth A total of 981 (71.0%) caregivers reported negative mental wellbeing changes in at least one child/youth that they were caring for, and 243 (17.6%) indicated that at least one child/youth had signs of MHA issues since the pandemic began (Table 4). In addition, 181 (13.1%) caregivers had at least one child/youth with unhealthy/excessive substance use, and 157 (11.4%) caregivers had at least one child/youth accessing MHA services or support (Table 4). --- Predictors of perceived changes in caregiving responsibilities We used logistic regression to analyse the relationship between sociodemographic characteristics and caregivers' and children/ youth's MHA concerns (independent variables) and perceived changes in caregiving responsibilities (dependent variable) (Table 5). Of the sociodemographic characteristics, only ethnicity significantly predicted changes in caregiving responsibilities. --- Discussion The results of this study suggest that ethnicity, caregiver strain, negative changes in mental health for caregivers, negative changes in mental health for their children/youth, more children/youth cared by per caregiver and dissatisfaction with social supports were all independently associated with negative perceptions of changes in caregiving responsibilities. The COVID-19 pandemic resulted in new and increased responsibilities for caregivers of children and youth. Caregivers helped children and youth with online schooling by providing technology to access online platforms and necessities along with support in doing so, while simultaneously working from home. 5 The burden of juggling multiple responsibilities while being isolated from social support may be associated with negative perceptions of caregiving responsibilities. Previous studies of the relationships between caregiver strain, mental health, number of children/youth, satisfaction with social support and caregiving responsibilities have reported similar results. 28,29 A prior study considered adult caregivers (aged 18+ years) caring for children under the age of 18 years in the USA during the pandemic and showed that there were significant links between parents' caregiving burden, mental health, and perceptions of children's stress, which were in turn significantly linked to child-parent closeness and conflict. 29 Complementary results in another study of caregivers of at least one child between the ages of 6 and 18 years during the pandemic in the USA showed that having a set routine could buffer negatively perceived changes as a result of the pandemic. 28 This shows that the caregivers were affected more by the idea of additional responsibilities than by the actual responsibility. Given these findings, further work is needed to explore how the severity of caregivers' and youth's MHA concerns, caregiver strain and social support affect perceived caregivers' responsibilities. Compared with caregivers of underrepresented groups, White caregivers were more likely to report perceived negative changes in caregiving responsibilities. There may be several explanations for this finding. First, it is possible that cultural experiences may affect cognitive processes. 30 Second, other variables in the model, such as caregiver strain, access to supports, social supports, etc., may have moderated the relationship between race and caregiving perceptions differently than expected. By contrast, a previous study, albeit focusing on caregivers caring for adult cancer patients during the pandemic, reported that caregivers of underrepresented groups were more likely to perceive higher negative perceptions in caregiving responsibilities. 31 Future research should explore the relationship between caregiving responsibilities and ethnicity in more depth. This might include exploring protective and precipitating factors that may be different between groups. In addition, whereas this study found that gender did not significantly predict perceived negative changes in caregiving responsibilities, a prior study found that caregivers who were female had more strain and burden. 31 Findings of this study suggest that many sociodemographic characteristics do not heavily influence perceived changes in caregiving responsibilities; these results are surprising in the context of the published work in the field. There could be different possible perceptions as to why there is a difference in findings, including differences in the sample population, survey service and percentage of females in the sample. The prior study looked at a sample population in the USA using Amazon Mechanical Turk with 42% females, 31 whereas the present study was conducted in Canada through AskingCanadians with 52% females. Thus, future exploration is warranted to further understand the relationships between gender and caregiving roles. Caregivers' perception of changes in mental well-being in their children/youth significantly predicted perceived negative changes in caregiving responsibilities. Previous studies of the relationship between changes in mental health and caregiving responsibilities have shown similar results, despite investigating various populations in different regions. 29,32 Caregiver burden, perceived child stress and conflict in child-parent relationships have all been found to be positively associated with one another in the USA. 29 A prior study found that as a result of the pandemic, primary school children in Turkey developed sleeping problems, anger issues, fidgeting, restlessness, appetite problems, sadness, etc., as perceived by their caregivers. 32 There was a significant association between these symptoms and perceived stress in caregivers. Children also seem to internalise the stress and burden faced by their caregivers, leading to the development of MHA signs. 19 --- Strengths and limitations Participants in this study were a representative sample of residents in Ontario, improving the generalisability of the findings to the population. However, the study was conducted through an online survey, limiting the sample to those with internet access. Furthermore, the responses regarding children and youth's MHA concerns and access to MHA services and/or support were based on the caregivers' perceptions, which might not have been accurate depictions; it is possible that children and youth may hide their symptoms or service involvement from their caregivers. Thus, it is also important to evaluate these factors with children and youth themselves directly wherever possible. In addition, caregivers were asked to compare their current caregiving responsibilities with those before the pandemic, which was over 2 years prior; this may have led to recall bias, affecting the validity of the data. It is also important to acknowledge that not all the changes in the lives of individual caregivers and their youths were solely the result of the pandemic; indeed, there may be many life events and intercurrent medical and social issues that affect the lives of caregivers, the measurement and the impact of which were beyond the scope of this study. Moreover, this study included an evaluation of a heterogeneous group of youth who were receiving care from caregivers. It is possible that certain subgroups (those with chronic illness, acute illness, physical disabilities, etc.) may have different levels of strain; it was beyond the scope of the current study to evaluate the differences between these different subgroups, but this may be a valuable issue to explore in future research. Last, the crosssectional nature of this study limits the conclusions to relationships observed at one particular time point in the COVID-19 pandemic. Future studies could explore perceived negative changes in caregiving responsibilities as a result of the COVID-19 pandemic longitudinally to better determine how these relationships evolved over time (in caregivers and youth pairs). --- Future implications This study examined the factors that contributed to perceived negative changes in caregiving responsibilities among adult caregivers (aged 18+ years) of children and youth (aged 0-25 years) in Ontario, Canada, during the COVID-19 pandemic. The findings show that certain factors (caregiver strain, negative changes in mental health for caregivers and their children/youth, high number of children/youth cared for per caregiver and dissatisfaction with social supports) were predictors of negatively perceived changes in caregiving responsibilities. Other factors including screening for risk of depression and anxiety, access to MHA services/supports and signs of MHA concerns in children/youth were not associated with the outcome. Ethnicity was the only sociodemographic factor that significantly predicted the outcome. These findings could lead to a greater understanding of the impact the pandemic has had on caregiving responsibilities and how this has affected different populations across Ontario. They may also help to inform healthcare providers who seek to support caregivers of children and youth by developing their understanding of the nature of the challenges experienced during the pandemic. --- Data availability Requests for access to deidentified data should be directed to the corresponding author ([email protected]). Data may be shared upon reasonable request and pending ethics approval. --- Author contributions M.C.: conception of work, study design, analysis and interpretation of data, drafted manuscript; R.M. and A.L.: study design, data interpretation, revision of manuscript. All authors read, critically revised and approved the final manuscript. --- Declaration of interest None.

The COVID-19 pandemic has created long-lasting changes in caregiving responsibilities, including but not limited to increased demands, loss of support, worsening mental and physical health, and increased financial worries. There is currently limited evidence regarding factors associated with perceived changes in caregiving responsibilities.This observational study aimed to investigate factors (sociodemographic characteristics of caregivers and mental health and/ or addiction concerns of the caregiver and their youth) that predict perceived negative changes in caregiving responsibilities among adult caregivers (aged 18+ years) of children and youth (aged 0-25 years) in Ontario, Canada, during the COVID-19 pandemic.Data were collected from 1381 caregivers of children and youth between January and March of 2022 through a representative cross-sectional survey completed online. Logistic regression was conducted to determine predictors contributing to perceived negative changes in caregiving responsibilities.Among the sociodemographic characteristics, only ethnicity significantly predicted outcome. Higher caregiver strain (odds ratio [OR] = 10.567, 95% CI = 6.614-16.882, P < 0.001), worsened personal mental health (OR = 1.945, 95% CI = 1.474-2.567, P < 0.001), a greater number of children/youth cared for per caregiver (OR = 1.368, 95% CI = 1.180-1.587, P < 0.001), dissatisfaction with the availability of social supports (OR = 1.768, 95% CI = 1.297-2.409, P < 0.001) and negative changes in mental wellbeing in at least one child/youth (OR = 2.277, 95% CI = 1.660-3.123, P < 0.001) predicted negative changes in caregiving responsibilities.These results support further exploration of the implications of negative perceptions of caregiving responsibilities and what processes might be implemented to improve these perceptions and the outcomes.

Background In a perfect world, all medical decisions would be based on evidence. While a lofty ideal to which most clinicians aspire, the context of medical practice makes that goal difficult to realize. On the one hand, there are tensions between clinical scientists and clinical practitioners, problems in the translation of research findings, and difficulties in the timing and sequencing of developing, assessing, adopting, and abandoning clinical practice [1][2][3][4][5][6]. On the other hand, there are conflicts within science, "evidence wars" [7] that weaken the power and influence of the evidence on healthcare practices, and much work that can be characterized as cleaning up the evidence base so that it is strong and pure enough to maximize impact (AGREE Project, Cochrane Collaboration, EQUATOR Network, GRADE Working Group). Narrowing the gap between research and practice is not simply a matter of convincing knowledgeable and conscientious practitioners to do the right thing, but rather acknowledging that clinicians' practice patterns are located within a social structural context so that even the strongest evidence's influence is mitigated by the conditions of the context in which health care is practiced. In their article on "Evidence-based de-implementation for contradicted, unproven, and aspiring healthcare practices," Prasad and Ioannidis call these factors "entrenched practices and other biases" [7]. We propose to disaggregate "entrenched practices and other biases." Following the lead of Prasad and Ioannidis [7], we delineate the extra-medical influences of historical, economic, political, and social contexts that should be considered in de-implementation and illustrate using the case example of the de-implementation of radical mastectomy. --- Discussion Historical context Some drugs, devices, and procedures have a long and deep practice history. They are entrenched in organizational infrastructure, and that inertia makes de-implementation of a routinized procedure particularly difficult. The case of radical mastectomy illustrates historical entrenchment. Radical mastectomy was developed as a treatment for breast disease at the end of the 1800s, when women presented with tumors that were reported to be up to 5 in. in diameter [8,9] that hemorrhaged, were infected by fungus, emitted exhausting discharges, caused much pain, and metastasized [10]. Halsted developed his radical techniqueexcision of the breast, pectoral (chest) muscles, axillary lymph nodes, and associated skin and subcutaneous tissue-in response to these conditions. Prior to the late 19th century, cancer was believed to be a blood disorder, not amenable to surgery [11]. That view gave way to the theory of cancer as a local disease. Physicians believed that once the breast succumbed to the disease, it was diseased in toto [12], making it necessary to remove the complete breast. Next, they theorized that breast cancer spread via the axilla, so surgeons began removing the axillary nodes, whether or not they evidenced disease [13]. Acceptance of the theory of local disease with centrifugal spread encouraged surgeons to operate early, without waiting for the traditional sign of ulceration of the tumor to make the diagnosis of cancer. In line with the theory of breast cancer as a local disease that spread centrifugally, Halsted [9] and Meyer [14] contended that cancer spread from the breast and axilla to the pectoral muscles, and therefore, they too should be removed. Halsted [9] also believed that cancer could be spread by cutting through diseased tissues, either because cancer cells were liberated or because a contaminated scalpel would infect new sites. This led to the technique of cutting a wide margin of healthy tissue along with the diseased portion of the breast and removing the entire breast and all associated tissue in one piece at once, akin to amputating a limb. Halsted's version of mastectomy was distinguished from that of his colleagues by his insistence on the routine performance of the procedure for the removal of the breast exactly as he had defined it. The glands and tissues specified by Halsted were no longer to be removed or not removed at the discretion of the individual surgeon determining their approach on a case-by-case basis; they were to be removed by him on a regular basis. The Halsted approach signaled the standardization of surgical treatment for breast cancer. Radical mastectomy became the unvarying response to every set of symptoms and every medical history ( [15]: page 64), and until 1968, the number of radical mastectomies in the USA increased each year. As early as 1941 in Scotland, McWhirter began treating his patients with simple mastectomy and radiation therapy [16,17]. Influenced by McWhirter, an American surgeon, George Crile Jr., began experimenting with lesser surgery and radiation treatment. Crile published his first clinical report in 1961 in the Annals of Surgery [18]. By 1965, there were four clinical reports or trials published in the literature. Figure 1 shows the accumulation of the English language evidence base comparing radical mastectomy to lesser procedures up to the 1990 US National Institutes of Health Consensus Development Conference on the treatment of early-stage breast cancer [19]. Of note is that all of the trials except for one, a 1972 clinical trial by Atkins et al. [20], found the lesser surgery to be equal to or better than radical mastectomy or found that lesser surgery with adjuvant therapy was superior to radical mastectomy. Results from the first US clinical trial of postoperative radiation were published in 1970. The number of clinical trials continued to grow, and pooled analysis [21] and meta-analysis [22] confirmed that lumpectomy (breast-conserving surgery to remove a tumor (lump) in a breast and a small amount of normal tissue around it [23]) plus radiation was just as effective as mastectomy. Radical mastectomy had a long reign as the treatment of choice for breast cancer. However, its de-implementation was eased because practitioners did not necessarily have to completely abandon the procedure; they had something they could do instead-modified radical mastectomy (surgery in which the breast, most or all of the lymph nodes under the arm, and the lining over the chest muscles are removed [23]). Figure 2 shows that as radical mastectomy declined, modified radical mastectomy increased in the USA. Therefore, de-implementation of radical mastectomy was facilitated by the development of radiotherapy as adjuvant to lesser surgery and mammography to detect breast cancer at an earlier stage when amenable to treatment with lesser surgery. The historical view of breast cancer also reveals how worldviews, theories championed by specialists, and individual clinicians' beliefs about causal mechanisms of disease mesh with practice decisions. When breast cancer was thought to be a blood disorder, surgery was not a viable option. When the paradigm shifted to view breast cancer as a local disease, surgery became the primary option and was used to support and justify the Halsted approach. Others have also noted how clinicians' beliefs about the body, pathophysiology, disease, and treatments can support change (both adoption and deimplementation) and inhibit change when it runs counter to the current worldview [24]. More recently, Gabbay and le May have highlighted the role of mindlines (collectively reinforced, internalized, tacit guidelines) in professional socialization [2], part of which includes clinicians' beliefs and worldviews. What is often difficult to tease apart, however, is whether worldviews/beliefs lead to practice change or whether changes in practice catalyze changes in worldviews to justify what they are proposing (sometimes referred to as persuasive knowledge use). In any case, paying attention to the history of the evolution of clinicians' worldviews and beliefs may provide important insight into the de-implementation process. --- Economic context Some drugs, devices, and procedures have a complex financial structure that can pit physicians, hospitals, the pharmaceutical industry, and device manufacturers against each other. If de-implementation has financial consequences for a sector that is powerful and profiting, economic entrenchment may be a barrier. The case of radical mastectomy illustrates economic forces of resistance to de-implementation. In fee-for-service payment systems, the more a provider does, the more profit is generated. Therefore, the de-implementation of radical mastectomy in the USA was up against the desire and need to maintain well-established revenue streams. In 1986, Mu<unk>oz and colleagues compared the cost of lumpectomy versus mastectomy for stage I or stage II breast cancer at Long Island Jewish Medical Center [25]. Hospital inpatient and surgeon fees were higher for mastectomy compared to lumpectomy, so one would assume that mastectomy would be the favored option. However, after a lumpectomy, the hospital charged for outpatient radiotherapy, and the services of a physician radiotherapist were also reimbursed. Therefore, when considering who is profiting 0 20,000 financially from the radical mastectomy slated for deimplementation, surgeons would lose financially, but hospitals and radiologists would benefit from deimplementation [25]. In 2000, Palit et al. returned to the question of the costs of breast-conserving surgery versus modified radical mastectomy, in light of the evolving nature of breast cancer treatment, especially the integration of post-mastectomy reconstructive surgery into treatment protocols [26]. They found that women who received breast-conserving surgery had a shorter length of stay and a lower cost of surgery than women who received modified radical mastectomy (with or without reconstruction) [26]. However, the addition of radiotherapy (typically 25 daily sessions) made the total cost of breastconserving surgery greater than that of modified radical mastectomy alone and 16% greater than that of modified radical mastectomy with reconstruction [26]. Palit et al. found that breast-conserving surgery was underutilized with respect to the number of women eligible and observed that the preference for modified radical mastectomy "appears to be guided primarily by physician attitudes and confusion regarding standard eligibility criteria" ( [26]: page 444). Therefore, from an economic perspective, it appears that modified radical mastectomy would be difficult to sustain were it not for the positioning of the surgeon, who performs the biopsy that confirms the diagnosis and tumor staging, and thereby has the benefit of a series of ongoing contacts with the woman patient and therefore the opportunity to exert influence in the choice of breast cancer treatment. Furthermore, if a government payer or a private insurance company continues to reimburse a drug, device, or procedure-even in the face of an evidence base that indicates the need for de-implementation, most likely there will be little incentive for de-implementation. --- The context of politics and specialties Some drugs, devices, and procedures have a wellestablished association with a particular practice specialty. Those seeking to promote the de-implementation of a technology must consider who "owns" the drug, device, or procedure and assume that those clinicians would want to keep what they have. With respect to radical mastectomy, surgeons, who historically sought to distinguish themselves from generalists who could only offer palliative care, were the specialists who offered a treatment. After perfecting the procedure of radical mastectomy, it became the standard treatment for breast cancer. Paradoxically, this routinization of a procedure in order to reduce uncertainty about when to use it resulted in "deskilling" or loss of clinical skills or judgment. Surgeons, in fact, were left with one approach to all permutations of breast cancer. If the surgery failed to stop the disease, treatment failure was attributed to the surgery coming too late or being too conservative [27]. Without diagnostic procedures or alternative therapies, radical mastectomy became entrenched as the only treatment for breast cancer, and surgeons were the specialists who treated it. Radical mastectomy reached its heyday in the USA during the 1950s, a time when radiation oncology was beginning to take hold in Europe [15]. Advances in diagnostic radiology resulted in mammograms that provided early detection, making it increasingly untenable for surgeons to continue to defend radical mastectomy as the standard treatment when faced with patients who had small tumors and a degree of malignancy that could now be staged by ever more sophisticated pathology analysis. Furthermore, by the late 1970s, surgeons were doing biopsies on an outpatient basis, and women patients were demanding a "two-step" procedure, i.e., separating diagnostic biopsy from breast cancer treatment. Biopsy allowed surgeons to hold their position as the point of first contact in the definitive diagnosis of breast cancer, but once through the gates and armed with a pathology report, women patients with early non-metastasized cancer could seek the treatment options that involved other specialties. --- Social context De-implementation does not necessarily occur within the protected confines of research science and clinical practice. Research and practice are situated in a broader social context, and under some circumstances, the timing and sequencing of developing, assessing, adopting, and abandoning drugs, devices, and procedures are subject to outside social forces that affect health services. The interest in de-implementing radical mastectomy extended beyond biomedical researchers and clinicians. At the same time that an evidence base was developing in the USA, the social movements of the 1970s and 1980s catalyzed the women's health movement. Movement participants challenged the authoritarian relations within the practice of medicine, insisted on patient participation in medical decision-making, recognized patients' psychological and social needs, and advocated that the patient's experience be acknowledged in the doctor-patient encounter. Lay activists encouraged women with breast cancer to seek consultation from radiation and medical oncologists and diagnostic biopsies from surgeons with a reputation for considering the possibility of lesser surgery. This sentiment was reflected in popular articles indexed in the Readers' Guide to Periodical Literature [28] that discussed radical mastectomy. These articles were published in high-circulation journals, such as Readers' Digest (17,829,000), McCall's (7,500,000), Good Housekeeping (5,600,000), Newsweek (3,000,000), Consumer Reports (2,150,000), and Glamour (1,946,078) [29]. The intensity of the viewpoints expressed is indicated by these select titles of some articles: "Breast Cancer Debate: Mastectomy study" [30], "Controversy over breast cancer: Radical mastectomy" [31], "Right to choose: Mastectomy?" [32], "Breast cancer: The retreat from radical surgery" [33], "I said no to my doctors" [34], and "Breast Cancer: Death to the Radical?" [35]. Furthermore, journalists often interviewed critics of radical mastectomy, such as surgeon George Crile Jr. [36], noted coordinator of the national randomized controlled trial of breast cancer treatments Bernard Fisher [37], and coordinator of cross-national trials Gianni Bonadonna [38]. The reporting on breast cancer in the popular literature indicates that consumers were aware that there was controversy within medicine and that there were treatment alternatives to radical mastectomy. This awareness may have led to some degree of patient pressure on surgeons. In a 1985 television interview on lumpectomy, surgeon Susan Love claimed that, "This treatment option was not developed by doctors and surgeons looking for a better way. It was women who said, 'I refuse mastectomy. You better find another way to treat me'" [39]. De-implementation came at a time when there was a convergence in clinicians' and patients' awareness of the scientific evidence, patients expressing their preferences for an alternative, and when there were other viable treatments to offer instead of radical mastectomy. While changing social forces strengthened women patients' role in de-implementing radical mastectomy, recently researchers have noted an upward trend in the rate of mastectomy. Katz et al. reported the relationship between patients' retrospective reporting of their involvement in breast cancer treatment decision-making and the actual treatment they received [40]. Their survey respondents reported that their concerns regarding the recurrence of disease and the effects of radiation gave rise to their preference for mastectomy over lumpectomy [40]. Gomez et al. found increasing rates of mastectomy in women diagnosed with ductal carcinoma in situ (DCIS) who were white, younger than 50 years at diagnosis, and living in the highest 20% of neighborhoods ranked by socioeconomic status [41]. The authors contended that this group of women, who we would expect to have high decision-making agency, most likely preferred mastectomy because of negative attitudes toward radiotherapy, positive expectations regarding reconstruction, their understanding of their risk of recurrence, and not wanting to undergo tamoxifen therapy for 5 years post-lumpectomy (as recommended by a 2009 practice guideline, tamoxifen therapy was not necessary for women who were treated by mastectomy [42]). While we laud these researchers for considering the patient's role in treatment decision-making, we note that their methodology did not employ direct observations of the physicianpatient discussions of treatment courses (such as can be found in Routine Complications by socio-linguist Candace West [43]). Therefore, it is possible that surgeons influence their patients' knowledge, attitudes, and beliefs regarding the risk of recurrence of breast cancer, the possible effects of radiation, the utilization of tamoxifen, and the promise of plastic surgery reconstruction. Given that a modified radical mastectomy is a quicker and easier surgery than breast-conserving surgery that in the USA commands higher surgeon fees [26], and is followed by subsequent reconstructive surgery that upper socioeconomic status women would be able to afford, a consideration of financial and professional forces would suggest that surgeons may be offering guidance that favors mastectomy over lumpectomy [26]. In Canada, where the costs of cancer surgery, radiation, and chemotherapy treatment are covered by provincially funded healthcare plans and provided to all citizens and landed immigrants, surgeons receive the same fee for either breast-conserving surgery or mastectomy (modified radical or otherwise). Plastic surgeons are the sole beneficiaries of a woman's choice of reconstruction after a modified radical mastectomy. A recent Canadian Institute for Health Information (CIHI) report found that mastectomy rates exceeded 50% for women who would have to travel 1.5 h or longer (each way) to reach a center offering radiotherapy [44]. In this case, it seems plausible to assume that women who live in rural or sub-rural areas are avoiding travel to the approximately 25 daily radiation therapy visits following breastconserving surgery, rather than expressing a preference for modified radical mastectomy surgery. Thus, the distance a woman has to travel between her home and a cancer center is a care delivery structural feature that can be conceptualized as a social force that sustains the use of modified radical mastectomy. --- Summary In sum, evidence did play a role in the case of de-implementing radical mastectomy-but it was not the star of the performance as would be hoped for from an evidence-based medicine perspective. We share Prasad and Ioannidis' ideals regarding the evolution of ever-better science to guide implementation and de-implementation [7]. However, there are many cases of factors other than science catalyzing and sustaining misguided implementation and of sound and proven science failing to catalyze de-implementation [27,[45][46][47]. Even when the empirical evidence is strong, scientific arenas may not necessarily be where the implementation/de-implementation decisions are made. Most scientists eschew advocacy, preferring to "let the science speak for itself" [48]. However, how can the science be heard, when there is a great deal of clatter from extrascientific historical, economic, professional, and social forces in the practice arena? Traditionally, the approach used to try to keep clinicians up to date has been continuing professional development or continuing medical education that relies on pedagogical theory. This underlying assumption of this approach is that if you educate people, they will be enlightened and will change their actions in response to the knowledge gained. In our experience, telling clinicians that the scientific evidence says what they are currently doing is no longer beneficial may not convince them to de-implement because it may not have been the scientific evidence that established the practice nor sustained its utilization [2]. In the field of implementation science, the role of psychological theories in understanding and predicting clinician behavior is influential. This should not be surprising given that ultimately individual clinicians (and their patients) make and follow through (or not) on decisions about care. While we acknowledge that these theories are helping to build our understanding of deimplementation, we believe that psychological theories are best utilized in the interpersonal realms of how we recruit, mobilize, involve, and retain clinicians on the teams that work for de-implementation. For example, Bowen and Graham have proposed an Engagement Paradigm in which scientists and clinicians collaborate to create evidence that is pre-designed to be clinically relevant and utilizable [49]. Furthermore, Ong et al. demonstrated that even when trying to precipitate behavior change (a key issue for psychology), psychological theory may offer little assistance [50]. Instead, they analyzed patient practices and strategies in the broader social context to explain why, for whom, and under what circumstances a treatment works [50]. To advance the science of de-implementation, we argue that the time is ripe for more in-depth analysis and understanding of all of the factors operating to sustain commitment to medications, devices, procedures, interventions, and tests that the evidence slates for deimplementation. There is a wealth of work from the disciplines of Science and technology studies, the History and philosophy of science, the Sociology of health and illness, and Medical Anthropology that addresses issues of de-implementation in medicine and beyond. Table 1 delineates these disciplines, what they study, issues commonly addressed, methodological approaches typically taken, and a sample publication for each. There is much to be learned from applying social science approaches to study both resistance to de-implementation (e.g., radical mastectomy [27] and episiotomy [46]), as well as appropriate and even rapid de-implementation (e.g., halting the prescribing of rofecoxib and hormone replacement therapy). Contrasting these findings with the results from studies of rapid (both appropriate and inappropriate) implementation (e.g., the rapid adoption of experimental HIV drugs [45]) will advance our understanding of de-implementation. Furthermore, as we consider the extra-scientific forces that sustain "entrenched practices and other biases" [7], we need to search literature not indexed in typical medical databases such as PubMed, CINAHL, and EMBASE for socio-historical analyses of de-implementation in health and other fields (see for example, a provocative study of CT scanners in radiology departments published in the management literature [55]). In sum, we contend that even when clinicians know what practices are in accord with scientific evidence, they often remain confined by a set of structural forces beyond their control and are not able to make changes. We propose to conceptualize de-implementation as systems change, leading to work for de-implementation at the systems level. We suggest intervening in the broader context in which clinicians work-the social, political, and economic realms-to change systems rather than trying to change individuals. We also argue that the time is right to further develop our de-implementation theories to incorporate the social science that considers historical, economic, professional, and social forces for change. --- Competing interests The authors declare that they have no competing interests. --- Authors' contributions TM and IDG contributed equally to the conception and design of the manuscript. TM drafted the manuscript and IDG revised it, and they both edited it critically for important intellectual content. Both authors read and approved the final version. --- Author details 1 The Sophie Davis School of Biomedical Education, City University of New York, 160 Convent Avenue, New York, NY 10031, USA. 2 School of Epidemiology, Public Health and Preventive Medicine, University of Ottawa, 451 Smyth Road, Ottawa, Ontario K1H 8M5, Canada.

Background: In their article on "Evidence-based de-implementation for contradicted, unproven, and aspiring healthcare practices," Prasad and Ioannidis (IS 9:1, 2014) referred to extra-scientific "entrenched practices and other biases" that hinder evidence-based de-implementation. Discussion: Using the case example of the de-implementation of radical mastectomy, we disaggregated "entrenched practices and other biases" and analyzed the historical, economic, professional, and social forces that presented resistance to de-implementation. We found that these extra-scientific factors operated to sustain a commitment to radical mastectomy, even after the evidence slated the procedure for de-implementation, because the factors holding radical mastectomy in place were beyond the control of individual clinicians. Summary: We propose to expand de-implementation theory through the inclusion of extra-scientific factors. If the outcome to which we aim is appropriate and timely de-implementation, social scientific analysis will illuminate the context within which the healthcare practitioner practices and, in doing so, facilitate de-implementation by pointing to avenues that lead to systems change. The implications of our analysis lead us to contend that intervening in the broader context in which clinicians work-the social, political, and economic realms-rather than focusing on healthcare professionals' behavior, may indeed be a fruitful approach to effect change.

I ndividuals of transgender experience (herein- after ITE), particularly women of trans experience, are disproportionately affected by HIV. In a meta-analysis of HIV prevalence studies in the United States, among those whose serostatuses were laboratory-confirmed, nearly 28 percent of women of trans experience overall, and 56 percent of African American women of trans experience, were HIV-positive (Herbst et al., 2008). A systematic review and meta-analysis of studies that assessed HIV infection burdens among transgender women found the pooled HIV prevalence was 19.1 percent in 11,066 transgender women worldwide. The study concluded that transgender women are a very high burden population for HIV and are in urgent need of prevention, treatment, and care services. The meta-analysis showed remarkable consistency and severity of the HIV disease burden among transgender women (Baral et al., 2013). Men of trans experience are also disproportionate affected by HIV (Clements-Nolle, Marx, Guzman, & Katz, 2001;Herbst et al., 2008), and research suggests that men of trans experience who have sex with men are at increased risk for HIV acquisition (Reisner & Murchison, 2016;Rowniak, Chesla, Rose, & Holzemer, 2011). A recent study with HIV-positive men of trans experience in the United States highlighted significant unmet social and health care needs. Approximately half the sample of men of trans experience were living in poverty and only 60 percent had sustained viral suppression (Lemons et al., 2018). Additional research is needed to understand the HIV risk behavior among women and men of trans experience and develop evidence-based HIV interventions to address their needs. It is also important to consider HIV risk and protective factors among nonbinary ITE, who have received minimal attention to date in research and programmatic interventions. Syndemics theory has been proposed to describe the synergistically related epidemics that cluster and arise from harmful conditions (Singer, 1994). Over the past two decades, it has been used to explain and analyze the burden of HIV/AIDS among sexual and gender minorities (Parsons,doi: 10.1093/hsw/hlz008 <unk> 2019 National Association of Social Workers Millar, Cain, & Gurung, 2018;Stall et al., 2003). In the context of transgender populations, several syndemic conditions have been theorized as contributing to the high HIV incidence and prevalence within this group, including experiences of discrimination and social exclusion, stigma, health care providers' insufficient competence, and elevated rates of sexual risk and substance use (De Santis, 2009;Garofalo, Deleon, Osmer, Doll, & Harper, 2006;Herbst et al., 2008;Kenagy, 2005;Sevelius, Reznick, Hart, & Schwarcz, 2009;Stotzer, 2009;Xavier, Hannold, & Bradford, 2007). In particular, many ITE lack resources and encounter prejudices and inconsistencies in care (Gridley et al., 2016). As such, a complex syndemic of psychosocial factors affect engagement in HIV risk behaviors, access to care, and utilization of care services if and when they are accessible. Despite the growing evidence on the effectiveness of HIV risk reduction interventions for other marginalized groups, very few biobehavioral approaches to reduce HIV risk and transmission among women of trans experience have been developed and tested, and none exist for men or nonbinary ITE. According to a recent review of the literature on behavioral interventions to prevent HIV transmission and acquisition for women of trans experience, only five U.S.-based HIV prevention interventions have been published in the peer-reviewed literature, all of which are early-phase nonrandomized interventions (Taylor, Bimbi, Joseph, Margolis, & Parsons, 2011). Culturally grounded, theory-driven HIV prevention interventions for ITE should adopt comprehensive approaches that incorporate inclusive health care, stable housing, sexual empowerment, and trauma support, and respond to the impact of stigma, discrimination, and social marginalization on HIV risk. Existing interventions have drawn on community strengths to develop appropriate interventions (Garofalo et al., 2012;Nuttbrock et al., 2013). Engaging women of trans experience in the development and implementation of interventions has shown particular promise. The Life Skills intervention, grounded in socioecological theory, was developed by and for young women of trans experience (Garofalo et al., 2012). Life Skills took a multifaceted approach that emphasized safer sex practices and confronted the stigmatizing experiences that leave many women of trans experience with limited employment opportunities and decreased sexual agency. It is important to note that the research and practitioner team collaborated with young participants to develop and refine the Life Skills intervention. Participants suggested incorporating more information concerning gender confirmation surgeries and ensuring cultural fluency. Training peer health advisers among sexual and gender minority communities has improved engagement in HIV prevention and care of at-risk populations. Studies suggest that lay health adviser interventions reach large numbers of at-risk community members and may benefit populations disproportionately affected by HIV (Rhodes et al., 2015;Sun et al., 2015). Peer-led interventions have also been found successful in combating depression and transphobic experiences that promote risk behaviors in younger transgender populations (Nuttbrock et al., 2013). This article extends the previous work on implementation science to address the needs of ITE by presenting and discussing the Transhealth Information Project (TIP), a peer-led six-session hybrid individual-and group-based HIV risk reduction intervention for ITE. Two community-based organizations in Philadelphia, the Gay and Lesbian Latino AIDS Education Initiative (GALAEI) and Prevention Point Philadelphia, launched the intervention in 2003. TIP remains the only multiservice, trans-specific HIV prevention intervention in the three HIV epicenters in the northeastearn U.S. corridor comprising Philadelphia; Camden, New Jersey; and Trenton, New Jersey; and it is led and staffed by ITE. In the following sections, we provide an overview of TIP's theoretical framework and approach for promoting and reinforcing a combination of biomedical, behavioral, and social and structural intervention components to promote HIV prevention and care; identify strategies for engagement of multiple stakeholders, including prison systems, in the provision of services for ITE; and demonstrate the scope and reach of this peer-led intervention in regard to linkage to social services that address behavioral factors and social and structural determinants of health. --- MATERIALS AND METHOD --- Theoretical Underpinnings of TIP The intervention incorporates core components of other HIV prevention interventions that have shown efficacy with individuals of trans experience and other marginalized populations such as men who have sex with men and racial or ethnic minorities. Such interventions include Choosing Life: Empowerment, Actions, Results (CLEAR); Comprehensive Risk Counseling Services; routine HIV testing and HIV care; and peer delivery of services. The theoretical bases of TIP are those of CLEAR: cognitive-behavioral therapy (CBT) and social action theory (SAT) (Lightfoot, Rotheram-Borus, & Tevendale, 2007;Rotheram-Borus et al., 2004). CLEAR applies cognitive-behavioral techniques to maintaining health, reducing the risk for HIV and other sexually transmitted infection (STI) transmission or reinfection, and improving the quality of life of youths and adults living with HIV/ AIDS. Strategies in the intervention include roleplaying as a means of learning new skills and improving old ones, building client's belief that he or she can change a behavior (self-efficacy); and instilling the belief that changing behaviors will result in a desired outcome (response efficacy). In the context of this intervention, CBT is intended to help the person reconceptualize his or her understanding of traumatic experiences, including transphobia and gendered violence, as well as his or her understanding of himself or herself and his or her ability to cope. SAT asserts that a person's ability to change behaviors that endanger his or her health is influenced by the individual's cognitive capability (ability to think, reason, imagine, and so on), and environmental factors and social interactions that encourage or discourage the change process. SAT incorporates the principles that are expressed in traditional sociocognitive models of health behavior change. These models include sociocognitive theory, the health belief model, the transtheoretical model (stages of change), and theories related to social context, interpersonal relationships, and environmental influences. The peer-navigator model is another core element of the intervention. Peer-led programs and interventions are led and conducted by people from the communities most affected by the disease or condition, and operate through organizations established and governed by these communities (Brown et al., 2018). TIP's peer-led model includes peer activities ranging from peer service delivery (such as peer-led rapid HIV testing or referral to peer-led needle and syringe program) to peer health promotion (such as peer-developed and peer-implemented stigma reduction campaigns or community development, including the Trans March in Philadelphia) and peer leadership (such as peers taking leadership roles in their community, their sector, or participating in policy and law reform). Building from this peerled model, TIP is fully staffed by men and women of transgender experience. In addition, the staff is diverse in terms of race and ethnicity. Overall, this study is guided by community-based participatory action research approaches. The term "ITE" emerged in our community advisory board meetings (a total of four for this particular project) composed of more than 15 ITE. A thorough reflection and discussions emerged regarding how best to refer to individuals within these communities. Some advisory board members emphasized that individuals who meet the definition of "transgender" do not always like or identify with this term. Moreover, whereas calling a woman who was assigned male at birth a "transgender woman" centers her transness, calling her a "woman of transgender experience" centers and reflects her womanhood. --- TIP's Content TIP is designed around three major priorities: leadership, structural interventions, and HIV risk reduction. Additional details are provided in Appendix 1. The priorities of leadership, structural intervention, and HIV risk reduction are embedded in the intervention sessions. Although every session is delivered to every client, the order is not fixed but rather determined by individuals' needs, goals, and readiness to participate. Overall, TIP focuses on reinforcing short-term and lifelong goals, engagement in problemsolving behaviors, and identifying triggers related to sexual risk and substance use. Since its original development, the intervention has been adapted and modified by community leaders in collaboration with other agency providers. These community members have refined TIP to include six sessions that seek to maximize potential of achieving long-term behavior change. Session details are provided in Appendix 2. --- RESULTS --- Participant and Client Characteristics Since its launch in 2003, TIP has provided services to more than 1,500 transgender individuals. In 2016, TIP provided 230 individual sessions to 56 ITE, including 11 HIV-positive individuals and 45 who faced an elevated risk of infection, representing a 31 percent increase compared with clients served in 2015. During the 2017 fiscal year, TIP provided services to a total of 80 ITE. There were also noteworthy changes in the demographics of the population served. More specifically, there has been an increase in ITE from Puerto Rico who were displaced by the recent hurricane disaster and now reside in the Philadelphia metropolitan area. In 2017, 97 percent of TIP clients self-identified as black or Latino. --- Participant Engagement and Intervention Delivery Participants are recruited from local HIV/AIDS service organizations, social media (for example, Facebook, Twitter), and through established partnerships, including local AIDS service organizations in the northeast corridor and the Philadelphia prison system. TIP peer-navigators approach ITE and speak to them about the TIP intervention, highlighting the comprehensive social and health services available, including employment support, legal aid, and HIV prevention and care services. TIP peer-navigators distribute business cards and encourage potential participants to reach out for more information. For online engagement, TIP peer-navigators use a passive approach by posting health information and community resources, including health fairs and employment opportunities, and reaching out to those who send private messages or post comments on the main page. At enrollment, all clients are assessed by a TIP peer-navigator based on their self-identified and selfprioritized needs for hormone therapy, housing assistance, substance abuse treatment, sexual risk reduction, sterile syringe exchange, mental health services, domestic violence intervention, medical care, partner services, and screening for STIs, including viral hepatitis and HIV. Once initial assessment is completed, TIP clients are scheduled for the intervention sessions. TIP peer-navigators provide the intervention on-site or at other facilities where ITE clients obtain services and where TIP has developed rich professional relationships, such as the Morris Home, the LGBTQ Home for Hope (formerly the Divine Light LGBTQ Wellness Center), the Addiction Recovery Program, and the Philadelphia prison system. HIV counseling and testing are offered to all TIP clients. TIP peer-navigators are certified HIV testers and can administer the test themselves. Clients reported feeling safe and affirmed having the test administered by someone who reflects their own lived experience. All clients who test negative are informed of and referred or linked to preexposure prophylaxis (PrEP). All clients who test positive for HIV are referred to a health facility of their choice. GALAEI has a strong working relationship with Drexel Medicine's Partnership Comprehensive Care Practice, and clients can typically access same-day medical care after being diagnosed. TIP also encompasses an online component. Online content is shared through an active Facebook page. The online component of the intervention intends to provide up-to-date information on clinical components of transgender care, HIV/AIDS information, sexual health, social justice activism and involvement, trans health empowerment content, and community mobilization. See Figures 1234for more information on online intervention content. --- DISCUSSION TIP has an established record of reaching and promoting service utilization beyond HIV prevention and care among ITE. TIP's utilization speaks to community interest and investment in comprehensive care, including services responsive to multiple syndemic factors, among ITE (including individuals who are also members of racial or ethnic minorities). The intervention's approach to preventing HIV among the uninfected and addressing it among the infected in this highly vulnerable population is grounded in the deployment of comprehensive services, including legal aid and housing in addition to medical care and gender-affirming support. All of TIP's clients are ITE from the northeast corridor, including Philadelphia; Camden, New Jersey; and Trenton, New Jersey, who are currently living with HIV or are at risk for HIV infection. Interagency and interdisciplinary collaboration for the Promoting the Intervention and Community Social Justice Initiatives promotion of health and wellness for ITE have been essential and instrumental in the growth of TIP. TIP's success in establishing community collaborations builds from trust and community activism. In particular, our Philadelphia prison system collaboration, building on educational collaborations and mutual interest in serving the needs of ITE, resulted in TIP being delivered in the prison system. To our knowledge this is one of the few interventions in the country being delivered in the prison system. TIP peer-navigators continue to work to prevent HIV transmission among hard-to-reach, at-risk ITE including those who are homeless, engage in survival sex work, use illicit substances, or inject silicone or hormones without medical supervision or reliable access to clean syringes. The program has a strong record of reaching such high-risk populations and linking them to HIV counseling, testing, and referrals; HIV primary care; and HIV prevention services, including access to biomedical prevention tools such as PrEP. The peer-leader approach of the TIP intervention, grounded in leveraging assets of the community and training and retaining well-trained peer leaders from the community itself, has contributed to TIP's impact in reaching high-risk populations. Other aspects of the TIP intervention address social and economic needs, including those that have been linked with physical and mental welfare. The leadership development emphasis of the intervention allows employment-specific skills training such as résumé writing and job interviewing, referrals to housing, and linkage to other communitybased services and supports such as food and utility assistance. The leadership development emphases emerged from a conscious effort to embed the best and most relevant components of HIV prevention strategies within a holistic approach that focuses on affirming, safe, trans-led programming that addresses the numerous structural barriers and challenges faced by transgender individuals every day. TIP peernavigators actively maintain an extensive knowledge of resources throughout the northeast corridor and provide referrals for a wide range of needs on a daily basis for transgender clients. The greatest strengths of TIP, both of which may be adapted and replicated elsewhere, concern its commitments to peer leadership and individualized prevention plans. TIP was designed by and for ITE and is implemented by and for ITE. Peer-navigators work with clients to develop an individualized prevention plan that coincides with identified shortterm and life goals. Goals can include, but are not limited to, increased condom, PrEP, or treatmentas-prevention use; increased behaviors that promote healthier feelings, thoughts, and behaviors; consistent daily routines to stay healthy; consistent emotional regulation; and effective coping with challenges of daily living. The intervention continues to reach considerable numbers of ITE who might otherwise lack access to care and other resources. Glaring gaps exist in the current intervention arsenal available to reduce HIV transmission, improve HIV care continuum outcomes among HIV-positive individuals, and buffer the impact of syndemics among ITE. TIP was directly informed by syndemics literature, and its comprehensive approach reflects an understanding that HIV infection and transmission are affected by numerous factors including discrimination and stigma, access to medical care, housing, employment, and more. The TIP intervention is promising; the communitybased approach for recruitment and engagement building on inclusion and service provision has enabled the increase in the number of ITE being served. Despite TIP's reach and engagement, its efficacy is yet to be tested. A pilot grant was recently secured through an academic partnership to obtain feasibility and acceptability data. If TIP is proven to be efficacious, the intervention can then be replicated in various communities to provide support, increase service utilization, and facilitate behavior change. Limitations to this research must be acknowledged. First, this pilot study lacks a control or comparison group, so observed findings must be interpreted with caution. A randomized trial of TIP's intervention can determine the efficacy of the intervention. Second, findings are restricted to linkage-to-service provision. TIP is housed at a local AIDS service organization staffed by five full-time community health workers and social services providers. The agency lacks the funding and infrastructure to implement pre-and postassessments, rigorous evaluation procedures, and thorough analysis of data. However, TIP's success in reaching, engaging, and linking ITE to comprehensive services to address social determinants of health is well documented here. Third, TIP does not systematically collect information from all individuals who receive services from the intervention, and therefore we lack comprehensive information about the participation rate, measures, and characteristics of participants. Fourth, attrition and follow-up are challenges for intervention delivery. Some of the participants do not have contact information given their lack of housing stability and financial constraints. Larger investments in HIV prevention and treatment for this population are warranted, including investments in building the capacity of AIDS service organizations and communitybased organizations to design, implement, and evaluate theory-based programs that respond to the needs of ITE. CONCLUSION ITE continue to be disproportionately affected by HIV/AIDS. These disparities in HIV/AIDS are attributed to a set of behavioral characteristics and social determinants of health, including discrimination, stigma, incarceration, and lack of access to comprehensive health care. These intertwined syndemics are complex and should be given priority to reduce HIV-related health disparities among ITE. TIP serves as an innovative and culturally appropriate homegrown intervention that effectively delivers care along the HIV continuum for HIV-positive and HIV-negative ITE. TIP has established a track record of reaching and engaging high-risk populations, and then linking them to comprehensive social services and HIV prevention and treatment. HSW current peer-navigators of the Transhealth Information Project for their contribution to service and research and the scientific contribution of their ongoing community-collaborative action board members. They would like to thank Drs. Mallory Johnson, Emily Arnold, and Ethan Levine for their feedback and support. --- Exploring Different Types of Communication • Learn key components of assertive communication and identify differences between aggressive, assertive, and passive communication

Individuals of transgender experience (ITE) in the United States face an elevated risk of HIV infection. Several conditions have been attributed to the high HIV incidence and prevalence within this group, including experiences of discrimination, unemployment, incarceration, stigma, and elevated rates of sexual risk and substance use. In response to these needs, the Gay and Lesbian Latino AIDS Education Initiative and Prevention Point Philadelphia, two local community-based organizations in Philadelphia, developed the Transhealth Information Project (TIP). TIP is a peer-led six-session hybrid individual-and group-based intervention emphasizing leadership, social and structural interventions, and HIV risk reduction that incorporates other evidence-based practices for HIV prevention and care. Since 2003, TIP has served over 1,500 ITE and linked them to HIV prevention and care services. TIP has an established record of reaching ITE and linking them to HIV prevention services and HIV primary care. TIP's utilization speaks to the need for interventions to respond to the complex, interacting syndemic factors that cumulatively determine HIV vulnerability among ITE.

INTRODUCTION Procrastination is a stable harmful tendency within individuals, defined as the voluntary but irrational delay of intended course of actions (Elliot, 2002;Steel, 2007). In addition, the procrastination also refers to an off behavior that keeps unnecessary delay and reap negative consequences caused by this delay per se (van Eerde, 2000Eerde,, 2003)). The absolute number of procrastinators is sizeable (Potts, 1987), with approximately 75% of college students considering themselves procrastinators and nearly half of them procrastinating consistently and problematically (Steel and Ferrari, 2013). Furthermore, procrastination significantly harms people's health, wellbeing, work efficiency and academic performance (Kachgal et al., 2001;Sirois, 2007;Balkis and Erdinç, 2017). An earlier meta-analysis thoroughly and explicitly reviewed the nature of procrastination and revealed how procrastination is associated with many variables, such as personality, mental health and demographic features ( van Eerde, 2003;Eerde, 2004;Steel, 2007). However, the conflicting results for the association between gender and procrastination were observed frequently in existing studies. For instance, Li (2013) and colleagues used largescale sample to report the strong effect for the gender differences in procrastination among Chinese students, with more procrastination in males (Li, 2013). In addition, this conclusion is also supported by the Turkish population (Nilufer, 2017). However, the inconsistent results reported show that there is no gender differences in procrastination (Ajayi, 2020;Wang, 2020). To make matters worse, a portion of studies provided evidence to claim more procrastination in females instead of males (Bian, 2017;Song et al., 2020). In this vein, so far there is no solid evidence to clarify this association. Furthermore, as the close linkage between socioeconomic status (SES) and the self-regulation (Miller et al., 2015), the role of SES on procrastination that caused by the failure of self-regulation also caught our eyes. Nevertheless, results for such relations were found heterogeneous: Yao (2020) reported a significant negative correlation between SES and procrastination (Yao, 2020), whereas provided null findings in other studies (Huang et al., 2017;Xing, 2019). Thus, the needs for a meta-analytic evidence to clarify the association between SES and procrastination emerged. Further, the association between education and procrastination is still sparked much interests for us, but remains inconsistent conclusions. Ferrari et al. (2009a) have demonstrated results for claiming the negative relationship between education and procrastination (Ferrari et al., 2009a). However, graduates students were found procrastinate more than students in high and middle schools (Wen, 2014;Li, 2019). Thus, it leads us to infer the association between the education and procrastination, as well to promote a need to clarify what the direction is for such influence. A considerable body of meta-analyses on procrastination has been conducted. Steel (2007) provided the correlation between procrastination and other psychological features, and revealed that task aversiveness, task delay, self-efficacy, impulsiveness, conscientiousness, self-control, distractibility, organization, and achievement motivation are strong predictors of procrastination. However, demographic features such as age are not significant predictors of procrastination (Steel, 2007). More recent metaanalyses have discussed demographic features of procrastination (Balkis and Erdinç, 2017;Krispenz et al., 2019) and focused on the relation between specific topics such as procrastination and time perspective (Sirois, 2014), academic performance (Kim and Seo, 2015;Saele et al., 2017) and intervention (van Eerde and Klingsieck, 2018). In addition to the general demographic features of procrastination and these specific features, other social factors should be probed, including SES, cultural differences, and educational background. Both 4-dimensional theoretical model and competing theory of SES have suggested that SES is defined as a measure of individuals' combined economic and social status, which may influence many aspects of personal behaviors, such as smoking, addiction and drinking (Pampel and Rogers, 2004;Cutler et al., 2008;Baker, 2014). A previous study suggested that individuals with lower SES procrastinate in treatment or hospitalization and/or do have apply for health insurance (Weissman et al., 1991). In addition, lower SES affects the procrastination tendency of college students, as students with lower SES are more worried about their financial situation, which triggers anxiety and leads to increased academic procrastination (Stöber and Joormann, 2001). Similarly, differences in SES can also explain to certain extent the tendency of individuals to use Facebook to avoid or procrastinate on tasks (Arnett, 2016). Furthermore, a recent study measured childhood SES and investigated its relationship with procrastination found that SES is closely related to procrastination via parenting style and conscientiousness trait (Shimamura et al., 2021). Such associations could be attributed to the personality-trait formation. In detailed, the high SES was found to make children prone to form conscientiousness trait that is significantly factor to persist procrastination (Noftle and Robins, 2007;O'Connor and Paunonen, 2007;Lubbers et al., 2010). Another pathway to explain why SES could influence procrastination is the variability of self-control ability. Existing evidence indicated that the low SES was a risk factor to make children and students posing less self-control and self-regulation ability (Johnson et al., 2011;Ng-Knight and Schoon, 2017). In addition, the bidirectional interplay of SES and procrastination could be ascribed to the mediated role of impulsivity. Gustavson et al. (2014Gustavson et al. (, 2015) ) have demonstrate the robust genetic association between procrastination and impulsivity, with high impulsivity in procrastinators (Gustavson et al., 2014(Gustavson et al.,, 2015)). Further, it is noteworthy that the high impulsivity could be a reliable predictors for low SES (Assari et al., 2018;Walsh et al., 2019). In this vein, it promotes us to infer that the high impulsivity may make participants prone to procrastinate more for causing poor academic or professional achievements, which in turn, bring about low SES. However, other studies suggested that college students ranging from lower to upper SES backgrounds showed no significant difference in their scores in an university program that aimed to measure academic procrastination (Pychyl et al., 2002). Hence, the relationship between SES and procrastination remains ambiguous, and it is worth exploring through meta-analysis methods. It has been proposed that cultural differences in norms and values exist in the perception of time that may affect individuals' evaluation of long-term consequences and risk avoidance, which may thus affect procrastination (Brislin and Kim, 2003). A previous study found that the prevalence of procrastination among British citizens is higher than that among American or Australian citizens, demonstrating that procrastination is more common in Westernized, individualistic, English-speaking countries than in other countries (Ferrari et al., 2005a). In addition, it was found that cultural differences between the United States and Russia are related to procrastination caused by the use of the Internet for social interaction (Doty et al., 2020). Likewise, the same conclusion was found in non-Anglo-Saxon population as well. For example, the problematic procrastination was observed more in Mongols than do of Chinese population (Wang, 2014). In contrast, a group of studies exploring adult procrastination in different cultures found that procrastination is common in every situation, and the procrastination patterns of arousal and avoidance show cross-cultural similarities rather than differences ( van Eerde, 2003;Ferrari et al., 2007;Klassen et al., 2009). This claim-that is-no significant cultural differences in procrastination was supported in non-Anglo-Saxon populations, such as Bourau, India and Tibet (Kuang, 2012;Song, 2014). In addition, a study surveying college students from Ukrainian and Slovak revealed that there were no statistically significant differences in procrastination between Ukrainian and Slovak students (Ko<unk>ková et al., 2020). Overall, it is worthy to explore there were cultural differences for procrastination in the current study. An interesting association for the procrastination is that delaying off courses irrationality would increase likelihood to remain singe in marital relationship, which caused to a reduction in the size of the family (Moore, 2004). Further, the procrastination was found to be a predictor for divorced rates and it would make individuals hard to obtain romantic relationship (Roberts et al., 2007;Steel, 2010b). A straightforward evidence provided by Steel and Ferrari (2013) indicated the negative association between the procrastination and family size. On the other hand, the high fertility rate was found prominently in individuals with less conscientiousness, which is the key predictor for procrastination (Bouchard, 2005). Thus, the current study is also interested in whether the family size could link to procrastination. Moreover, educational background has long been regarded as a potential indicator of procrastination tendencies (Ferrari et al., 1995(Ferrari et al.,, 2009b;;Steel and Ferrari, 2013). Liberal arts students and science students are exposed to different teaching structures and content (Jacobsen, 2006;Van der Wende, 2011) and therefore may differ in their level of procrastination. Conscientiousness is a strong predictor of both academic performance and procrastination (Noftle and Robins, 2007;O'Connor and Paunonen, 2007;Steel, 2007). Later research found a negative relationship between procrastination and academic math performance comparable in strength with that of conscientiousness (Lubbers et al., 2010). Thus, we may deduce that science students procrastinate less than liberal arts students. However, another study investigated whether procrastination is related to student majors and found no significant difference in procrastination tendencies among students of different majors (Zarick and Stonebraker, 2009). Thus, what the association between educational background and procrastination is could be examined by using meta-analysis in the current study. As conflicting results for the association between procrastination and some demographic characteristics were consistently found, the potential factors to impact these results have also sparked much interests in current study (Pychyl et al., 2002;<unk>zer et al., 2009;Steel and Ferrari, 2013). Svartdal and Steel (2017) have pioneered the examination and revision for five mainstreaming scales, and indicated the variability of psychometric quality for different measurements (Svartdal and Steel, 2017). In addition to this field, results based by heterogeneous measuring tools were found in elsewhere (Dawis, 2000;Marsh et al., 2013). In this vein, it lead us to assume that the conflicting results may be moderated by different scales. Further, procrastination type is also noteworthy to be an alternative moderator. On the basis of temporal motivation theory, the main factor to promote one postponing off is the inadequate motivations (Steel, 2007). A robust body of studies provided solid evidence to claim the interaction effect of gender and motivation, which demonstrated that female posed high intrinsic motivation than male in academic activities (e.g., reading, L2 learning, Rusillo and Arias, 2004;Kissau, 2006;Hakan and Münire, 2014). Despite no straightforward evidence, such interaction may bring about the specific sex-differences between academic procrastination and other ones. Lastly, the educational stage stress should be taken into account to explain the heterogeneous results for sex-differences in procrastination. It is well-known to us for the close association between perceived stress and procrastination, with more stresses for stronger procrastination tendency (Stead et al., 2010;Sirois and Tosti, 2012). On the other hand, more stresses were perceived for students in high educational stages than others, and make undergraduates more prone to procrastinate (Pascoe et al., 2020). Given the significant gender-difference (Barnett et al., 1987), we are interested in probing into how the educational stage may influences procrastination and even the interaction of gender and procrastination. In the current study, we aim to integrate the results of previous studies of the relationship between procrastination and gender, SES, cultural differences, and educational background and to identify factors that influence their relationship. As aforementioned, we hypothesize that the males may procrastinate more than females. In addition, participants with low SES, western cultural contexts, or majoring in arts may be incline to procrastinate than others. Further, we assumed that the association between procrastination and these demographic architectures would be moderated by measurements/types of procrastination and educational stages. --- MATERIALS AND METHODS --- Participants and Study Design This study aimed to investigate the sociodemographic characteristics of procrastination by using a meta-analysis that included gender, socioeconomic status, multiculturalism, family size, and educational background. All the types of procrastination would be in the scope of the current study, such as general procrastination, academic procrastination and decision-making procrastination. Taking into account potential moderated variables or hierarchical natures, the sub-metaanalysis and moderation analyses were used to probe whether these effects were moderated by measurements, procrastination types or other factors. Finally, jackknife analysis was used to validate the robustness of the pooled meta-analytic effects. On balance, this study strove to demonstrate the sociodemographic characteristics of procrastination and unveil the factors that may moderate these effects. This study has been approved by Institutional Review Board (IRB) of Faculty of Educational Science, Sichuan Normal University (China). --- Systematic Meta-Analytic Protocol The protocol for performing the sociodemographic meta-analysis in the current study was preregistered in the Open Science Framework (OSF) repository beforehand1. This study fully adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for literature searching and data extraction (see Supplemental Information). --- Data Source and Search Strategy The databases to be searched to acquire meta-analytic data were Web of Science (WoS), Science Direct (Elsevier), ProQuest, and Google Scholar. In addition to these international databases, Chinese academic databases were screened as well, including the China National Knowledge Infrastructure (CNKI)2, Wangfang3, Vip Consult Center4 and China Biology Medicine disk (CBMdisc5 ). Searches in these databases were limited to peer-reviewed empirical research published from Jan 2000 to Jan 2021. Despite lack of peer review, dissertations for the full-time Chinese doctoral and master's degrees were also included in the preliminary data pool once they were determined to be eligible by a modified Newcastle-Ottawa quality assessment (see more details below). Studies published in preprint form were excluded from the data pool for meta-analysis. --- Retrieval Procedure As a strikingly productive tool, Boolean logic expression (BLE) was drawn upon for literature retrieval. To avoid missing the target, no elimination operator was used in the search. The full search expression in searching the international databases was as follows: (procrastination OR procrastinator OR procrastinate) AND (gender OR sex) OR (socioeconomic status) OR (country OR Han) OR (family size OR single child OR double child) OR (education OR educational backgrounds OR STEM). Likewise, such BLE was also adopted to retrieve target studies in the Chinese databases using Mandarin. To cover all the alternatives for, no exclusion criteria were used in the first searching procedure. Furthermore, to strengthen the search accuracy, the elimination operator was used for re-retrieval. Finally, reference tracking to a hub paper that was cited frequently was undertaken to validate the convergence of this retrieval, such as Steel (2007) and Steel and Ferrari (2013). Specifically, the references cited in hub papers would be reviewed manually, one by one, to validate whether there were missing target papers and to scrutinize whether these potential papers were worth checking for eligibility in the current study. --- Inclusion and Exclusion To screen available studies for quantitative meta-analysis, the inclusion and exclusion criteria are given here. First, the aims of the screened study should be in line with those of the current study. That is, papers seeking to build upon the link between procrastination and sociodemographic features could be included. Second, these studies should provide adequate effect sizes for meta-analyses, including t values, sample sizes, and descriptive sample information. Third, the procrastination should be measured by widely-used or board-certified scales (e.g., general procrastination scale, and pure procrastination scale). Thus, these studies measuring procrastination from self-made questionnaire or self-report interview would be excluded. Last, the minimum sample size was limited to 30 in each included study to ensure statistical power. --- Data Extraction, Data Coding and Statistics On the basis of PRISMA protocol, the fundamental information and data were extracted by two independent researchers, including the authors, title, t-statistics, moderators, measurements, and sample size. Further, data coding was performed independently by them. Subsequently, they exchanged data extraction and coding records, and further re-did these procedures to examine for inconsistent results. Finally, once these data were checked for no errors, one researcher inputted them into the CMA software, and another one would doublecheck whether there were typo independently to validate the correctness of pooling effect size. Comprehensive meta-analysis (CMA, V3.1) was drawn for meta-analytic statistics in the current study. First, the raw t-statistic (i.e., t-value and sample size) for each included study was estimated for the weighted r value and corresponding 95% confidence intervals (CI). Furthermore, all the estimated r values were pooled into either a fixed-effect model (FEM) or randomeffect model (REM), which was determined by the heterogeneity of the data pool. In this vein, I 2 and Q 2 tests were used to evaluate the between-study heterogeneity, with <unk> 20% for low variation, 50% for medium variation and 75% for high variation. A FEM should be adopted if low or medium heterogeneity is detected, while an EFM should be more suitable when high heterogeneity is found (Borenstein et al., 2011). Given the adverse effects of publication bias, Egger's test and fail-safe N tests were performed to estimate the effects of potential publication bias, with statistical significance in Egger's test at p <unk> 0.05 and N > 5k + 10 for predominant publication bias (Fragkos et al., 2014). In addition, the jackknife test was adopted to validate the robustness of the meta-analytic results (Radua and Mataix-Cols, 2009;Frodl and Skokauskas, 2012). This process adopted a leave-one-paper-out (LOPO) scheme to iterate the meta-analysis and further examine whether the targeted significant effects could be maintained. --- Jackknife Analysis and Newcastle-Ottawa Quality Assessment The Jackknife analysis is to examine the robustness of metaanalytic results by using iteration procedure. In detailed, the meta-analysis would be redone by removing one study included in the datapool, subsequently. This procedure would be iterated until each one was removed once. If the significance of results for all the iteration kept consistent with this meta-analysis, these results would be considered robust enough. Meanwhile, the Newcastle-Ottawa quality assessment was developed to examine the quality of included studies, including five items: balanced gender ratio, sample size, validity, and quality of scales. More details can be found elsewhere (Wells et al., 2014). --- RESULTS All the coding data and unfolded documents (results) have been uploaded to the OSF repository (see text footnote 1) to promote reproducibility and transparency. --- Included Study and Fundamental Information On the basis of PRISMA protocol, we attempted to search studies on gender-differences in procrastination, and a total of 20,552 studies were retrieved. Afterward, 13,622 studies were retained after duplication checks. Furthermore, 8,755 studies were removed as they fell outside our research aims (e.g., literature review, opinion article) when screening abstracts. Fulltext examination was conducted to determine the eligibility of these 3,324 studies. Last, 193 studies were ruled out for formal quantitative meta-analysis because of either a lack of statistical information or non-standardized measurements. In summary, this final meta-analytic model included 193 papers concerning general and academic procrastination (n = 102,484) (for more details, see Figure 1). Notwithstanding that, it is unexpected that the vast majority of included studies is derived from mainland China (k = 13 and n = 3,146 for other countries, k = 180 and n = 97,604 for Chinese population). Thus, the sample bias should be mentioned for the following analyses. Other metaanalytic models for socioeconomic status, country, family size, and educational background can be found in the SI. --- Synthesized Main Findings --- Males Procrastinate More Than Females A total of 193 papers (k = 193) were included to pool effects for revealing gender differences in both general and academic procrastination, with 102,484 participants [47,901 males (46.73%), 54,583 females (53.27%)]. The findings derived from both Q and I 2 tests showed a high level of between-study heterogeneity [Q (192) = 1,266.78, p <unk> 0.001, I 2 = 84.84] and thus indicated that the REM is more suitable here. The REM results demonstrated that the procrastination tendency was significantly higher in males than in females (r = 0.042, 95% CI: 0.023-0.056, z = 4.785, p <unk> 0.001) (see Supplementary Table 1 and Figure 2). No publication bias was found in this meta-analysis (Egger's test t = 0.38, p = 0.70; Begg test tau = 0.02, p = 0.65; fail-safe N = 7,226) (see the funnel plot in Supplementary Figure 1). The results of the modified Newcastle-Ottawa quality control assessment demonstrated good quality for the included studies and showed high scorer reliability based on the Spearman test (Qs = 4.48, r = 0.99, p <unk> 0.001) (see Supplementary Table 1). --- No Significant Differences in These Procrastination Tendencies Based on Socioeconomic Status Forty papers (k = 40) were pooled into a meta-analytic model to probe whether there were differences in procrastination tendencies based on socioeconomic status [N = 21,478; 9,540 males (44.41%), 11,938 (55.59%)]. As the between-study heterogeneity was quite high, the REM was adopted for this meta-analysis [Q (39) = 106.74, p <unk> 0.001, I 2 = 63.43]. The results revealed no significant effects of socioeconomic status on procrastination tendencies (r = 0.019, 95% CI: -0.004-0.041, z = 1.627, p = 0.104) (see Figure 3 and Supplementary Table 2). No significant publication bias was found in this meta-analytic model according to Egger's test and fail-safe N test (Egger's test t = 0.56, p = 0.57; Begg test tau = 0.005, p = 0.962; fail-safe N = 36) (see the funnel plot in Supplementary Figure 2). The included studies were observed to be of good quality on the basis of a modified Newcastle-Ottawa quality control assessment (Qs = 4.53, r = 1.00, p <unk> 0.001) (see Supplementary Table 3). --- No Significant Multicultural Differences in These Procrastination Tendencies It should be noted that only Chinese populations were included for this meta-analysis aiming at the cultural differences in procrastination. As the majority and minorities of China posed independent cultures, participants in different ethnic group were considered to undergo different cultural contexts in the current study. Thus, these results derived from this analysis should be considered exploratory and primary. By comparing procrastination between majority and minorities, the six papers (k = 6) were included in the meta-analysis to clarify whether procrastination tendencies would vary with respect to multicultural contexts [N = 3,091, 1,047 males (33.87%), 2,044 females (66.12%)]. No significant between-study heterogeneity was found in this meta-analysis [Q (5) = 9.02, p = 0.108, I 2 = 44.57]. Instead of the REM, the FEM was deployed to estimate the total effects, and a null finding was revealed (r = 0.002, 95% CI: -0.048-0.050, z = 0.055, p = 0.956) (see Figure 4 and Supplementary Table 4). No significant publication bias was found in this model, either in Egger's test or in the fail-safe N test (Egger's test t = 0.44, p = 0.68; Begg test tau = -0.33, p = 0.452; fail-safe N = 0) (see the funnel plot in Supplementary Figure 3). Likewise, all the included studies were assessed as eligible (Qs = 3.71, r = 0.80, p <unk> 0.05) (see Supplementary Table 5). --- No Evidence to Support the Impact of Family Size on These Procrastination Tendencies To gain further insights regarding the roles of family size in procrastination tendencies, 61 studies were included in the metaanalytic model. Both Q and I 2 tests were performed to detect potential between-study heterogeneity, and the results indicated that the REM is more suitable [Q (60) = 182.25, p <unk> 0.001, I 2 = 67.08]. Thus, the REM was used but yielded a null finding (k = 61, r = 0.011, 95% CI: -0.008-0.031, z = 1.122, p = 0.262) (see Figure 5 and Supplementary Table 6). Additionally, there was no significant publication bias in this meta-analytic model (Egger's test t = 1.19, p = 0.23; Begg test tau = 0.08, p = 0.347; fail-safe N = 0) (see the funnel plot in Supplementary Figure 4). All the included studies were well validated in terms of quality (Qs = 4.52, r = 1.00, p <unk> 0.001) (see Supplementary Table 7). --- No Significant Differences in These Procrastination Tendencies Based on Educational Background In the current study, educational background was divided into two types: literature/arts and science. To reveal whether this sociodemographic feature could lead to differences in procrastination tendencies, 42 papers (k = 42) were used for a meta-analysis. As between-study heterogeneity was high, the REM was thus used for pooling total effects [Q (41) = 440.05, p <unk> 0.001, I 2 = 90.68]. The results showed a null effect for this association (r = -0.010, 95% CI: -0.055-0.034, z = -0.46, p = 0.643) (see Figure 6 and Supplementary Table 8). As examined by Egger's and fail-safe N tests, no significant publication bias existed in this model (Egger's test t = 0.37, p = 0.70; Begg test tau = -0.03, p = 0.76 fail-safe N = 0) (see the funnel plot in Supplementary Figure 5). The included studies were well validated in terms of quality (Qs = 4.62, r = 0.99, p <unk> 0.001) (see Supplementary Table 9). --- Synthesized Results of Moderated and Sub-Group Analysis --- Gender Difference Was Moderated by Various Measurements High between-study heterogeneity has long been acknowledged to be implicated to potential moderators. In this vein, the current study conducted moderated analysis by using mixed effects analysis. Results indicated that the gender difference of procrastination tendency was moderated by various measurements significantly (point estimate = 0.042, 95% CI: 0.027-0.057, z = 5.474, p <unk> 0.001). Sub-group analysis was conducted to clarify this moderated effects, and showed the high gender differences effects in Aitken Procrastination Inventory (API, point estimate = 0.035, 95% CI: -0.011-0.081), The Procrastination Assessment Scale-Students (PASS, point estimate = 0.056, 95% CI: 0.029-0.083), and the Academic Procrastination Questionnaire-Middle School Student (APQ-MSS, point estimate = 0.072, 95% CI: 0.031-0.112), as well appeared low effects in General Procrastination Scale (GPS, point estimate = 0.018, 95% CI: -0.019-0.056) and Tuckman Procrastination Scale (TPS, point estimate = 0.026, 95% CI: -0.027-0.078). --- Gender Differences Had a Stronger Effect on Academic Procrastination Than on General Procrastination Further moderated analysis revealed the moderating role of procrastination types, including academic procrastination and general procrastination (point estimate = 0.042, 95% CI: 0.026-0.058, z = 5.139, p <unk> 0.001). Subgroup analysis demonstrated prominently larger effects on academic procrastination (point estimate = 0.047, 95% CI: 0.030-0.065) than on general procrastination (point estimate = 0.019, 95% CI: -0.019 to -0.057). --- Graduate Students Showed Stronger Gender Differences Related to These Procrastination Tendencies As large effects on academic procrastination were found, we performed a sub-sub-meta-analysis to probe whether such effects would be moderated by the different stages of education, including primary school, junior school, high school, undergraduate and postgraduate levels. The results illustrated the significant moderating effects of educational stages on gender differences related to procrastination tendencies (point estimate = 0.041, 95% CI: 0.026-0.056, z = 5.300, p <unk> 0.001). Further post hoc analysis demonstrated larger effects for undergraduate students (point estimate = 0.040, 95% CI: 0.018-0.062) and graduate students (point estimate = 0.138, 95% CI: 0.027-0.246) than students of other education levels (see Table 1). FIGURE 3 | Forest plot for the meta-analytic results toward socioeconomic status differences of procrastination. --- Results of Jackknife Examination of Robustness All the statistical results were validated to pass the jackknife test for examining robustness, showing no outliers (see Supplementary Table 10). --- DISCUSSION The current study performed sociodemographic meta-analyses to synthesize the results of previous studies on the relationship between general/academic procrastination and gender, SES, cultural differences, and educational background, and to explore potential factors that affect this relationship. By including 193 quantitative studies with 102,484 participants, the results showed that the males procrastinate more than females with moderate effect size (r = 0.042, 95% CI: 0.023-0.056, z = 4.785, p <unk> 0.001) for general and academic procrastination. In addition, by using a small sample size (n = 21,478, k = 40 for SES; n = 3,091, k = 6 for multi-cultural contexts; n = 32,096, k = 61 for family size; n = 21,767, k = 42 for educational stage), no significant effects were found to support the sociodemographic association between procrastination and SES, multi-cultural contexts, family size, educational stage (r = 0.02, p = 0.10 FIGURE 4 | Forest plot for the meta-analytic results toward multi-cultural differences of procrastination. for SES; r = 0.002, p = 0.97 for multi-cultural contexts; r = 0.01, p = 0.26 for family size; r = -0.01, p = 0.64 for educational stage). Further, we found that various measurements of procrastination, procrastination types, and educational stages significantly moderated this relationship (Q = 0.04, p <unk> 0.001 for different measurements; Q = 0.04, p <unk> 0.001 for procrastination types; Q = 0.04, p <unk> 0.001 for educational stages). Post hoc analysis demonstrated the high effect size for genderdifferences of procrastination in API (Q = 0.035, p <unk> 0.001) and PASS (Q = 0.056, p <unk> 0.001) and revealed the low effect size by using GPS (Q = 0.018, p >

Procrastination describes a ubiquitous scenario in which individuals voluntarily postpone scheduled activities at the expense of adverse consequences. Steel (2007) pioneered a meta-analysis to explicitly reveal the nature of procrastination and sparked intensive research on its demographic characteristics. However, conflicting and heterogeneous findings reported in the existing literature make it difficult to draw reliable conclusions. In addition, there is still room to further investigate on more sociodemographic features that include socioeconomic status, cultural differences and procrastination education. To this end, we performed quantitative sociodemographic meta-analyses (k = 193, total n = 106,764) to fill this gap. It was found that the general tendency and academic procrastination tendency of males were stronger than females (r = 0.04, 95% CI: 0.02-0.05). No significant effects of differences in socioeconomic status (i.e., poor or rich), multiculturalism (i.e., Han nation or minorities), nationality (i.e., China or other countries), family size (i.e., one child or > 1 child), and educational background (i.e., science or arts/literature) were found to affect procrastination tendencies. Furthermore, it was noteworthy that the gender differences in procrastination tendencies were prominently moderated by measurements, which has a greater effect on the Aitken Procrastination Inventory (API) (r = 0.035, 95% CI: -0.01-0.08) than on the General Procrastination Scale (GPS) (r = 0.018, 95% CI: -0.01-0.05). In conclusion, this study provides robust evidence that males tended to procrastinate more than females in general and academic profiles, and further indicates that procrastination tendencies do not vary based on sociodemographic situations, including socioeconomic status, multiculturalism, nationality, family size, and educational background.

, family size, educational stage (r = 0.02, p = 0.10 FIGURE 4 | Forest plot for the meta-analytic results toward multi-cultural differences of procrastination. for SES; r = 0.002, p = 0.97 for multi-cultural contexts; r = 0.01, p = 0.26 for family size; r = -0.01, p = 0.64 for educational stage). Further, we found that various measurements of procrastination, procrastination types, and educational stages significantly moderated this relationship (Q = 0.04, p <unk> 0.001 for different measurements; Q = 0.04, p <unk> 0.001 for procrastination types; Q = 0.04, p <unk> 0.001 for educational stages). Post hoc analysis demonstrated the high effect size for genderdifferences of procrastination in API (Q = 0.035, p <unk> 0.001) and PASS (Q = 0.056, p <unk> 0.001) and revealed the low effect size by using GPS (Q = 0.018, p > 0.05) and TPS (Q = 0.026, p > 0.05). Also, the effect size for gender-differences of procrastination was found in academic procrastination (Q = 0.042, p <unk> 0.001) compared to general procrastination (Q = 0.019, p > 0.05). Lastly, graduate students were found higher effect size than others significantly (Q = 0.138, p <unk> 0.001). Thus, this study could lead us to draw a conclusion that males procrastinate more than females in both general and academic procrastination, especially in Chinese contexts. Further, this relationship may be moderated by the measurements, type of procrastination and academic status in the almost Chinese samples. Further, this relationship may be moderated by the measurements, type of procrastination and academic status. On the other hand, by using small-size samples, there were no enough evidence to claim the sociodemographic association of procrastination for SES, multi-cultural contexts, family size and educational stages. --- Gender Is Significantly Correlated With General and Academic Procrastination In particular, this study provided robust statistical evidence that males procrastinate more than females (n = 102,484, k = 193; r = 0.042, 95%, p <unk> 0.001) in both general and academic procrastination, which was consistent with previous investigations suggesting a relationship between them (Pychyl et al., 2000;Steel, 2007;Gröpel and Steel, 2008). There might be promising evidence suggesting a causal role of demographic features (i.e., gender) in procrastination. Males were found to possess a lower level of self-control, which is a key determinant of procrastination (Tewksbury and Higgins, 2006;Ward et al., 2018). As a result, males may tend to procrastinate more due to a lack of goal-directed processing ability and an inability to suppress tempting stimuli (Pychyl et al., 2000;Ferrari, 2001;Steel, 2007;Steel and Klingsieck, 2016). Similarly, males also have a higher level of impulsivity than do females (Cross et al., 2011). A large number of studies have demonstrated that procrastination is positively associated with impulsivity from the behavior, neural variance and behavioral genetics perspectives (Steel and König, 2006;Gustavson et al., 2014;Liu and Feng, 2017), suggesting that males may procrastinate more than females as a result of intrinsic neurobiological factors. Furthermore, a previous meta-analysis found that females score higher on effortful control than males and that effortful control is closely related to procrastination as well (Else-Quest et al., 2006;Lian et al., 2018). This may explain why females procrastinate less than males. On the other hand, existing studies have provided insights into the evolutionary origins of procrastination, suggesting that procrastination can be considered a strategy frequently used throughout life to deal with unpredictable circumstances (Steel, 2007;Chen and Chang, 2016;Chen and Qu, 2017). Compared to females, males were identified to be more sensitive to unpredictable environments, in which they are able to adapt and quickly develop a strategy to succeed using their past experience (Del Giudice, 2009;Jonason et al., 2017). Thus, males' procrastination tendencies may be stronger than those of females as an evolutionary consequence. Notwithstanding that, the external ecological validity should be mentioned to discuss above results. Notably, there were no any statistical considerations aiming at the ecological validity of this meta-analysis though the substantial heterogeneity for these included studies was found (Andrade, 2018). Despite statistical significance, above explanations to support our findings that the males procrastinate more than females were largely grounded post hoc evidence. There were evidence not enough to validate whether this conclusion could be generalizable elsewhere. Given that, extending this conclusion should be quite careful. On balance, the results demonstrated that males procrastinate more than females, not only because of their low levels of self-control and effortful control and high levels of impulsivity but also because of human evolutionary influences. Null Findings Were Observed for the Association Between Procrastination and SES, Multicultural Differences, Family Size and Educational Background However, the hypotheses in present study has not been fully confirmed. For instance, no significant correlation was found between SES and procrastination in the limited sample size (n = 21,478, k = 40, r = 0.02, p = 0.10). One possible explanation is that SES mediates procrastination through other factors, such as parenting style and self-efficacy (Wäschle et al., 2014). Another explanation may be that the high heterogeneity confounded the meta-analytic effects as unaccountable random factors. In addition, the results showed no significant correlation between multicultural differences and procrastination. As reviewed by Steel (2007), procrastination is a personality-like trait and is relatively stable in cross-cultural contexts. Furthermore, there was no significant association between family size and procrastination. On the one hand, procrastination may manifest in postponing childbirth (Steel, 2010b;Steel and Ferrari, 2013;Schippers et al., 2015), which may complicate the association between procrastination and the number of children. On the other hand, unplanned pregnancies are associated with impulsiveness, which is a strong predictor of procrastination (Kahn et al., 2002). Thus, we posited that the association between family size and procrastination may develop in a more complicated, non-linear manner. In addition, no correlation was found between educational background and procrastination, which was consistent with previous research suggesting that majoring in liberal arts or science is not a determinant of procrastination (Seker, 2015). Collectively, despite the failure to validate all the hypotheses, the current study provides robust meta-analytic evidence to substantiate the association between gender and procrastination. --- Heterogeneous Psychometric Tools, Types and Educational Stages Biased the Conclusion Regarding the Association Between Gender and Procrastination Existing studies have not well established the connection between gender and procrastination. Some studies demonstrated that gender was significantly correlated with procrastination (Gröpel and Steel, 2008), while others revealed a null correlation between them (Ferrari, 1989;Ferrari et al., 2005b). Thus, it is valuable to explain why there were different results for the association between procrastination and gender. To this end, the current study performed moderation analysis to clarify the potential factors resulting in inconsistent results. The results of the moderation analysis showed that the heterogeneity in the measurement tools, procrastination types, and educational stages could moderate the association between gender and procrastination. First, the current study noted the different sensitivity levels of tools for measuring procrastination, with strong effects for the association between gender and procrastination in terms of API, PASS and APQ scores and weaker effects on GPS and TPS scores. Thus, to obtain reliable results, widely used, revised and well-established scales or questionnaires, such as the GPS and PPS (i.e., pure procrastination scale), should be adopted. In addition, the psychometric properties of different scales should be validated robustly before attempting to measure procrastination. Despite the solid theoretical basis, there was weak evidence to support the psychometric robustness of some of the scales. Another important finding derived from this submeta-analysis was that the observed association between procrastination and gender was influenced by procrastination type, with the largest effects on academic procrastination. It has been suggested that females show a fear of strangers and unfamiliar events at an earlier age than males (Archer, 1991). Additionally, female students tend not to procrastinate in their academic tasks because of a fear of achieving low course grades (<unk>zer et al., 2009). On the other hand, male students more frequently reported that they procrastinated in their studies due to risk taking and resisting control (Lippa, 2002). In addition, male students are more impulsive than adults, suggesting that they may be more inclined to delay academic tasks (Steinberg et al., 2008;Duckworth et al., 2013). One more alternative explanation worthy to note that the procrastination type moderated the gender-difference of procrastination by age. Beutel et al. (2016) provided solid evidence to demonstrate that the gender-differences of procrastination were found only in the young population (ages from 14 to 29) instead of the overall large-scale sample (Beutel et al., 2016). It may indicate that the academic procrastination was frequently found in young students and it thus let the gender-differences of procrastination more obvious than others. Overall, these findings may indirectly indicate that gender differences more strongly affect general procrastination than academic procrastination. Last, the sub-meta-analysis further revealed that the observed relationship between procrastination and gender is influenced by educational stage, and the influence of undergraduate students is stronger than that of students at other levels. Previous studies have suggested that undergraduate/graduate students procrastinate more than high school or primary school students, because undergraduate/graduate students have more freedom in terms of time and content to complete academic tasks than high school students do (Milgram et al., 1993;<unk>zer and Saçkes, 2011). Thus, compared to students at other educational stages, undergraduate students were likely to be influenced by more distractors and devalued the utility of rewards for completing academic tasks, which made those students more inclined to procrastinate. Additionally, taking into account neuroendocrine factors, females are more fearful and avoidant than males as part of hormonal factors in the late teens (Coates and Wolfe, 1995). In this vein, as procrastination has been proven to be correlated with avoidant motivation (e.g., Steel, 2007Steel,, 2010a)), the moderating role of educational stage on the association between gender and procrastination could be explained by the neuroendocrine variation between males and females. --- Limitation and Future Directions Although this study here revealed demographic characteristics of procrastination, several limitations should warrant cautions. It is worthy to note that a portion of meta-analyses just included the limited number of existing studies, such as the cross-cultural differences of procrastination. Thus, given the marginal sample size, these conclusions should be preliminary and exploratory, as well should be interpreted more cautiously. Another one should bear in mind that the included studies mainly focus on Chinese population (93%). Given the significant sample biases, it should warrant some cautions when these conclusions would be extended elsewhere. Also, it is merit to test the generalizability of the current study in more diverse samples. It is worthy to note that a portion of meta-analyses just included the limited number of existing studies, such as the cross-cultural differences of procrastination. Further, the robustness of results that derived from cross-cultural differences was challenged by sample representation as well. Taken both reasons together, the conclusions for this analysis should be preliminary and exploratory, as well should be interpreted more cautiously. In addition, we found considerable heterogeneity in these meta-analyses. Although we discovered some moderating factors, more factors are worth studying to account for this heterogeneity. Furthermore, the sample representation was somewhat inadequate (e.g., much more Chinese participants than others), which suggests the need for future research using a representative sample to achieve unbiased results. Lastly, as limited by existing literature, it lacks a portion of types of procrastination in meta-analytic model, such as bedtime procrastination and health procrastination. Thus, it is merit to include multiple procrastination types for providing more robust evidence in future study. --- CONCLUSION In summary, this meta-analysis suggested that males procrastinate more than females. In addition, the current study revealed no significant association between procrastination and other sociodemographic features, including socioeconomic status, cultural differences, family size and educational background. Overall, the main strength of our meta-analysis study yields insights into the sociodemographic characteristics of procrastination and the factors that may moderate these effects. Given the association between procrastination and societal ailments (e.g., delayed medical treatment), identifying sociodemographic characteristics associated with procrastination would be valuable for directing public policies aimed at prevention. --- DATA AVAILABILITY STATEMENT All the raw data, scripts, figures and materials can be accessible at OSF repository (https://osf.io/c928r/). --- ETHICS STATEMENT The studies involving human participants were reviewed and approved by IRB of Faculty of Educational Science, Sichuan Normal University (China). The patients/participants provided their written informed consent to participate in this study. --- AUTHOR CONTRIBUTIONS DL: conceptualization, methodology, software, writing-original draft and visualization, supervision, project administration, and funding acquisition. YH and YT: writing-review and editing, methodology, and validation. All authors contributed to the article and approved the submitted version. --- SUPPLEMENTARY MATERIAL The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpsyg. 2021.719425/full#supplementary-material --- Conflict of Interest: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Publisher's Note: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Procrastination describes a ubiquitous scenario in which individuals voluntarily postpone scheduled activities at the expense of adverse consequences. Steel (2007) pioneered a meta-analysis to explicitly reveal the nature of procrastination and sparked intensive research on its demographic characteristics. However, conflicting and heterogeneous findings reported in the existing literature make it difficult to draw reliable conclusions. In addition, there is still room to further investigate on more sociodemographic features that include socioeconomic status, cultural differences and procrastination education. To this end, we performed quantitative sociodemographic meta-analyses (k = 193, total n = 106,764) to fill this gap. It was found that the general tendency and academic procrastination tendency of males were stronger than females (r = 0.04, 95% CI: 0.02-0.05). No significant effects of differences in socioeconomic status (i.e., poor or rich), multiculturalism (i.e., Han nation or minorities), nationality (i.e., China or other countries), family size (i.e., one child or > 1 child), and educational background (i.e., science or arts/literature) were found to affect procrastination tendencies. Furthermore, it was noteworthy that the gender differences in procrastination tendencies were prominently moderated by measurements, which has a greater effect on the Aitken Procrastination Inventory (API) (r = 0.035, 95% CI: -0.01-0.08) than on the General Procrastination Scale (GPS) (r = 0.018, 95% CI: -0.01-0.05). In conclusion, this study provides robust evidence that males tended to procrastinate more than females in general and academic profiles, and further indicates that procrastination tendencies do not vary based on sociodemographic situations, including socioeconomic status, multiculturalism, nationality, family size, and educational background.

Introduction Political instability and extremist insurgencies present significant challenges in collecting vital statistics and conducting regular public health surveillance. For example, the 2018-2019 Kivu Ebola epidemic that is currently occurring in the Democratic Republic of Congo has been difficult to control, in large part due to political violence. Insurgent groups have attacked response personnel, health care providers, and clinics, which interrupted core response activities like identification of Ebola victims, tracing of suspected contacts, and vaccination of at-risk persons. Since the outbreak was declared on 1 August 2018, more than 100 violent attacks have been recorded in the region, including 42 separate attacks on health care facilities and the murder of a World Health Organization (WHO) epidemiologist [1,2]. In Nigeria, Boko Haram has also severely disrupted health services especially in the Northeastern region. Borno-one of the states severely impacted by the insurgency, has lost more than 40% of its healthcare facilities to the insurgency leaving only a third of the remaining facilities fully functional [3]. An estimated 48 health care workers were killed in the state and more than 250 injured. Between 2015 and 2017, 35% of the physician workforce had been lost to other states [3]. Targeting of health facilities has been employed as a tactic of war in Burundi, which severely constrained the supply of health 2 of 11 services [4,5] and in Nicaragua between 1982 and 1987, where more than 20% of health units were destroyed [6]. Several studies have addressed the potential for excess morbidity and mortality from diseases during complex emergencies, which frequently include population movement and resettlement, environmental degradation, food shortages, and limited access to healthcare [7][8][9][10][11][12][13]. Insurgencies and other conflicts can intensify these issues by preventing the timely collection of surveillance data and the implementation of control measures [4]. Longer-term political and military conflicts can also be associated with intimate partner violence and limit progress towards the eradication of diseases or sustainable development goals [14][15][16]. Boko Haram, an Islamic sect that seeks to create a "pure" Islamic state (envisaged to be more fair and transparent) in Nigeria, was founded in 2002. [17]. In 2010, the scope of attacks in the north-east of Nigeria increased, and included bombings, mass shootings, and executions [17], culminating in the abduction of 276 school girls in 2014 [18]. On 20 January 2012, more than 250 people were killed in bomb attacks in Kano State [19]. According to Human Rights Watch [20], more than 935 people died from terrorism attributed to Boko Haram between 2009 and 2012. Insurgencies like Boko Haram affect the health status of residents, particularly children, living in impacted regions in both direct and indirect ways [21,22]. Documented direct effects of the insurgency have included physical [23] and mental health impacts such as Posttraumatic Stress Disorder [24]. Indirect effects include lack of access to both community and community health resources (e.g., adequate shelter, pipe-borne water, food, basic sanitation, and social services); damage to infrastructures like health facilities, roads, markets, water supply, and sewage systems; and disruption of supply chains for food, vaccines, and medication [23,24]. As a result, children exposed to conflicts like this often suffer from higher rates of vaccine-preventable diseases (e.g., measles, polio, hepatitis, and tuberculosis), severe malnutrition, diarrheal disease, and acute respiratory infections. According to the 2013 DHS, the northeastern region of Nigeria has the worst maternal and child health indices in the country [25]. In 2013, more than 50% of the country's 53 polio cases were from Boko Haram-affected states [26]. In 2014, only a single case of polio was identified in Nigeria, and that was from Borno State, located in the region impacted by Boko Haram [26]. While the acute impacts of Boko Haram attacks on morbidity and mortality can be documented [27], quantifying the longer-term health impacts of this type of political insurgency is difficult. Rapid needs assessments, such as those that can be carried out after a natural disaster or in refugee camps [28] are not possible due to the inability to send teams to the region because of instability on the ground in the regions that are most impacted by Boko Haram activities. Collecting primary data from households to estimate health impacts, as was done after military conflicts in Iraq and Afghanistan [29] is difficult in the region because of sporadic and unpredictable violent attacks (e.g., suicide bombings). Likewise, using data obtained from press accounts may result in inaccurate estimates-first because these data are often incomplete, and second, because these data may not record deaths that are the indirect result of the insurgency. Therefore, in an attempt to address the challenges inherent in collecting data on the health impacts of Boko Haram activities, we used Demographic and Health Surveys (DHS) data to investigate the determinants of under-five mortality in an armed conflict setting in Northern Nigeria. --- Materials and Methods --- Data Sources Publicly available data were obtained from the 2008 and 2018 Nigeria DHS. Sponsored by the U.S. Agency for International Development (USAID), DHS is a nationally representative household survey that provides data in the areas of population, health, and nutrition [25,30,31]. The survey is conducted every 5 years in more than 90 countries to allow for comparisons over time and across countries. Additional information on DHS is available elsewhere [30,31]. --- Survey Methodology A two-stage cluster sampling method was employed by DHS in Nigeria. The country was stratified into six geopolitical zones with urban and rural designations in each zone: North Central, North East, North West, South East, South West, and South South. In each stratum, enumeration areas were determined using data from the most recent national census. In the first stage, primary sampling units were selected from the enumeration areas of each stratum. In the second stage, households within each primary sampling unit were selected. In each selected household, women with children at five or less years old were asked about the health of their children. Following data collection, sampling weights were generated for each primary sampling unit and used in data analysis. A detailed explanation of the survey methodology can be found elsewhere [30,31]. --- Study (Boko Haram-Exposed) and Control (Boko Haram-Non Exposed) States For this study, Boko Haram states were defined as those with high Boko Haram activity based on media reports, and included Borno, Yobe, Adamawa, Gombe, and Bauchi (Figure 1). Control states shared similar socio-demographic characteristics, including religion and culture, with the study states except for exposure to the insurgency, and included Sokoto, Jigawa, Niger, Nassarawa, Taraba, Kwara, Benue, Kogi, and Kebbi (Figure 2). Four Northern Nigerian states (Kano, Kaduna, Plateau, and Zamfara) were excluded from the study as they were frequently affected by non-Boko Haram conflicts, like tribal and religious clashes. sored by the U.S. Agency for International Development (USAID), DHS is a na representative household survey that provides data in the areas of population, hea nutrition [25,30,31]. The survey is conducted every 5 years in more than 90 coun allow for comparisons over time and across countries. Additional information on available elsewhere [30,31]. --- Survey Methodology A two-stage cluster sampling method was employed by DHS in Nigeria. The was stratified into six geopolitical zones with urban and rural designations in eac North Central, North East, North West, South East, South West, and South South. stratum, enumeration areas were determined using data from the most recent n census. In the first stage, primary sampling units were selected from the enumera eas of each stratum. In the second stage, households within each primary sampl were selected. In each selected household, women with children at five or less y were asked about the health of their children. Following data collection, sampling were generated for each primary sampling unit and used in data analysis. A deta planation of the survey methodology can be found elsewhere [30,31]. --- Study (Boko Haram-Exposed) and Control (Boko Haram -Non Exposed) States For this study, Boko Haram states were defined as those with high Boko Ha tivity based on media reports, and included Borno, Yobe, Adamawa, Gombe, and (Figure 1). Control states shared similar socio-demographic characteristics, inclu ligion and culture, with the study states except for exposure to the insurgency, cluded Sokoto, Jigawa, Niger, Nassarawa, Taraba, Kwara, Benue, Kogi, and Kebbi 2). Four Northern Nigerian states (Kano, Kaduna, Plateau, and Zamfara) were ex from the study as they were frequently affected by non-Boko Haram conflicts, lik and religious clashes. --- Variables The outcome variable of interest in this study was 'child death,' which was binary (1 =yes, 0 = no). The predictor variables/covariates were: residence in Boko Haram state (1 = yes; 2 = no), sex (1 = male; 2 = female), twin (1= child is a twin; 2= child is not a twin), wealth index of family (1 = poorest; 2 = poorer; 3 = middle; 4 = richer; 5 = richest), place of residence (1 = urban; 2 = rural), parity of mother (1 = one childbirth; 2 = two childbirths; 3 = three childbirths; 4 = four childbirths; 5 = five childbirths; six = at least six childbirths), highest educational level of mother (1 = none; 2 = primary; 3 = secondary; 4 = higher), and age of mother (1 = less than 35 years; 2 = at least 35 years). The variable, wealth index, was derived from data on household ownership of assets, types of water access and sanitation facilities; and materials used for housing construction. Principal component analysis was then used to measure all the households on a continuous scale of wealth, placing them into five quintiles (1 = poorest; 2 = poorer; 3 = middle; 4 = richer; 5 = richest). --- Statistical Analysis DHS data are clustered by design, with individual-level variables (for mother and child) nested within household-level variables. This violates the independence assumption required of an ordinary least squares regression model, as the odds that child death equals one, rather than zero, may vary from one cluster to another, resulting in a Type I error. We fitted two-level mixed effects logistic regression model to account for this potential limitation. A mixed effects logistic regression is suitable for modeling a binary outcome variable in clustered or longitudinal data [32]. The model is parametrized to permit flexibility in the violation of dependence within and between clusters [32]. We used the xtmelogit command to model the log odds of child death-as a linear combination of the predictor variables. Sex, twin, age of mother, parity of mother, and highest educational --- Variables The outcome variable of interest in this study was 'child death,' which was binary (1 = yes, 0 = no). The predictor variables/covariates were: residence in Boko Haram state (1 = yes; 2 = no), sex (1 = male; 2 = female), twin (1 = child is a twin; 2 = child is not a twin), wealth index of family (1 = poorest; 2 = poorer; 3 = middle; 4 = richer; 5 = richest), place of residence (1 = urban; 2 = rural), parity of mother (1 = one childbirth; 2 = two childbirths; 3 = three childbirths; 4 = four childbirths; 5 = five childbirths; six = at least six childbirths), highest educational level of mother (1 = none; 2 = primary; 3 = secondary; 4 = higher), and age of mother (1 = less than 35 years; 2 = at least 35 years). The variable, wealth index, was derived from data on household ownership of assets, types of water access and sanitation facilities; and materials used for housing construction. Principal component analysis was then used to measure all the households on a continuous scale of wealth, placing them into five quintiles (1 = poorest; 2 = poorer; 3 = middle; 4 = richer; 5 = richest). --- Statistical Analysis DHS data are clustered by design, with individual-level variables (for mother and child) nested within household-level variables. This violates the independence assumption required of an ordinary least squares regression model, as the odds that child death equals one, rather than zero, may vary from one cluster to another, resulting in a Type I error. We fitted two-level mixed effects logistic regression model to account for this potential limitation. A mixed effects logistic regression is suitable for modeling a binary outcome variable in clustered or longitudinal data [32]. The model is parametrized to permit flexibility in the violation of dependence within and between clusters [32]. We used the xtmelogit command to model the log odds of child death-as a linear combination of the predictor variables. Sex, twin, age of mother, parity of mother, and highest educational level of mother were included as individual-level categorical predictor variables. Household-level variables comprised the wealth index of family and place of residence. Results were reported as adjusted odds ratios (aOR) and 95% confidence intervals (CI). All statistical analyses were conducted using Stata 17.0 (College Station, TX, USA). --- Results A total of 107,976 births were included in the study, with 49,627 occurring in the five years preceding 2008, before Boko Haram, and 58,349 between 2013 and 2018, after Boko Haram. A larger proportion of respondents in Boko Haram states were educated and wealthy in 2018 as compared to 2008 (higher education: 2.6% versus 1.1%; Richest quintile: 3.8% versus 2.2%) (Table 1). Compared to control states, mothers in Boko Haram-impacted states were younger, poorer, less educated, and lived in a rural location. Across both 2008 and 2018 time periods, the proportion of mothers with parity of at least six was higher in Boko Haram states compared to controls, e.g., in 2008, 64.4% of mothers in Boko Haram impacted states had experienced at least 6 childbirths versus 57.2 % of mothers in control states. States exposed to Boko Haram had a higher proportion of child deaths in the period preceding the insurgency than after (24.0% versus 15.8% respectively) (Table 1). --- Predictors of Under-Five Mortality Residence in a Boko Haram state was not significantly associated with under-five mortality (OR = 1.06; 95% CI: 1.00, 1.10). Twins were more than three times likely to die than singletons (OR = 3.18; 95% CI: 2.96, 3.42), and the odds of child death increased with the number of children ever born to a mother. For example, a mother that experienced two childbirths was 1.4 times as likely to record under-five child mortality than a mother who had given birth once (OR = 1.44; 95% CI: 1.24, 1.67). Likewise, a mother that has given birth at least six times was nearly three times as likely to record under-five mortality than one who had only conceived once (OR = 2.94; 95% CI: 2.57, 3.36). Wealth was found to be protective, as children raised in a wealthier family were less likely to die before their fifth birthday than the socioeconomically deprived (Richest versus Poorest: OR = 0.42; 95% CI: 0.37, 0.47) (Table 2). --- Discussion This study used DHS data to examine the determinants of under-five mortality in an armed conflict setting in Northern Nigeria. The study used available secondary data because reliable primary data collection is difficult in conflict-affected areas, where health data are badly needed due to the potential for conflict to have both direct and indirect impacts on physical and mental health of residents. Using the available data, residence in an armed conflict setting was not significantly associated with under-five mortality. Twin gestation, wealth index of family, study period, age of mother, highest educational status of the mother, sex of the child, mother's religion, and parity of mother predicted deaths before the fifth birthday. This study did not find a significant association between residence in an armed conflict setting and under-five mortality. There are potential explanations for this observation. During the sampling phase of the DHS survey, some clusters considered high risk of insurgent activity were excluded due to safety concerns [25,33]. This exclusion might have selectively biased the results seen in this study. For example, if respondents more likely to need access to health care impacted by Boko Haram activities were less likely to be included, this could bias estimates towards the null. For example, the share of the 'richer' quintile in the Boko Haram region was 10.2% in 2018 as opposed to 8.2% in 2008. In 2018, the 'richest' quintile in the Boko Haram region was 3.8% as opposed to 2.8% in 2008, which might indicate selection bias (Table 1). Second, DHS only interviewed alive mothers at the time of the survey [31]. Thus, deceased mothers of deceased children were excluded from the study, which might also bias estimates towards the null. Third, Boko Haram predominantly targeted the state, its institutions (e.g., police stations, army barracks), and public places (e.g., markets and places of worship)-locations with higher adult populations than children under five years of age [33]. Consequently, deaths resulting from direct exposure to the insurgent attacks would be expected to be lower in underfives than in older age groups. Fourth, given that children under-five have better innate immunity and higher immunologic memory to survive recurrent infections compared to younger children (like infants and neonates) [34], they would be expected to better tolerate the indirect health impact of the insurgency (like vaccine preventable diseases, diarrheal disease, acute respiratory infections, and severe malnutrition), which might otherwise increase their likelihood of mortality. Fifth, the rising positive social change occurring in Northern Nigeria, such as increasing age at marriage and higher educational attainment, could help reduce under-five mortality even in an insurgency era [35,36]. The positive association between twin births and under-five mortality may be explained by the higher incidence of pre-term birth, intrauterine growth restriction, twin-twin transfusion syndrome, and congenital abnormalities seen in twin gestation and therefore a higher rate of morbidity and mortality [37,38]. As much as a 3-fold increase in the perinatal risk of death has been estimated to occur in a twin, compared to a singleton pregnancy [39]. The finding that wealth index was negatively associated with under-five mortality is consistent with a previous study, which found that 16% of children in poor households die before their fifth birthday, compared to 8% of children born in wealthy households [40]. The influence of wealth on child morbidity and mortality may be explained by the effect of income on proximate determinants of health-food, water, clothing/bedding, housing, fuel/energy, transportation, hygiene/preventive care, sickness care, and information [41]. The positive association between parity and child mortality is well documented in the literature [42,43]. Various theories have been proposed to explain this observation, ranging from biologic factors affecting the mother to sociologic factors that affect access to care and utilization of health care services [44]. The biologic theory posits that increased physical and caloric stress associated with high parity results in a'maternal depletion syndrome', which subsequently affects the health and wellbeing of the child [44]. The sociologic theory postulates that children of high parity mothers suffer from poor nutritional status, and subsequently higher mortality, due to reduced parental investment and competition between siblings for finite resources [45]. This study has several limitations. As previously discussed, eight clusters in areas designated as exposed to Boko Haram were deliberately omitted from the DHS survey in 2013 due to security concerns [25], which might have selectively biased the study results. Future studies should be conducted using primary data, where it can be safely collected to assess this research question. Other approaches to collecting valid public health data in areas impacted by violence should also be considered. Second, the migration history of families was not captured in the data. It is possible that families with higher child mortality might have migrated to other neighboring countries (e.g., Chad) or the Nigerian States that were perceived to be relatively safer and were therefore not included in this study. Third, historical data on childbirth and death were obtained by DHS using self-reported questionnaires, which might introduce recall bias, although the death of a child before age 5 is unlikely to be misremembered by a survey respondent. However, misreporting of child death is still possible-mothers may not wish to discuss the death of a child and thus decide not to mention it during a survey. Notwithstanding these limitations, our study has considerable strengths. Our findings are an important contribution to the disaster literature, which is grossly deficient in the determinants of under-five mortality in armed conflict settings. The predictors identified by our study could be used to direct evidence-based intervention efforts for improving child survival in disaster settings. Second, we employed a multilevel mixed effects logistic regression that is suitable for clustered DHS data. The use of ordinary least squares regression would have disregarded the independence assumption, resulting in a Type 1 error, and limiting the validity of our results. Finally, the data for this study were derived from DHS, which has been proven to be an accurate and reliable source of nationally representative data for developing countries [46]. --- Conclusions Residents of armed conflict settings are not uniformly impacted by under-five mortality. In the face of limited public health resources, culturally attentive interventions should be directed at twin children, poor households, multiparous women, and mothers aged over 35 years, with little or no western education. In addition, further studies should consider collecting primary or rapid assessment data from Boko Haram-afflicted areas to re-examine the impact of the insurgency on physical health outcomes. --- Data Availability Statement: Data for this study can be accessed from https://dhsprogram.com/ data/ (accessed on 26 October 2022). --- Author Contributions: Conceptualization, I.M.K., K.W.S. and J.A.H.; methodology, I.M.K., K.W.S. and J.A.H.; software, I.M.K. and J.A.H.; validation, J.A.H. and K.W.S.; formal analysis, I.M.K. and K.W.S.; investigation, I.M.K. and K.W.S.; resources, I.M.K., K.W.S. and J.A.H.; data curation, I.M.K.; writing-original draft preparation, I.M.K.; writing-review and editing, I.M.K., J.A.H. and K.W.S.; visualization, I.M.K.; supervision, J.A.H. All authors have read and agreed to the published version of the manuscript. --- Conflicts of Interest: The authors declare no conflict of interest.

Insurgencies like Boko Haram may impact the physical health and well-being of adults and children living in geographic areas under their political control. However, it is difficult to obtain reliable health data in conflict-affected areas. This study explored the potential to use data from the Demographic and Health Surveys (DHS) to examine the determinants of under-five mortality in Northern Nigeria. Data were derived from DHS conducted before and after the start of the Boko Haram insurgency in 2009. A multi-level mixed effects logistic regression model was used to identify predictors of under-five mortality in an armed conflict setting. Results were reported as adjusted odds ratios (aOR) and 95% confidence intervals (CI). Residence in an armed conflict setting was not significantly associated with under-five mortality (aOR = 1.06; 95% CI: 1.00, 1.10). However, twin gestation (aOR = 3.18; 95% CI:2.96, 3.42), wealth index of family (richest versus poorest: aOR = 0.42; 95% CI: 0.37, 0.47), religion of mother (Islam versus Christianity: aOR = 1.50; 95% CI: 1.43, 1.57); highest educational level of mother (higher versus none: aOR = 0.33; 95% CI: 0.29, 0.37), and parity of mother, significantly predicted death before the fifth birthday. Repeated studies are needed to assess the impact of Boko Haram insurgency on physical health outcomes, particularly in areas where primary data collection is difficult or impossible.

Introduction Digitization is the process of converting analogue information into digital formats and incorporating digital technology into other fields. With the advent of digitalisation, there have been changes in how individuals work, interact, and get information [1][2][3][4][5][6][7][8]. Digitalisation could be a powerful tool to advance gender equality and sustainable development [9][10][11][12][13] by increasing access to education [14][15][16][17][18][19][20], healthcare [21][22], and economic opportunities, especially for women and marginalized groups. Likewise, gender equality is essential for harnessing the full potential of digitalisation and achieving green [23][24][25][26][27][28][29][30][31] and inclusive [32][33][34][35] economic development. As digitalisation continues to transform the workplace, it is crucial to understand the implications for gender equality in career aspirations. As a result of digitalisation, there is a greater need for digital skills across all industries. According to research, there are gender differences in the acquisition of digital skills because women commonly encounter barriers, such as biases and preconceptions that prohibit them from adopting technology [36]. Gender inequality would persist if women were less likely to seek employment in fields where a strong demand for digital skills exists. The way that people pick their jobs is significantly influenced by gender stereotypes. According to established gender stereotypes and biases, males are often linked with technology-related businesses, whilst women are frequently urged to pursue caring or humanities-related occupations. According to [37], these stereotypes may influence people's career aspirations and contribute to gender inequality in the digital sector. Gender stereotypes need to be destroyed to motivate more women to pursue professions in technology and related fields. The "digital gender divide" refers to the disparity between men and women in terms of access, use, and ownership of digital technology. For women, particularly in developing countries, accessing and using digital tools and resources might be more challenging [38]. Women may not be able to take advantage of digital possibilities or pursue jobs in industries that significantly rely on technology because of the gender gap in the digital world. Digitalisation affects gender and career intention in both positive and negative ways [39][40][41]. While the digital era presents new opportunities for job advancement, it also widens the gender divide that already exists. Closing the digital gender gap, helping women develop their digital skills, and dispelling gender stereotypes are all necessary for promoting gender equality in career aspirations. To address these issues and build an inclusive digital workforce, further research and legislative reforms are required to give more in-depth insights into the factors that contribute to gender disparities in career intention. With the help of network analysis, this paper intends to investigate gender, career intention, and digitization as they relate to the intellectual relationships and structural networks among research elements. Thus, demonstrating the intellectual and bibliometric architecture of the research field. To achieve this goal, the remaining paper structure is as follows: The literature review is presented in Section 2, the study's data and methodology are described in Section 3, the empirical findings are discussed in Section 4, and the research's conclusions are presented in Section 5. --- Literature Review Because of the rapid advancement of digital technology, the workplace has experienced considerable changes [40,41]. As digitalisation permeates society, it is crucial to examine how these developments affect people's intentions for their careers and if gender plays a role in these Mavis Mensah Senyah Virtual Economics, Vol. 6, No. 3,2023 choices. The link between gender, career goal, and digitalisation is examined in this review of the literature by using previously conducted research. The advancement of digital technologies has led to the creation of new employment opportunities, notably in technology-driven industries. As a result of digitization, jobs have been generated in sectors such as data analytics, artificial intelligence, and software development [42,43]. These possibilities can appeal to people with a range of skill sets and interests [44][45][46][47][48][49][50]. People need to have a certain level of digital literacy and aptitude to flourish in the job market in the digital age. The authors of two studies [51,52] found that individuals with higher levels of digital competence are more likely to plan to work in the digital industry. The acquisition of digital skills can have a significant impact on the pursuit of digital careers and professional goals. Gender stereotypes have a huge impact on people's career goals. Society's expectations and perceptions of gender roles have an impact on career decisions, which leads to occupational segregation [53]. People, particularly women, may be discouraged from pursuing careers in technology and closely related industries by conventional gender stereotypes. The career objectives of men and women differ, according to several studies. Men are more likely than women, according to studies, to indicate interest in technical and engineering-related occupations [54]. Several things, including social expectations, self-efficacy views, and exposure to role models, might be blamed for these discrepancies. Despite the growing need for digital skills, there are still gender inequalities in the digital industry. Per [55] and [56], women are underrepresented in leadership roles and tech-related occupations. Stereotypes, prejudices, and structural barriers that keep women from pursuing and advancing in digital occupations are the cause of these discrepancies. Several variables influence gendered professional ambitions in the digital era. These include gender biases, educational and cultural backgrounds, mentorship opportunities, and workplace dynamics [57]. These problems must be resolved in order to create an environment that encourages people of both genders to pursue careers in technology. To ensure that people of all genders have equal access to high-quality education and digital skills training, efforts have been made to promote STEM (science, technology, engineering, and mathematics) education among girls and women and support their participation in technology-related fields [58,59]. It may be extremely important for educational institutions and training programs to play a part in giving people the digital skills they require to seek digital careers. Efforts must be made to dispel stereotypes and misconceptions about gender and digital jobs. The establishment of genderneutral performance assessments, inclusive recruiting processes, and unconscious bias training are just a few examples of the practices and policies that organizations may use to enhance diversity and inclusion [60,61]. It is possible to create a more welcoming environment by dismantling gendered stereotypes in career counselling and school courses. In recent years, the scientific community has paid increased attention to gender, career intention, and digitization. The following are the primary literary axes of the body of literature: gender and career intention: [62][63][64][65][66][67]; among others. Several scientific papers [68][69][70][71][72][73][74][75]; among others also aimed at analysing digitalisation in the dynamics of gender and career intention. The abovementioned show that deeper research should be realised on gendered career intentions in the context of digitalisation using bibliometric analysis despite extensive research on these concepts. --- Methods The use of bibliometric analysis in social science research has significantly increased in recent years [76][77][78][79][80][81][82]. Indicators of the popularity of bibliometric analysis include the development, accessibility, and availability of tools like VOS viewer, Publish or Perish, CiteSpace, Bibliometrix, CRExplorer, and databases like Scopus, PubMed, Web of Science, Microsoft Academic Search. Additionally, information science has replaced commercial research, and bibliometric analysis in scientific development is rapidly growing in this field [77]. The popularity of the bibliometric technique in social science research is not a reflection of a particular trend but rather demonstrates how successfully it can be applied to manage enormous volumes of scientific data to produce high-impact research [77][78][79][80][81]. The scientific community has received the Scopus database favourably since it offers thorough coverage and is easy to use [83][84][85][86]. The study employed the Scopus online database for the analysis since it offers the biggest database, the most well-known research papers, and a VOS viewer compatibility for visualization analysis. The analysis takes into account author performance, research subjects, productivity, and output performance. Performance analysis examines how elements of scholarly research have affected a certain field of study [87]. This approach fits the bibliometric research's reputation for naturally descriptive analyses [77][78][79][80][81]. The study applies the following criteria and filters for papers'--- Results and Discussion Researchers employ bibliometric analysis for several purposes, including to identify new patterns in the performance of articles and journals, cooperation styles, and research elements, as well as to investigate the conceptual framework of a particular subject in the existing literature [77,89]. In order to identify patterns and co-citations of a specific theme by year, author, country, journal, and research constituent [82], bibliometric analyses cover a large number of published research articles. Organizations can make data-driven policy decisions as a result. Stakeholders will learn about the trend in gender and career intentions using trend analysis. Figure 1 illustrates the evolution of 187 documents refined from 202, identified in the Scopus database between 2001 and May 2023, a period of 22 years published in English. The results show an upward growth trend in the publication of works on studies involving digitalization, gender, and career intentions. --- Table 1. Metrics for performance analysis --- Metric Description Result --- Publication years The earliest and latest publication year found in the currently selected results. --- 2001-2023 --- Citation years Number of years from the earliest year found in the currently selected results to the year of the search (usually the current year). The total publications/number of active years of publication (TP/NAY) 8.9 Total Citations (TC) The sum of the citation counts across all currently selected results. 1596 --- Average citations (AC)/year The average number of citations per year (i.e., Citations / Citation years) 72.55 --- Cites/paper The sum of the citation counts across all papers, divided by the total number of papers. The median and mode are also calculated and available separately. --- Authors/paper The average number of authors per paper, calculated as the sum of the author counts across all papers, divided by the total number of papers. The median and mode are also calculated and available separately. --- h-index (h) h Number of documents cited at least h times (a measure of influence) 19 g-index (g) g Number of documents cited at least g2 times (a measure of impact) --- Figure 1. The annual number of publications on gender, digitalisation, and career intentions (2001-May 2023) Source: generated by the author from the Scopus database using Excel. The analysis indicates a progressive growth from 2001 to May 2023 The increasing use of digital technologies has changed a variety of societal factors, including opportunities and job options [90]. New disciplines and industries have emerged as a result of the digitalisation era, presenting opportunities and challenges for gendered career ambitions. Also, with the rise of the information economy, remote work, and flexible job options, the workforce dynamics have changed in the digital age [91]. By modifying conventional career routes and providing for more flexibility in work-life balance, these changes have had an impact on gendered professional ambitions, which can have an impact on people's goals and decisions. Figure 2 shows that, with a minimum of 5 research outputs, the United States of America has 60 publicationsthe largest among this research element. Source: compiled by the author based on the Scopus database using Excel. This suggests that research publication on gender, career intentions, and digitalisation in the US has received considerable attention. With 29 research outputs, the United Kingdom follows the United States as a notable player in this field. This shows that scientists are actively investigating gender, career intentions, and digitalisation in the UK. Spain and Germany follow with 16 and 15 research outputs respectively. The country of origin of 16 research outputs is undefined. There is more clarity on which countries show more interest, explore, and use knowledge to promote better living standards. Research needs funding and sponsorship. Sponsorship and funding promote the study and the growth of knowledge as stakeholders utilize research findings. Figure 3 is a graph showing the trend for sponsorship growth. Funding sponsors like the National Science Foundation have funded 12 publications, the European Commission has 7 documents under their belt, followed by the European Regional Development Fund and Ministerio de Ciencia e Innovación with 3 each, among others. The remaining nine organizations sponsor an average of 1. The dominant keywords in 2016 were education and technology; gender and STEM in 2019, and in 2020, they were diversity and digital technology. Table 2 presents the most cited publications devoted to research on digitalisation, gender, and career intentions. Source: Generated with the data from Scopus database using Harzing's Publish or Perish. An essential method for scientific mapping, citation analysis measures the influence of published works by counting the number of citations they have received. It operates under the premise that citations replicate intellectual contributions and have an impact on research horizons [92,93]. This analysis revealed that [72] has the highest citation (309), followed by 216 citations from [68]. The implication is that all things being equal, [72], on average, has the most significant influence in terms of contributions to the research area. The results of a co-occurrence analysis of keyword networks related to the intersection of digitalization, gender, and career intentions. The analysis identified four clusters that reflect the main study streams in publications within this domain. Cluster one consisted of twelve keywords: computer games, computer programming, computer science, computer science education, digital storage, education computing, engineering education, gender stereotypes, information systems, interactive computer graphics, students, and teaching. This cluster appears to focus on the educational and pedagogical aspects of digitalization concerning computer science and technology-related fields [94][95][96][97]. The inclusion of gender stereotypes indicates a likely investigation into how gender biases and preconceptions impact students' experiences and career intentions in these fields. Understanding the relationship between gender stereotypes and educational settings in computer science and related disciplines is crucial for promoting inclusivity and diversifying the talent pool in technology-driven industries. Cluster two consisted of eleven keywords like career choice, career, decision making, education, female, gender, male, medical education, middle-aged, training, and gender equity. This cluster seems to centre on the career intentions and decision-making processes of individuals with a particular focus on gender dynamics. The inclusion of "gender equity" suggests an exploration of the gender imbalances and disparities that may exist in career choices and opportunities [98][99][100] --- Conclusions The current study's findings demonstrated that concerns relating to digitalisation, gender, and career ambitions were constantly urgent. The volume of publications is trending positively. The results made it possible to draw attention to the numerous sponsors (the National Science Foundation supported 12 publications, and the European Commission financed 7 papers, among others) and nations (the US with 60 publications and the UK with 29 publications) looking into the issues of digitalisation, gender, and career ambitions. It is undeniably evident that the majority of this research is coming from the West. The co-occurrence analysis of keyword networks identified four clusters with distinct clusters which highlights the multidimensional nature of the research and emphasizes the importance of understanding gender dynamics in the context of digitalization and career intentions. Over the last 20 years, there has been an increase in interest in publications on gender, career intentions, and digitalisation, indicating a greater understanding of how these three topics interact and have an influence on numerous facets of society. There has been a large amount of study and intellectual engagement in examining the linkages between digitalisation, career ambitions, and gender, as seen by the approximately 200 papers that have been published in this area. The 8.5 annual growth rate suggests that interest in this subject has been rising over time. This increased rate shows that more academics and researchers are interested in researching how digitalisation affects career goals and how gender plays a factor in this. The steady increase in publications shows a steadfast interest and emphasizes the significance of this field of research. The fact that these papers have generated 1596 citations further underlines the significance and relevance of the research done in this area. Citations are proof that a piece of work has been acknowledged and used as a source by other scholars, demonstrating its value and contribution to the larger academic community. The field is receiving a moderate amount of attention and --- Data Availability Statement: The data sources used in this study are available in the Scopus database. --- Mavis Mensah Senyah Virtual Economics, Vol. 6

The study aims to investigate the relationship between digitalisation, gender stereotypes, and digital skills on gender discrepancies in career intentions. The study uses the VOS viewer tool to present data from the Scopus database and perform analyses on gender and career intention and science mapping. Harzing's Publish or Perish software is used to analyse the performance of publications involving digitalisation, gender, and career intentions. The study uses the VOS viewer tool to present data from the Scopus database and perform analyses on gender and career intention and science mapping. The study shows that in about two decades of the emergence of interest in publications around digitalisation, career intentions, and gender, about two hundred publications have been produced. An annual growth rate of 8.5; 1596 citations emerged from these publications with a yearly average of 72.55. The results showed a persistent increase in research by nations, authors, funding organizations, etc., with the US emerging as the nation with the highest research output. The study concludes that there is a persistent increase in research on digitalisation, gender, and career intentions. The US has the highest research output in this area. The study highlights the need for more research in this area to address gender discrepancies in career intentions. The study also suggests that policymakers should focus on providing equal opportunities for women to access digital skills and education to bridge the digital gender divide.

There are more than 600 First Nations communities and 60 Indigenous languages in Canada (1). With an international reputation for multiculturalism and inclusion, there are a myriad venues for celebrating Indigenous cultures throughout Canada. However, Canada has a long history of colonialism and exclusion that continues to impact Indigenous people (2). This is manifest in the health, social and economic disparities between Indigenous and non-Indigenous Canadians (3,4). Narratives around these disparities are widespread and contribute to negative perceptions about Indigenous communities (5)(6)(7)(8). Stereotyping and discrimination can have significant impacts on the health and well-being of Indigenous people, including Indigenous childrens' cultural identity development (9,10). This is important because strong cultural identity can be a protective factor against suicide for Indigenous youth (11,12) and is associated with more favourable developmental outcomes (13,14). Important to fostering strong cultural identity is affirming Indigenous cultures (15). Paediatrics & Child Health, 2018, Vol. 23,No. 4 However, attitudes and policies regarding Indigenous people have, for the most part, been negatively oriented (16). These attitudes extend to the Canadian health care system, where systemic racism has been identified as a significant challenge facing Indigenous people (17). Addressing negative attitudes and stereotypes will require education and awareness-raising to elevate understanding on the part of non-Indigenous people, including health care providers (18,19). It is critical for Indigenous and non-Indigenous people to work together on such initiatives, and to involve young people in opportunities to learn about their cultures and represent the strengths of their communities in meaningful ways (20). --- THE MASKWACIS COMMUNITY The Maskwacis community is located in Alberta, Canada, and consists of four Plains Cree Nations with a combined population of approximately 15,000. National media attention has focused on crime and gang violence in Maskwacis, largely ignoring the rich and vibrant Cree culture that threads the four Nations together. In Maskwacis, Cree history, culture and language are featured prominently in traditional community ceremonies and cultural events that regularly take place, and the Cree language is taught in Maskwacis schools. Community elders view their culture and language as a means to combat the social and public health problems that Maskwacis community members face due to the destructive impact of colonization, residential schools and forced assimilation. --- RESEARCH QUESTIONS AND PURPOSE In the face of negative media attention, community members and elders identified the importance of promoting community strengths and reframing perceptions of Maskwacis. Toward this aim, two research questions were addressed: 1) How do youth in Maskwacis view their community strengths? and 2) To what extent can photographs be used as a tool for reframing perceptions of an Indigenous community? --- RESEARCH METHODS AND PROCESSES --- Community-based participatory research The current study followed a community-based participatory research approach (CBPR) (21,22). CBPR is well suited for research with Indigenous communities, because it yields equitable researcher-participant participation and power relations (23)(24)(25). Methods for the current study were approved by Human Research Ethics Board 1 at the University of Alberta. --- Participants Two elders and 11 youth (age 15 to 19) were engaged. Maskwacis school staff identified students for project participation. The project coordinator visited students at their schools to discuss the project. Next, the project coordinator and a community partner visited students' homes to garner parental consent for participation. Parents were offered a small gift in appreciation. --- Training Photovoice begins by providing participants with cameras and photography training. For the current project, two experienced Indigenous photographers provided technical training, while two community elders mentored youth to recognize their community's positive assets. Youth also worked with the project coordinator over a series of ten sessions to explore their community and capture photographs. --- Photo exhibits Part of the photovoice process involves communicating messages to policymakers, and a primary method for such communication involves holding photo exhibits (26). Four exhibits were initially held at an Elders Centre in the Maskwacis community, the Wetaskiwin community hospital, the University of Alberta Hospital and the University of Alberta Faculty of Extension campus. Brief feedback questionnaires were distributed at each exhibit. As a result of media coverage and extended public interest, two additional exhibits were held by request at the Maskwacis Cultural College (MCC) and the Red Deer Community Hospital. --- Data analysis Data collection and analysis occurred concurrently for photos and accompanying descriptions. Wang and Burris' (1997) methods for analyzing photovoice data were used. These include 1) selecting photos, where participants choose photos that they deem most representative of their experiences; 2) contextualizing, where participants provide descriptions of their photos; and 3) codifying, where data are sorted into codes, categories and themes. Selecting photos and contextualizing occurred during group sessions with the youth, after which the project coordinator began codifying. Emerging codes and themes were then discussed during subsequent group meetings. Codifying was carried out in alignment with thematic analysis (27). This involved producing initial codes, looking for overarching themes among codes, refining and naming themes and completing final analysis. Thematic analysis was carried out for photos and accompanying descriptions as well as qualitative questionnaire data. Simple descriptive statistics were used to analyze numerical questionnaire data. --- FINDINGS --- Photos and descriptions Participants captured more than 200 photos and selected 40 as most representative of their experiences. Youth also provided descriptions for the 40 photos. Analysis of the visual content of photos and accompanying verbal descriptions resulted in four themes, depicted in Figure 1. --- Culture Eleven photos fit within the culture theme. According to participants, it is important for people to know that there is thriving traditional culture in Maskwacis. With respect to Figure 2, a participant commented that cultural pride is a strength of the community. Youth also captured images that provide examples of the Cree language in and around Maskwacis (Figure 3). In addition, participants felt that including Cree language on signs entering and exiting the neighbouring town of Wetaskiwin demonstrated respect. As one participant stated, 'this sign showcases the community's willingness towards reviving our Cree language.' Participants also agreed that tipis represent an important aspect of the Cree culture, and therefore, a strength of the community (Figure 4). Participants additionally captured images of cultural symbols including beading and berry picking. Looking to the future Ten photos were categorized as looking to the future. According to participants, schools in Maskwacis represent a significant strength (Figure 5). One participant noted that, 'our schools are a way of giving us a future.' Participants also captured photos of youth. As one participant described, 'youth in our community are amazingly talented... We have bright futures if we want them.' In addition, participants captured images that they described as metaphorically representing a future orientation. With regard to Figure 6, a youth participant expressed that, 'growing up here, we have a long road ahead of us, but we always have the choice to take the road to success.' Another youth captured an image of a moon at night and described that, 'in order for there to be light, you also need some darkness. So challenges and hard times can lead to growth.' With regard to Figure 7, 'This picture represents that we can all grow upward and thrive.' --- Natural world Twelve photos aligned with the natural world theme. As participants described in relation to Figure 8,'some people think that Maskwacis is a scary place at night. These pictures show it can be beautiful here at night.' Participants also captured photos of ponds, reeds and wild plants, expressing that, 'in our community, we have a deep appreciation for Mother Earth.' With regard to Figure 9, a participant explained that, 'we see all creatures and parts of nature as being important, not just humans.' In describing Figure 10, a participant noted that, 'in Maskwacis, there's lots of space for exploring the outdoors.' Relationships Six photos aligned with the theme of relationships. In describing images related to this theme, a participant explained that, in Maskwacis, there are opportunities for engaging in games and social activities that build relationships. As one youth described, 'I like how we can participate in traditional activi-ties... if we want to, and doing these activities can help us make friends.' In describing Figure 11, a participant noted that, 'this tree represents a family and community and how we all share the same strong core and deep running roots.' Another stated that, 'this picture is all about love, which we have a lot of in our community.' In response to Figure 12, 'Maskwacis can be a really fun place to be together.' Finally, in describing Figure 13, a youth noted that, 'people who live here visit each other a lot in their homes, and we all share Maskwacis as our home.' --- Questionnaires In total, 392 questionnaires were distributed at photo exhibits. Participants responded to questions related to the photos and the Maskwacis community. Data analysis did not result in significant differences among responses between data collection sites. Accordingly, questionnaire data from all photo exhibit sites are presented together. --- Previous ideas about Maskwacis Survey respondents were asked to indicate previous ideas about the community. Data analysis resulted in three categories for this survey item; namely, 1) troubling social issues, 2) working through challenges and 3) positive perceptions. Responses aligned with the category of troubling social issues (n=116) generally indicated that participants perceived the ---. the gang problems and abuse problems on the news...' Within the category of working through challenges (n=45), respondents reported perceiving Maskwacis as a community that was striving to overcome negative social issues. As one respondent from the Extension exhibit expressed, 'I knew it was a community that is working at making itself a better place to be.' A respondent from the Wetaskiwin hospital described Maskwacis as 'a...place trying to rebuild.'Responses aligned with the category of positive perceptions (n=30) mainly described Maskwacis as a place with a strong culture and resilient members. One respondent from the Wetaskiwin hospital described Maskwacis as 'a close-knit community based on family and relationships.' At the Maskwacis event, a respondent stated that, 'The media portrays us as a place of trouble and loss, but this is not reality. There's a lot of talent in our community.' --- Changes in ideas Respondents were asked whether viewing the photo exhibit had changed their ideas about Maskwacis. Ninety-three per cent of respondents (n=236) indicated that viewing the photos had improved their views of Maskwacis. Of those respondents who indicated that their views had not changed, most stated that they previously held positive views of the community, and that the photos had reinforced these views. Eighteen per cent of respondents (n=46) indicated that viewing the photos helped them understand that Maskwacis was rich with culture and pride. As one respondent from the Wetaskiwin hospital stated, 'I didn't realize this was such a proud and traditional community.' Seventeen per cent of respondents (n=42) indicated that the photos made them aware of the talent and creativity in the community. Fifteen per cent of respondents (n=39) felt that the pictures made Maskwacis seem more relatable to them. As one respondent expressed, 'Makes me feel "like" the youth rather than seeing the community as an "other."' An additional 15% (n=39) referenced how viewing the photos made them aware of the beauty in Maskwacis. According to a University Hospital respondent, 'It looks like people who live here see their community as beautiful. The photos make me think the community is beautiful too.' Thirteen per cent of respondents (n=34) mentioned that the photos helped them to view the community in a hopeful and uplifting light. As a respondent from the Maskwacis event noted, 'The photos gave me hope that our youth are willing and ready to change the image of our community.' Another 9% of respondents (n=22) indicated that the photos made them think of Maskwacis as a place where opportunities were provided for youth to express themselves and participate in community projects. Finally, 6% of respondents (n=14) noted that viewing the photos had positively changed their views of the community, without providing further detail. --- DISCUSSION The current photovoice project, which aimed to positively reframe perceptions of the Maskwacis community, was undertaken in response to concerns from Indigenous community members about the negative perceptions that are widely held about their community. This was not to ignore the current social issues that Maskwacis struggles with and that can largely be traced to colonialism and residential school experiences. Rather, this project aimed to provide a more complete picture of the community by asking youth to uniquely focus on its strengths. To this end, two research questions were posednamely, 1) How do youth in Maskwacis view their community's strengths? and 2) To what extent can photos be used as a tool for reframing perceptions of an Indigenous community? With respect to understanding how youth in Maskwacis view their community strengths, the content of photos captured by youth was analyzed using thematic analysis. This resulted in four themes that described the strengths of Maskwacis, reflecting the community's strong relationships, commitment to culture, the beautiful natural world that is a part of Maskwacis and the community's ability to look toward the future. Although this study was conducted specifically with Maskwacis, findings prompt consideration of the strengths and resilience of other Indigenous communities facing similar social and health issues. In addition, findings provide foundational support for the use of photographs to reframe perceptions of an Indigenous community. The majority (93%) of survey respondents provided examples of ways that viewing the photos had positively changed their perceptions of the Maskwacis community. These examples focused on culture and pride, talent and creativity, relatability, beauty, hopefulness and youth opportunities. In addition, an unprecedented number of people viewed the exhibits, attended photo exhibit openings and responded to surveys. Furthermore, the project received media coverage that was more widespread than anticipated. In total, nine news articles were published on the current project by local media outlets. As a result, a total of six photo exhibits were held, given that media coverage resulted in additional requests for exhibits. This suggests an appetite among Indigenous and non-Indigenous community members to appreciate and understand positive perspectives of Indigenous communities. Moreover, there is increasing recognition from researchers, service providers and policymakers regarding the critical importance of culturally responsive care in improving the health outcomes of Indigenous patients (28). In order to provide culturally responsive care, it is necessary for paediatric health care providers to maintain insight into issues, both medical and sociopolitical, affecting the health of Indigenous patients. The current study can contribute to elevating such insight and is therefore highly relevant to paediatric health care providers. Furthermore, youth involved in the current project anecdotally described the benefits of learning photography skills, being guided by elders to look for community strengths, as well as the power and appropriateness of imagery among Indigenous peoples. As such, this project opens up possibilities for related future research, including the impact of project participation on youth themselves.

In the face of negative media attention, community members and Elders from the First Nation community of Maskwacis identified the importance of promoting community strengths and reframing perceptions of their community. Two research questions were addressed: (1) How do youth in Maskwacis view their community strengths? and ( 2) To what extent can photographs be used as a tool for reframing perceptions of an Indigenous community? A community-based participatory research approach was used for the current photovoice project, through which two Elders and eleven youth were engaged. Indigenous photographers trained youth participants, and Elders mentored youth to capture photographs that represented their community's strengths and resilience. Youth selected forty photos to display at a total of six photo exhibits, where feedback was gathered from 392 attendees using brief questionnaires. The content of their photos was analyzed using thematic analysis. This resulted in four themes that described the strengths of Maskwacis, reflecting the community's strong relationships, commitment to culture, the beautiful natural world that is a part of Maskwacis, and the community's ability to look toward the future. In addition, findings provide foundational support for the use of photographs to reframe perceptions of an Indigenous community. The majority (93%) of survey respondents provided examples of ways that viewing the photos had positively changed their perceptions of Maskwacis. This study prompts consideration of the strengths and resilience of other Indigenous communities facing similar social and health issues. Therefore, findings are highly relevant to paediatric health care providers seeking to provide culturally responsive care.

Introduction --- The historical context The route to engaged research undoubtedly varies according to academic discipline and the journey is characterised by the relationship between knowledge, its producers and those affected by the research. Here, we briefly explore 70 years of development in relation to the sciences and their publics as a way of illustrating the significance of the research we have conducted. After World War II ended, progress towards the 'endless frontier' of research-and especially scientific research-was seen as the means by which nations would ensure their peoples' future health, prosperity and security. [1] In return for these goods, the people's role was to sanction research funding; effectively, to endorse researchers' licence to practise. [2] Further discussion of the reasoning behind the development of a relationship between professional scientists and the public centred on the idea that part of the duty of being a scientific researcher was a responsibility to improve understanding, communicate the technological, humanitarian and economic benefits of science and contribute to better-quality public and private decisionmaking. [3] In this formulation, the relationship between scientists and the public was largely conceptualised as educative: the scientists' main purpose for communicating was to school a scientifically illiterate public. [4] Despite the obligations placed on scientists and the perceived economic value of increasing public understanding, by the mid-twentieth century the scientific establishment was becoming concerned by scientists':'mistrust, lack of understanding and often unwillingness and inability to communicate adequately' and their tendency to'retreat into their shells, frowning on those who ventured onto the public stage'. [3] (para 6.1) One of the significant implications of this judgement was that without scientists' active participation in society, funding for scientific research could be politically vulnerable. [5] Alongside these economic considerations grew concerns that the role of scientists as specialised keepers of knowledge, healers of ignorance and assuagers of deficits was neglecting the considerable, if informal and localised, understanding and expertise possessed by members of the public. [6] Survey evidence indicated people were willing to learn more about science and to discuss its ethical and social implications, [7] often because they considered scientists would benefit from listening more 'to what ordinary people think'. [8] (p. 52) This implied that scientific researchers needed to do more than simply tell people what they were doing; they also needed to listen to people and respond, even if they considered their antagonism, fears or hopes to be ill-founded. [9] Further evidence, predominantly from qualitative accounts, showed that researchers could gain from drawing on the knowledge and experiences of other expert stakeholders and members of the public, in effect valuing different kinds of expertise as additional resources to inform the overall research endeavour; the collaborative sum would be greater than researchers working only with other academic researchers. [10] Over time, the rhetoric shifted from understanding to engagement. This movement is encapsulated in the definition developed by the UK's National Co-ordinating Centre for Public Engagement (NCCPE), which defines engagement as 'a two-way process, involving interacting and listening, with the goal of generating mutual benefit'. [11] The complexities of public engagement with research Public engagement with research takes multifarious and diverse forms. The New Economics Foundation's Participation Works! [12] detailed 21 participatory mechanisms; Wilsdon and Willis [13] recorded activities including deliberative polling, focus groups, citizens' juries, consensus conferences, stakeholder dialogues, Internet dialogues and deliberative mapping; Rowe and Frewer [14] identified around a hundred participatory activities; Mesure [15] approximately 1500 initiatives; DIUS [16] noted growing numbers of science and discovery centres, museums, cafes scientifiques and festivals. Bauer and Jensen [17] extended engagement to include public lecturing, giving interviews, writing popular books or articles and collaborating with non-governmental organisations and Davies added 'volunteering activities, participatory social research, [and] even informal conversations about research outside the university setting, as well as more familiar activities such as giving talks to school groups or holding university open days'. [18] (p.726) These many activities suggest that at least to some degree researchers and citizens are willing to engage through dialogue and other forms of participation, but nevertheless, some important questions remain. First come the practical challenges: who should (and should not) engage, when and how often, how should the engagement be organised, for what purpose(s), how will success or failure be measured and how should the engagement (and contributions of the participants) be funded? Second is the disruptive challenge of acknowledging and accrediting expertise within existing systems of academic validation, for example peer-reviewed publication. For some, accepting the idea that 'publics, not only the scientists or experts, can make useful and valuable contributions to discussions and decisions about science and technology', [19] (p.13) challenges hard-won expertise and raises questions about how researchers can ensure that the contributions of all participants in a research project are recognised in ways that are meaningful and useful. Third, how might researchers' professional identities be affected by adopting a more open and distributed approach to research? Researchers may well have spent many years developing their knowledge, skills and craft; being seen by peers and others as an expert is an important part of professional identity: Involving the public inevitably means researchers have to give up some of their power [and] although many researchers have recognised that this shift is essential for projects to become genuinely collaborative, no one has reported finding it easy. [20] (p.66) A fourth challenge relates to the question of who owns research; for researchers, their professional value may depend on their being in control of ideas, data and intellectual property. Many agencies have an interest in the value of research and its outputs: researchers themselves, their institutions, funders, industry, etc.; extending ownership by inviting wider engagement may be to the advantage of the research but could be seen as diluting the return on investment. Fifth, despite descriptions of dialogue as a two-way communication, the extent of genuine reciprocity and dialogue in public engagement activities is debated. [21] In practice, tension persists between assumptions about deference, expertise and scientific privilege [22]-the language of dialogue, debate and lay agency-and evidence for the appropriateness of, and public preference for, listening rather than talking in certain circumstances. [23,24] As Holliman et al., argued: It is possible to imagine 'deficit desire'-audiences who actively seek the linear lecture from a trustworthy expert with little expectation or interest to challenge the scientific perspective presented to them-as well as 'dialogue fatigue'-those who actively avoid forms of engaged democratic citizenship in relation to science. And they could be the same person. [25] (p.276) In other words, genuine engagement requires not only that publics are given input into the topics of research but also that the methodologies for engaging are considered collaboratively. This is not to say that engagement is synonymous with democracy, at least not straightforwardly. Engaged research requires different forms of expertise, whose relevance will wax and wane depending on the research and on the point in the research cycle where engagement happens. A final challenge for researchers and institutions in the UK is that the concept of engagement has the potential to become entwined with the concept of measuring the impact of research. In 2014, the Research Excellence Framework (REF) reported on the quality of research in UK higher education institutions. [26] Although public engagement with research is often conceived as the reification of a co-operative relationship among institutes of higher education (HEIs), the communities within which they sit and stakeholders with an interest in the outcomes of the research, for example through HEIs sustaining local economies, [27] it has, to varying degrees, also been identified both as a mechanism for providing evidence of the impact of research, and as itself being a form of impact. [28] In the REF, the efficacy of the work of academics within UK universities was assessed to determine the reach and significance of the engaged research's public value and impact. As we noted above, in the past, public engagement with research has served multiple purposes, for example as a vehicle for increasing openness and transparency, as a means for maintaining support for the use of public money for research, [2] enhancing citizens' discussion of scientific issues, [7,8] incorporating public concerns and skills in research [10] or encouraging more students to take up specific subjects such as the sciences. [29] However, the potential long-term effect of the REF could instead be to transform public engagement into the 'engine powering the conceptual, critical and methodological framings and motivations of impact'. [27] (p.120) Our counter to this challenge is to foster and encourage progressive approaches that ultimately improve the quality of research in ways that are meaningful to all participants. --- Context of this research Many interventions across the UK higher education sector have been designed to address the relationship between universities and other sectors of society. From 2008-2011, RCUK, the UK Higher Education Funding Councils, and the Wellcome Trust funded the Beacons for Public Engagement initiative. Six 'beacons', university-based collaborative centres, were established to support, recognise, reward and build capacity for public engagement. The NCCPE, also established under the Beacons for Public Engagement initiative, was set up to co-ordinate and share learning among the Beacons and across UK higher education institutions and research institutes. In 2011, RCUK issued a more narrowly-focussed call to fund eight 'Public Engagement with Research' (PER) Catalysts over three years (completing March 2015). The PER Catalyst initiative invited universities to draw on and develop the learning from the Beacons programme, with the overarching aim of embedding a culture within universities where public engagement with research is strategically planned, systematically assessed, valued and recognised. The Open University (OU) received funding under the PER Catalyst for action research to develop and implement strategies that promote structured and equitable mechanisms for effective and sustainable engagement with a range of publics, stakeholders and user communities. To achieve this, the project team proposed a programme of organisational change in which OU researchers were given opportunities to be involved and engaged as co-researchers. This programme of organisational change is informed by an action research approach, [30] initially involving rigorous research-based inquiry into how the university currently functions with respect to public engagement with research. Findings from this initial diagnostic exploration are the focus of this paper. In conceptualising the plans for our project, 'An open research university', we found limited data on researchers' conceptualisation of engagement, what activities they consider constitute engagement, what communities they believe they are engaging with and what communities they would (or would not) like to engage with. [31,32] How do researchers' views on engagement relate to the operation of public engagement with research and research practice within the university? We have used the findings from this study, in combination with other data, to map researcher practices across the OU, investigating the processes of innovation that have led to the introduction of public engagement with research. We have collated existing resources and documented current practices to support researchers in several academic domains. This helped us to identify areas where additional interventions that have an impact on institutional systems and processes could be considered; we describe some in our Conclusions. --- Method Our findings are derived from OU researchers' responses to the 2013 Vitae Careers in Research (CROS) and the Principal Investigators and Research Leaders (PIRLS) online surveys. Vitae (www.vitae.ac.uk) is a not-for-profit organisation supporting the professional and career development of post-graduate researchers and research staff in UK higher education institutions and research institutes. It is funded by Research Councils UK, UK HE funding bodies and institutional subscription. CROS and PIRLS are biennial comparative surveys run by a group of UK universities to form a 'benchmarking club'. [33] Each survey has a set of common core questions that is asked in all participating universities; the CROS question set is slightly different to that of PIRLS. In addition, universities have the option to add questions that are asked only within their own institution and the results of which are only available to that university. The OU's CROS and PIRLS surveys were administered via the Open University's staff networks. The surveys were run in May 2013 and were open for two weeks. --- Ethics statement The research protocol was approved by the Open University's Human Research Ethics Committee and the surveys were hosted on the Bristol Online Surveys platform, which provides a secure environment for the administration and analysis of online surveys. [34,35] Neither the CROS nor PIRLS surveys asked respondents to give personal information such as names or contact details, although they were asked for data such as length of service and academic discipline. The institution-specific questions additionally stipulated that respondents should not include details that could inadvertently identify them, such as URLs or the names and locations of activities. All the institution-specific questions were free-text responses. Participants were deemed to have consented when they submitted the questionnaire. The invitation to respond to the surveys included the following information: "Your responses will be anonymous and you will not be identified or identifiable. You can complete [the survey] if [information identifying which survey (CROS/PIRLS) was appropriate]. It should only take about 20 minutes to complete; the information you provide will bring long-lasting benefit to you and your peers. Participation in the survey is entirely voluntary and you can discontinue at any point up until the questionnaire is submitted. This is because no names are taken, so we cannot distinguish between individual participants' data. All data is confidential and will be stored securely. By submitting or returning the questionnaire, you consent to take part in the survey and for your data to be included. If you agree to take part in this survey, please click on the link below." --- CROS survey The CROS survey had 27 core questions, of which some were in several parts. A number of these were related to public engagement and the impact of research, including: --- PIRLS survey The PIRLS survey had 19 core questions, of which some were in several parts. As with CROS, some were related to public engagement and research impact: --- Institution-specific questions Professional development is an integral aspect of the action phases of our PER Catalyst project. In part to inform the requirements of future training needs, we consulted the OU's Research Career Development Team at an early stage in the project to develop four institution-specific questions that were added to both the CROS and PIRLS surveys. These questions explored researchers' understandings of publics and public engagement with research: 1. The Open University is working to create a culture in which public engagement with research is embedded within strategic planning for research and the operational practices of researchers at all levels. In fewer than 150 words, how would you define 'public engagement with research'? 2. Please describe, in general terms, a successful activity involving public engagement with research in which you have participated and the criteria you used for judging the success of the activity. 4. Please describe your three most important reasons for engaging with the non-academic communities or people that you listed in question 3a. --- Coding and analysis Responses for the selected questions from the core sets were downloaded and imported into SPSS21, which was used to store and analyse the data. The majority of responses were categorical variables that could be summarised as univariate measures. Respondents were free to choose whether or not to answer any particular question. The free-text responses were analysed as qualitative data. We acknowledge that by using free-response questions to collect qualitative data, we did not have the opportunity to probe respondents to extract deeper meaning and validate responses using their own concepts, frames of reference and vocabulary. Using web-based surveys offers the opportunity to ask questions of a larger number of respondents than can feasibly be managed by interview. However, using surveys means we lose contextual information, such as the age and sex of respondents and the social context in which they are completing the survey. [36,37] In the project as a whole, we used a mixed methods approach, triangulating quantitative with qualitative approaches (a series of semi-structured interviews) to complement each other and support validation. [32] However, in this paper we are not reporting on the interview data. The analysis began with the inductive construction of coding categories. The coding categories were emergent; that is, they were formulated during analysis and grounded in the data, rather than being defined beforehand. Although we acknowledge that no one can approach analysis with a completely open mind, a data-grounded approach to analysis allows the researcher to start with fewer preconceptions. [38] To minimise bias that might result from an individual's personal conceits and existing notions, two researchers independently printed, read and scrutinised the free-text responses for key concepts, actions, relationships and meanings. The resulting codes were applied to the responses, then refined and edited to remove overlaps, confusions and repetitions. This process was repeated until no further changes were made to the codes. Finally, the two researchers agreed a set of codes and definitions (see Tables 1,2, 3 and 4). --- Findings: Exploring the Landscape of Public Engagement with Research In this section, we discuss how researchers define public engagement with research, what activities they consider to be engagement and the communities with which they do (or do not) choose to engage. We also discuss some related data on researchers' experiences of training for, and participation in, public engagement with research. --- Response rates For both surveys, the response rate of OU researchers was broadly in line with the response rate of the national survey. Sixty-eight UK universities took part in the CROS survey. A total of 8216 complete, non-duplicate responses were received, an overall response rate of approximately 26% of potential respondents. [34] The Open University CROS survey received 57 responses (34% response rate). Forty-nine UK universities took part in the PIRLS survey. A total of 4837 complete responses were received, representing a response rate of 28%. [35] The Open University PIRLS survey received 114 responses (22% response rate). Unless specifically stated, our findings are based on researchers' responses (57 + 114 = 171 in total). --- Defining and practising public engagement with research We begin with the premise that language and practice are inextricably enmeshed. Therefore, in this section, we will compare researchers' personal definitions of public engagement with research (see Table 1) with the types of public engagement activity they describe they undertake (see Table 2). Answers to the institution-specific questions showed that researchers vary in their conceptualisation of public engagement with research (see Table 1). Most respondents (Table 1: 73%, n = 125) offered some kind of definition or description of public engagement with research, ranging from the personal ('I enjoy giving public lectures') to the utilitarian ('I'm paid to do it'), to the philosophical ('involving the public, as in Habermas's conception of the public sphere'). A very small number (1%, n = 2) described engagement negatively, for example 'a distraction from core business'. Thirty-nine per cent (n = 63) did not reply, did not provide a definition or said that they 'didn't know'. This latter finding suggests some researchers remain unaware of the recent shift in official rhetoric about public engagement with research. Dissemination and communication. The most common definitions of 'public engagement with research' focussed on the dissemination, communication or presentation of research. Thirty-two per cent (Table 1: 32% n = 54) of researchers offered definitions that fitted within 'dissemination', for example: 'The dissemination of both the goals and results of research to wider audiences, including non-academic' 'Disseminating our research results to public and other institutions in a clear and understandable way so that they can be used to help them achieve their aims' (emphasis added) 'Dissemination of ideas and practical potential of research, and transference of these ideas into common sense understanding of the world by the public' Dissemination is clearly a central code in this analysis. We interpret this to mean the respondents are conceptualising engagement in a mode that Irwin described as first-order: deficit-model, one-way, top-down communication. [10] Ideas are 'transferred' into public understanding, information is 'conveyed' and talks are 'aimed'. As Davies argued of scientists, some persist in perceiving engagement as difficult, dangerous and framed within an over-arching context of one-way transfer. [39] On the other hand, characterising linear one-way communication as somehow inferior to dialogue ignores people's desire for information and the enablement that can arise from understanding. [23,25] There is also evidence of some conflation of terminology, which is also reflected in other studies. [31,32,40] Several responses included mixed terminology (see, for example the emphasis above), which could indicate the incorporation of dialogic methods within wider dissemination strategies, indicating the dominant dissemination view of engagement is reinforced by researchers' descriptions of engagement activities in which they had been involved (see Table 2). Approximately half (Table 2: 53%, n = 90) of the researchers were able to describe an activity they considered to be engagement. Although only 8% (n = 13) of researchers said they didn't do any public engagement activities, a further 32% (n = 55) left the response blank, which may or may not indicate non-participation. Those who did respond described a variety of activities, by no means all of which could be said to be framed by the NCCPE's [11] vision of dialogue and mutuality. For example, 19% (n = 33) of researchers described researcher-led dissemination activities: 'We took part in an RCUK event to showcase our work to the public' 'Conveying to non-expert colleagues and to the wider public the content, scope and significance of my research' 'Informing and explaining to the public about developments, potential developments and applications in research' This seems to demonstrate that the language and rhetoric of engagement is being operationalised in specific and quite conservative ways; researchers 'have begun to talk the talk of engagement, but have not started to walk the walk'. [41] (p.61) Our challenge, as an action research project, is to identify where this is so, and to invite researchers to explore more imaginative and relevant mechanisms to engage a wider set of stakeholders. Partnership, dialogue and collaboration. Partnership, dialogue and collaboration are mentioned in both researchers' definitions of public engagement with research (Table 1) and their descriptions of activities (Table 2). For example, 11% (n = 19) of researchers' definitions (Table 1) could be coded as the 'collaboration' of different communities-researchers, non-academics, members of the public, stakeholders, policy-makers, etc.-in the process of research, for example: 'Involving the public from the outset in the design, consultation, evaluation and distribution of a research project' 'Collaborating with key stakeholders in the planning, execution and dissemination of research, ensuring that findings [are] presented in way relevant and accessible to key audiences' 'Collaborating with 3rd sector, industry, media, government sectors to enable knowledge and skills exchange' Collaboration was sometimes linked with dialogue; a small group of researchers (Table 1; 8%, n = 14) employed 'dialogue' or its equivalent in their definitions: 'A process of dialogue and shared learning between researchers and others, those being researched as well as wider audiences and communities. In some contexts the co-production of research may be a useful model' 'Engaging in and being open to dialogue with non-academic publics, not solely by giving talks but by listening/participating as well' These descriptions were also reflected in the description of activities. The next most common description of engagement activities (Table 2: 14%, n = 23) was of 'partnerships' involving working with practitioners and co-production of research with diverse groups: 'Public events that helped to shape/reshape the trajectory of a research project at its outset' 'Working closely with advisory groups and learning from them; feeding back [...] in an iterative fashion to policy actors/professionals as part of the research process' These are progressive concepts and operations of engagement, in which socially-relevant knowledge is both produced with and communicated among different groups of social actors. [42] However, we believe these views were influenced by researchers' discipline; although responses of this nature were received from across the university, 11 of the 19 (Table 1) that fell into this category came from the social sciences. In these comments, the process of dialogue has a value in itself, allowing researchers to be open about their values and purposes when developing their investigations [16] and to demonstrate transparency, accountability and public value. [43] We note, however, that while listening, sharing and exchanging are important components of dialogue for this group, in practice the label of 'dialogue' was applied to a range of very different practices and strategies, not all of which are distinctively different from one-way communication or significantly reduce the sender's role in controlling communication. [23] Why undertake public engagement with research? Researchers were asked to state up to three reasons why they took part in public engagement with research. We received 246 responses to this question; in Table 3, percentages have therefore been calculated against n = 246. Sixty-nine per cent (n = 118) of the 171 respondents provided at least one reason for participation; 28 more (Table 2, n = 90) than described an activity in which they had taken part. Researchers gave a range of reasons for engagement, including education and communication, collaboration and dialogue and to improve the quality of research. The biggest group of responses (Table 3: 15%, n = 37) focussed on the idea that public engagement enabled researchers to influence policy or policy-makers, or drive social change: 'To effect social change' 'To have an impact on public policy and therefore indirectly on service provision' 'Policy may change in favour of the marginalised' This is borne out by responses to question 25a of the CROS survey (see Method; note that. this question was only asked in the CROS survey). Thirty-two of the 57 respondents claimed they had engaged with policy-makers or end-users of research. However, there were some nuances in participation: two-thirds of researchers funded by EU frameworks said they had not undertaken any public engagement activities but would like to. This group was also the most likely to say they had engaged with policy-makers, indicating they are drawing a distinction between the 'general' public and the specific public of policy-makers. This disconnection may also be a result of the conservative operation and conceptualisation of engagement we described earlier; if researchers conceptualise public engagement with research as dissemination, they might not regard non-dissemination-based activities (such as discussions with policymakers) as engagement. The next largest group of comments (12%, n = 29) covered the idea that engaging with nonacademic communities was a way to increase the quality of research by supporting more accurate research, increasing the relevance of research and relating it to the real world: 'To keep research useful To keep research practical To potentially make a real difference' 'Purely academic perspectives are often narrow or miss aspects which are apparent to professions in the field, or those individuals who experience it. Engaging these perspectives provides a more accurate picture.' Earlier research uncovered reasons for undertaking public engagement which include recognising a duty to communicate research and its social and ethical implications; [44,45] desire to further a career; [46] and the belief that engagement may lead to research that is more'socially, economically and environmentally viable'. [47] (p.1) All these reasons were mentioned by OU researchers, albeit in relatively small numbers (see Table 3). Beyond these instrumental reasons, previous research has identified other factors as significant in affecting researchers' willingness to participate in public engagement with research that are relevant to the action phases of our Catalyst for PER project. These include researchers' previous experience and personal perceptions of their capability, [46] their urge to do what their colleagues are doing [24] and having had some training in public engagement. [45] Although we did not find evidence of all these factors in our survey (see Table 3), nevertheless, responses to question 19 in the CROS survey (n = 57) are relevant. 2 Thirteen of the respondents had had some training in public engagement with research, whilst more than half (n = 33) had no experience of relevant training, but would like to have the opportunity. Those most likely to have had training were mid-career researchers (n = 4), with the highest demand for training coming from experienced researchers (n = 14). --- Assessing the quality of public engagement with research Of the 171 researchers surveyed approximately half (53%, n = 92) gave no response to the question about what they saw as a successful activity and what criteria they used to make this assessment (Table 4). Thirty-nine responses were variants of 'none' or 'no criteria'. Of the 40 substantial responses, 14 respondents provided more than one suggestion (n = 54). Where criteria were offered, 18 measured success by post-event follow-up (such as emails from participants, or media interest) and 15 judged it by the behaviour of participants at the event. Only five responses mentioned some form of formal or semi-formal evaluation. These findings, which are supported by previous research [31] and calls for routine assessments of engagement, [48] suggest that one of the key areas for development for the higher education sector is to extend the focus on developing and delivering activities to include measures of quality; in effect, to nurture a culture in which reflective practice in engaged research is valued, supported and rewarded. For example, the RCUK's Concordat for Engaging the Public with Research states that, 'Research organisations should themselves monitor and evaluate the efficacy of the public engagement they support...' [48]. However, we acknowledge there can be challenges to this approach. Anecdotally, we have encountered researchers who value the idea of engagement as something undertaken voluntarily and out-of-hours, because it allows them to retain control over their activity, which could be undermined if explicit formal evaluation were demanded. [49] Who are the publics of public engagement with research? The final group of institution-specific questions asked researchers for their views about the communities with which they engaged. Researchers gave a wide range of responses in their answers, including media professionals, teachers, parents, school pupils, learned societies, healthcare professionals, patients, policy-makers, non-governmental organisations, companies, industrial partners, charities, community groups, voluntary and third sector organisations, university students and anonymous audiences such as viewers, listeners and readers of mass media. This diversity highlights the challenge of identifying the publics in engaged research, and the need for resources to support the processes of public formation. [50] In all, 70% (n = 121) named at least one public and 41% (n = 70) named more than one; only 1% (n = 2) said they had not engaged with any public. Twenty-nine per cent (n = 50) left the response blank. This finding is reinforced by the researchers' responses concerning groups with which they would not wish to engage. Half (55%, n = 95) left the response blank, while by far the largest single response (29%, n = 50) was some version of 'none', indicating researchers' general willingness to work with a range of audiences. However, this is not quite borne out by responses to the question of which groups they would like to engage with, where responses were very often the same as responses to the question of which groups they had engaged with. Twenty-one per cent (n = 25) simply said 'the same' (or equivalent) and 60% (n = 70) named at least one of the same communities. There may be many reasons for this. First, respondents may be unaware of the existence of different communities; as we noted earlier, terms that have currency in the field of public engagement research may not be readily exchangeable in other fields. For researchers in different fields, the term 'public' is constructed in many different ways, to the extent that language use has shifted from one 'public' to multiple 'publics' that form, re-form and overlap, depending on their interests, backgrounds, experiences and preoccupations. A restricted view of available publics may also be an artefact of researchers' views of what constitutes public engagement; if engagement is construed as one-way transmission, this restricts audiences to those who are able and/or wish to be receivers. [31] Second, there may be restrictions on the kinds of audiences with whom researchers are able to interact; for example, funders may introduce restrictions that limit researchers to interacting with certain groups (e.g. only supporting working with industrial partners) or researchers may be working with very defined groups (e.g. specific patient groups). Overall, these results indicate another key challenge: to explore and better understand the nature of 'the public' in public engagement with research and to develop interventions that support the development of meaningful research collaborations. [50] --- Conclusions: Enacting Engaged Research This study is part of a wider action research project with the overarching aim of embedding public engagement with research into the Open University's research culture. We have used the findings from this study, in combination with data from a second strand of research using semi-structured interviews (not reported in this paper), to shape and inform a series of interventions designed to achieve this aim. It follows that the findings, which have been discussed with senior executives at the institution, [51] our Advisory Panel, [52] the other Catalyst for PER projects, and commented on by our funders, [53] have been instrumental in identifying priority areas for intervention. Furthermore, they have been used to support professional development opportunities for researchers at all levels from postgraduate researchers to professors. [54] Here we describe three key interventions: defining engaged research, assessing and showcasing excellence, and creating a culture of reflective practice through evaluation and the development of learning resources. --- Co-producing a definition of engaged research In our discussion of researchers' definitions of public engagement with research, we noted some of the confusion between dissemination, dialogue and collaboration. We also offered evidence of the wide range and types of activities and publics that researchers considered to be engagement, noting the diversity both in approaches and publics. Our solution was to collaboratively produce, with researchers and Senior Executives, a definition of engaged research that could shape and inform future strategy and practice, based on evidence of the various ways that researchers and research teams from across a wide range of academic disciplines are already interacting with various kinds of 'public': Engaged research encompasses the different ways that researchers meaningfully interact with various stakeholders over any or all stages of a research process, from issue formulation, the production or co-creation of new knowledge, to knowledge evaluation and dissemination. This definition was discussed, revised and approved by the Open University's Research Committee in July 2014 [55] and subsequently approved by Senate in November 2014). It is an intervention designed to catalyse change. It has been developed in discussion with academics from across the institution, with the intention that it can be applied across all academic domains. It extends respondents' characterisation of public engagement as predominantly a communication activity and something that is presented after the research has been completed. It addresses the challenges we discussed in the Introduction: first, the practical challenges-the who, when, how and why of engagement; second, the valuing of stakeholders' expertise; third, the development of identity within the research process; fourth, the ownership of knowledge; and fifth, the interaction between researchers and stakeholders. (In this definition, stakeholders may include user communities, and members of the public or groups who come into existence or develop an identity in relationship to the research process.) Co-creating this definition with academics, whilst also being informed by research, has enabled us to begin to clarify the rationales and opportunities for broadening and deepening future engagement and in particular to address the confusion about the different ways that the 'impact agenda' [27] of innovation, enterprise, knowledge transfer and knowledge exchange connects with public engagement with research. We argue that this concept of engaged research is useful because it provides a basis for exploring the mechanisms by which a wide variety of economic and academic benefits, as well as social benefits, can be produced throughout

Notwithstanding that 'public engagement' is conceptualised differently internationally and in different academic disciplines, higher education institutions largely accept the importance of public engagement with research. However, there is limited evidence on how researchers conceptualise engagement, their views on what constitutes engagement and the communities they would (or would not) like to engage with. This paper presents the results of a survey of researchers in the Open University that sought to gather data to fill these gaps. This research was part of an action research project designed to embed engagement in the routine practices of researchers at all levels. The findings indicate that researchers have a relatively narrow view of public engagement with research and the communities with which they interact. It also identified that very few strategically evaluate their public engagement activities. We conclude by discussing some of the interventions we have introduced with the aim of broadening and deepening future researcher engagement.

all academic domains. It extends respondents' characterisation of public engagement as predominantly a communication activity and something that is presented after the research has been completed. It addresses the challenges we discussed in the Introduction: first, the practical challenges-the who, when, how and why of engagement; second, the valuing of stakeholders' expertise; third, the development of identity within the research process; fourth, the ownership of knowledge; and fifth, the interaction between researchers and stakeholders. (In this definition, stakeholders may include user communities, and members of the public or groups who come into existence or develop an identity in relationship to the research process.) Co-creating this definition with academics, whilst also being informed by research, has enabled us to begin to clarify the rationales and opportunities for broadening and deepening future engagement and in particular to address the confusion about the different ways that the 'impact agenda' [27] of innovation, enterprise, knowledge transfer and knowledge exchange connects with public engagement with research. We argue that this concept of engaged research is useful because it provides a basis for exploring the mechanisms by which a wide variety of economic and academic benefits, as well as social benefits, can be produced throughout the research process. It shifts the focus from assessing the benefits that flow from completed academic research to considering how the boundaries of academic research practice can become more permeable to participation and partnership working by people and agencies that have not traditionally been considered part of the research community. This participation can be linked to achieving various kinds of impact over time but can also usefully be considered as providing value in its own right, since the methodologies that are used to generate impact can be assessed whether or not impacts are ultimately achieved. --- Assessing quality and showcasing excellence in engaged research Gaining institutional approval for this research-informed definition has also allowed us to explore the nature and purposes of engaged research as they play out in different academic disciplines. In turn, this has given us licence to introduce other measures, including assessments of quality. Through our work we have encountered lingering echoes of the view that being involved in engagement might be bad for a career and cause researchers, or their work, to be taken less seriously by colleagues. [45] Whilst this perception of public engagement as an activity for those less fitted to an academic career was repudiated by, for example, Bentley and Kyvik, [56] who conducted a meta-analysis of scientists' activities in 13 countries, showing that researchers who participated in public engagement published, on average, significantly more academic publications than those who did not, we argue that there is still a job to be done to recognise and reward excellence in engaged research. At a strategic level, we have used evidence from the research we report here to inform contributions to an institutional review of promotion criteria; operationally, we have introduced an Engaging Research Awards Scheme. Conceptualising this Awards Scheme in the light of the considerable diversity we encountered throughout our research and as we supported researchers in their engagement planning called for clear definition of the types of activities and publics that the scheme would assess. We therefore begin the process of collaboratively producing the definition of engaged research (above). Having addressed the issue of definition, we were able to develop a set of criteria for assessing the submissions, further clarifying what counts as excellence in engaged research. The criteria can be found at http://www.open.ac.uk/blogs/per/wpcontent/uploads/2014/03/Engaging-Research-Awards-Scheme_assessment_protocol_2014. pdf. The winners of the inaugural scheme are listed on http://www.open.ac.uk/research/main/ news/ou-announces-winners-its-first-engaging-research-awards-scheme. --- Co-creating a culture of reflective practice Our approach to embedding engaged research within the culture of our institution has been to combine research with interventions, working across different academic disciplines and with researchers at all levels. Such an approach has strengths but also limitations, not least the degree to which the research findings and interventions can be directly linked. One can and should inform the other, but those running action research projects also have to take into account changing circumstances that are outside their control. In our case, as a small multi-disciplinary project, we have also had to judge when our interventions could be useful, and when to let our colleagues work with their stakeholders to progress their agenda. It follows that, ultimately, for engagement and an 'open research' approach to become genuinely embedded within the research culture of a large and diffuse organisational structure such as a university, researchers and their publics must come to value the processes of engaged research and create a culture of reflective practice and mutual respect as partnerships develop and shift over time. This demands a combination of generic resources and bespoke support, only some of which can be provided centrally from an action research project such as ours. To address these requirements, we worked with researchers throughout the research cycle, from developing grant applications, to managing engagement during projects, and finally to considering engagement that is sustainable beyond the projects' lifetimes. We are also developing open access resources as part of a learning programme, including a set that considers the issues and challenges facing researchers in developing digital engagement practices. Through our research and interventions we have found that very few researchers undertake formal evaluation of engaged research, which may be a reflection of the view that engagement activities are regarded as informal and personal and therefore either less amenable to evaluation or not needing to be evaluated. [31] Thus, most researchers assessed the success of their activities in terms of short-term, informal measures of success, such as receiving positive feedback at the event. However, researchers do offer reasons for undertaking engagement that-if they are to be valued-require effective evaluation. For example, enthusing future generations about science/ the arts/etc. was offered as a reason for engagement. However, without long-term evaluation it will be difficult to know whether such enthusing is successful. We have responded to the data on measuring the quality of engaged research, supporting a seed funding scheme for active researchers wanting to explore the generation and systematic collection of evidence of the impacts from engaged research, demonstrating effects, changes and/or mutual benefits to those participating. Three awards were made under this scheme; the projects will be completed by spring 2015. --- Final thoughts Scientists have widely 'acknowledged the benefits to scientists [emphasis in original] of communicating their work with the public', [57] (p.194) although data for researchers in other fields is scarcer. There is longitudinal evidence to support the view that the majority of scientists have a positive attitude to participating in engagement activities [44,45,46] but we have only partial and indirect knowledge of why publics engage. [58] Furthermore, for many UK academics the concept of public engagement with research has become intricately entangled with instrumental demands for economic and social impact. [27] The challenge for the higher education sector is to explore whether an action research-informed approach, such as the one we have outlined in this paper, can produce a more progressive agenda for engaged research, one that is meaningful and relevant for both researchers and publics. --- Data are available from the Open University's institutional research data catalogue http://oro.open.ac.uk/41364/. --- Author Contributions Analyzed the data: AG GD RH AA. Wrote the paper: AG GD RH AA. Designed the institution-specific survey questions: AG RH AA.

Notwithstanding that 'public engagement' is conceptualised differently internationally and in different academic disciplines, higher education institutions largely accept the importance of public engagement with research. However, there is limited evidence on how researchers conceptualise engagement, their views on what constitutes engagement and the communities they would (or would not) like to engage with. This paper presents the results of a survey of researchers in the Open University that sought to gather data to fill these gaps. This research was part of an action research project designed to embed engagement in the routine practices of researchers at all levels. The findings indicate that researchers have a relatively narrow view of public engagement with research and the communities with which they interact. It also identified that very few strategically evaluate their public engagement activities. We conclude by discussing some of the interventions we have introduced with the aim of broadening and deepening future researcher engagement.

Introduction There are significant racial disparities in breast cancer incidence, care, and outcomes, with Black women experiencing higher incidence of aggressive breast cancer subtypes, lower quality of care, and higher mortality rates than white women [1][2][3]. These disparities persist despite efforts to eliminate them. An increasing number of scholars and scientists identify systemic racism as the fundamental cause of health disparities [4][5][6]. This causal effect is exerted through numerous interrelated pathways, including residential segregation, economic deprivation, healthcare quality and access, social and environmental exposures, and environmentally conditioned health behavior (e.g., smoking). In order to be effective, efforts to prevent breast cancer and eliminate disparities must therefore acknowledge and work to address racism as a system. This approach differs from traditional intervention paradigms, which seek to isolate factors in a linear causal chain resulting in a single health outcome, in that it works from the assumptions that multiple interrelated causes can exert a multiplicative influence over and above the sum of their effects, that numerous outcomes are likely to trace back to the same interrelated set of causes, and that disruptions in one subsystem will likely not change outcomes. Focus is therefore shifted from finding a single cause to finding those subsystems, or nodes in the network of related causes, that mediate the relationships between numerous other subsystems and the outcome(s) of interest. Those component subsystems that are necessary for the maintenance of systemic equilibrium can be understood as leverage points [4]. With respect to health disparities generally, and breast cancer disparities specifically, residential racial segregation is one such leverage point. It has historically produced inequities in access to opportunities and resources [7]. Place of residence largely determines access to high quality education, which influences educational attainment and subsequent labor market opportunities [8]. Additionally, place of residence affects access to credit markets, housing, health care, and the quality and nature of interactions with law enforcement and the criminal justice system [4]. People of color in the United States continue to be concentrated to a great degree in communities with lower levels of resources and opportunities relative to white communities, with resulting inequities and disparities in outcomes associated with health and well-being. Residential racial segregation has facilitated inequitable access and exposure with regard to the social, economic, and environmental determinants of health. It has served as a fulcrum-a leverage point-in the race discrimination system, influencing the operation of every major subsystem, and producing racial inequities and disparities. In this article, we will outline an approach to harnessing the power of place, such that it may serve as a leverage point to amend, rather than contribute to racial disparities in breast cancer. We begin with an overview of Black/white racial breast cancer disparities, followed by a brief outline of the state of the scientific literature on individual risk factors for, and protective factors against, breast cancer, with particular focus on those risk and protective factors that disproportionately impact historically marginalized groups. We then discuss various potential mechanisms by which racial discrimination may exert a causal influence on breast cancer disparities by conditioning the population distribution of these risk and protective factors, with the caveat that these mechanisms are not yet well understood. In order to re-conceptualize this web of interrelated influences in a more actionable framework, we echo other scholars and scientists in identifying systemic racism as the underlying cause that links all of the interrelated mechanisms of disparity together. We use concepts from systems theory to demonstrate that residential segregation is a key leverage point in the racial discrimination system. We then put forth a systems-oriented, community-engaged approach to breast cancer prevention and disparities research and outline opportunities for researchers, policymakers, and interventionists to harness the power of place by partnering directly with community stakeholders in coordinating breast cancer prevention efforts in under-resourced minority communities. While a detailed description of Black/white racial breast cancer disparities is beyond the scope of this paper, there is a large body of scientific data indicating that breast cancer and its consequences are differentially distributed by race. Because these differences adversely impact marginalized communities, and are related to injustice, they are called health disparities. In our usage, therefore, the term disparity should be taken to mean more than simply difference, but a difference which has its source in inequity or injustice. Racial breast cancer disparities can be understood as falling into three broad categories: disparities in incidence, disparities in care, and disparities in outcome. Historically, Black women have had lower rates of breast cancer than white women. However, recent literature indicates that breast cancer incidence is rising among Black women, and in 2012, overall breast cancer incidence for Black women reached levels equaling those of White women [2]. Though this on its own does not illuminate an absolute Black/white breast cancer disparity, it does show a disparity in change over time wherein Black women do not share equally in improvements occurring across the broader population. Further, in a study of 375,761 cases of breast cancer, Black women were more likely than white women to be diagnosed with triple-negative breast cancer, which is more aggressive and difficult to treat than more common breast cancer subtypes [9]. In that same study, the overall incidence rate for breast cancer among women less than 44 years old was highest among Black women. Compared with white women, Black women disproportionately experience longer waiting periods after abnormal screening, treatment delays, lower likelihood of receiving guideline-concordant cancer care [1], lower quality of care [10], and early treatment termination [11], which has been linked to negative outcomes. Though mortality rates for breast cancer have steadily decreased in recent decades, outcomes have improved less for Black women than for white women [12]. For sex-specific cancers, including breast cancer, racial disparities in survival have been shown in a randomized trial of nearly 20,000 subjects to remain significant when controlling for other demographic factors, as well as clinical and treatment variables [13]. In fact, as breast cancer incidence among Black women has increased, mortality rates for breast cancer have also continued to be higher among Black women than among any other racial group [3,14]. In 2012, breast cancer mortality among black women was 42% higher than it was among white women [2]. --- The Fundamental Causes of Breast Cancer Disparities Breast cancer disparities persist despite the development and implementation of numerous interventions seeking to eliminate them [15]. There are several comprehensive and wide-ranging reviews of the current state of scientific research on factors that contribute to breast cancer incidence, outcomes, and disparities [3,16,17]. Overall, this literature indicates that risk and protective factors are distributed differently by race, with racial minorities more likely than whites to be exposed to many risk factors. Specifically, it points to the need for prevention efforts aimed at reducing alcohol consumption, tobacco use, and chronic stress, as well as chemical, occupational, and ionizing radiation exposures, and promoting breastfeeding, healthy diet patterns, physical activity, and healthy vitamin D levels [3,16,17]. Though our ultimate aim is to highlight the complex interrelationships between the environmental determinants of breast cancer outcomes, we will begin with a brief discussion of some individual risk and protective factors that are differentially distributed by race. We will then turn our attention to the ways in which racial residential segregation ensures the inequitable distribution of these determinants at the population level. --- Risk and Protective Factors for Breast Cancer Alcohol is a known risk factor for breast cancer, and there is a dose response relationship between the two, with higher consumption corresponding to greater breast cancer risk [18]. Though there are some conflicting results, there is limited evidence that high levels of alcohol consumption are associated with greater breast cancer risk among Black women than for any other racial group, and that higher levels of alcohol consumption increase risk of more aggressive subtypes of breast cancer, which also disproportionately impact Black women [19,20]. The carcinogenic properties of tobacco smoke are well known. Though smoking is more strongly associated with cancers of the mouth, throat, and lungs, it is also a risk factor for breast cancer, particularly among women who started smoking at a young age or before carrying a pregnancy to term [16,17]. Smoking is a risk factor for luminal breast cancer in particular, and there is some evidence that that risk is especially elevated for Black women [21]. Further, another study found that though race and class were not a predictor of tobacco spending, vulnerable populations may experience worse breast cancer outcomes attributed to tobacco addiction than others [22]. Environmental pollutants and toxic exposures account for between 7% and 19% of the world's cancer cases [16,23]. Exposures to toxic substances in the air, in the water, and through industrial agricultural pesticides contribute significantly to those figures, and there is burgeoning evidence that people of color are disproportionately impacted by those exposures [24]. There is a broad array of chemical compounds in common household products that are of concern for breast cancer preventionists, including: bisphenol A (BPA); heavy metals, such as cadmium; polybrominated diphenyl ethers (PBDEs, AKA flame retardants); phthalates; alkylphenols; per-and polyfluoroalkyls (PFAs); pesticides; herbicides; solvents; aromatic amines; and parabens [16]. There are various mechanisms by which these compounds may increase breast cancer risk, including endocrine disruption, promotion of tumor growth, and adverse epigenetic impacts. There is an emerging body of evidence to suggest that people of color are exposed to carcinogens in consumer products more frequently and in higher doses than whites [25,26]. Ionizing radiation is a risk factor for various cancers, including breast cancer, and a common mechanism of exposure is in medical imaging, such as CT scans and X-rays [16]. One area of potential concern regarding racial disparities in breast cancer incidence is that, though Black women are less likely to receive testing for BRCA1/2 mutations, exposure to ionizing radiation through medical diagnostics was found in at least one study to increase risk of breast cancer by 90% among women with that mutation [27]. Occupational factors may relate with breast cancer incidence and outcomes through numerous possible mechanisms, including: "chemical exposure; stress, including around job security and fair wages, threats or acts of sexual and physical violence, and lack of power to advocate for oneself; challenges with time and accommodation for breastfeeding; light-at-night exposure; and many other issues" [16] (p. 204). These factors are of particular concern for women of color, who likely experience added stressors related to discrimination which may exacerbate their deleterious effects [28,29]. Though the biological mechanisms linking breast cancer and diet are not well understood, the extant literature indicates that diet can function both as a risk factor and as a protective factor for breast cancer. Though some conflicting findings complicate the picture, the bulk of relevant studies show a positive association between consumption of red and processed meat and breast cancer [30][31][32]. A recent meta-analysis also found a positive association between adolescent fat intake and breast cancer risk later in life [33]. On the other hand, the preponderance of scientific evidence suggests that healthy eating habits, such as the Mediterranean diet, have a protective effect against breast cancer, and improve overall health [16]. Due to the relatively lower quality of food sources in lowincome communities of color, these diet-related risk factors may contribute significantly to disparities [34]. A recent meta-analysis including 13,907 breast cancer cases from 27 studies on the relationship between breastfeeding and breast cancer found that there was an inverse association between the two, with longer breastfeeding duration related with lower relative risk compared with shorter duration (RR = 0.471, 95% CI, 0.368-0.602) [35]. At the same time, one California based study found that Black women are less likely to exclusively breastfeed their infants than white women [36]. This may be related to the fact that Black women are more likely to be employed in occupations with little to no flexibility or support services for new mothers, such as paid family leave, or experience other structural impediments related to discrimination that make breastfeeding more difficult. Vigorous physical activity is another likely protective factor against breast cancer [15,16]. These effects have been observed in studies that focused on Black women specifically [37,38]. There are reasons to believe, however, that women of color may have lower levels of physical activity, and that this disparity in activity can be explained by the fact that they disproportionately experience structural and logistical barriers to physical activity, such as unsafe neighborhoods and a lack of access to greenspace [39][40][41]. There is some evidence to suggest that vitamin D deficiency may be related to breast cancer disparities. In one study, low levels of vitamin D were associated with an elevated risk of breast cancer of 23% in Black women [42]. Black women are at elevated risk of triple negative breast cancer, a particularly aggressive subtype, and are 10 times more likely to experience vitamin D deficiency than white women. This has led some researchers to suggest that some of that elevated risk may be attributable to vitamin D deficiency [43]. --- The Role of Social Determinants Williams et al. note that though both racial discrimination and racial breast cancer disparities are well documented, the mechanisms that link discrimination to disparities are not well understood [3]. There are numerous plausible ways to begin tying the individual risk and protective factors together with discrimination into a coherent theoretical framework of racial breast cancer disparities. We briefly outline and assess several possibilities below. Though they are presented separately here, the causal mechanisms are not truly separable. In reality, they are deeply interconnected. At the individual level, the positive association between low SES and disease does not generally hold for breast cancer, but striking patterns emerge when examining incidence, severity of breast cancer subtype, race/ethnicity, and SES together [3,44,45]. Discrimination can be understood as a stressor, and there is burgeoning research examining the impact of this specific type of stress as it relates to overall health. For example, experiences of discrimination have been linked to poorer health, and higher levels of biological and behavioral indicators of disease risk, such as inflammation, obesity, and smoking [46,47]. Further, there is a growing body of literature demonstrating the neurobiological impacts of stress and trauma in early life, including methylation of the nuclear receptor 3C1, "the gene that codes for the glucocorticoid receptor on the hypothalamic-pituitary-adrenal axis" [3] (p. 2141). Increased methylation in this nuclear receptor "represents a unique record of past adverse psychosocial experience," and has been shown to be associated with numerous negative health outcomes, including breast cancer [3] (p. 2141). Further, Geronimus et al. put forth the concept of weathering to encapsulate the effect of chronic stressors experienced by Black women in America [28,29]. One example of this effect is reflected in a process of accelerated aging. By measuring telomere length, Geronimus and colleagues were able to determine that, compared with white women, Black women in their sample were biologically 7.5 years older on average. Another key concept related to stress across the life course is allostatic load, a biological measure of overall wear and tear on the body due to stress. As Williams et al. put it: "the concept of allostatic load has been used to capture the biological dysregulation across multiple physiological systems that result from the cumulative burden of repeated stressors" [3] (p. 2141). Allostatic load has been found in multiple studies to be higher in people of color than among whites, and to be associated with poor health [29]. Place of residence impacts breast cancer risk significantly. In a recent systematic review of 17 studies focusing on racial residential segregation and cancer disparities, 70% of studies showed a statistically significant association between segregation and health disparities. The authors state: "residing in segregated African-American areas was associated with higher odds of later-stage diagnosis of breast and lung cancers, higher mortality rates and lower survival rates from breast and lung cancers, and higher cumulative cancer risks associated with exposure to ambient air toxins" [48] (p. 1195). Racial residential segregation also ensures that place-based risk factors are inequitably distributed by race. Given that individual and environmental exposures to breast cancer risk factors are impacted by place of residence, it is plausible that the disparities in incidence, care, and outcomes outlined by Landrine et al. are the result of the differential distribution of these exposures between segregated neighborhoods [48]. Again, though these myriad factors are presented separately above, the upstream contributors to breast cancer disparities are not actually separable in reality. The treatment above is not nearly complete, as it only alludes to a system of racial inequity, discrimination, and disparity that has deep historical roots and implications for nearly every domain in which health and welfare can be measured. Though there are many plausible mechanisms by which racial breast cancer disparities may arise, empirical investigation into those mechanism is in its early stages. For example, questions remain regarding the linkages between socioeconomic factors such as labor market opportunity, educational access, and credit access, indicators of health risk such as high allostatic load, and breast cancer risk. It is clear, however, that the impacts of both historical and ongoing racial discrimination are pervasive and highly interrelated. The high degree of interrelation between possible causes complicates matters for researchers seeking a parsimonious explanatory model, and interventionists seeking a small number of variables on which to focus programs. A clearer understanding of these causes and their interrelationships could aid in efforts to prevent breast cancer, and reduce disparities. --- How Systems Thinking Can Identify Leverage Points Arising out of the biological sciences, with applications across a broad range of fields, including economics, bioinformatics, meteorology, and others, systems theory is designed to model complex, interrelated networks of component subsystems. It may therefore prove useful in developing conceptual models to explain the complex and highly interrelated causes of breast cancer disparities, and to clarify why those disparities have thus far proven resistant to intervention. Most discussions of breast cancer disparities and other health inequities are confined to the healthcare or public health systems, and stop short of articulating the broader, more fundamental problem of systemic racism that contributes to inequities across a diverse array of outcomes. In the following section, we will outline a framework for research and intervention in breast cancer disparities that is explicitly informed by an acknowledgement of the interlinked nature of disparities across multiple domains, including large-scale policy frameworks, local-level social and environmental factors, and the biological sequelae of discrimination [4]. Much like racial breast cancer disparities themselves, the biological and social contributors underlying those disparities do not arise in a vacuum, but out of a long history of systemic racism and discrimination. There are significant disparities between Black and white Americans in the criminal justice system, including policing practices, incarceration, and jury participation, which have been linked not only to explicit legal discrimination, but also to implicit bias [49]. Residential segregation, which persists today despite no longer having the force of law, is a significant driver of negative outcomes across numerous domains, including health, through its impact on educational access, employment opportunities, and exposure to environmental hazards [7]. Systemic racism, and racial residential segregation in particular, also ensure the inequitable distribution of each of the risk and protective factors for breast cancer outlined above. For example, place of residence impacts the quality of available foods, exposure to environmental toxins through residential and occupational sources, likelihood of tobacco and alcohol use, and opportunities for physical activity (through access to safe outdoor spaces) [15,22,35,[40][41][42][50][51][52]. Further, though socioeconomic status, education, and employment status are each associated with both race and health outcomes, racial disparities in health outcomes persist even among Black Americans who are high SES, highly educated, and gainfully employed [53]. This suggests that racism contributes independently to health disparities over and above the effects of these mediators. There is increasing scholarly and scientific consensus that racism itself is a fundamental cause of health disparities. Yearby argues that the social determinants of health framework articulated in Healthy People 2020 is insufficient precisely because it does not give a primary place to racism as an upstream factor contributing to all other social determinants [6]. Williams et al. provide an overview of the social context of breast cancer disparities among Black women, outlining the burgeoning body of research on the biological effects of discrimination which ultimately lead to elevated breast cancer risk [3]. Reskin gives a broad outline of racial discrimination in American society, and makes a compelling case that each of the subsystems in which disparities exist are connected by the broader system of racism within which they are situated [4]. According to her model, racial disparities in outcomes across numerous domains are the result of a single, integrated system of racial discrimination. She argues that ad hoc, sector-specific interventions to reduce disparities will likely fail to disrupt the systemic equilibrium that maintains inequity due to the systemic property of robustness. This literature indicates that, in order for racial breast cancer disparities to be eliminated, researchers, policymakers, and interventionists must address racism as a system. In the following section, we will introduce key concepts from systems theory, and discuss ways in which they can be used as tools to reconceptualize racial breast cancer disparities in light of the system of racial discrimination within which they arise. --- Emergence Established systems exhibit the property of emergence. In a network of interrelated subsystems, the system as a whole will begin to exert an influence over and above the sum of the additive effects of those component subsystems. As Reskin explains, within the race discrimination system, this means that, in addition to the effects of individual experiences of discrimination and explicitly traceable instances of institutional racial bias, the system of racial discrimination itself exerts influence in maintaining the equilibrium of racial disparity [4]. In other words, the interrelationships, or feedback loops, between factors like Adverse Childhood Experiences (ACEs) and childhood poverty, environmental exposures, and healthcare quality and access, produce effects all of their own, independent of those exerted by each subsystem individually. With regard to breast cancer specifically, it is possible that the interactive effects between tobacco use, poor diet, and low physical activity result in a greater increase in breast cancer risk than the additive effects of each risk factor would indicate. These behavioral risk factors may also interact with the effects of ACEs to further increase breast cancer risk. Given the disproportionate impact of each of these environmentally conditioned risk factors on Black women, breast cancer disparities between Black and white women may be understood as an emergent effect of systemic racism, operationalized through racial residential segregation. Such emergent effects are generally robust to changes in a given subsystem. --- Robustness Established systems exhibit the property of robustness. This means that systems tend toward a state of equilibrium, and that that state of equilibrium will generally persist even in the face of significant disruptions within individual subsystems. The emergent effects of the system as a whole will ensure that the equilibrium is reasserted. In the case of racial disparities, this may explain the resistance of disparities to targeted interventions [4]. For example, a bias-reduction intervention targeting the healthcare subsystem that does not address risk factors related to chemical exposure, physical activity, or ACEs may not disrupt systemic equilibrium sufficiently to significantly reduce disparities. This is, in part, due to the fact that each of these subsystems is related to the others in complex ways. --- Complexity In systems theory, complexity is a formal designation which refers to a system in which there are a large number of component subsystems, and a large degree of interrelatedness between those component subsystems [54]. In the study of racial disparities in breast cancer, there are several layers of complexity: (1) the complexity of the problem itself; (2) the complexity of the interconnected web of causes that contribute to the problem; and (3) the complexity of the tasks associated with researching and intervening on the problem. For example, as mentioned above, behavioral risk factors, such as smoking, diet, and physical activity, may interact with one another in producing breast cancer risk. At the same time, they may also interact with social environmental exposures (i.e., violence, ACEs), or residential and occupational exposures to endocrine disrupting substances. Each of these layers of complexity contributes to breast cancer risk and to the intransigence of disparity. As mentioned above, complexity can pose a challenge for researchers and for interventionists, as parsimonious explanatory models and predictable results are difficult to achieve. --- Leverage Points In the context of complex systems, a full representation of all subsystems and all their interrelationships would yield a model insufficiently parsimonious for practical use. We propose that researchers, policymakers, and interventionists should conduct a critical review of the complex relationships between upstream causes of breast cancer and racial breast cancer disparities to identify systemic leverage points. Drawing on Reskin, we define leverage points as systemic nodes (or subsystems) where multiple other subsystems intersect, modifications in which have the potential to exert significantly amplified effects across the entire system [4]. As long as interventions to reduce racial breast cancer disparities confine themselves to only one or two subsystems within the broader context of racial inequity, the systemic quality of robustness will likely result in the reassertion of equilibrium, and the persistence of disparity. In contrast, intervention approaches to racial breast cancer disparities that explicitly address leverage points have the potential to impact numerous subsystems simultaneously and disrupt that equilibrium. Thus, leverage points should be understood as ideal areas of intervention and research. --- Place as a Leverage Point in The System of Racial Discrimination As Williams et al. state, "Residential segregation has been identified as a leverage point or fundamental causal mechanism by which institutional racism creates and sustains racial economic inequities" [5] (p. 117). Despite much intervention, residential racial segregation continues to be pervasive in the US. It should be noted, however, that racially homogenous communities are not innately problematic. Rather, the problem is that segregation is driven by racism. Racism has been defined as, "an organized societal system in which the dominant racial group uses its power to devalue, disempower, and differentially allocate societal resources and opportunities to groups defined as inferior" [55] (p. 2). For example, the Plessy v. Ferguson decision is correctly derided for upholding the constitutionality of segregation, and for maintaining the fiction that there could be "separate but equal" facilities for Black and white Americans. This decision was unjust precisely because of the reality that communities in which Black Americans lived, and their accompanying infrastructure, suffered from radical, intergenerational economic deprivation and social oppression. The consequences of that deprivation persist to this day, and they include disparities in health. Given that residential segregation has facilitated the production of disparate health outcomes, "place" is of central importance in shaping health trajectories of individuals and communities. The relationship between place, race, and breast cancer is already the subject of attention from health scientists and interventionists. State level authorities and researchers are actively working to understand and address racism, social and environmental exposures, and inequities in community infrastructure, based on the understanding that each of these represent critical upstream factors that condition the distribution of an array of other risk factors for breast cancer [16,56]. In fact, racism has significant detrimental effects on economic development, physical infrastructure, and the social environment in communities of color. Therefore, programs that address residential segregation and identify place as a key leverage point will have the advantage of addressing each of these areas at once. Experiences of discrimination occur within the social and spatial context of the communities and workplaces within which people of color live, and these communities and workplaces disproportionately play host to the many risk factors related to social and physical environments outlined above. We argue that place is a systemic leverage point through which numerous other subsystems exert a deleterious effect on communities of color, and contribute to breast cancer disparities. Though this framing is by no means comprehensive, it does suggest that the effectiveness of efforts to reduce disparities could potentially be maximized through a focus on place, since it is a modifiable point at which all of the other systems intersect. --- Directions for Future Research There are three major areas of research that could help further the work of harnessing the power of place to reduce racial disparities in breast cancer. First, research is needed that traces the full causal pathway between the environment (both social and physical) and breast cancer outcomes. Second, further research is needed in the design and im-plementation of community-based (or place-based) intervention frameworks to address the upstream factors that contribute to racial breast cancer disparities in residentially segregated communities. Third, research should be conducted to explore the relationships between racial discrimination and other forms of discrimination as they pertain to breast cancer disparities. Though a significant and growing body of literature traces the associations between numerous social, environmental, and biological factors and breast cancer incidence, the causal mechanisms that underlie the observed associations are often not well understood. For instance, though studies near uniformly show a positive association between breast density and breast cancer risk, the biological mechanism through which this association is operationalized is unknown [17]. It is possible that there is a confounding effect exerted by environmental factors associated with both breast density and breast cancer risk. Further, between and among the broad array of factors that are known to be associated with breast cancer risk, there are complex interrelationships that make rigorous assessment of a single causal mechanism extremely difficult. For example, communities of color tend to have lower SES than white communities. In those communities, risks of exposure to carcinogenic substances tends to be elevated, access to safe outdoor spaces conducive to physical activity is more limited, and quality foods are harder to come by [16,25,35,[40][41][42]. There is a need to better understand how social factors, such as explicit and implicit discrimination, SES, occupational exposures, and access to nutrition, interact with factors in the built environment, such as chemical and other environmental exposures, light at night, and accessibility of green space, to exert influence on human physiology in ways that elevate breast cancer risk [16]. In turn, we must explore further how racial differences in the distribution of allostatic load, adverse epigenetic changes in the HPA axis, and endocrine disruption stemming from environmental stressors and exposures relate to racial disparities in breast cancer incidence and outcomes [5,49,50]. Given the elevated impact of certain environmental exposures on breast cancer risk during critical developmental phases early in life, intervening in the social and built environment is particularly promising for breast cancer prevention over the long term [16]. However, empirical investigation is needed to explore which environmental exposures drive breast cancer disparities, and what intervention approaches would be effective in reducing the risk they represent. Overall, more research is urgently needed on the ways in which social factors, such as poverty and violence, interact with correlated environmental exposures, such as industrial pollutants and residential chemical exposures, to influence breast cancer incidence in women of color. Tracing the full causal pathway from environment to incidence could inform the development of targeted multilevel programs to disrupt the equilibrium of the system of racial disparity. According to Williams et al., "we lack the empirical evidence to identify which mechanisms of segregation (e.g., educational opportunity, labor market, housing quality) should be tackled first, would have the largest impact, and is most likely to trigger ripple effects to other pathways" [5] (p. 117). Several types of research endeavors could further this work, and systems thinking can be applied at numerous levels. For example, though we argue in this paper that, in the broader system of racial breast cancer disparities, "place" functions as a leverage point, it is possible to use the same process to identify other leverage points. Further, if "place" is selected as the key leverage point, and a specific community of practice is identified, there are likely further leverage points that should be identified within that community which function as highly interconnected nodes through which multiple subsystems exert influence on breast cancer disparities. These community-based leverage points may be community centers, health centers, the transportation system, the labor market, municipal authorities, or any number of other possibilities. Identifying these leverage points will require significant input from community stakeholders who can act as experts on the specific strengths and challenges of that community. One possible area of research would therefore engage community stakeholders in the development of community-specific concept maps of key systemic interrelationships to identify leverage points, whereas another would pilot interventions informed by these maps to coordinate community efforts toward addressing those leverage points. When promising intervention frameworks are available, larger implementation studies should be carried out to determine the most effective strategies to ensure program adoption, fidelity, sustainment, and effectiveness [57]. Overall, a robust body of literature is needed that explores the complex interrelationships of subsystems contributing to breast cancer disparities, and tests the efficacy of place-based interventions to address key leverage points. Though the focus of this discussion has been on racial discrimination as a determinant of breast cancer disparities, a third area of needed research would explore the intersectional relationships between racial discrimination and other forms of discrimination. For example, gender discrimination has been shown to be negatively associated with mammography screening, and it is plausible that this reduction in screening may disproportionately impact Black women [58]. Therefore, further research should be conducted exploring the intersectional impacts of racial and gender discrimination, and particularly the possible moderating impact of gender discrimination on the relationship between racism and breast cancer screening behaviors. This research may harness the power of place through community engagement approaches similar to those outlined under research areas one and two above. --- Application to Breast Cancer Prevention Eliminating racial breast cancer disparities will require significant community investment and coordinated efforts across multiple systems to prevent breast cancer in communities of color. The upstream causes of disparities are tied to place, and these efforts will need to be guided by community stakeholders to address community-specific concerns. In this paper, we provide a conceptual framework for one approach to harness the power of place to reduce breast cancer disparities. The potential applications of systems theory, and related leverage points, to breast cancer prevention are numerous. Traditional approaches tend to ignore these interrelationships in favor of linear models that are perceived to be more parsimonious and actionable. However, they often obfuscate the reality that emergent, systemic effects maintain the equilibrium of disparity, even in the face of significant investment in interventions targeting subsystems. Given that the various subsystems that make up the causal foundations of racial breast cancer disparities are spread across a diverse array of interrelated domains of practice and expertise, we argue that efforts to reduce these disparities should: (1) engage stakeholders across multiple sectors (subsystems), especially community members; (2) be informed by appropriate concept maps of the complex systems at play in the specific community; and (3) represent a coordinated effort across these sectors to address key leverage points at which multiple subsystems intersect. If we are to offset the negative effects of systemic racism to reduce and ultimately eliminate racial inequities in health broadly, and breast cancer specifically, there must be a greater focus on creating "communities of opportunity" [55]. This term describes "the transformation of local communities (that had been historically disadvantaged because of racism and its related systematic under-investments), into places that provide opportunities in education, labor markets, housing

Racial disparities in breast cancer present a vexing and complex challenge for public health. A diverse array of factors contributes to disparities in breast cancer incidence and outcomes, and, thus far, efforts to improve racial equity have yielded mixed results. Systems theory offers a model that is well-suited to addressing complex issues. In particular, the concept of a systemic leverage point offers a clue that may assist researchers, policymakers, and interventionists in formulating innovative and comprehensive approaches to eliminating racial disparities in breast cancer. Naming systemic racism as a fundamental cause of disparities, we use systems theory to identify residential segregation as a key leverage point and a driver of racial inequities across the social, economic, and environmental determinants of health. We call on researchers, policymakers, and interventionists to use a systems-informed, community-based participatory approach, aimed at harnessing the power of place, to engage directly with community stakeholders in coordinating efforts to prevent breast cancer, and work toward eliminating disparities in communities of color.

linear models that are perceived to be more parsimonious and actionable. However, they often obfuscate the reality that emergent, systemic effects maintain the equilibrium of disparity, even in the face of significant investment in interventions targeting subsystems. Given that the various subsystems that make up the causal foundations of racial breast cancer disparities are spread across a diverse array of interrelated domains of practice and expertise, we argue that efforts to reduce these disparities should: (1) engage stakeholders across multiple sectors (subsystems), especially community members; (2) be informed by appropriate concept maps of the complex systems at play in the specific community; and (3) represent a coordinated effort across these sectors to address key leverage points at which multiple subsystems intersect. If we are to offset the negative effects of systemic racism to reduce and ultimately eliminate racial inequities in health broadly, and breast cancer specifically, there must be a greater focus on creating "communities of opportunity" [55]. This term describes "the transformation of local communities (that had been historically disadvantaged because of racism and its related systematic under-investments), into places that provide opportunities in education, labor markets, housing markets, credit markets, health care and all other domains that drive well-being" [55] (p. 3). If investments are made to restructure place to provide equitable access to opportunities and resources, outcomes related to health and well-being will improve, disparities will be reduced, and a positive trajectory toward equitable outcomes can be established. An effective place-based initiative should focus on changing the systems to prioritize the provision of opportunities and resources that shape health and well-being to better serve the population. Such initiatives may include efforts aimed at early childhood development, improving neighborhood environments and housing conditions, reducing childhood poverty, enhancing income and employment opportunities, and access to high quality health care. Though any successful effort will require significant community investment, the costs should be offset by the fact that interventions on the upstream factors contributing to breast cancer disparities are likely to have a positive impact on health and welfare across multiple other domains, including cardiovascular and mental health, crime and violence, and overall quality of life. It is possible, through a process of strategic and sustained investment in historically marginalized communities, that intergenerational deprivation may be permanently eliminated, and the resultant disparities significantly reduced. Finally, it should be noted that marginalized communities are not monolithic, and there is significant heterogeneity both between and among the various communities who experience health disparities. Taking a place-based, community-engaged approach to breast cancer prevention is therefore advantageous, given that it will allow for the development of research questions and intervention strategies that address the specific strengths and challenges of each community. Though much of the focus in this essay has been on racial breast cancer disparities between Black and white women, the methods outlined here have significant applicability in the task of eliminating disparities in Latinx, indigenous, and other racial/ethnic populations. --- Conflicts of Interest: The authors declare no conflict of interest.

Racial disparities in breast cancer present a vexing and complex challenge for public health. A diverse array of factors contributes to disparities in breast cancer incidence and outcomes, and, thus far, efforts to improve racial equity have yielded mixed results. Systems theory offers a model that is well-suited to addressing complex issues. In particular, the concept of a systemic leverage point offers a clue that may assist researchers, policymakers, and interventionists in formulating innovative and comprehensive approaches to eliminating racial disparities in breast cancer. Naming systemic racism as a fundamental cause of disparities, we use systems theory to identify residential segregation as a key leverage point and a driver of racial inequities across the social, economic, and environmental determinants of health. We call on researchers, policymakers, and interventionists to use a systems-informed, community-based participatory approach, aimed at harnessing the power of place, to engage directly with community stakeholders in coordinating efforts to prevent breast cancer, and work toward eliminating disparities in communities of color.

Introduction Sleep has been recognized as an important health behavior associated with health outcomes. Research indicates the mean-predicted 10-year cardiovascular risk to be lowest among adults who reported sleeping 7 h per night and increased as participants reported sleeping fewer and more hours [1]. In a recent synthesis of reviews, the most favorable associations between sleep duration and health outcomes follow U-shaped dose-response curves with the lowest risk typically observed at approximately 7-8 h per day [2]. Despite the concluding recommendation of 7-8 h/day of sleep for adults, it is important to recognize an observed large inter-individual variability that may affect how the recommendations apply to a given person. National Sleep Foundation recommendations, which are evidence-based and consensus-driven [3], corroborate this U-shaped dose-response with elevated sleep insufficiency risks at <unk> 7 h/day. Hence, insufficient sleep patterns, defined as fewer than 7 h/day of sleep, can increase risk of health problems and chronic conditions including cognitive issues, increased inflammation, diabetes, cardiovascular disease, and overall mortality [4][5][6][7][8][9]. Sleep has also been associated with cultural factors and experiences. Culture may be defined as the shared values, norms, and codes that collectively shape a group's beliefs, attitudes, and behavior through their interaction in and with their environments [10]. As a dynamic experience, acculturation may influence health behaviors that impact sleep patterns. In a narrative review of the literature, Airhihenbuwa and colleagues argue that sleep is a cultural experience with positive, negative, and unique elements that deserve additional attention and represents a dominant aspect of life that varies among diverse groups or peoples [10]. International sleep duration comparisons reveal substantial cross-cultural variability. For example, individuals in East Asian (especially Japanese) cultures tend to sleep less than individuals in Western cultures [11]. As a result, one might expect Japan to experience higher prevalence of health risks, chronic conditions, and lower longevity than, for example, North American countries (U.S. or Canada). Yet, this is not the case, as evidenced by various indicators such as obesity, diabetes, and life expectancy [12][13][14]. In fact, recent evidence suggests that despite sleeping significantly less than European Canadians, Japanese Canadians slept less efficiently, yet reported being less tired and having better health. Furthermore, acculturation effects were observed as, relative to European Canadians, Japanese Canadians showed weaker associations between sleep and physical health, yet Asian Canadians' sleep behaviors were similar to those of European Canadians [15]. Hence, despite serving fundamental biological functions, sleep behaviors may be influenced by both initial childhood culture as well as acculturation when assimilating to a new place. Many health statistics in the U.S. are based on surveillance measures tracked at the county or state level. Yet regional cultures may blur county or state borders. Woodard used historical research techniques to track and map "first settler effects" [16] and competing North American colonization streams [17]. He identified 11 major regional cultures on the continent north of the 22nd parallel, plus two smaller enclaves of regional cultures (i.e., Spanish Caribbean and Greater Polynesia) lying largely or entirely outside this space [18]. This American Nations model has been applied to explain differences in entrepreneurship [19], economic development [20], mortality [21], gender wage gaps [22], personality characteristics [23], gun violence [18], and voting behavior [24]. An in-depth description of the regional cultures, such as delineated by Woodard [16], is beyond the scope of this report. Instead, here we present the unique principles and identities of the 13 American Nations in Table 1 including some overarching observations related to these regional cultures. Historically, people who founded the settler-colonial societies in various areas of what is now the U.S. came from different parts of Europe and brought with them different cultures with distinct religious, ethnographic, ideological, economic and political characteristics. These colonial projects competed in colonial times, even when they belonged to the same (English and, later, British) empire. As outlined in Table 1, each region had their own fundamental principles, and these were often contradictory. A nation may be defined as a group of people who share the same culture. Based on this premise, the American Nations represent a group of 11 major nations and two smaller enclaves (which we recognize in this report as well). In our attempts to correlate regional cultures to healthrelated outcomes, we recently published the geographic distribution of unhealthy living characteristics according to the American Nations model and observed once again significant heterogeneity [25]. Characteristics described have included physical inactivity, obesity, diabetes, and adequate access to exercise opportunities, but not sleep. We hypothesize that the American Nations regional cultures will show substantial variance in sleep duration, with individualistic regions -where the neglect of public goods generates stress, poor health and socioeconomic outcomes and shortened life expectancy -exhibiting shortened sleep times. Hence, it is the purpose of this study to describe insufficient sleep patterns across 13 regional cultures that collectively make up the American Nations. --- Methods --- Data sources The most recently available U.S. data (2020) on the prevalence of insufficient sleep were obtained from the County Health Rankings & Roadmaps (CHRR) program of the University of Wisconsin Population Health Institute (https://www.countyhealthrankings.org/) [26]. The CHRR uses the Behavioral Risk Factor Surveillance System (BRFSS) (2020: https://www.cdc.gov/brfss/index. html) from the Centers for Disease Control and Prevention (CDC) as the primary source of insufficient sleep data. The BRFSS is a state-based random digit dial telephone survey that is conducted annually in all states, the District of Columbia, and U.S. territories. Data obtained from the BRFSS are representative of each state's total non-institutionalized population over 18 years of age and have included more than 400,000 annual respondents with landline telephones or cellphones since 2011. Data are weighted using iterative proportional fitting (also called "raking") methods to reflect population distributions. The CDC produced county estimates using singleyear BRFSS data and a multilevel modeling approach based on respondent answers and their age, sex, and race/ethnicity, combined with county-level poverty. For those counties where there were no or limited data, the modeling approach borrowed information from the entire BRFSS sample as well as Census Vintage population estimates, and parametric bootstrapping method produced standard errors and confidence intervals for those point estimates [17]. The CHRR program defines insufficient sleep as the percentage of adults who self-report that they sleep less than 7 h per night on average. The insufficient sleep data are age-adjusted. Age is a non-modifiable risk factor, and insufficient sleep is among the health behaviors that may be associated with different age groups in the population [2,3]. The CHRR program reports an age-adjusted rate to fairly compare counties with differing age structures. Using zip-code data, the CHRR database was linked to the American Nation dataset from the Nationhood Lab, Pell Center for International Relations and Public Policy, Table 1 The Identities of the 'American Nations' Salve Regina University (https://www.nationhoodlab. org/) [18]. Percentages for the insufficient sleep metric were then estimated using the population of each American Nation listed in Table 1. --- Statistical analysis Analysis of variance (ANOVA) was used to compare differences in insufficient sleep prevalence among the American Nations. When appropriate, follow-up post hoc analyses were conducted using Bonferroni multiple comparisons to determine significant mean differences judged at the p <unk> 0.05 level of significance. --- Study ethics The HealthPartners Institute Research Subjects Protection Program determined that this study is exempt from IRB review and ongoing oversight under 45 CFR Part 46 as it involves the analysis of existing, publicly available data sets. --- Results Table Statistical differences between the 13 American Nations were observed as indicated by a significant overall ANOVA test (F = 188.083 (df = 12), MS = 1478.271, p <unk> 0.001). Bonferroni post-hoc comparisons indicate statistical differences between multiple American Nations and are summarized in Table 2. These observations are supported by Fig. 1 which illustrates rounded mean data in a geographical map format in which the countylevel prevalence indicators are merged according to the regional clusters that form the American Nations. This Figure clearly shows a grouping of higher sleep insufficiency regional cultures that include New France, Deep South, Greater Appalachia, Spanish Caribbean, Tidewater, and New Netherland. Higher insufficient sleep prevalence is also noted for the Hawaiian region (i.e., Greater Polynesia). These ANOVA results indicate that differences between the 13 regional cultures that make up the American Nations show significant variability. Follow-up post hoc results indicate that regional cultures geographically positioned in the South-Eastern U.S., along the Appalachian mountain range, and the eastern seaboard south of New York report significantly higher sleep insufficiency as compared to other American Nations. Variation in sleep outcomes may indicate that regional cultures where community and social cohesion principles are supported tend to report lower levels of sleep insufficiency. However, despite significant differences among the various regional cultures, all American Nations exhibit substantial opportunity for improvement since the most favorable insufficient sleep indicator was 30.1% for Left Coast. --- Discussion The U.S. has experienced a sustained lack of improvement in population-based healthy longevity when it is compared to other countries [27]. Part of this trend is a continued lack of improving healthy living characteristics, such as physical activity, obesity, and nutrition-related factors. Sleep is another health-related behavior that has emerged as an important variable in health-related outcomes and overall longevity [1][2][3][4][5][6][7][8][9][10][11][12][13][14]. The data presented herein indicate that room for improvement of insufficient sleep exists across all regional cultures identified in the American Nations model and that some regions-such as those that include Greater Polynesia, New France, the Deep South, or Spanish Caribbean-may need to consider prioritization of sleep hygiene as a health concern. Furthermore, the insufficient sleep trend noted in this report is consistent with those reported earlier for physical inactivity, obesity, and diabetes [25]. Factors and circumstances centered around the Midwest and the South-Central states appear to be responsible for the observed patterns, an observation corroborated by data presented on life expectancy trends over the past century [27]. These observations are corroborated by earlier reports including an analysis at the county-level by Grandner and colleagues [28] on the geographic distribution of insufficient sleep across the U.S. Insufficient sleep hotspots were found in 84 counties spread across 12 states including Alabama, Arkansas, Georgia, Illinois, Kentucky, Louisiana, Missouri, Ohio, Tennessee, Texas, Virginia, and West Virginia. Collectively, these states are represented within the American Nations' regional cultures of the Deep South, Midlands, Greater Appalachia, Tidewater, and a single hotspot county within El Norte. The current investigation extends these findings beyond geographic variability into the realm of regional cultures. Knowledge of an underlying set of values and principles of a specific geographic regions may aid in the development of public health campaigns. As such, this report presents novel data that, in combination with county and state-level data on health, policy, and socioeconomic contexts, regional cultural differences may be useful and important in crafting powerful public health campaigns. Moving beyond a general description of the American Nations, we highlight Yankeedom and the Deep South as two examples of distinctly different regional cultures. Yankeedom was founded by Calvinists who settled the Massachusetts Bay colony and believed they had been chosen by God to bring into being about a more perfect society that would be a "light on a hill" to guide humanity. To fulfill their mission as a covenanted people, the Puritans emphasized education, local political control, the interests of the community (over that of individuals), and the establishment of strong institutions to guide their shared "errand in the wilderness. " It is a culture that imbued strong support and faith in government and other shared institutions understood to be extensions of the community and primary tools for the furtherment of its health and goals. Like the other regional cultures, Yankeedom spread across its own nearly exclusive settlement By contrast, the Deep South was founded by English slave lords from Barbados who transplanted their fully-formed West Indies stye slave plantation culture to the subtropical lowlands of a new colony on the North American mainland that was for a time called "Carolina in the West Indies. " This culture -which eventually spread across the majority of South Carolina, Georgia, Alabama, Mississippi and Florida and parts of North Carolina, Arkansas, Tennessee, and Texas -was organized to benefit the planter oligarchy, which shunned investments in public goods and taxes or regulations affecting land and capital. Ideologically they embraced the republics of classical Antiquity, where democracy was the privilege (or "liberty") of the few and subjugation and slavery the natural and necessary lot of the many. In the Antebellum period their leaders denounced the assertion of natural human equality in the Declaration of Independence and argued that the U.S. should be understood as a collection of Anglo-Saxon ethnostates protected by a federal umbrella. When their race-based slave system was destroyed in the 1860s they created an apartheid regime backed by legal and extralegal violence that endured for another century and into the living memory of millions of the region's residents today. These two regions -the superpowers of the intraregional political struggle that has formed the backdrop for so much of U.S. history since the 1830s -represent opposing preferences regarding social policy. Yankeedom is communitarian, relatively comfortable with strong government, working through institutions, the regulation of economic and social actors, and levying taxes to pay for stronger social services, public institutions, and social safety net protections. The Deep South is individualistic, seeking to maximize the economic and political autonomy of the region's "winners" by keeping taxation, regulations, social services, welfare provisions and the government weak. Much as Putnam [29] found in examining differences between regions in northern and southern Italy, these "American Nations" strongly differ in levels of social and civic trust and also in health outcomes, including per capita COVID-19 deaths [30], gun deaths [31], and life expectancy [32]. Greater Appalachia and, to a lesser extent, the Far West, are also individualistic sides of the spectrum (though for different reasons); New Netherland and Left Coast are reliably communitarian. As there is "no evidence to indicate that, at the population level, the need for sleep differs across ethnicity or countries, " prominent researchers have said they believe "that a significant portion of the variation in sleep patterns seen across countries is likely to be cultural in origin" [33]. County-level sleep deprivation hotspots in the U.S. have previously been shown to have higher proportions of "younger individuals of lower socioeconomic status and poorer health" and also to have high rates of obesity, diabetes, cardiovascular and lung diseases and several types of cancer" [28]. Our results in this present study show substantial variation in sleep duration, with individualistic regions -where the neglect of public goods generates stress, poor health and socioeconomic outcomes and shortened life expectancy -exhibiting shortened sleep times. Furthermore, we previously documented unhealthy living characteristics, including sleep patterns, across the U.S. as part of an effort to describe the syndemic relationship between unhealthy living behaviors, chronic diseases, and coronavirus disease 2019 (COVID-19) [34]. We presented data at the county or state level across the U.S. for COVID-19 death rates, physical inactivity, cigarette use, obesity, nutrition-related variables, sleep, and heart disease death rates and noted significant heterogeneity among the prevalence estimates. Interventions designed to increase the likelihood of adopting healthier behaviors may benefit from messaging and communications that align with the cultural predisposition of the population. When regional cultures blur state and county lines, an urgent need may exist to (re-)conceptualize and (re)design messaging and health promoting interventions according to regions where cultural disposition is similar. The map presented in this brief report provides a novel approach to demonstrating the geographical considerations of insufficient sleep as a population and public health concern in the context of a cultural paradigm. Mental health, stress and anxiety have been recognized as strong factors for sleep apnea or sleep insufficiency [2,3]. Data from the CHRR indicates that median household income, cost of living, and long commuting time are among the main indicators of poor sleep-observations that apply to areas within Greater Polynesia (such as in Honolulu County which is one of the least-rested counties that also reports long work commuting times due to multiple jobs needed to support households and a 3-6 hours' time lag behind the rest of the U.S.). Obesity is another factor associated with sleep apnea [2,3] and our previous research [25] has documented high obesity prevalence in similar county-level hotspots as noted by Grandner and colleagues [28]. Mental health concerns are related to social cohesion and feelings of belonging and such factors may find stronger support in communities that have strong cultural drivers that include strength of community (e.g., Yankeedom), tolerance (e.g., New Netherland), or support for social reform (e.g., the Left Coast) as compared to regional cultures that rely on individualism (e.g., the Deep South) or reject the idea of social reform (e.g., Greater Appalachia). More specific and targeted efforts to tailor messages to specific regions across the country may result in stronger responses to calls for action. This type of approach may also spur more innovation in intervention designs that result in higher uptake of programs and increased likelihood of success in achieving projected results. Underlying cultural differences between the American Nations may also be related to the type of policies adopted by the counties and states that make up these American Nations. Future research should recognize the complexity of these interrelated issues and delving into reviews of such policies and related practices may be well-justified from a public health perspective. --- Strengths, limitations, and conclusions The current analysis has several strengths that should be balanced against several limitations. Certainly, the results are limited by the fact that sleep insufficiency data comes from the BRFSS and is based on self-report. However, the BRFSS is a long-standing data source for population health behavior trends and has been tested for validity and reliability [35,36]. Another limitation reflects the nature of the clusters of county-level data that make up the American Nations. Each county has an estimated rate, and some county "hot spots" rates may be dissipated by other counties within the same American Nation. As such, differences between the 13 American Nations reflect averages across the clusters of counties that define the regions. Thus, specificity may have been limited in the analyses of county clusters reported here when compared to variation between individual counties (e.g., as compared to Grandner et al. [28]). However, this limitation should be balanced against the novel approach that the clustering provides as it pulls cultural influences together, thereby allowing for an innovative perspective on the influence of regional cultures on sleep behavior. We have presented a point of view on insufficient sleep and its association with regional cultures as defined by the American Nations model [16] which applies the well-accepted anthropological framework of "fist settler effects" [17]. As far as we know, this is the first study to report on the influence of the American Nations regional cultures on sleep insufficiency. Cultural aspects of sleep should be considered alongside and integrated with its relationships to biological, physiological, social, economic, and political considerations. The interplay between insufficient sleep, other health-related behaviors, disease outcomes, and a host of cultural, economic, social, and physical concerns needs to recognize the complexity of this dynamic situation. Therefore, we hope that additional research will address questions that place health-related behaviors such as sleep, in context of cultural considerations and call for a systems science approach to address the complexity of these challenges [32,33,37,38]. Such approaches may also provide opportunities for increased use of dialogue and open debate in meeting the cultural challenges of our time and allow for innovative interventional designs that align with regional cultural dispositions of the population. --- Data Availability The datasets generated and/or analyzed during the current study are available from the County Health Rankings & Roadmaps program at the University of Wisconsin Population Health Institute (https://www.countyhealthrankings. org/) and the Behavioral Risk Factor Surveillance System (2020: https:// www.cdc.gov/brfss/index.html) from the Centers for Disease Control and Prevention. The American Nations dataset from the Nationhood Lab, Pell Center for International Relations and Public Policy, Salve Regina University (https://www.nationhoodlab.org/) are available from the upon reasonable request from Mr. Colin Woodard. --- Author contributions NP interpreted the data, wrote the first draft of the manuscript, and was responsible for creating the final draft. RA merged the data files and calculated prevalence statistics, and reviewed manuscript drafts. DL reviewed manuscript draft and provided editorial feedback. CW provided data on the American Nations, was responsible for creating the American Nations map, and reviewed manuscript drafts. All authors read and approved the final manuscript. --- Declarations Ethics approval and consent to participate The HealthPartners Institute Research Subjects Protection Program determined that this study is exempt from IRB review and ongoing oversight under 45 CFR Part 46 as it involves the analysis of existing, publicly available data sets. --- Consent for publication Not applicable. --- Competing interests The authors declare that they have no competing interests. --- Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Background Insufficient sleep can increase the risk of health problems and chronic conditions including cognitive problems, increased inflammation, diabetes and cardiovascular disease, and overall mortality. In this report, insufficient sleep, as a cultural experience, was tracked across the United States according to the American Nations model of U.S. regional cultures. Methods County Health Rankings & Roadmaps program data relating to insufficient sleep were matched at the zipcode level with the American Nations dataset from the Nationhood Lab. Percentages for the insufficient sleep metric were then estimated using the population of each of the American Nations.High levels of sleep insufficiency across all regional cultures indicate considerable room for improvement and a potential need to prioritize sleep hygiene as a health concern. ANOVA results show significant differences among the 13 American Nations and indicate variation in sleep outcomes between cultural regions. Follow-up post hoc analyses appear to support the notion that those regional cultures that place a higher value on social investments tend to report lower levels of sleep insufficiency. Furthermore, the insufficient sleep pattern noted in this report is consistent with those reported earlier for physical inactivity, obesity, and diabetes. Conclusions Factors and circumstances occurring in the Midwest and the South-Central states appear to be responsible for the observed patterns. Future research should consider reviews of policies and related practices adopted across the American Nations to identify causal patterns.

Introduction The widespread availability of positioning technologies like GPS in smartphones and other mobile devices has promoted the use of real-time location updates in mobile apps and location based services. Location-Based Social Networks (LBSNs), like Foursquare, Gowalla and Facebook Places, provide a platform for updating one's location, that is viewable by friends, by checking into a set of venues in the geographical proximity of the user's current location. An emerging category of LBSNs are Location Based Social Discovery (LBSD) networks, that are specifically designed to Copyright c 2013, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. enable establishment of (new) connections to nearby users. The popularity of LBSNs has also attracted research interest as they offer a new source of information that enables studies of Internet users' online and offline behaviors. While the check-in-and content sharing-based LBSN have been studied extensively with a focus on various topics, including users' behavior (Li and Chen 2009), (Scellato and Mascolo 2011), mobility prediction (Scellato, Noulas, and Mascolo 2011), (Cho, Myers, and Leskovec 2011) and social link recommendations (Crandall et al. 2010), (Cranshaw et al. 2010), the emerging LBSD apps are yet to be studied due to the young age of such services and unavailability of extensive datasets. On the other hand, location information has been used in re-identification studies, to evaluate potential for unique identification of users based on e.g. their home-work locations, or a set of unique locations they have visited in the past. Such studies are commonly based on mobile phone datasets (Wang et al. 2011). In this paper, we investigate a new LBSD mobile network by studying the dataset collected from an increasingly popular social discovery application, "Momo"1, launched in late 2011. Momo provides two OSN related functions: social discovery, which enables a user to discover surrounding people based on the geographical distance between them, and instant messaging, that allows users to (subsequently) communicate. By default the application updates the user's location to the server (unless users explicitly opt-out of status updates), hence a rich set of spatio-temporal information about the users is captured in our dataset. Our motivation for this study is two-fold: first, we wish to characterize the evolution of a new LBSD network, both in regards to population growth and to the way the application is utilized by the newly signed-up users. Then, as the spatiotemporal information on how individuals utilize this application is available, we are interested in evaluating the potential for re-identification of users based on this data and using a representative set of similarity metrics. The re-identification scenarios logically follow from the current trend of user profiling by mobile carrier or providers, mobile analytic companies, location-based service providers and potentially by application developers. These entities possess pieces of information covering different aspects of the users' life and would be highly likely to have an interest in enriching their data by exchanging and aggregating the information from other parties. In most of the cases, mobile users present in different databases can be linked by one or multiple unique identifiers like Android ID, Apple ID and accounts that are associated with the mobile device. However these unique identifiers may not be available or applicable in some circumstances, for instance, if we try to link an individual who uses two or more devices, or the device identifiers have been changed during a system upgrade. We have collected approximately 150 million location updates in a period of 38 days, from 19/5/2012 to 28/6/2012, and over 8 million user profiles. We analyze this dataset to provide insights about the new network comprising LBSD application users, growth in user numbers and the way it is used in regards to the user activities i.e. social discovery and messaging. Our contributions are as follows. We characterize user activity in a new LBSD mobile network and show the differences between Momo and a check-in-based LBSN (Gowalla), with Momo having a higher user activity and a lower number of locations visited by users. We conclude that the latter is a likely consequence of the incentives given to users of check-in-based networks to accumulate a larger number of visited locations than what they would be normally inclined to do. We analyze the evolution of the number of users and the way the LBSD application is utilized, including a study of the newly signed users during the monitoring period. We observe that the new users have a significantly higher activity in the first week after signing up for the service and that the vast majority of users, after that period, has a relatively low level of activity that relates to social discovery. We evaluate the potential for re-identification of users, utilizing temporal, spatial, or spatial-temporal characteristics of their in-application activity and a number of selected metrics. We show that, overall, the spatial data provides the greatest accuracy in re-identifying users while the temporal data may be of limited value when used in isolation. On the other hand, using spatio-temporal data, particularly when a large data volume is available, also has good reidentification potential. We also show the relevance of varying level of available data, i.e. the user's level of activity and lifetime in the network, on the re-identification accuracy. The rest of the paper is organized as follows. First, we outline relevant related work in Section 2. We describe the data collection and our dataset in Section 3, followed by the characteristics of user's activities in Section 4. In Section 5 we analyze the evolution of the user numbers and the way users utilize the application. Section 6 evaluates the potential for re-identification based on different data types and volume available. We conclude and outline future work in Section 7. --- Related Work We provide an overview of research work related to various aspects of our LBSD app study. A number of previous studies based on measurement of user's activities in LBSNs focus on characterizing check-in-based social networks like Gowalla, Brightkite and Foursquare. (Li and Chen 2009), (Scellato and Mascolo 2011), (Cheng et al. 2011) analyze the use of LBSNs in regards to the volume of check-ins and their spatio-temporal characteristics. Cho et al (Cho, Myers, and Leskovec 2011) analyze the social and spatial characteristics of Gowalla and Brightkite check-ins, aided by (related) cell phone traces. They first show the link between human movement and, respectively, social relationships and periodic user behavior (social relationships can account for between 10-30% of all human movement, while 50-70% of movement relates to periodic user behavior) and develop a mobility model that incorporates their findings. Similarly, (Scellato, Noulas, and Mascolo 2011) propose a new mobility model based on features of visited locations. Works including (Crandall et al. 2010), (Cranshaw et al. 2010) study predictions (recommendations) of social links. LBSN evolution was studied in (Allamanis, Scellato, and Mascolo 2012), with researchers proposing a new model of network growth based on a combination of social and spatial factors. Instant messaging (IM) applications were characterized in (Leskovec and Horvitz 2008), however this was done based on fixed user data. Our paper studies the characteristics and the evolution of an emerging LBSD app that combines social discovery and IM. We demonstrate the differences between the check-inbased LBSN and the LBSD (Momo) and characterize evolution of both individual user's traffic and growth in overall population, highlighting the link with major promotional activities of Momo. We show how the mode of use shifts for most users, with time, from social discovery to IM use. We additionally study re-identification in LBSD apps. The extensive research work on re-identification spans a number of fields of study. Related to spatial information, researchers in (Golle and Partridge 2009) have analyzed the U.S. Census data and have shown that on average, close to 20 individuals from the datatset share the same home or work locations, and that 5% of people in the dataset can be uniquely identified by home-work location pairs. A related work by Zang et al (Zang and Bolot 2011) generalized the use of (home-work) location pairs to an approach that uses top N locations to evaluate the uniqueness of US cellphone users. Both works strongly support the case for using location information to derive quasi-identifiers for reidentification of users. A number of research works e.g. (Mohammed, Fung, and Debbabi 2009), (Bonchi, Lakshmanan, and Wang 2011) and (Shokri et al. 2011) have raised the privacy issues in publishing location data and have focused on theoretical analysis of obfuscation algorithms. Further, (Li et al. 2008) exploit location history and evaluate the merits of different metrics to re-identify users from a dataset collected from a (small) number of cellphone and GPS device users. We note that the location related reidentification research works that are based on experimental data, have focused on mobility traces from cellphone or GPS devices, while the different nature of LBSN/LBSD data (human driven updates related to activities in the social network) necessitates a study that specifically addresses this environment. We evaluate the potential of re-identifying users by their temporal or spatial patterns (or a combination of those) in the LBSD app. As users reveal their location in a noncontinuous manner (as opposed to having a dataset that includes user's full trajectory in a measurement time period), the mobility trace and location sequence approaches were not applicable in our study. We examine a number of basic user similarity metrics that have been proposed in uniqueness measurement (Zang and Bolot 2011) and link prediction studies (Wang et al. 2011), (Crandall et al. 2010). We stress that our focus is on evaluating the potential for user reidentification in a LBSD application, rather than on proposing new re-identification algorithms. --- Data Collection and Datasets In this section we first briefly introduce the Momo LBSD application and the associated network, and then describe our data collection methodology. We also provide a comparative analysis of the basic characteristics of the LBSD network and outline key features of Momo, when compared to a popular check-in based LBSN, i.e. Gowalla. Specifically, we highlight the main differences between the two networks from the usage pattern and user behavior perspectives. --- LBSD Application: Momo Momo is a location-based social discovery application which allows users to discover people located in close geographical proximity and to connect with them using IM. The Chinese-based company launched its first iOS application in August 2011 and an Android version was rolled out in December 2011. Since then, the service has accumulated over 10 million registered users2. New users in Momo have to create a profile containing basic personal information (name, age, gender and icon photo) and some optional attributes (e.g. occupation, company, school, interest, etc.). The application is given permission to access the device's location information, which is extracted either directly from the cellphone GPS or via Google Mobile Map API3. When a user launches the application, a location update is sent to the Momo server. User can obtain a list of nearby users from the server (discovery function). These users are then displayed in the application according to their proximity to the user location, i.e. ranked in terms of distance to the device's current location. User status and location is simultaneously publicly revealed to other nearby users (and friends4 ), unless the "invisible" option is selected by the user. However, we have observed that only about 6% of the total number of users in our dataset have their location hidden from public access. Finally, individuals can establish a direct connection to any of the discovered users, who can be added to their Friends lists. This friendship relationship is never made public. --- Data Collection The Momo mobile clients communicate with the server(s) via a set of network APIs. The profile API allows a client application to fetch a full profile of a user, identified by a numerical identifier. We crawled user profiles by selecting an exhaustive set of IDs, that were accessed sequentially during the data collection. We used the profile API to collect our first dataset, Profiles, comprising of user profiles. Our second dataset, Updates, was collected using the nearby API, which upon request provides a list of nearby users. By varying the geographical coordinate parameters over time, we collected an extensive set of real-time location updates originating from different monitored areas. Collecting user updates for such a large and dynamic system is a challenging task. Due to the high number of updates during different periods of the day, and the diversity of possible world-wide user locations, we have optimized our crawling strategy by considering the following trade-off: while we were interested in minimizing the number of requests to be sent to nearby APIs, we also had an aim to avoid missing any of the user updates within a monitored area. The nearby API dictates that the server will only respond with up to a maximum of 30 users per update request, and a maximum of 40 requests can be made from any single set of coordinates, thus limiting the coverage area. To maximize the amount of collected data while working within these constrains, we have designed a dynamic and distributed crawling mechanism that operates in two steps. First, the crawler schedules a set of monitoring points and their corresponding coverage areas. This step is based on a modified version of the 2-dimensional closest point search algorithm in lattices (Agrell et al. 2000), which in our case aims to discover the minimum number of monitoring points by estimating the total number of active users in a given initial lattice area. This is achieved by recursively selecting potential monitoring points and verifying whether or not they are capable of discovering all other users activity within their currently defined monitoring area. Specifically, by simply requesting the last retrieved set of of nearby users of each selected monitoring point, we can decide whether the monitoring point can reach the edge of its assigned lattice. If the number of expected users exceeds the crawling capability of a single crawling (monitoring) point, i.e. more than 30 <unk> 40 = 1200 users, the crawling area is further divided into four equal size lattices. Specifically, for each of the selected cities globally, we monitor a lattice area of 60 <unk> 60 km 2 around the city center. The crawler scheduling process is performed every hour. The output of this step is a set of task parameters consisting of target coordinates, expiration time and the area to be covered. Second, each of the scheduled crawling points requests the close users' locations from the server. Every 15 minutes, the crawling points selected at the first step (i.e. chosen hourly) update the location of the users observed in their vicinity. It should be noted that in case of multiple updates made between two different requests, the crawler can only record the last update. Although this prevents our crawler from guaranteeing that all user updates are collected, we believe that the chosen request period of 15 minutes is already fine grained enough to allow us to capture the vast majority of updates. We have indeed tried different request periods, and empirically decided that such a choice is a good compromise between the number of recorded updates and the number of requests to the server (e.g. compared to a 2 minute requests period, less than 1% of the user updates are missed by having the period set to 15 minutes). Figure 1 shows the example allocation of crawling points and user density at 6 AM and 11 PM, around Shanghai Area. The populated areas can be easily identified from the density map. During peak hours, the crawling range can be as small as 0.5 km around the populated area, which means that up to 1200 users can be discovered within 1 km 2 area. The collected Profiles dataset consists of 8 million user profiles, each containing the user's most recent update time and associated GPS location. These profiles also include the publicly available user's personal information, namely: nickname, gender, age, interest, occupation and identities in other popular social networks. Every record in the Updates dataset consists of a user ID, timestamp and GPS coordinates (latitude and longitude). This dataset contains 150 million updates from 3.3 million active users in 48 cities (including cities from Australia, Canada, China, France, Germany, Italy, Japan, Korea, Singapore, Spain, US, and UK), over a period of 38 days during May-June, 2012. The Updates dataset contains approximately 65%5 of all active users in the entire Momo network, present at the time --- Gowalla Dataset Gowalla is a popular location-based social networking service that allows users to check-in their current location and share it with friends. The properties of the Gowalla LBSN have been studied in a number of research works e.g. (Cho, Myers, and Leskovec 2011).The dataset used in this work, consisting of more than 6 million check-ins from more than 196K users over the period of Feb. 2009 -Oct. 2010, was published in the SNAP website 6. We use this dataset to evaluate the main differences between the usage and behavioral patterns of the two services: Momo as an LBSD network and Gowalla as a check-in-based LBSN. --- Characterizing User Activity We analyze user activity in the Momo network and provide a comparison of various related metrics to those derived from the Gowalla dataset. Activity Distribution We start by examining the extent of difference in the user activity distribution in both networks. For the Momo service, we use the Updates dataset. In order to have a meaningful comparison of the two datasets in regards to the measurement duration, we use a 38-day portion of the Gowalla dataset (the same duration of the Momo dataset). In addition, as the location updates of Momo do not correspond to a specific "spots" as they do in Gowalla, we define a unique location as a 1 <unk> 1km 2 grid, which is then considered as a specific location for the Updates dataset. Table 1 shows the user activity distribution function based on different parameters, with values shown for the 5, 25, 50, 75 and 95 percentiles. We define active days, as the number of days (out of the total number of observed days) on which users have checked-in at least once. Overall, the distribution statistics show that Momo users are more active than Gowalla users. While both Gowalla and Momo users do reveal almost the same number of unique locations per day, interestingly Gowalla users exhibit a higher number of total unique locations across the full observation period. Momo users, however, are more active, with almost two times higher median number of updates per day compared to Gowalla users. The latter also have a significantly lower number of location updates than the Momo users, with the median number of updates being only seven, while 50% of Momo users checked in at least 14 times. Notably, both networks have a similar distribution of active days for more than half of the user population. However, a considerable fraction of users in Momo publicly reveal a higher number itoring period (obtained from Profiles dataset) 6 http://snap.stanford.edu/data/loc-gowalla.html of unique locations compared to Gowalla users. While several reasons might explain the higher total number of visited locations for Gowalla users, we believe this is mainly due to the different nature of the two applications, where in Gowalla the users are incentivized to check-in at different locations that would accumulate in their history. Momo users, even though more active, seem unwilling to publicly share a number of locations they visit, and do not simply use the LBSD application in diverse locations. The two applications are designed for different purposes, and as such the user data collected from both services consequently reflects this difference in nature. --- Correlation between daily activity and active days In Figure 2 we show the number of updates per day as function of the number of active days in the network, for both Momo and Gowalla. We can observe, for both services, that the increased loyalty to the applications results in a higher level of user activity. We then calculate a positive Spearmen coefficient of 0.625 in Momo and 0.552 in Gowalla, which indicates a high correlation between the number of active days and the number of updates per day. We again observe that the number of check-ins per day for Momo users is higher than for Gowalla, which is also illustrated by a sharper correlation slope. --- Distribution of updates as a function of activity level Table 1 suggests a high skew in the distribution of updates, across users. To explore this further, Figure 3(a) depicts the cumulative distribution function of updates as a function of the top n% active users for both Momo and Gowalla datasets. Both networks show a similar distribution, with top 20% active users contributing 72% (resp. 75%) of total check-ins in Gowalla (resp. in Momo). This observation is in accordance with the Pareto principle ("a minor proportion of causes generate a major proportion of effects"). Distribution of updates/check-ins as a function of the number of locations Similarly to exploring the relation between users and updates, we examine the number of locations that attract the majority of check-ins. Figure 3 shows the cumulative distribution of updates as a function of the top n% locations shared by users, for both Momo and Gowalla datasets. In Gowalla, the check-in is associated with both geographical coordinates as well as with points of interests, called "spots", that could be e.g. a coffee shop or a library. As illustrated in the figure, Gowalla check-ins are distributed in a larger number of spots than grids, i.e. 20% spots attract less than 70% of check-ins while more than 85% of check-ins are done in 20% of grids. This can be explained by the fact that a number of spots can be aggregated in a fewer number of 1 <unk> 1km 2 grids. We also note that Momo users' updates are distributed in more grids than Gowalla check-ins, which implies that Gowalla users are more likely to check-in within popular areas e.g. city centers. Daily and weekly activity patterns Figure 4 shows the daily and weekly activity patterns of Momo and Gowalla users. For a meaningful comparison, we choose users from the same timezone. As previously observed in (Cho, Myers, and Leskovec 2011), during weekdays, the Gowalla curve shown in 4(a) exhibits two noticeable peaks around 12-2 pm and 6-8 pm, whereas the Momo activity curve shows a single peak around 10-11 pm. We also observe that there is less activity during late night and early mornings in Gowalla daily patterns, as opposed to the Momo users which exhibit a relatively high level of activity during these periods. Again, the different nature and the intended use of the two applications can explain such a different daily usage pattern. Gowalla users have limited opportunities to check-in to a place other than their home at late night and early morning hours, while Momo users can use the service in a meaningful way, by either discovering nearby people and/or chatting with friends from home; in both cases, their location is updated automatically when they refresh the user list or bring the application to the foreground. We also highlight the steady weekly pattern of user activity in Momo, with close to no variation between the weekdays and the weekends. Gowalla, on the other hand, shows a "typical" weekends/weekdays variation, with users being more active during weekends. This again reinforces our observation that the user activity level in Gowalla, and as such their location updates, seem to be closely linked with their visits to locations they have an incentive to check-in to. 9 - 2 0 1 1 1 0 - 2 0 1 1 1 1 - 2 0 1 1 1 2 - 2 0 1 1 0 1 - 2 0 1 2 0 2 - 2 0 1 2 0 3 - 2 0 1 2 0 4 - 2 0 1 2 0 5 - 2 0 1 2 0 6 - 2 0 1 2 0 7 - 2 0 1 2 total users (CDF) Date 0 10k 20k 30k 40k 50k 60k 70k 80k 90k 100k 0 8 -2 0 1 1 0 9 -2 0 1 1 1 0 -2 0 1 1 1 1 -2 0 1 1 1 2 -2 0 1 1 0 1 -2 0 1 2 0 2 -2 0 1 2 0 3 -2 0 1 2 0 4 -2 0 1 2 0 5 -2 0 1 2 0 6 -2 0 1 2 0 7 -2 0 1 2 new users Date (1)(2) (3) (4) (5) (6) (7) --- Network and Usage Evolution Next, we study the evolution of the Momo LBSD network, including the growth of user numbers and identifying events that have contributed to significant spikes in growth. We also analyze how the new users' activity evolves through time, by studying user retention rate, user behavior and the temporal evolution of the mobility patterns of Momo users. We note that, due to limitations of the available information from the Momo APIs, we can only monitor in real-time the behavior of active users within the data collection period (38 days) and in the monitored geographical areas, as per the contents of the Updates dataset. We also have access to the last record of activity and the account age for all Momo users, as per the Profiles dataset. --- User Volume Growth and Trigger Events Since its launch date in August 2011, Momo has achieved significant growth, with a population size of more than 10 million users reached in less than one year 7. Figure 5 shows the daily user growth between August 2011 and July 2012 7 http://www.cnetnews.com.cn/2012/0803/2104590.shtml (based on the data from the Profiles dataset, which includes the sign-up dates of users). We observe that the application first experienced a fivemonth long slow start with, on the average, less than 5K new users per day. The pace of user growth became faster in early 2012 and received two major bursts around the end of April and the beginning of June 2012. To understand the external factors that drove the application growth, we have identified the events corresponding to the specific spikes of user growth, by associating them with various Momo company related information: (1) From 21/10/2011 to 23/10/2011, Momo launched an advertising campaign on Weibo (a Chinese popular micro-blog platform), promoting the product in 27 influential accounts that had millions of followers. (2) On 15/12/2011, the company released the first Android app version. Although we do not observe a sudden user growth after the Android version release, we can see that the user volume growth climbed steadily from the time of this event. (3) On 07/01/2012, Momo was awarded the "Best Social App 2011" by Geekpark.com8. (4) From 23/01/2012 to 29/01/2012, we observe a burst of user growth during the 7-day Chinese New Year public holiday. (5) On 17/02/2012, another popular advertising campaign was launched, receiving 15k re-posts soon after its release. ( 6) 27/04/2012, which corresponds to the most significant growth spike that occurred following a popular (funny) video mentioning the Momo application9. The video received one million views in 10 hours, and the volume of Momo's daily new users almost doubled on the following day. ( 7) 02/06/2012, Mainstream media "City Weekly" magazine (Southern Metropolis Daily) published a cover article on targeted social applications, with a section dedicated to introducing the Momo application 10. Figure 5 suggests that although the two advertising campaigns in social media had immediate positive results, both online and traditional off-line media coverage (events ( 6) and ( 7)) attracted a steady and more significant user growth. Notably, we observe that online media coverage seemed to have a longer lasting period in which a high number of new users was attracted. This can be explained by the easy to access, long-tailed viral dissemination and reproducible characteristics of Internet content, as opposed to one-off conventional printed media content. Likewise, the burst of user growth around the Chinese New Year indicates that special events like public holidays --- The Evolution of User Activity We analyze a number of metrics related to the level of user activities in the Momo network. Retention rate. This is one of the most important performance metrics that measures the percentage of retained users after a period of time. According to mobile analytic company Flurry's report 11, the average 90-day retention rate for social networking applications is around 34%. Although we cannot compute the absolute retention rate including all Momo users, based on the "join date" from user profiles, we are able to derive the retention rate of users for a specific account age, as shown in Figure 6(a). Similar to Flurry's approach, the "retained" users are defined as the users who login to the application at least once in the past 7 days. Figure 6(a) indicates that the percentage of retained users sharply drops in the first two months, and becomes stable after three months. The trend suggests that the users who stayed in the network for more than three months are likely to continue using the application for a longer time period. The 30days retention rate is approximately 44.8% and is reduced to 33.1% after 90 days. Interestingly, we observed a reduction of retention rate around the account age of 160 days, which corresponds to users who joined during the Chinese New Year public holiday. Engagement over time. We are interested in analyzing the activity level of users over time, demonstrating their involvement in using the application. Figure 6(b) shows the engagement of users with different account ages up-to 300 11 http://blog.flurry.com/bid/90743/App-Engagement-The-Matrix-Reloaded days, where the engagement is measured by the frequency of use per week. Figure 6(b) clearly shows that the newly joined users have a much higher activity level compared to more experienced users. The decline of activity slows down around 60 days and remains at the level of around 20 updates per week. Probability of active discovery. In the Momo application, the mobile client updates its status and location when the user turns on the application, or when the user actively discovers nearby users, i.e. refreshes the user list. In the case of active discovery, multiple updates with short inter-arrival times can be observed from the server side. An active discovery session is defined as a set of consecutive updates, with each pair of updates occurring within a 30 minutes time interval. We measure users' discovery behavior by the active discovery probability, that is computed by the number of active discovery sessions over the total number of sessions. As shown in Figure 6(c), we observe that, for new users, as high as 20% of sessions are active discovery session, and that the average discovery probability decreases steadily as the account age of users increases, being reduced to an average of 15% for the account age of 300 days. This trend suggests that users spend less time on social discovery as they establish connections with (a sufficient number of) nearby users over time. Unique location v.s. account age. Figure 6(d) shows the average number of unique locations visited by users with a specific account age. We can observe that more experienced users tend to use the application in more locations over the same period of time, compared to recently signed-up users. --- User Re-identification The extensive amount of information generated by users of Momo, including the locations they have visited when using the service, frequency and time of use, can also be utilised for service personalization, by profiling users in order to offer them e.g. recommendations on places to visit, or people to connect to. Considering such scenarios where specific users are targeted, in this section we evaluate the potential for re-identifying users, based on the pervasive spatial and temporal information that may be collected e.g. by the service providers. As different types of information may be collected by different service providers, we evaluate their capability to reidentify users utilizing three different levels of background knowledge: using only the check-in time stamps, based on the subset of locations they have shared with the service and combining both temporal and spatial patterns of LBSD application use. To better understand how the availability of data may affect the linkage performance, we also evaluate the performance of re-identification with different data collection period durations (e.g. comparing periods of 1 to 19 days of location information). We also vary the level of user activities and study whether it impacts the linkability of user patterns. Next, we first introduce the methodology adopted to assess the re-identification capabilities, followed by the details of the similarity metrics used in our study. --- Methodology We first divide the Updates dataset records into two sets: a training set R and a test set R. Each set comprises an equal number of observation days and depending on the selected strategy, the two sets contain either the user check-in time, the check-in location or both. Then, for each user i we extract all records r i (resp. ri ) from the training set R (resp. the test set R) corresponding to the user activity during that period. The goal of a re-identification classifier is to predict the linkage between user records in R and R, assuming the identities in both sets are unknown or anonymized. The classifier first computes the similarity score between r i and all user records in R. We have decided to adopt two different approaches: (i) a unique re-identification case where the classifier identifies the "best candidate" record rj from R, i.e. the record with highest similarity to r i (amongst all records). (ii) a set of candidate records R pre i, predicted as being potentially linked to r i as they have a similarity score higher than a predefined threshold. We define k as the size of the candidate records set, i.e. k = |R pre i |. The value of k will depend on the threshold value, as a more restrictive threshold will result in a smaller set of candidates. In the following, we introduce the similarity metrics used in this study. Temporal Cosine Similarity (TCS) : Considering the time-based vectors r i and r j, we compute their cosine similarity to capture the temporal similarity of the usage pattern. Let P (i, t) <unk> f (i,t) n(i) be the probability that user i uses the application during a period of time t (in our case periods are defined as hours) and f (i, t) is the cumulative frequency of updates in period t across the record duration. n(i) is user i's total number of updates. Then r i = [P (i, 0), P (i, 1),..., P (i, 23)] and r j = [P (j, 0), P (j, 1),..., P (j, 23)]. The TCS score between user i and j is computed as: T CS(i, j) = cos(r i, r j ) = r i • r j r i <unk> r j Spatial co-location rate (SCR) : Considering the spatial usage patterns of users, we extract from

With more people accessing Online Social Networks (OSN) using their mobile devices, location-based features have become an important part of the social networking. In this paper, we present the first measurement study of a new category of location-based online social networking services, a location-based social discovery (LBSD) network, that enables users to discover and communicate with nearby people. Unlike popular check-in-based social networks, LBSD allows users to publicly reveal their locations without being associated to a specific "venue" and their usage is not influenced by the incentive mechanisms of the underlying virtual community. By analyzing over 8 million user profiles and around 150 million location updates collected from a popular new LBSD network, we first present the characteristics of spatialtemporal usage patterns of the observed users, showing that 40% of updates are from the user's primary location and 80% are from their top 10 locations. We identify events that trigger bursts of growth in subscriber numbers, showing the importance of social media marketing. Finally, we investigate how usage patterns may be utilized to re-identify individuals with e.g. different identifiers or from datasets belonging to different online services. We evaluate re-identification by usage, spatial and spatial-temporal patterns and using a number of metrics and show that the best results can be achieved using location data, with a high accuracy: our experiments demonstrate that we can re-identify up-to 85% of users with a precision of 77% using monitored spatial data. Overall, we find that although users exhibit strong periodic behavior in their usage pattern and movements, the success rate of reidentification is highly dependent on the level of activeness and the lifetime of the users in the network.

we compute their cosine similarity to capture the temporal similarity of the usage pattern. Let P (i, t) <unk> f (i,t) n(i) be the probability that user i uses the application during a period of time t (in our case periods are defined as hours) and f (i, t) is the cumulative frequency of updates in period t across the record duration. n(i) is user i's total number of updates. Then r i = [P (i, 0), P (i, 1),..., P (i, 23)] and r j = [P (j, 0), P (j, 1),..., P (j, 23)]. The TCS score between user i and j is computed as: T CS(i, j) = cos(r i, r j ) = r i • r j r i <unk> r j Spatial co-location rate (SCR) : Considering the spatial usage patterns of users, we extract from r i and r j, the two sets L r i and L r i as the set of unique locations visited by user i and j. We then compute the Jaccard-index to measure the similarity of unique locations visited by both users. SCR(i, j) = |L r i <unk> L r j | |L r i <unk> L r j | Spatial top co-location rate (STCR) : Considering users' top N locations, we aim to measure the similarity between popular co-locations. For each user i (resp. j), we only consider the locations from vector r i (resp. r j ) and build a vectors L i (resp. L j ), ranked according to the frequency of visited locations. The STCR score is then computed as: ST CR(i, j) = | N k=1 L i <unk> N k=1 L j | N In the case where any of the two vectors contain less than N elements, we set N = min(|L i |, |L j |). In our experiment we consider top 10 locations, i.e. N = 10. Spatial cosine similarity (SCS) : This metric captures the similarity between two users location frequency patterns. Let P (i, l) <unk> f (i,l) n(i) be the probability that user i utilizes the application in location l; f (i, l) is the cumulative frequency of updates in location l across the recorded period and n(i) is user i's total number of updates. For all locations in both sets, i.e., L = L(i) <unk> L(j), we then present location vectors of users i and j as r i = [P (i, 0), P (i, 1),...., P (i, l)] and r j = [P (j, 0), P (j, 1),...., P (j, l)]. The SCS metric is measured as: SCS(i, j) = cos(r i, r j ) = r i • r j r i <unk> r j Spatio-temporal co-location rate (StCR) : This metric is motivated by the observation that users may visit different locations on specific hours of the day. We first divide the location-based check-in vector r i of each user i into T subvectors, corresponding to different periods of time duration T. We then compute the spatial co-location rate of each of the T sub-vectors between any two users. The StCR score of users i and j is defined as: StCR(i, j) = --- Performance Evaluation Binary classifiers are commonly evaluated in terms of precision and recall. These are defined using the possible outcomes of a classifier, i.e. two records are deemed to be either linked or non-linked, and the resulting validity of the classification, i.e. two records will be matched if they belong to the same user, and non-matched otherwise. Table 2 shows the possible combinations, used to compute the Precision and Recall values as: Recall = T P T P +F N and P recision = T P T P +F P. For most cases, precision and recall provide a good indication of the performance for varying classifier parameters. However, when there is a large difference of e.g. a very small precision and a high (uncomparable) recall value, we use an alternative evaluation method by defining a successful prediction as the correct identification of a matched record that can be re-identified either uniquely or within a set of k candidate records. We then use the success rate as the % of successful predictions over the total number of linked records. We now evaluate the performance of different similarity metrics. Temporal approach Our initial experiments using TCS have shown a low accuracy of unique re-identification, therefore we consider the performance for a resulting candidate set. Figures 7(a) and 7(b) show the success rate of re-identification with a varying umber of active days and updates per day. We can observe that when using only a single day of user's data, the classifier provides almost random results. For the entire 19 days, the classifier can narrow down the candidate set to 10% of the sample size 60% of the time, which considering the sample size is not a promising result. Spatial approach The precision-recall curves of SCR, STCR and SCS are shown in Figure 8(a). The SCR model has extremely low precision with all thresholds, which suggests that the probability of co-location is quite high amongst users and the low recall also suggests many users are inconsistent in location updates. By considering the visited location frequency, STCR and SCS metrics outperform SCR. STCR metric only considers the top 10 most relevant user's locations, ignoring the infrequent locations that are not likely to be revisited, and hence increases the linking probability. The SCS metric compares the direction of location vectors, i.e. cosine similarity. The frequently visited locations contribute more to the direction of the vectors, while the influence of infrequent or in-transit locations is suppressed. We can observe that SCS shows a solid performance, we can re-identify up-to 85% of users with a precision of 77%. Spatio-temporal approach Figure 8(b) shows the precision-recall curves for StCR, StTCR and StCS reidentification results. We can observe that StCR and StTCR perform better than the original spatial-only approach, while the performance of StCS degrades compared to SCS. The improvement suggest that users are commonly co-located and may share the top locations, however within a time dependency. On the other hand, SCS computes the cosine similarity of the overall location frequency; segmenting the pattern into 24 sub-patterns, i.e. via StCS, results in a less identifiable pattern. We now consider how a varying amount of data available for re-identification would affect the accuracy of various similarity metrics. Considering the available quantity of data Tables 3 and4 show the success rate of uniquely re-identifying users, for the various similarity metrics and based on a selected number of active day and user activity levels (updates/day). With the exception of SCR, all other metrics have an improved performance with a longer monitored period. When only using a single day of records (i.e. one day each in the training and testing sets), approximately half of the users can be correctly re-identified. The success rate increases to 86 % when using a week of data records, and almost all users (97.7%) 4 indicate that users who have higher level of activity have a higher chance to be re-identified. --- Conclusion In this paper, we have characterized a new LBSD network, in terms of overall properties (when compared to more established LBSNs), and the way it is being used by it's subscribers. We have then analyzed the network evolution, identifying specific events contributing to extraordinary user growth, and the evolution of user's activity patterns. Finally, we have evaluated the potential for service providers who may monitor and retain data on user's activities, to reidentify LBSD users who may have e.g. multiple accounts or a number of mobile devices, by comparing the performance of a number of similarity metrics. Our results show that a significant quantify of data is required for re-identification. This indicates that re-identification may be applicable primarily to applications that are heavily used, preferably on a daily basis, e.g. from the social, gaming or business productivity categories.

With more people accessing Online Social Networks (OSN) using their mobile devices, location-based features have become an important part of the social networking. In this paper, we present the first measurement study of a new category of location-based online social networking services, a location-based social discovery (LBSD) network, that enables users to discover and communicate with nearby people. Unlike popular check-in-based social networks, LBSD allows users to publicly reveal their locations without being associated to a specific "venue" and their usage is not influenced by the incentive mechanisms of the underlying virtual community. By analyzing over 8 million user profiles and around 150 million location updates collected from a popular new LBSD network, we first present the characteristics of spatialtemporal usage patterns of the observed users, showing that 40% of updates are from the user's primary location and 80% are from their top 10 locations. We identify events that trigger bursts of growth in subscriber numbers, showing the importance of social media marketing. Finally, we investigate how usage patterns may be utilized to re-identify individuals with e.g. different identifiers or from datasets belonging to different online services. We evaluate re-identification by usage, spatial and spatial-temporal patterns and using a number of metrics and show that the best results can be achieved using location data, with a high accuracy: our experiments demonstrate that we can re-identify up-to 85% of users with a precision of 77% using monitored spatial data. Overall, we find that although users exhibit strong periodic behavior in their usage pattern and movements, the success rate of reidentification is highly dependent on the level of activeness and the lifetime of the users in the network.

Literature Review A gender stereotype is a preconceived view where differences exist between men's and women's attributes, characteristics, and/or roles (Martin et al., 1990). Gender stereotypes perpetuate gender inequalities and have a long-lasting, negative impact on women. Children's TV shows are currently under scrutiny for reinforcing gender stereotypes through various gender stereotype portrayals (Kim et al., 2016). These portrayals include, but are not limited to, an obvious distinction in appearance and roles based on gender, and the use of different terminology when addressing male and female characters (Leaper et al., 2002). Early childhood (from birth to age 8) is defined as a crucial development period in a child's life (Worthman et al., 2016). Thus, gender stereotypes portrayed in children's TV shows can be extremely detrimental to a child's upbringing. A child spends an enormous amount of time dedicating their day to watching TV. Therefore, gender stereotypes in children's TV can unintentionally affect a child's views and opinions on how boys and girls should be perceived. Studies have also shown that children perceive media messages about gender as an illustration of society as a whole (Signorielli, 1990). So, it is crucial to identify the TV programs that negatively affect children's views on gender. Children develop gender stereotyping patterns from key domains of stylistic appearance, activities, social role, etc. (Walsh & Leaper, 2020). This paper examines similar domains in children's TV such as appearance, activities, occupation, and specific terminology used when describing characters. In addition, this paper considers the prominence and talkativeness of male and female characters. Currently, very few research papers actually examine gender stereotypes in children's TV shows. Instead, they mainly examine gender stereotypes in middle-aged and teenage TV shows. This literature review identifies gaps that will help measure gender stereotype patterns in children's TV shows. Considering the prominent gaps just discussed, this paper aims to answer the following question: Are gender stereotypes present in children's TV shows, and if so how do they compare in the US and India, and has the level of gender stereotype patterns changed over the past decade? --- Method --- Selecting Children's TV Shows This study (1) identifies gender stereotypes in children's TV shows, (2) examines the differences in gender stereotypes in the United States and India, and (3) examines whether or not gender stereotype patterns have changed over the past decade. When selecting the children's TV shows, the researcher first began with a list of shows that are/were produced for children aged 2 to 10 years old. From this category, the researcher ensured that the selected TV shows had released at least 30 episodes prior to 2011 and 2021. To analyze the differences or similarities in gender stereotype patterns over the past decade, the researcher set two cutoff years for 2010-2011 and 2020-2021. The researcher specifically chose these years because she believed they would gather more recent/applicable results. To measure the relevancy of these TV shows in a child's life, the TV show must have released at least 5 new episodes during the years 2010-2011 (for the analysis of gender stereotypes in a decade-old children's TV show), and 5 new episodes during years 2020-2021 (for the analysis of gender stereotypes in currently airing children's TV shows). Based on these criteria, the researcher selected the 4 following currently airing and decade-old TV shows: • United States: Arthur, Thomas & Friends • India: Chhota Bheem, Bapu The length of episodes varies from 15 to 25 minutes. Most of the episodes are available through online platforms, such as YouTube. All of the selected episodes are also animated. --- Coding Instrument To measure the variability of gender stereotypes, the researcher randomly selected 2 episodes from each TV show aired online and assessed 5 gender-stereotyped domains: (1) the prominence of female and male characters, (2) terms used when describing characters, (3) activities & occupation, (4) talkativeness, and (5) appearance. The purpose for choosing these specific domains is as follows: ( --- 1) Number of Female and Male Characters Research has shown that male characters exceed female characters by a 2 to 1 ratio in children's shows in various countries (Götz et al., 2008). In 2014, Shannon Case, a graduate research assistant at Georgetown University, discovered that the highest-rated children's TV shows featured around 60.2% male characters and 39.8% female characters. Since female figures are underrepresented compared to male characters, children may be persuaded to believe that boys are inherently superior to girls in society. To test this idea, the researcher tallied the number of female and male characters for both American and Indian children's TV shows. The researcher determined the characters' gender based on terms (e.g., pronouns) used, appearance, and voice. (2) Terminology Used When Describing Characters Women are socially characterized as emotional, while boys are characterized as physical and aggressive. Therefore, the researcher coded feminine-stereotyped terminology with words similar to inept, selfish, indecisive, passive, etc. The researcher coded masculine-stereotyped terminology with words similar to athletic, versatile, analytical, etc. ( --- 3) Activities & Occupation Gender stereotypes assign genders to specific activities and occupations. Both female and male characters display stereotypical gender-role behavior with males having greater representation than females in children's TV shows. Female characters mainly participate in nurturance and non-physical activities, while male characters participate in physical activities (Barner, 1999). Based on these findings, the researcher considered masculine-stereotyped activities as participating in math, construction, sports, and playing with trucks and/or cars, etc. On the other hand, the researcher considered feminine-stereotyped activities like painting, singing, drawing, playing with dolls, etc. Through this coding scheme, the researcher was able to identify whether female characters were generally neglected and ignored and placed inferior to male characters. --- (4) Talkativeness It is a prominent notion that women talk differently than men in both subject and manner. Society continuously characterizes women as more talkative and involved, and men as more silent and unattached. In children's TV shows, the talking times may assert a difference in prominence among characters. For example, if a male or a female character holds a greater talking time, their prominence in the show, as a result, will increase. Therefore, the researcher measured and compared the talking time of female and male characters using a timer. (5) Appearance Characters' appearances in children's TV shows are represented stereotypically in three main ways: colors, clothing/accessories, and body type. Firstly, colors are commonly used to assume a certain gender; "feminine" colors are considered to be pink and purple and "masculine" colors are considered to be blue and gray. Secondly, clothing/accessories are used to differentiate between genders. Dresses and jewelry, such as earrings, are correlated with females, and boys are correlated with traditional masculine-stereotyped clothing, such as shirts and suits (Lindemuth et al., 2011). According to Caroline J. Lindemuth and colleagues, researchers at the University of the South, 44.3% of females and 0% of males wore a dress or skirt, and 67.5% of females and 0.5% of males wore 2 or more earrings. Lastly, specific body types are used to separate genders. Females are usually depicted as thin and slim, while males are depicted as broad and muscular. According to Ashton Gerding and Nancy Signorielli, researchers at the University of Delaware, female characters had thin body types, while male characters were portrayed with more diverse body types. Therefore, the researcher coded characters' types of clothing based on color. Colors similar to blue and gray were associated with a masculine-stereotyped appearance. Colors similar to pink and purple were associated with a feminine-stereotyped appearance. The researcher also associated dresses and jewelry with a feminine-stereotyped appearance and shirts and suits with a masculine-stereotyped appearance. Lastly, the researcher associated thin and slim body types with a feminine-stereotyped appearance, and muscular body types with a masculine-stereotyped appearance. The researcher also used momentary time-sampling (MTS) to analyze the difference in behaviors of female and male characters. MTS is a behavior assessment method that divides an observation time into intervals and tallies behaviors as an occurrence (Saudargas & Zanolli, 1990). Therefore, the researcher broke each episode down into ten-second intervals. During each interval, female and male stereotyped behaviors were identified as an occurrence. Next, to find the average mean of gender stereotypes present in children's TV shows, the researcher then divided the number of occurrences obtained from MTS by the number of female and male characters present in each episode. For example, if a variable (such as feminine-stereotyped appearance) appeared 50 times in an episode with ten female characters, the average mean of that variable for female characters would be 5.0. After gathering the means for each genderstereotyped domain, the researcher compared the results to come to a conclusion. When analyzing the gender-stereotyped domains, the researcher might accidentally overlook or misinterpret certain behaviors. Since the researcher has never used MTS before, she re-evaluated each TV episode 3 times to eliminate any potential errors while collecting data. The researcher chose to collect quantitative data because a range of numerical values was needed to efficiently compare gender stereotypes in the US and India, and the change in their presence over the past decade. Through these values, the researcher was able to make a distinction between gender stereotype patterns and was also able to see by how much a specific stereotype increased or decreased. Overall, quantitative data helped the researcher answer the how and where portion of her research question. Simply looking at the qualitative data would not have been enough to make a clear distinction between the differences in gender-stereotyped patterns in the children's TV industry. --- Results --- Number of Female and Male Characters When determining each character's gender, the researcher viewed the tone of voice, characteristics, etc. Every television show analyzed was male dominant, meaning there were more male characters than female characters present (Tables 1 &2). Additionally, as seen from Table 4, when comparing American TV shows to Indian TV shows, Indian TV shows held an overall greater gender-stereotyped terminology than American TV shows. Additionally, as seen from Table 4, when comparing American TV shows to Indian TV shows, Indian TV shows held an overall greater presence of gender-stereotyped activities & occupation than American TV shows. --- Terminology Used When Describing Characters --- Activities & Occupation --- Talkativeness For American children's TV shows, the talking time for female characters decreased from 2010-2011 (x<unk> = 12.20) to 2020-2021 (x<unk> = 7.25). On the other hand, the talking time for male characters increased from 2010-2011 (x<unk> = 6.88) to 2020-2021 (x<unk> = 7.88). For Indian children's TV shows, there was an increase in the talking times for both female characters from 2010-2011 (x<unk> = 3.65) to 2020-2021 (x<unk> = 5.06), and male characters from 2010-2011 (x<unk> = 6.21) to 2020-2021 (x<unk> = 6.75). Overall, male characters talked drastically more than female characters for each episode. Additionally, as seen from Table 4, when comparing American TV shows to Indian TV shows, female characters held greater talking times in American TV shows than in Indian TV shows. --- Appearance For American children's TV shows, there was a decrease in feminine-stereotyped appearance from 2010-2011 (x<unk> = 8.60) to 2020-2021 (x<unk> = 5.81), and a slight decrease in masculine-stereotyped appearance from the 2010-2011 (x<unk> = 9.88) to the year 2020-2021 (x<unk> = 9.58). For Indian children's TV shows, there was a drastic increase in both feminine-stereotyped appearance from 2010-2011 (x<unk> = 10.37) to 2020-2021 (x<unk> = 13.29), and masculine-stereotyped appearance from 2010-2011 (x<unk> = 6.92) to 2020-2021 (x<unk> = 9.54). Additionally, as seen from Table 4, when comparing American TV shows to Indian TV shows, Indian TV shows held an overall greater presence of gender-stereotyped appearances than American TV shows. --- Discussion The researcher analyzed eight children's TV shows from various local networks (PBS, Nick Jr., etc.). The television industry is highly responsible for influencing a child and their beliefs. Therefore, the researcher needed to uncover various ways that TV shows can negatively affect a child's perception of gender. Additionally, the researcher yearned to find out whether certain gender-stereotyped patterns occur more in countries with different socioeconomic statuses (e.g., India and the U.S.). Lastly, it is important to discover whether the children's TV industry is addressing such gender-stereotyped patterns in their shows. If they are, then there is accumulated change and hope for future generations to not be influenced into believing such stereotypes. If they are not, then this study aims to bring attention to this urging topic. To the researcher's knowledge, her study is the first national comparative content analysis to be conducted for children's TV shows. In her analysis, the researcher did not include side characters (e.g., characters that did not have a major speaking role or characters that were placed in the episode for a maximum of 2 seconds). If she did not do this, then there would have been an abundance of unnecessary female and male characters considered in the analysis, which would have created a calculation error when finding the average means. --- Number of Female and Male Characters In the majority of all TV shows analyzed (both American and Indian TV shows), there were twice as many male characters than female characters present. These results resonate with previous studies conducted (Götz et al., 2008). These results send children the message that men are superior to women and that they also hold greater importance in society. --- Terminology Used When Describing Characters In the analysis of American children's TV shows from 2010-2011 to 2020-2021, there was a decrease in femininestereotyped terminology and an increase in masculine-stereotyped terminology. In the analysis of Indian children's TV shows from 2010-2011 to 2020-2021, there was a decrease in both feminine-and masculine-stereotyped terminology. Overall, these results are promising as they show a change or decrease in the presence of gender-stereotyped domains over the past decade. Additionally, when compared to American TV shows, Indian TV shows held a greater mean of gender-stereotyped terminology. These results send a false message to children on how boys and girls should behave. If these stereotype patterns persist, specifically in Indian TV shows, it will train children to address/characterize each other based on gender, which is incorrect. --- Activities & Occupation In the analysis of American children's TV shows from 2010-2011 to 2020-2021, there was a decrease in femininestereotyped activities & occupation, and an increase in masculine-stereotyped activities & occupation. In the analysis of Indian children's TV shows from 2010-2011 to 2020-2021, there was a decrease in both feminine-and masculinestereotyped activities & occupation. In general, these results show that as time passes, change in the children's TV industry is occurring and there is less pressure on children to follow specific gender stereotype ideals. Additionally, when compared to American TV shows, Indian TV shows held a greater mean of gender-stereotyped activities & occupations. Overall, this informs us that the Indian TV industry requires more reform than the American TV sector. It also demonstrates the ongoing pressure that children in India are subjected to in terms of society's expectations. --- Talkativeness In the analysis of American children's TV shows from 2010-2011 to 2020-2021, the talking time for female characters decreased, while the talking time for male characters increased. This result counters the stereotype that women are more talkative and involved, while men are more silent and unattached. This is a negative trend that has been occurring over the past decade with female characters receiving less screen time and recognition. --- Appearance In the analysis of American children's TV shows from 2010-2011 to 2020-2021, there was a decrease in both feminine-and masculine-stereotyped appearance. On the other hand, for Indian children's TV shows from 2010-2011 to 2020-2021, there was an increase in both feminine-and masculine-stereotyped appearances. These results depict a positive trend in the American TV industry and a negative trend in the Indian TV industry over the past decade. While there is evident change occurring in the American TV sector, this is not the case for the Indian TV sector. Additionally, when compared to American TV shows, Indian TV shows held a greater mean of gender-stereotyped appearances. Specifically, in Indian children's TV shows, male characters were depicted with muscular body types and colors similar to blue and gray, while female characters were depicted with thin/slim body types and pink and purple colors. If these gender-stereotyped patterns continue, they can have a negative impact on a child's perception of how they should dress and express themselves. --- Conclusion Television shows have gradually become more prominent in children's lives. However, gender stereotypes are currently being questioned in children's television shows due to varied gender stereotype depictions (Kim et al., 2016). These gender stereotypes inadvertently influence a child's perceptions of how boys and girls should be treated, and perpetuate unnecessary inequality in today's society. Additionally, it is known that countries of different socioeconomic levels vary in the progress of gender equality (Rozario et al., 2017). For example, countries like India and the U.S. have different levels of gender stereotype domains present in their children's TV shows. This variation ultimately affects how a majority of girls and boys act as they grow older. Therefore, it was important to analyze how gender is represented in the TV industry, and ultimately identify if change is needed within specific children's TV shows. Out of the five gender-stereotyped domains analyzed, three of them (the prominence of female and male characters, activities and occupation, and appearance) showed that there is still a drastic change needed in the children's TV industry. As discussed previously, while there was promising evidence that gender stereotype patterns have been decreasing over the past decade, there is still a need for greater equality. For example, when comparing American and Indian TV shows to the five gender-stereotyped domains, four of them (terminology, activities and occupation, talkativeness, and appearance) showed that Indian TV shows hold greater gender-stereotyped patterns when compared to American TV shows. These results hopefully urge the TV industry to continue decreasing the number of gender stereotypes present in children's TV shows. --- Limitations and Future Directions The researcher also understands that her research study contains limitations. First, the researcher examined a small number of TV shows and episodes. In total, the researcher only analyzed 4 TV shows and 8 episodes (2 episodes per TV show). However, there are many other TV shows that play(ed) a prominent role in a child's life (e.g., Bob the Builder, Angelina Ballerina, Ek Tha Jungle). Additionally, the researcher only coded for a total of 5 gender-stereotyped domains. There are many additional coding variables that could have been analyzed when conducting this study. For example, when coding for gender-stereotyped appearance, the researcher could have split the category further to incorporate other gender-stereotyped factors. When conducting a similar study, the researcher recommends that future researchers analyze a greater number of children's TV shows and episodes. Finally, she advises researchers to broaden the scope of their coding variables to have a better understanding of the many gender stereotypes present in children's TV shows.

Gender stereotypes negatively affect people's views and opinions on gender and, thus, bolster gender inequalities. Recently, the children's TV industry has been reinforcing gender stereotypes throughout its shows, which can be highly detrimental to a child's upbringing. In addition, countries of different socioeconomic levels vary in the progress of gender equality. This discrepancy creates a barrier in identifying where exactly change needs to be initiated in the children's TV industry. Therefore, this study 1) identifies gender stereotypes in children's TV shows, 2) examines the differences in gender stereotypes in the United States and India, and 3) examines whether or not gender stereotype patterns have changed over the past decade. Based on specified criteria, the 4 following TV shows were selected: 1) Arthur, 2) Thomas & Friends, 3) Chhota Bheem, and 4) Bapu and assessed on 5 gender-stereotyped domains: 1) the prominence of female and male characters, 2) terms used when describing characters, 3) activities & occupation, 4) talkativeness, and 5) appearance. To measure the variability of gender stereotypes in each show, momentary-time sampling, and average mean were also used. Overall, findings show that gender stereotype patterns are present in children's TV shows, have decreased over the past decade, and have a greater presence in Indian TV shows than in American TV shows. These results show a need for drastic change and equity worldwide in the children's TV industry. A research study was conducted to explore gender stereotypes in preschool children's TV in 2020. In their paper, authors Abigail Walsh and Campbell Leaper, psychologists and professors at UC Santa Cruz, examined gender representations in TV targeted at preschool-age children. The authors' sample programs were three randomly selected episodes from 34 TV preschool shows. These TV shows ranged across four different networks: Nick Jr., Disney Jr., and PBS. Walsh and Leaper analyzed the three selected episodes in conjunction with seven specific gender stereotype domains: (1) Talking, (2) Supportive Speech, (3) Suggestions, (4) Aggressions, (5) Activities, (6) Appearance, and (7) Prominence of male and female lead characters. They collected data relating to the seven domains and concluded that "analyses revealed ways that gender continues to be represented in biased and stereotypical ways in contemporary television programs aimed at preschool-age children" (Walsh & Leaper, 2020). However, Walsh and Leaper's study solely examined TV shows in the U.S. when countless shows selected were also consumed in other countries. Aysen Bakir, a professor in the Marketing department at Illinois State University, examined gender roles in children's TV commercials in the U.S. and India. According to Bakir's study, female voiceovers were used in 25.3% of Indian commercials and 30.3% of American commercials. On the other hand, male voiceovers were used in 33.8% of Indian commercials and 59.1% of American commercials. Indian and American TV commercials showed an overall greater representation of males than females. Countries of different socioeconomic levels vary in the progress of gender equality (Rozario et al., 2017). Therefore, a national comparison between two countries, such as India and the U.S., will help identify whether prominent issues lie within children's TV shows and where change should be initiated in the children's TV industry. In 1995, another study by Teresa L. Thompson and Eugenia Zerbinos, professors at the University of Dayton, compared gender-stereotyped patterns in TV shows that were decades apart. Thompson and Zerbinos conducted a study that updated research done in the 1970s surrounding gender stereotypes in cartoons. Thompson and Zerbinos' study identified that males were given more representation than females. On average, there were more male characters than female characters in each episode, with only 55% of cartoons holding one to five female lead characters, and 100% of cartoons holding one to five male lead characters. Compared to data in the 1970s, post-1980 female characters had a slightly greater talking time and frequently answered more questions. However, the data was nearly identical, meaning there were no drastic changes in gender stereotype patterns. Thompson and Zerbinos' study was conducted over 20 years ago. So, it is unlikely that their data precisely represents gender stereotypes in current children's TV shows. A recent study would be a more accurate representation of gender-stereotyped portrayals. For example, in a study done in 1999 by Mark R. Barner, an assistant professor of Communication Studies at Niagara University, female characters mainly participated in nurturance and non-physical activities, like painting, while male characters participated in physical activities, like construction. Additionally, Barner stated that female characters were generally neglected, ignored, and placed inferior to male characters. For the past decade, females have continued to be underrepresented on TV shows, including cartoons and commercials. Very few studies have addressed whether or not such female portrayal has improved or deteriorated in recent years. Therefore, it is necessary to identify whether or not gender stereotype patterns have increased, decreased, or stayed the same in the past decade. Gender stereotypes are prevalent issues in society, and their possible application in children's TV shows must be addressed. Television is the most persuasive tool in the developmental period of a child's life. To protect future generations from luring themselves into such a dangerous trap, it is necessary to fill in major gaps within the portrayal of gender stereotypes in children's TV shows. These gaps include analyzing gender stereotypes from a nationwide perspective, specifically in countries with different socioeconomic levels, such as India and the U.S. Additionally, to view whether or not the TV industry is addressing its gender stereotype portrayals in children's TV shows, it is necessary to conduct a comparative analysis from decade-old TV shows to current airing TV shows. This will provide a clear understanding of where change, if any, is needed to take place.

Introduction Sex workers and their clients remain at high risk of contracting sexually transmitted infections (STIs) (Dias 2015;Mc Grath-Lone et al. 2014). Indeed, commercial sex work represents an important channel for the diffusion of STIs (Shannon et al. 2014), and their prevalence is estimated to be high among female sex workers in European countries whilst HIV prevalence is generally low (Platt et al. 2011). This also appears to be the case in the UK, where STIs tend to be more prevalent among both female sex workers and men paying for sex, although an increasing diversity of sex markets has led to different risk levels of STI infection (Dias 2015). Clients remain at greater risk of acquiring STIs and contributing to their transmission (Jones et al. 2015). A study of indoor-working female sex workers in London found that migrant sex workers tend to see more clients and are less likely to use contraception than UK-born ones, although both groups reported more consistent condom use for penetrative sex than oral sex (Platt et al. 2011). There remains, however, little recent epidemiological knowledge as to the specific protective behaviours used by female sex workers in the UK or European contexts. The spread of infectious diseases across networks can be modelled modelled as simple contagion, i.e. contagion 'for which a single activated source can be sufficient for transmission' (Guilbeault, Becker, and Centola 2018, 4). Effective STI prevention requires understanding the structure and composition of the sexual networks across which infections are transmitted through simple contagion, i.e. the set of individuals and the sexual relationships among them (Chami et al. 2017;Centola 2018). However, obtaining a complete map of commercial sex networks using traditional data collection methods (e.g. contact tracing or census data) presents several challenges (Zhang and Centola 2019;Klovdahl 2005), and studies addressing structural network characteristics usually draw on relatively small, geographically limited populations (Shushtari et al. 2018). Internet-based sex markets have grown in recent years (Sanders et al. 2018). They offer sex workers and their clients new ways to communicate with each other, and provide new opportunities to develop effective interventions to target large populations and reduce STI diffusion (Hsieh et al. 2014). The analysis presented in this article builds on the literature on sexual networks. It draws on user-generated data from a popular website dedicated to review sex workers' services to create an original empirical dataset of the commercial sex network in the UK, with implications for STI prevention. --- Sexual networks and STIs The study of sexual networks plays a crucial role in understanding both the rate and the extent of STI diffusion (Newman 2002). First, network structural characteristics such as network cohesion, average path and tendency toward clustering, and connectivity can tell us how quickly or how far STIs might spread across communities (Campbell and Salathé 2013). For example, in a highly dense network or a network where individuals have high contact rates, diseases spread quickly as most of the members are closely connected to each other (Doherty et al. 2005). STI spread is also accelerated by relatively short paths between any two individuals, and the tendency toward clustering, i.e. the tendency of an individual's contacts to have contacts among each other and to cluster into densely connected groups (Guilbeault, Becker, and Centola 2018). The presence of many unconnected pairs instead affects the incidence of STIs as it limits the extent of their diffusion (Doherty et al. 2005). Second, actors' positions within the network increase their risks of contracting STIs and their likelihood of spreading infection (May and Lloyd 2001). Rothenberg et al. (2007) showed, in a community of teenagers in rural Georgia, that participants with syphilis had a higher degree and betweenness centrality than participants without syphilis, where degree centrality measures an actor's number of relationships with other network participants, and betweenness centrality measures the extent to which an actor lies on the shortest path between any pair of network participants (Wasserman and Faust 1994). The latter, in particular, is useful to identify bridging people, i.e. individuals who are more likely to facilitate STI spread across communities (Youm 2015). The distribution of centrality scores across network participants is also relevant. Sexual networks are usually characterised by a small number of very active individuals, and a large number of actors with only one or few sexual contacts. A positively skewed degree distribution increases the rate of STI diffusion (Newman 2002) but also makes the immunisation of these highly connected individuals particularly effective (Doherty et al. 2005). Finally, assortative mixing by degree, i.e. the tendency of individuals to interact with others with similar level of activity, increases STI diffusion rate, but limits the extent of spreading (May and Lloyd 2001). Disassortative mixing, on the other hand, occurs when there are contacts between highly connected and less-connected actors, and increases the extent of STI diffusion (Youm 2015). Despite the value of sexual networks for understanding STI risk, few studies have considered the network structure, position and composition of commercial sex networks (Schrager et al. 2013;Latkin et al. 2011). A systematic review of social network analyses of female sex workers and HIV risk behaviours found only four studies addressing structural network characteristics, each of which drew on relatively small, geographically limited populations (Shushtari et al. 2018). There are two main reasons for these gaps. First, these studies require the use of complete network design, i.e. the collection of data from all the members of a community, which is expensive, time consuming and raises ethical concerns as it asks participants to name their sexual partners (Klovdahl 2005). Second, hidden populations such as sex workers and their clients are typically hard to reach (Valente and Pitts 2017). Thus, prior studies have been small-scale in nature or have relied on contact tracing to study egocentric networks. Both approaches miss the complexity and heterogeneity of commercial sexual networks at regional and even national levels (Hao et al. 2015;Klovdahl 2005). --- Sex work and digital technologies The internet has had a transformative impact on the sex industry and, consequently, on the way we research it. Digital platforms changed the way sex workers and clients interact before the in-person activity takes place (Sanders et al. 2018;Stewart Cunningham et al. 2018). Before the advent of the internet, sex workers could reach clients by streetwalking specific urban areas, or working in brothels, massage parlours and walk-ups (Scott Cunningham and Kendall 2011;Crotty and Bouché 2018). Today sex workers can advertise their services, and be contacted by clients, via advertising platforms, agencyowned or personal websites, and social media platforms (e.g. Twitter or Facebook) (Grov et al. 2017;Gezinski et al. 2016). Since the beginning of the 2000s, there has also been a large diffusion of customer review websites, where clients can write and share detailed descriptions of their experiences with sex workers (Crotty and Bouché 2018;Gezinski et al. 2016). These reviews can play an important role in potential clients' decision-making (Sanders et al. 2018). In some cases, a negative review is enough for putting a sex worker out of business. On the contrary, positive reviews can help build trust among clients that the sex worker is genuine and 'professional' (Sanders et al. 2019;Noack-Lundberg et al. 2019). Analysing a Brazilian online community over six years, Rocha et al. (2010) found that a good review is a predictor of the future popularity of the sex worker. This means that reviews from clients can alter sex workers' centrality within commercial sex networks by attracting both many local clients and sex tourists from other cities. The direct consequence of this is that customers' online activities such as forum discussions and reviews can shape offline interactions between sex workers and their clients. Contemporaneously, the study finds a strong influence of offline factors, such as urbanity and geography, on the network structure. Hsieh et al. (2014) find similar results regarding the relevance of geography from the analysis of online communities in Brazil and the US. The authors find that the travelling of clients and sex workers to different locations can explain around 50% of their centrality in the network. Therefore, the study of online commercial sex networks is important for at least two reasons. First, given their popularity among both sex workers and clients, these platforms are a valuable source of information for understanding the structure of online communities (Rocha, Liljeros, and Holme 2010;2011;Hsieh, Ková<unk>k, and Logan 2014). For instance, data from these platforms can be used to identify key players in the virtual community, i.e. popular sex workers and active clients (Zhang and Centola 2019). Second, and perhaps more importantly, as online and offline networks overlap and shape each other, online data can provide an insight into the structure and composition of offline commercial sex networks (Rocha, Liljeros, and Holme 2010;Hsieh, Ková<unk>k, and Logan 2014). Sexual contacts extracted from popular online communities can compensate for the lack of traditional data on complete networks, and can be analysed to suggest prevention strategies based on network properties. --- The current study This study contributes to the research on sexual networks by examining a national commercial sex network of off-street female sex workers (i.e. sex workers in commercial venues) and their male clients. While abundant research exists on both the application of social network analysis for public health interventions and online sex communities, the linkage between these two dimensions is still under-explored. This study uses internetmediated data as an alternative approach to sequenced sampling to collect large scale sexual contact data. Specifically, we collected sexual contact information from the largest online community dedicated to reviewing female sex workers' services in the UK. The study builds on previous research on online socio-sexual networks (Hsieh, Ková<unk>k, and Logan 2014;Rocha, Liljeros, and Holme 2011) by answering the following questions. What are the main structural characteristics of this online sex community? How are direct sexual contacts distributed across sex workers and their clients? How can network structural characteristics and individual positions suggest effective STI prevention strategies? --- Method --- Data and procedure The online community from which we collected our data is openly accessible to anyone, although visitors need to confirm that they are older than eighteen the first time they access the website. It was created in 1999 for the exchange of information between sex workers and clients. The website specifically focuses on the off-street section of the sex market, which includes both sex workers working independently or for third parties. The off-street sex market represents the largest sector of the sex market in England and Wales, with some figures showing that up to three quarter of sex workers work in various indoor settings (Home Affairs Committee 2016). The platform offers several services such as a message board, escort advertisements, and web camming. However, its main function is reporting male clients' reviews of female sex workers. Each review contains dyadic information about client's username and sex worker's name, date and time, city, venue (e.g. escort agency, massage parlour), duration of the encounter, price paid, and three written accounts describing each of the venue, the sex worker and the intercourse. While clients need to login into the platform to provide a review, the website is free to search for service providers, reviews and sex worker's profiles. Note that reviews refer exclusively to face-to-face encounters, and not to services such as web camming or instant messaging. We developed a crawling and scraping software to collect this information from the online community. The software automatically and daily accesses, crawls, fetches and stores this information to a database. We analysed data reported between January 2014 and December 2017. We identified clients and sex workers by their usernames. Each client and each sex worker formed a vertex in the network. The identification of unique clients was straightforward as this relied on unique account names. However, identification of unique sex workers was more complicated as they were identified by generic, client-reported street names such as Bethan or Cleo. Our approach to identifying unique sex workers was conservative. We assumed that two or more reviews reporting the same name, for instance Bethan, referred to the same sex worker if they also reported the same venue and city. For instance, Review #1 and Review #2 refer to the same sex worker, Bethan, if they both report the same venue, i.e. Venue Y, and the same city, i.e. London. Two reviews reporting the same name and the same city but different venues, for instance Venue X and Y, refer instead to two different sex workers. To check accuracy and validity of our approach to the identification of sex workers, we inspected a random sample of 500 reviews reporting the same names but for different venues and cities to ensure that this approach was valid. Because reviews include descriptions of sex workers' appearance, we used this additional information to understand if, for instance, the Bethan in London is the same working in Cardiff. Despite sex workers sometimes working in more than one geographical area regularly or for a short period of time (Sanders et al. 2018), in none of these checks were we able to identify with confidence if a sex worker was working in two different cities or venues at the same time, suggesting the validity of our conservative approach. We also performed an additional check that our process did not merge separate sex workers into one. For 200 sex workers with more than 4 reviews we used the information about the sex worker's appearance to check whether they reported significant differences. For instance, descriptions reporting different ethnicity, physique, age, etc. for the same identified sex worker would point out a mistake in our coding process. In all the checks performed, we were confident that, from the information provided, the descriptions referred to the same sex worker, confirming the soundness of our approach. Finally, we transformed reviews into a map of a national off-street commercial sex network. Specifically, we established a link between Client A and Sex Worker B, every time Client A posted a review about Sex Worker B. We abstracted the frequency of each link assuming that once formed the link is persistent; that is, multiple reviews for the same sex worker from the same client formed a single connection. This process resulted in a binary, bipartite network as clients cannot directly connect to other clients, and sex workers cannot directly connect to other sex workers. The resulting network is likely to be a specific sub-set of the British off-street sex market of female sex workers and their male clients. Mapping the online sex industry in the UK is a challenging endeavour, and data collected from advertising or other online platforms are never complete (Sanders et al. 2018). Reviews are written by a minority of clients (Sanders et al. 2018) who seem to be among the most experienced or active ones. Indeed, clients often refer to previous experiences when writing their reviews, and use the same, unique argot (Holt and Blevins 2007). --- Measures We quantified network parameters using social network analysis measures.1 First, we identified the main structural characteristics of the commercial sex network which are known to affect STI diffusion (Hsieh, Ková<unk>k, and Logan 2014). Degree assortativity is the tendency of vertices with similar number of links to preferentially associate with each other (Newman 2003). The giant component is the largest connected subset of vertices in the network. For this subset, we can also calculate the average geodesic distance, i.e. the mean shortest path between any two vertices, and the diameter, i.e. the maximum distance between two vertices of the giant component (Wasserman and Faust 1994). We also calculated the clustering coefficient using Opsahl's approach (2011) to detect clustering in a two-mode network. Second, we calculated three different centrality scores for each vertex (sex workers and clients) in the network: degree centrality, betweenness centrality, and closeness centrality. Degree centrality measures the number of individuals with which each person in the network is connected. Betweenness centrality instead measures the number of times a vertex is along the shortest paths between any two other vertices in the network (Wasserman and Faust 1994). This measure is often used to identify individuals who are central because they are brokers; that is, they control communications or facilitate the exchange of resources within the network. Finally, closeness centrality is the inverse of farness, which measures the distance of an individual from every other individual in the network. We use Opsahl et al.'s (2010) approach, which enables us to calculate a closeness score for each individual in the network despite the commercial sex network having several disconnected components. In the last part of the analysis we tested the impact of different vertex removal strategies on the level of connectivity of the network, and on individuals' closeness to all other vertices in the network. Connectivity was measured using Krackhardt's connectedness score, which is equal to the fraction of all dyads (namely the combination of two vertices) connected through an undirected path (Krackhardt 1994). Individuals' closeness was measured by calculating the average closeness centrality score for all vertices in the network. By using Opsahl et al.'s (2010) approach, once again we circumvent the issue of having several disconnected components. Vertices were removed 1) randomly, 2) based on degree centrality scores, 3) based on betweenness centrality scores, and 4) based on closeness centrality scores. For random removal, we simulated vertex removal 100 times, and reported the average connectedness score and closeness score for the 100 replications. Following Hsieh et al.'s approach (2014), vertex removal was also based on the role of individuals within the network. Each of the three strategies mentioned above was thus applied to 1) clients, 2) sex workers, and 3) both clients and sex workers. The twelve resulting removal strategies approximate the impacts of different strategies which might focus on individuals with a specific role or position within the network. Although we gradually removed all vertices from the network, following Hsieh et al.'s example (2014), we report connectedness scores and average closeness scores only for the first 5% as we expect STI prevention strategies to reach only a limited number of individuals and to be effective when a small fraction of individuals are targeted (Newman 2002). Finally, it is worth mentioning that for the purpose of this study, targeting vertices does not necessarily mean physically removing people from the network but rather using immunisation strategies to prevent STI transmission. --- Results --- FIGURE 1 ABOUT HERE TABLE 1 ABOUT HERE --- General network characteristics The network is composed of 6,426 edges, or connections between sex workers and clients. Because each vertex corresponds to one sex worker or one client, 6,477 people were represented in the network, of which most are sex workers (60%). In addition, the network covers 1,656 venues and 328 geographical locations (see Table 1). Greater London and South East England, with respectively 3,154 and 1,502, are the two geographical areas concentrating the highest number of reviews of commercial sex encounters. A key feature of networks is the 'giant component', i. This means that, if we assume that the adoption of safe sex practices is equally distributed in the network, buying or performing sex in these two cities put people at a potential higher risk of contracting STIs, given the higher interconnectivity among vertices in the giant component. Indeed, the average geodesic distance, which represents the shortest path between any two vertices, is relatively short; on average, a person can reach any other person in the giant component using eight intermediaries (or nine steps). Their tendency to form clusters is, however, relatively low, given the clustering coefficient ranges from 0 (low clustering) and 1 (high clustering), and the giant component has a coefficient of 0.12. From an STI diffusion perspective, the size of the giant component is the maximum number of people who can be reached by an STI outbreak. Whilst the commercial sex network is likely to be more connected than our data suggest, the disproportionate number of clients and sex workers in London and Milton Keynes suggests that these two cities play a relevant role in the commercial sex network and may be key for STIs prevention. Degree assortativity refers to the correlation of the number of links that any two connected individuals have. A positive assortativity value indicates that active clients tend to meet with popular sex workers, and clients who have used the service only once or a few times tend to meet with unpopular or less popular sex workers (or sex workers with a limited number of selected clients). In our network, however, the negative assortativity score (-0.12, p<unk>0.001) indicates that less active clients tend to buy services from popular (or highly reviewed) sex workers, and more active clients tend to by services from less popular sex workers. --- TABLE 2 ABOUT HERE FIGURE 2 ABOUT HERE --- Vertex centrality Table 2 compares degree, betweenness, and closeness scores of clients and sex workers. The mean degree centrality score is 1.98, i.e. individuals in the network have, on average, approximately two direct links with other network members. When we consider the role of individuals in the network, we observe that clients have, on average, more direct links than sex workers (mean degree is 2.46 and 1.66, respectively). Clients' degree centrality scores also show wider variation than sex workers' scores (standard deviation is 5.03 and 2.09, respectively). Similar considerations apply to betweenness centrality, suggesting that clients are more central than sex workers both locally (degree) and globally (betweenness). There is less variation in clients' and sex workers' closeness scores. Figure 2 shows the cumulative degree, betweenness, and closeness distribution for clients and sex workers. As in other sexual networks (Rocha, Liljeros, and Holme 2011;Hsieh, Ková<unk>k, and Logan 2014), there is a small number of very active clients and popular sex workers, and a majority of actors with only one or a few connections with other individuals in the network, suggesting that immunisation strategies targeting the most central people in the network is likely to limit STI diffusion. --- Vertices removal to model prevention strategies The overall network connectivity is 0.345, i.e. 34.5% of all dyads are connected through an undirected path. Network connectivity decreases as we remove vertices randomly, based on vertices' degree, betweenness or closeness scores, or based on the role individuals have in the sex network (clients or sex workers). Of the twelve removal strategies tested, random removal strategies were the least effective (Figure 3). By removing 5% of the vertices randomly and irrespective of their role, the final network connectivity is 0.31. The random removal of clients is slightly more effective, with 28% of the dyads still connected through an undirected path after the removal of 5% of the vertices. Interventions based on centrality scores are instead much more effective. The removal of just 2% of sex workers based on their degree or betweenness centrality leads the network connectivity scores to 0.19 and 0.16, respectively. However, strategies focusing on clients seems to have a more disruptive effect compared to those focusing on sex workers. Network connectivity drops to 0.09 (degree-based) or 0.08 (betweenness-based) by removing 2% of the clients according to their centrality scores. Connectivity drops to 0.21 and 0.27 after closeness-based removal of clients and sex workers, respectively. Finally, type-independent strategies based on centrality scores are the most effective. After the removal of 5% of vertices with the highest betweenness centrality scores, connectedness drops to 0.009, whilst role-independent removal based on degree centrality scores leads to connectedness scores of 0.001, signalling the most effective outbreak response strategy for this online community. Connectivity after the removal of 5% of vertices with the highest closeness scores remains at 0.14. The effectiveness of different types of vertices removal is similar when assessed using the average closeness centrality of the individuals in the network. The overall average closeness is 290. It drops to 243 after the random removal of 5% of vertices, and to 66 after closeness-based removal of the same percentage of vertices. Degree-and betweenness-based removals appear to be the most effective, with the average closeness score dropping to 2 and 9, respectively. Again, strategies focusing on clients seems to have a more disruptive effect compared to those focusing on sex workers, whether removal is based on vertices' degree, betweenness, or closeness centrality scores. --- FIGURE 3 ABOUT HERE FIGURE 4 ABOUT HERE --- Discussion We present, for the first time in the UK, a social network analysis of sex workers and clients using internet-mediated data. Similar to other studies of online sexual networks, we found a small degree of disassortativity in our network (Hsieh, Ková<unk>k, and Logan 2014;Rocha, Liljeros, and Holme 2010). Clients that have used the service only one or a few times tend to visit highly reviewed sex workers whereas clients with many reviews tend to buy sex from less popular sex workers. In practical terms, this may suggest that relatively new clients start out by visiting the most popular sex workers before gradually expanding their visits to less commonly reviewed ones. We also found that the network was dominated by a giant component, which included 59% of nodes in the network and which was geographically located in London and its suburbs. A relatively high number of reviews in London is expected, but the predominant role of Milton Keynes in this online community comes as a surprise. We have identified two possible reasons for the high concentration of this online community in Milton Keynes. First, the offline commercial sex market in Milton Keynes is flourishing and very active. Milton Keynes is located 50 miles northwest of London, has a population of about 250,000 residents and is one of the fastest growing cities in the UK. It is conveniently located near several towns and cities (e.g. Northampton, Bedford, Luton, Leicester, Cambridge, and Oxford), and more than 7 million people live within an hour drive from Milton Keynes. It is an economically prosperous area characterised by the presence of many companies, and a large share of high-skilled jobs and, consequentially, high wages. All these factors contribute to Milton Keynes being a convenient place for sex workers to set up their businesses, and for clients to access such services. Second, clients active around Milton Keynes may be unusually engaged with this online community, i.e. they tend to write more reviews than clients located elsewhere. This might be due to the intermediary role that agencies play between clients and sex workers in the area. When sex workers work through an agency, the latter takes care of the advertising, clients' screenings and bookings. If a few agencies have acquired a prominent role in Milton Keynes, they may be actively and successfully encouraging clients to write a review after the sexual encounter (Sanders et al. 2019). Irrespective of which hypothesis may be correct, the relevance of Milton Keynes in the online community does not necessarily reflect its role in the offline sex industry. The town could be an important hub for the commercial sex industry in the UK, but online reviews might exaggerate its role. Future research, employing more traditional methods such as surveys, interviews and ethnography, might want to deepen the insight provided by our study. The analysis also showed the role that clients and sex workers have in the network. Clients have a more central role both locally and across the network, and buying or selling sex in the two main hubs, London and Milton Keynes, is associated with high centrality scores, increasing their potential risk of contracting STIs. Counterintuitively, our analysis suggested that clients have more sexual contacts than sex workers. This last finding is likely to be artefactual given our data collection method and conservative approach in identifying sex workers. Nonetheless, the results provide an original insight into the behaviours of clients and their implication for STI transmission. Clients with high degree centrality scores, i.e. those with many sexual partners, travelled to several locations in the UK to buy sex. This may suggest that 1) prolific clients may travel looking for exclusive or potentially 'niche' services; 2) people that travel frequently for personal reasons can buy sex in different locations. Either way, prolific clients are geographically mobile and more likely than sex workers to bridge distant parts of the network (Soothill and Sanders 2005). While just a small group of clients travelled to more than two locations, some of them covered long distances. For instance, the most prolific one had 125 sexual partners across 28 different locations. Finally, a simulation of possible preventive strategies indicated that targeting vertices using network parameters is more effective than random targeting. Interventions focusing on the most active members-whether sex workers or clients-may be thus the most effective strategy to reduce disease spread and protect the community. Notably, there was little discernible difference in whether members were targeted based on their betweenness centrality (i.e. their role as brokers) or their degree centrality (i.e. the number of links in the network) whereas closeness-based removal underperformed compared to the other two centrality measures. The results of this study can inform effective interventions to prevent STI transmission through sex work. They can, in fact, be used to identify individuals that are at higher risk of contracting STIs given their position in the network. While not every sexual contact leads to STI diffusion, every contact increases the probability of transmission (Zhang and Centola 2019). This probability is heightened when sexual contacts involve central (i.e. active) individuals and risky sexual practices. If central actors contract an infection, it is highly likely that many others in the network will contract it as well. The immunization of highly connected vertices can reduce the risk that others in the network will be infected. Even if an infection enters the network, it will not easily reach other people, if highly connected members resort to safer sex practices (Valente 2017;Chami et al. 2017). Whilst the effectiveness of targeting highly connected nodes over random targeting or targeting everyone is well known (e.g. Newman 2002), this study provides evidence that this may be true for online commercial sex networks, too. Naturally, our approach presents some limitations. First, despite the continued growth in web applications for the mediation of client-sex worker relationships, our data are still likely to represent a sub-section of the entire sample of sex workers and clients engaged in off-street commercial sex in the UK, and our findings may be subject to selection bias. Whilst some network measures are likely to be relatively robust despite the bias in our data (e.g. degree distribution), others should be interpreted with caution (e.g. centrality scores for clients and sex workers).We also know very little about clients posting these reviews, and whether they correspond with the typical social media platform user (i.e. male, young, and in managerial, administrative, or professional occupation) (Sloan et al. 2013). Second, these reviews are self-posted and completely anonymous. As with any self-report survey, there exists a possibility of fictive data. Clients can post reviews of sex workers they have never met and may not post reviews of other encounters. However, clients and sex workers can flag-up fraudulent reviews to the moderator who can, in turn, remove the reviews from the website. The number of posted reviews is used to rank clients according to their reputation and experience, which disincentives underreporting and the use of multiple accounts. Similarly, sex workers would lose their accumulated social capital and established reputation if they used multiple names for different locations and venues (Holt and Blevins 2007).2 Third, we analysed a snapshot of the commercial sex network using reviews posted between 2014-2017 without taking time into consideration. Future research should model the dynamics of the commercial sex networks to assess how past reviews affect the creation of new links, and how the network evolves over time, with implications for STI diffusion and prevention. Future research could also explore how network structure and composition facilitate the diffusion of behavioural norms, including safe-sex practices (Argento et al. 2016) as online communities can play a key role in sharing STI prevention messages with hard-to-reach populations (Minichiello et al. 2015). This would require an understanding of how the diffusion of safe-sex practices-which is better understood as complex contagion, i.e. contagion that requires multiple contacts and social reinforcement (Guilbeault, Becker, and Centola 2018) -can be effectively achieved in online communities of sex workers and their clients, and how multiple contagions (e.g. STI diffusion and STI prevention campaigns) would interact in the network. --- Conclusions The relevance of online technologies in today's sex industry makes the use of large-scale online data increasingly important to understand commercial sex networks. Online data about commercial sex contacts can provide insights into the structural and geographical characteristics of sex networks that would be otherwise prohibitive to obtain using traditional data collection methods. The results of this study show that the giant component of this online network clustered around a major conurbation. But perhaps more surprisingly, an exurb of London, Milton Keynes, played a significant role in the giant component. This suggests a role for considering possibly 'unexpected' geographical dimensions of socio-sexual networks, including possible contextual features influencing popularity of sex work services. Our findings have several implications for public health policy and practice. Previous modelling studies of HIV prevention in sex workers have shown that small, incremental improvements in coverage of biomedical interventions (periodic condom inundation, uptake of pre-exposure prophylaxis) can effect substantial improvements in HIV incidence (Poteat et al. 2015). Indeed, under the most effective strategies simulated in this study, a low level of'vertex removal', representing coverage of interventions to block STI transmission, yielded substantial network effects. Future analyses should seek to understand individual, geographical, and temporal dimensions of the network as well as explore the possibility of using this community to improve population health. For example, what factors account for vertex centrality, and how did major public events (e.g. Olympic games) shape network configuration? In addition, simulation of outbreak control strategies should seek to understand the potential impacts of structural, rather than individual, interventions on network STI diffusion. Our analysis was only able to draw on random vertex selection to simulate outbreak control effectiveness. Future analyses could draw on temporal and probabilistic models to consider more nuanced intervention strategies. --- Figures --- Ethics approval This research has

Pl e a s e n o t e: C h a n g e s m a d e a s a r e s ul t of p u blis hi n g p r o c e s s e s s u c h a s c o py-e di ti n g, fo r m a t ti n g a n d p a g e n u m b e r s m a y n o t b e r efl e c t e d in t hi s v e r sio n. Fo r t h e d efi nitiv e v e r sio n of t hi s p u blic a tio n, pl e a s e r ef e r t o t h e p u blis h e d s o u r c e . You a r e a d vis e d t o c o n s ul t t h e p u blis h e r's v e r sio n if yo u wis h t o ci t e t hi s p a p er. This ve r sio n is b ei n g m a d e a v ail a bl e in a c c o r d a n c e wi t h p u blis h e r p olici e s. S e e h t t p://o r c a . cf. a c. u k/ p olici e s. h t ml fo r u s a g e p olici e s. Co py ri g h t a n d m o r al ri g h t s fo r p u blic a tio n s m a d e a v ail a bl e in ORCA a r e r e t ai n e d by t h e c o py ri g h t h ol d e r s .

public health policy and practice. Previous modelling studies of HIV prevention in sex workers have shown that small, incremental improvements in coverage of biomedical interventions (periodic condom inundation, uptake of pre-exposure prophylaxis) can effect substantial improvements in HIV incidence (Poteat et al. 2015). Indeed, under the most effective strategies simulated in this study, a low level of'vertex removal', representing coverage of interventions to block STI transmission, yielded substantial network effects. Future analyses should seek to understand individual, geographical, and temporal dimensions of the network as well as explore the possibility of using this community to improve population health. For example, what factors account for vertex centrality, and how did major public events (e.g. Olympic games) shape network configuration? In addition, simulation of outbreak control strategies should seek to understand the potential impacts of structural, rather than individual, interventions on network STI diffusion. Our analysis was only able to draw on random vertex selection to simulate outbreak control effectiveness. Future analyses could draw on temporal and probabilistic models to consider more nuanced intervention strategies. --- Figures --- Ethics approval This research has been approved by the School of Social Sciences Research Ethics Committee of Cardiff University (REF: SREC/2745).

Pl e a s e n o t e: C h a n g e s m a d e a s a r e s ul t of p u blis hi n g p r o c e s s e s s u c h a s c o py-e di ti n g, fo r m a t ti n g a n d p a g e n u m b e r s m a y n o t b e r efl e c t e d in t hi s v e r sio n. Fo r t h e d efi nitiv e v e r sio n of t hi s p u blic a tio n, pl e a s e r ef e r t o t h e p u blis h e d s o u r c e . You a r e a d vis e d t o c o n s ul t t h e p u blis h e r's v e r sio n if yo u wis h t o ci t e t hi s p a p er. This ve r sio n is b ei n g m a d e a v ail a bl e in a c c o r d a n c e wi t h p u blis h e r p olici e s. S e e h t t p://o r c a . cf. a c. u k/ p olici e s. h t ml fo r u s a g e p olici e s. Co py ri g h t a n d m o r al ri g h t s fo r p u blic a tio n s m a d e a v ail a bl e in ORCA a r e r e t ai n e d by t h e c o py ri g h t h ol d e r s .

Introduction In the last two decades, Cape Verdean governments have intensified the investment on the National Reproductive Health Program, aiming to provide universal and qualified services to adolescents, young people, and adults, in partnership with schools and youth centers (i.e., local associations where several sporting, cultural and community activities are developed) [1]. Around 20% of the Cape Verdean population is young (15-24 years old) and specific attention has been given to Sexual and Reproductive Health (SRH) of this group (e.g., through the creation of SRH Centers for teenagers and young people, and the inclusion of sex education in the scholar curriculum) [2]. Nevertheless, some data indicate the persistence of SRH problems, especially among the youngest population, namely: (a) premature initiation of sexual life and considerable high rates of early and unplanned pregnancies; (b) adoption of risk behaviors associated with sexual life initiation, such as alcohol consumption; (c) estimates of high rates of illegal abortions, linked to physical, economic, and cultural obstacles to its implementation inside the national health service; (d) insufficient knowledge about the HIV infection despite some improvements on recent indicators about the use of condoms [1][2][3][4]. The prevalence of teenage pregnancies and fertility-related health problems are among the most serious social problems in Cape Verde. According to the Cape Verdean Demographic and Reproductive Health Survey II (2005), about 19% of young women aged between 15 and 19 years old got pregnant at least one time, 15.2% were already mothers, and 3.7% were pregnant. There was a poor knowledge on and little use of contraceptives; likewise, the use of a contraceptive methods in the first sexual intercourse has been reported by only 20% girls and 18% boys [5]. Taking these epidemiological scenarios into account, we propose to develop, implement, and evaluate a community-based intervention program to promote better SRH related behaviors, and, ultimately, SRH-related wellbeing and quality of life across the life course of young Cape Verdeans (i.e., 15-24 years; [6]). The implementation and effectiveness assessment will be conducted in the city of Mindelo, as a pilot for national implementation. The intervention will be tailored for relevant subgroups (with identified health risk behaviors) and will be gender-and age-sensitive. This health promotion program will be also based on successful and already tested health interventions for improving SRH of young people (including Cape Verdean youngsters) [7], adapted to the needs and cultural-embedded values of the Cape Verdean young population, taking into account the perceptions of local health and educational professionals. A key point will be the cooperation with the community, through a social prescribing approach, i.e., an innovative intervention in primary healthcare units that allows a partnership to work with sources of support within the community. Within this perspective, the wider contextual determinants of health are important factors to address, as they that have a particular relevance on SRH behaviors [8]. Moreover, there is a growing evidence of mobile health interventions (mHealth) as cost-effective tools for improving youth SRH related behaviors, namely in low-to-middle income countries, especially because its specific advantages on minimizing youngsters' barriers on accessing SRH services [9]. Therefore, the coproduction and use of digital health solutions will be another key component of this program, to be implemented according to the complex interventions model [10]. The idea is that the intervention could benefit from both recognized advantages of social prescribing and digital tools on promoting health behavior change and the long-term wellbeing/quality of life. This intervention program will adapt to the Cape Verdean sociocultural context a set of good health promotion practices specifically developed in Portugal, benefiting from previous close partnership between academic and health institutions from Portugal and Cape Verde (e.g., Projeto Bairro Feliz (Happy Neighborhood's Project)), mostly involving immigrants and young people) [11]. A main assumption of this project regards to the potential of health and educational professionals on minimizing barriers to access to health services and on increasing communities' health literacy and their capacity to avoid risky behaviors, through individual's education and fostering health-oriented social activities. The behavioral change wheel will support the intervention development [12,13]. Considering that SRH is key to improve the quality of life across the life course [14], the proposed intervention will focus on promoting lifelong SRH-related wellbeing and quality of life. Due to the several definitions of quality of life and wellbeing, we will embrace a comprehensive approach, which includes not only the physical, mental, and social health but also life satisfaction (overall quality of life) [15]. Similarly, the SRH-related wellbeing is here conceptualized and will be assessed according to a socioecological approach, i.e., as intraindividually experienced SRH-related wellbeing but also socially and structurally influenced [16]. Thus, we intend to improve the youth's quality of life and wellbeing through this community-based intervention program, considering short-term outcomes at a multidimensional perspective: Hypothesis H1 (H1). At the individual level and self-care domain-increase in youth's perceived general self-efficacy, assertiveness, sense of coherence, and self-esteem related to SRH behaviors and decision taking; Hypothesis H2 (H2). Related to mental health and wellbeing domain-improvement in stress coping/resilience skills related to SRH; Hypothesis H3 (H3). At interpersonal and social domain-higher perceived social support (and related satisfaction) and improvements in the social cohesion relate to SRH. --- Hypothesis H4 (H4). In the subjective life domain-higher levels of happiness and satisfaction with life throughout adolescence and adulthood; Several of these short-term expected outcomes (e.g., sense of coherence, self-esteem, perceived stress) were also assessed in other salutogenic health promotion programs [17], due to be considered as relevant determinants of health behavior change. Figure S1 details its specific role in the context of the proposed intervention, by schematically represent and detail the I-Decide conceptual rational. According to the theory behind the proposed intervention, improvements on youngsters' adherence to healthy sexual and reproductive behaviors, cognitive function, health literacy (intermediate-term), and also in their physical health and psychosocial well-being (long-term) are also expected. --- Materials and Methods Considering the complex challenge of health behaviors change interventions, namely, on the SRH area, an Intervention Mapping plan will systematically guide the program's development and coordination [18,19]. The steps of this Intervention Mapping protocol will correspond to the proposed four main components of our program (Table S1), as identified and described below. Overall, the intervention will be developed within a timeline of three years, in which needs assessment and program's protocol development (Components 1 and 2) will be implemented subsequently. Afterwards, the program's implementation and evaluation (Components 3 and 4) will occur simultaneously, as an evidence-based local pilot intervention that could be afterwards extended to a national implementation. --- Component 1-Needs Assessment of SRH among Cape Verdean Youngsters The aim of this component is to identify the most relevant SRH-related needs, specific to youngsters' groups and target behaviors to address in the Cape Verdean context, through focus groups (FG) sessions with main stakeholders (i.e., Cape Verdean youngsters and local health and educational providers) [10]. It starts at the beginning of the program and is expected to develop for 9 months. To support the sociocultural adaptation of good practices tested in Portugal [11], two FG will be conducted in the country: one FG with 15-24 years old Cape Verdeans immigrants who already benefited from SRH interventions and the other with health professionals involved in those interventions. We also foresee a minimum (aiming to achieve data saturation) of 4 FG with youngsters and 4 FG with educational and health professionals (e.g., GPs, nurses, psychologists, high school teachers), in Cape Verde. Each FG will include 8 participants [20]. Sampling will follow intentional theoretical criteria: socioeconomic strata, educational levels, gender, age groups. Recruitment will be conducted by snowball methods while ensuring participants' heterogeneity. Data from FG (audio-recorded and verbatim transcribed) will be analyzed through descriptive thematic analysis [21], assisted by MAXQDA, and visually depicted in a Logic Model, which schematically represents the individual and environmental determinants of Cape Verdean SRH risk behaviors in relation to the health problem(s) intended to be reduced/eliminated (e.g., strong cultural/societal values around family and childbearing that contribute to early or unintended pregnancy) [18]. In addition, based on the results of the needs assessment, the proposed program's outcomes (e.g., perceived social support, self-efficacy, stress coping/resilience, assertiveness, happiness, satisfaction with life) will be confirmed and better specified, in order to assess the effectiveness of the intervention (Component 4). --- Component 2-Delphi Procedure for Designing the Intervention Protocol and Components A Delphi expert-consensus technique will be used, as an approach to enhance effective decision-making in health care. The Delphi methods has some advantages regarding other methods for obtaining expert opinion and consensus (e.g., expert round tables) [22,23]. This component is expected to last 9 months and aims to map actions responding to SRH-related needs, detected in Component 1 [10]. The Delphi panel will include social workers, public health professionals (e.g., nurses, general practitioners, psychologists), academics, members from national and international nongovernmental organizations (NGOs), and policy makers (e.g., working on the national health and youth ministries). Panel experts should have, at least, five years of professional experience on the youth SRH area in Cape Verde. The sample size for constructing a Delphi panel will be oriented by the effort to maximize expertise about the Delphi subject of interest (homogeneity criteria, also ensuring to involve all relevant experts in the specific area), together with different and complementary perspectives (heterogeneity criteria). Accordingly, the recruitment will be conducted by using snowball methods, being expected that the professionals recognized by their peers as experts on SRH will be identified and then invited to participate in the project. Due to the snowball sampling procedure, it is not possible to indicate a sample size that may be considered inclusive enough. Ultimately, the sample size will be defined by the panel itself. Anyway, a sample size of at least 25 experts will be required for guaranteeing a stability of answers (maximum of 50 items), as proposed by Ralitsa et al. [24]. The Delphi method will follow a minimum of two rounds. The panelists will fill in a questionnaire about good practices in SRH. Data collection form will include two types of questions: 1. Related to participants' agreement about different paradigms, models, contents and skills to promote among (youth) SRH, using a 5-point Likert scale (1 = totally disagree to 5 = totally agree); 2. Regarding operational aspects of the intervention (e.g., number, frequency, duration and locals of activities), using single choice answer formats. Throughout the questionnaire, the experts will be also encouraged to make comments, propose new items, and/or reformulate the items they attributed a low score to (i.e., "1-, 2 or 3" on a 5-point scale; "Not relevant" in a binary response). At the final of each round, the item's consensus level will be calculated in order to decide about its approval (with predefined consensus criteria for each indicator). Once again, questions about participants' sociodemographic and professional-related information will be collected and considered for the analysis. This component will result in the design of the intervention program's protocol, with the identification of the theory-based behavior change methods that better apply to the previously identified SRH problems of our target group and the decision on how to better apply them [10,18,25]. Thus, the draft of the program will emerge theoretically, but specially, from this Delphi study will be fine-tuned a-posteriori by local community partners and by the target group, minimizing resistance to its implementation, according to the proposed community-based participatory approach [26]. --- Component 3-Program's Implementation: Social Prescribing and Digital Health Activities The implementation of the intervention will be informed by the work developed within the previous components and is expected to last 12 months. The baseline framework of the intervention is supported by a combination of Social Prescribing and digital health literacy activities [8][9][10]. As a starting point, primary health care professionals (e.g., nurses, general practitioners, psychologists) will assess young patients during their interventions/consultations and will refer them to a facilitator/navigator (e.g., a Cape Verdean doctoral student) that will then provide individual support to access nonmedical/community resources of SRH-related health behaviors (e.g., educational professionals, local associations) [27]. A minimum of 2500 youngsters and of 100 professionals will be invited to participate in at least one of the activities of the program. With this procedure is assumed that it will be possible to retain at least 400 youngsters in the implementation, which will allow for the detection of relatively small differences of the intervention. The intervention will consist in the involvement of the young Cape Verdeans in prescribed social and digital health related activities, which will be mainly defined in cocollaboration both with the target group and the community partners, due to the proposed community-based participatory approach [26]. Considering other developed and tested promotion programs for young people in SRH area, it may include activities such as the involvement in creative art groups and in a final Portuguese-Cape Verdean exhibition about health SRH-related competences, enhancing the quality of life across the life course and positive learning about SRH salutogenic experiences. In the involved community institutions, they can collaborate in the coproduction of some digital information about SRH, to be included in a SMS and web-based question-and-answer service, promoting content and critical SRH literacy. In addition, in the coproduction of m-health promotion tools that could be afterward used by themselves or by other young Cape Verdean peer (e.g., a free mobile quiz game about SRH contents in which participants could win credits to make calls) and that facilitate knowledge sharing and SRH behavior change [9,28]. Activities related with the program's dissemination are also planned, aiming to facilitate networking and communication among the scientific community, health professionals, and the general public [29]. These outreach initiatives will start three months after the beginning of the project because it will imply the production of materials, such as: a logo and an intervention's brochure (visit card of the program); a program's website with informative materials with an interactive component (social media profiles on Facebook, LinkedIn, Instagram, Twitter and ResearchGate). --- Component 4-Program's Evaluation with RE-AIM Framework The program's evaluation will be based on the RE-AIM framework [30], which will measure the impact of this community-based intervention in five different dimensions: 1. Reach-assessment of the percentage of youths contacting with the program (getting aware with it, getting involved in any way with it); 2. Effectiveness-assessed with the noncontrolled community trial; 3. Adoption-will evaluate some process-related outcomes, detailed below; 4. Implementation-will evaluate the extent to which the intervention is adequate and is accomplished across the intervention (fidelity); 5. Maintenance of qualities/properties-will evaluate the extent to which the intervention will be sustained over time. This task will occur simultaneously to the program's implementation (i.e., during 12 months), but with three specific key moments: at baseline time and at two different follow-up moments (3 months and 6 months after being involved) [7]. Considering the Social Prescribing component of the program, participants will include youngsters that were previously referred to the nonmedical sources of SRH-related health behaviors (e.g., local associations) and so they will be recruited in these community contexts [27]. Expecting that only 1 out of 5 individuals will accept to participate, the baseline sample size will be composed of about 500 youngsters. Considering additionally a 20% potential dropout between baseline and the second follow-up, we estimate a minimum final sample size of 400 youngsters. According to the previous contextualization, our proposal is to develop an intervention focused on the improvement of the long-term young Cape Verdeans' SRH-related wellbeing and quality of life, by decreasing their SRH problems and barriers to SRH behavior change, also by increasing their SRH related literacy. This medium-and long-term impact will be assessed in a short-term by the outcomes detailed above (i.e., H1 to H4). The proposed shortterm outcomes will be assessed by specific translated and/or cross-culturally validated scales. To assess outcomes at the individual level and self-care domain (H1), the following measures will be used: 1. Portuguese version of the Self-Efficacy Scale, a 15-item measure that evaluates the general self-efficacy, throughout three specific subdimensions-initiation and persistence (6 items; e.g., I'm a self-confident person), efficacy to face the adversity (5 items; e.g., I easily give up on things) and social efficacy (4 items; e.g., My friendships have been achieved through my personal abilities to make new friends). Participants' answers are rated on a 7-points Likert Scale (1-totally disagree to 7-totally agree) [31]. --- 2. Portuguese version of the Antonovsky's Sense of Coherence Scale (SOC 13) [32], a measure that assesses the main construct of the salutogenic theory, the sense of coherence, through three high interrelated subdimensions-comprehensibility (5 items; e.g., Do you usually feel that the things that happen to you in your daily life are hard to understand?), manageability (4 items; e.g., Do you usually see a solution to problems and difficulties that other people find hopeless?) and meaningfulness (4 items; e.g., Do you usually feel that your daily life is a source of personal satisfaction?). It also uses a 7-point scale that is anchored by two phrases [33]. --- 3. Portuguese version of the Rosenberg Self-Esteem Scale (RSES), a 10-items unidimensional measure that evaluates the adolescent and adults' self-esteem. Examples of typical items are: "On the whole, I'm satisfied with myself " or "I feel I'm a person of worth". Participants' answers are rated on a 4-points Likert Scale (0-strongly disagree to 4-strongly agree) [34]. To assess outcomes at social and interpersonal level (H2), three different instruments will be used, namely: 1. Portuguese version of the Multidimensional Scale of Perceived Social Support, a 12-items measure that assesses the subjective social support at three main areasfamily (4 items; e.g., My family truly tries to help me), friends (4 items; e.g., I can count on my friends when something goes wrong), and other significant (4 items; e.g., There is a special person with whom I can share my happiness and sadness) [35]. Participants' answers are rated on a 7-points Likert scale (1-totally disagree to 7-totally agree). --- 2. Portuguese brief version of Satisfaction with Social Support Scale, also a 12-items measure composed by two specific subdimensions-satisfaction with social support (7 items; e.g., I'm satisfied with the number of friends that I have) and need for activities related to social support (5 items; e.g., I miss social activities that satisfy me). The young Cape Verdeans must indicate the degree to which they agree with each affirmation (if it applies to them) on a Likert scale with five answer options [36]. --- 3. A Social-Cohesion Index (VALCOS), a macro index of social cohesion that covers indicators of several dimensions of social life, including health and subjective wellbeing [37]. To measure outcomes related to mental health care (H3) we will use the Portuguese version of the Brief Resilient Coping Scale, composed by four items (e.g., I look for creative ways to overcome difficult situations; I actively look for ways to replace losses that I have in life) and where the participants' answers are assessed by a 5-points Likert-type scale (1-Almost never to 5-Often) [38]. Lastly, subjective wellbeing domain (H4) data will be assessed through the Portuguese version of the Lyubomirsky and Lepper Subjective Happiness Scale. This is a 4-times measure, with a 7-points visual scale anchored in two antagonistic statements, which express the participants' level of happiness (e.g., Generally, I consider myself as a: 1-not very happy person to 5-very happy person) [39]. The program's adoption will be assessed by the percentage of health professionals conducting social prescribing and the number of youngsters getting a social prescribing. Similar to other Social Prescribing interventions, implementation and reach assessment will be measured by the youth's level of engagement with the intervention and by their satisfaction with the program [40]. Finally, the extent to which the intervention is sustainablemaintenance-will be evaluated by the willingness from involved health and educational professionals, as well as from community resources stakeholders, in training peers from other regions. As presented in Figure S1, several contextual, social, and cultural conditions from young Cape Verdeans may influence their SRH behaviors (i.e., gender, ethnicity, and socioeconomic conditions, such as education and household income). Thus, they will be controlled in all the sampling and analysis procedures. --- Statistical Analysis The main interest in this study relies on comparing individuals' characteristics between baseline and follow up moment subsequent to intervention. Considering generic continuous characteristics in classic parametric context, a sample size of n = 393 will allow detecting a minimum standardized difference of 0.2, with 80% statistical power and a 95% confidence level. To assess statistical significance of intervention, the paired-sample T test or the nonparametric Wilcoxon signed-rank test will be used. Additionally, multivariable analyses will be performed using general linear models (GLM) with normal distributed errors to study the factors and covariates that associate with the differences of individuals characteristics before and after intervention. --- Ethical Considerations The intervention will be conducted in accordance with the Declaration of Helsinki, and the presented protocol was approved by the Cape Verdean National Commission of Data Protection (process number 296/2020/CNPD). After its approval, the protocol was also submitted to the Cape Verdean Health Research Ethics Committee, as it was defined by the country. Participants will give their written consent to participate after being fully informed about each component's objectives and activities, anonymity/confidentiality of the data collected, and the voluntary nature of their participation. In the case of participants who are 15 to 17 years old, the written consent from parents or a legal representative will be also asked. --- Expected Results and Outputs As previously detailed, the main results of the proposed program are the long-term promotion of young Cape Verdeans SRH-related wellbeing and quality of life. As intermediate results, and constituting the short-term indicators of the major results, a positive impact is also expected in crosscutting youth competencies from different dimensions such as: individual level and self-care domain (i.e., perceived general self-efficacy, sense of coherence, and self-esteem); social and interpersonal level (i.e., perceived social support and related satisfaction), social cohesion and assertiveness); mental health (i.e., stress/coping resilience skills); subjective wellbeing domain (i.e., happiness and satisfaction with life). The program's implementation will also result in different forms of data integration and presentation. It is expected that the intermediate results from each component will contribute to the production of technical reports and scientific papers. The preliminary findings from these papers will be also communicated and disseminated through the team's participation in national and international conferences in the fields of SRH and public health. Another important output will be the production of advanced (sustainabilityoriented) health professionals training materials, namely a standard operational procedure manual for the reimplementation of the intervention program. In addition to the production of training materials and standard operating procedures for the intervention implementation, the sustainability of this intervention program will be also enhanced thanks to the potential for replication of the study with other populations and wide use of the materials produced. Mainly resulting from the program's implementation (Component 3), a biennial Portuguese-Cape Verdean exhibition about SRH-related literacy, co-organized by health professionals, educational professionals, and youngsters, will also constitute a relevant intervention's output. Similarly, the joint production of digital health literacy materials is expected, including at least one free mobile quiz game about SRH and quality of life across the life course. This program output could be used afterwards by intervention users or adapted to their peers (including those from other Cape Verdean regions or other African countries). --- Discussion The proposed program can have a meaningful impact on the consolidation of the intervention and research capacity in Cape Verde. It has a plan of action for the establishment of networks of researchers, health and educational professionals, and research cooperation between Portugal and Cape Verde, namely through the translation to the Cape Verdean sociocultural context of good health promotion practices specifically developed and tested in Portugal. Accordingly, this involvement will target improvements in research and intervention competencies for at least a Cape Verdean promising researcher, which may guarantee the portability (to other cities/regions of Cape Verde) and sustainability of the program over time. The extent to which the intervention will be sustained over time will be also evaluated by measuring the willingness of the involved health professionals/head of community resources in training their peers to replicate the intervention in other Cape Verdean regions; especially because its implementation and assessment will be conducted in the city of Mindelo but has a pilot for national implementation. A standard procedure manual, developed as advanced and sustainability-oriented health and educational professionals training materials, will not only allow the adequate evaluation of the intervention effectiveness but also the adequate reimplementation of the good practices that will prove as effective in Cape Verde sociocultural context throughout this project. Potential future achievements are also planned, demonstrating the added value of the proposed cooperation, namely through some of the social prescribing and digital health literacy activities that will be implemented throughout the program. Accordingly, one of the expected outputs is that a biennial exhibition about SRH-related literacy could be co-organized by health professionals, educational professionals, and youngsters both from Portugal and Cape Verde. Besides this, it is expected that a quiz mobile game about SRH and quality of life across the life course, coproduced with the Cape Verdean young participants in the first phase of the intervention, could be afterward used by themselves or adapted to their peer (including those from other African countries or regions). Globally, the proposed intervention is focused on having a meaningful impact on the quality of life in Cape Verde, namely considering that the young Cape Verdeans universal access to SRH is key to improve their quality of life across a life course perspective [5]. This SRH promotion targets the development of social cohesion perception, throughout the articulation and strait cooperation between health and educational resources with other public or private agents (e.g., arts-related organization), within the framework of social prescribing. The visibility of such cohesion between stakeholders is expected to have a positive impact also in social support perception [41]. Both social prescribing and digital health related initiatives will target the promotion of crosscutting youth competences (e.g., health related locus of control, stress coping/resilience, assertiveness, self-esteem) that has positive impact in a wide variety of human life course activities, beyond SRH, that, ultimately, may contribute to highly relevant gains of quality of life at individual and community levels. Summing up, we will embrace a comprehensive approach of well-being and quality of life, bearing in mind not only the young Cape Verdeans' individual, mental, and social health but also their life satisfaction and some of their contextual elements (overall quality of life). Some limitations of this study are related with the specific context of one island (small, social, economic, culturally specific characteristics) where the study will be developed. Another limitation concerns the risk of dropout. However, in future research the extension to other islands is possible in order to replicate the study in different contexts and with a more powerful sample to confirm outcomes. Nevertheless, the replication study from this intervention has a high potential outside of Cabo Verde related with the language context in other African country's (e.g., S<unk>o Tomé e Principe, Angola) and Brazil, with gains of quality of life at individual and community levels. --- Conclusions This study presents an added value as a health promotion intervention with a comprehensive approach of well-being and quality of life across the life course. The framework of social prescribing and digital health literacy that supports this SRH intervention will allow social cohesion, throughout the cooperation between health and educational, from public or private sectors, including relevant stakeholders. A positive impact in social support perception is also expected, which is relevant from the perspective of sustainability. It is projected that the implementation and effectiveness assessment will embrace a pilot for national implementation, allowing to benefit several Cape Verdean's local communities, health services, educational programs, and policymakers. --- Data Availability Statement: Non-applicable, due to be an intervention protocol, without data at this moment. Is expected that more relevant results will be disseminated through scientific papers. --- Supplementary Materials: The following are available online at https://www.mdpi.com/1660-460 1/18/3/850/s1, Figure S1. I-Decide Conceptual Map; Table S1. I-Decide Intervention Mapping Plan. --- Institutional Review Board Statement: The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Cape Verdean National Commission of Data Protection (process number 296/2020/CNPD). Was also submitted to the Cape Verdean Health Research Ethics Committee. Informed Consent Statement: Informed consent will be obtained from all subjects involved in the study. --- Conflicts of Interest: The authors declare no conflict of interest.

Cape Verdean governments have intensified the investment on the National Reproductive Health Program, aiming to provide universal and qualified services, especially to the youngest people. Nevertheless, data suggest that some health challenges remain in this group (e.g., high rates of early/unplanned pregnancies, illegal abortions, sexual risk behaviors). In this paper, we present a protocol of a community-based social prescribing and digital intervention to promote wellbeing and quality of life across the life course of young Cape Verdeans, with a specific focus on Sexual and Reproductive Health (SRH) related behaviors. The intervention program, to be developed in three years, will follow an Intervention Mapping approach, namely regarding needs assessment and study's protocol. The program's implementation and evaluation will occur simultaneously. The main expected result is the development of a sustainable training program implemented in coproduction with Cape Verdeans from Mindelo (in São Vicente island), with replicable potential in other Cape Verdean regions. The intervention will contribute to SRH-related literacy through the digital health literacy materials and to quality of life across the young's life course.

larger social systems as well as the networks of intervention recipients will substantially improve the implementation process. This paper briefly introduces the SNA field and then outlines and details ways that SNA can contribute to the implementation process for evidence-based programs, practices, policies, and principles (programs will be used as shorthand for all of these). We begin with a discussion of why SNA matters for program implementation, briefly introduce the core elements of SNA and use social network diagrams for exploratory as well as confirmatory use, characterize an implementation process and the stages of program implementation, and suggest research measures, techniques, procedures, and tools that can be used to apply SNA to implementation. Throughout we draw on examples from our work and that of others. We conclude with a discussion of the potential benefits of this approach to program implementation. --- Why SNA Matters for Implementation Getting evidence-based programs into practice has increasingly been recognized as a concern in many domains of public health and medicine [4,5]. Research has shown that there is a considerable lag between an invention or innovation and its routine use in a clinical or applied setting [6]. There are many challenges in scaling up proven programs so that they reach the many people in need [7][8][9]. There are a number of social processes that are necessary in getting programs adopted, implemented, and sustained. Three that relate directly to social networks and program effects are: (1) partnerships between researchers, community, policy makers, and practitioners that support implementation [10][11][12], (2) intervention agents (i.e., those who deliver the program), implementation agents, and intermediaries (i.e., those who support the delivery of the program) [13], and (3) the social context of how people receive the program [9]. Many behavior change programs are created or designed in academic or research settings where they are tested under tightly controlled laboratory conditions (i.e. efficacy studies), and then are tested in realworld settings via academic-community partnerships, with the academic partner mostly responsible for maintaining rigorous research methods (i.e. effectiveness studies). Programs that meet high standards for rigorous designs, impact, and replicability in effectiveness trials can earn the label "evidence-based" and be considered by some policy makers and institutions as worthy of consideration for large-scale implementation [14]. This is the traditional translational pipeline flowing from efficacy and effectiveness to implementation research that has driven a large proportion of work in this field. Examples of this research pipeline can be seen in the Blueprints Project, (http://www.colorado.edu/cspv/blueprints/) which has taken highestlevel programs and supported their movement into other communities, service delivery agencies, and school systems. Other programs are developed by communities or organizations outside of research settings, and then implemented on large scale without much formal evaluation. Only later does impact on the population become of major evaluation interest. Such programs may continue to be revised in response to effectiveness findings, even as they continue to be implemented at a large scale. Hybrid designs [15,16], where effectiveness and implementation are examined together in the same study, provide a flexible way to examine and improve the program itself as well as its delivery system, in the context of local as well as global ecological factors and influences. Partnerships are vital to the successful adoption, implementation and sustainability of successful programs. Indeed, evidence-based programs that have progressed to implementation and translation stages report that effective partnerships with community-based, school, or implementing agencies are critical to their success [11,17,18]. Understanding which partnerships can be created and maintained can be accomplished via social network analysis. For example, Valente and colleagues [10] showed that an academic-community intervention to improve cancer screening rates among Asian Pacific Islander communities in southern California was successful at increasing linkage between community based organizations (CBOs) and themselves, and links from CBOs to universities, but not between universities or from universities to CBOs. A second factor related to effective implementation is the appropriate choice of who (or which organization) delivers the program. Many programs use haphazard, invalid, or convenience methods to identify who delivers the program [19]. Yet interventions implemented by community-identified leaders are often more effective than those by non-leaders [20,21]. In general, interventions delivered by people from the community of the beneficiaries of the program will be more effective than those delivered by outside agencies that are less connected to program recipients [22,23]. In an implementation research project led by Chamberlain and colleagues, Palinkas and others [24] collected social network data of key implementation agents. This implementation project compared and tested two alternative implementation strategies for scaling up an evidence-based practice known as Multidimensional Treatment Foster Care (MTFC) in a group of California and Ohio counties. This study found that counties whose service systems leaders were identified by others as a source of information and advice exhibited significantly greater progress in implementing MTFC than counties lacking such leaders. Such networks of policymakers and practitioners are a central feature of many models of dissemination and implementation [1,25,26]. This project also showed that a type of learning collaborative, Community Development Teams, designed to strengthen the network between counties as a way of resolving implementation challenges, (1) strengthened the network [27], (2) led to no improvement in speed or stage of implementation [15], and (3) improved the quality and quantity of implementation [15]. A third factor is the social context in which the program is received. Studies have shown that networks can mediate intervention effects. That is, intervention effects may vary as a function of the recipients' social networks [28,29]. For example, Shin and others [29] demonstrated that children with friends who were physically inactive gained more from an obesity prevention program than those with physically active friends. Wyman and others [30] hypothesized that a peer-led suicide prevention program would have differential effects based on how peripheral or isolated a youth was. In sum, network theories and techniques can help understand factors related to successful program implementation. --- What Is SNT/SNA? Social network theory and analysis (SNTA) is a field of research that has emerged over the past 100 years from a niche discipline to applications spanning many fields of the social, physical, and biological sciences. There are core tenets and principles as well as widely used software, visualization, and analytic tools. In the applications presented in this paper, we focus primarily on social networks of individuals during development, design, implementation, and monitoring of behavior change programs, but acknowledge these networks could also be of organizations, agencies, coalitions, and so on. We review four measures used in this description but the interested reader can consult other resources to learn of the many other measures commonly used in social network research [21,[31][32][33]. Social network analysis is conducted by recording data on who is connected to whom. These relations can be many and varied; and can be derived from survey information (e.g. who is friends with whom) or archival traces such as email exchanges, joint purchasing behavior, joint authorship behavior, or GPS co-location information, to name but a few. The network data are used to derive individual network measures such as the number of links each individual has, and network level measures such as network density, a count of the number of links present expressed as a proportion of all links possible. There are numerous network measures available and the measures introduced here are meant to be illustrative, not exhaustive. In this regard, a few simple definitions will suffice. Components are the subgroups in networks. A component is the set of all people in the network that are reachable via any number of steps. A network can have one or multiple components. Density is the number of links in the network expressed as a proportion of all links possible. Individual centrality measures the extent a person occupies a prominent or important position in the network; people on the periphery are many steps away from the center. Isolates are people with no links in the network, yet they are within the network boundary. Centralization assesses the extent links are focused around one or a few people. Reciprocity occurs when a relationship in one direction also occurs in the opposite direction: Bob and Mary report knowing one another. Transitivity occurs when two people who are connected to one another are also connected to a third person: Friends of friends are friends. We will use the network depicted in Fig 1 to illustrate some network principles. These data represent the extent to which 37 members of an organization know each other. Each circle is a person and the number inside the circle his/her ID number. Lines connecting people depict who knew whom with an arrow signifying a direction in that relationship. For example, 19 and 24 report knowing each other, but 15 reported knowing 36 yet 36 did not report knowing 15. --- Frameworks for the Stages of Implementation In order to apply social network theory and methods to implementation, we note first that implementation can be considered a developmental process, a transition through known stages [1,34]. Here, we use a four stage model of implementation progress [1], which is similar to those used in evaluation frameworks [35] and in the diffusion of innovations [25,36] The first stage, exploration, involves a broad assessment of the community to gain a deep understanding of the community's needs, vision, and opportunities for change. While this phase is carried out somewhat differently by different prevention systems, there are some common elements in having communities and organizations decide what program would meet their needs, how adaptations may be needed, and how an effective delivery system can be built within the home organization. At some level, all action steps involve developing a comprehensive mapping of the community boundaries and population to be served, and the organizations and political constituencies that are present. A first use of social networks in implementation involves the exploration of these constituents' mutual self-interests and relationships to one another as this provides the attractive forces to establish, grow, and sustain a coordinated coalition. Such information can be used to assess the social capital [37] available that can be mobilized as a resource to support program implementation and sustainment. Expressing and maintaining this self-interest is necessary as coalition members and their constituents often have different or competing interests as well. Information about each individual stakeholder's interests, their vision for their community, and their connections to other constituencies can be assembled from face-to-face interviews (including recommendations of whom they think the coalition should talk to next), their public relations with others, and digital footprints, particularly a summary of email and phone contacts. Indeed, this mapping has also been described by Kellam [23] as a critical first stage before a community undertakes the implementation of an evidence-based program. During this initial assessment, the development team should (1) understand barriers and facilitators to the behavior(s) of interest, (2) identify additional community partners, leaders, and gatekeepers who can help develop and/or implement the program, (3) identify ecological, and delivery system issues that may affect program implementation, and (4) gather any data needed to benchmark both the behaviors (such as the rate of obesity or substance use), and community and organizational factors that can affect implementation (e.g., demographics, resources, culture, and context). The second stage involves the adoption of an existing program or the creation of an intervention to address the behavior problem. There are many evidence-based programs available for use in prevention and treatment of physical, mental and behavioral health problems (see for example: http://www.cdc.gov/hiv/prevention/programs/ebis/). Most programs, however, need to be adapted to local settings, which is considered in the exploration phase and developed during the adoption phase. If a program is adopted, the implementing agents and/or agency may need to adapt the existing fidelity monitoring tool designed for that program to accurately track the delivery and implementation of the adapted program. Social network data may be important in the adoption stage so as to understand who makes the decision on which interventions to adopt and whether and how they need to be adapted. For instance, social networks have been viewed as an important characteristic of community coalitions [38][39][40][41]. Feinberg and colleagues [42] found network cohesion to be positively associated and network centralization to be negatively associated with community readiness to engage in the Communities That Care community-based prevention coalition. Bess and colleagues [38] found that initial coalition participation in a youth violence prevention program was associated with a pre-existing network of inter-organizational relations. The third stage is program implementation with fidelity, which involves delivering the program with adherence and competence in real-world settings [43]. Program implementation research focuses on how the program is delivered and research is designed to determine if variation in program outcomes are associated with variations in program execution. Increasingly, researchers are learning that program effects demonstrated under idealized or laboratory conditions often do not translate into sustainable programs in practice [44,45]. A key area where networks can be of use in this phase is examining the process of monitoring fidelity of the intervention agent(s), reviewing by a supervisor, and providing feedback in a timely fashion. A dynamic network, measuring the timing and degree of information exchange between a single intervention agent and a supervisor, as well as between intervention agents and the intended beneficiaries of the program (e.g., attendance at meetings, homework completion), could be used to measure qualities expected to predict improved outcomes. The final stage is sustainment and monitoring which occurs as we attempt to determine if the program continues to be implemented as intended over time and is continuing to have the anticipated effects. Monitoring long-term outcomes of programs is often neglected under the assumption that if the program works, then it will continue to work and benefit the communities receiving the program. The reality is that programs change and drift, the relevance of programs change, populations and communities served by these programs change, and the people implementing the programs change [46]. Any and all of these changes can affect a program's delivery, reach, fidelity, and impacts among intended audiences. Current theories suggest that sustainability is affected not only by which communities in a network are represented in a supportive network, but also how early this network begins, whether there is a "champion" and/or whether the network represents the power structures in organizations who have the capacity to address long term financial and administrative viability [47,48]. Network analyses tools provide ways to monitor programs so their continued sustainability can be assessed. --- Network Measures and Metrics Useful for Program Implementation Exploration and Needs Assessment There are at least five actions interventionists can perform using social network techniques during the exploration /needs assessment phase: (1) determine if there is a network, (2) identify isolated or marginal individuals or groups, (3) identify individuals or groups to engage in program design, (4) determine if there are subgroups that might need to be brought together, and (5) determine if individual or group attributes are associated with network ties or structures. First, network measures can be used to determine if there is a network. In some settings, there may be no formal or informal network in place to implement an intervention program. In such cases, it may be important to form a network. For example, many underserved communities do not have existing neighborhood councils, community groups, or informal networking opportunities that can provide support and/or advice on health topics and to help establish the importance of the program. Sometimes forming a network is easy and public health specialists or community activists merely need to find a place and time to meet. Other times, however, this scientific model of implementing evidence-based programs may be at odds with the communities' values, experiences, and resources. It is critical to establish a balanced partnership from the beginning [23], in order to avoid an imbalance in power favoring an academic over a community perspective which would then make it difficult to sustain programs over time [49]. Thus a developmental perspective on the network is important as is the topology and the link to organizations' power structure. In these cases, sometimes forming a network relies on a charismatic and highly motivated leader, and when this happens, it should be documented as program replication may require identifying such a person. There is tension here, however. If programs rely on this charismatic leader to get established and maintained, the program is at risk because it depends on this person's commitment, availability, and resources. To successfully transition to later stages of implementation, the program will require a deeper commitment by other community members, leaders, and stakeholders. Often the program will need to become institutionalized somehow. Therefore it is important to document how the original network was formed and what strategies and tactics were employed to develop and grow the network. Critical to successful network formation is the "Goldilocks Principle" that it has just the right amount of density; centralization/decentralization; core-peripheriness; variation among stakeholders; and high cohesiveness/low fragmentation. Determining the right levels of density, centralization, clustering and so on can be difficult and will be context specific. Since these metrics vary from zero to one, a helpful guide would be to consider levels below 0.30 as low and, depending on the context, amenable to be increased. Levels in the 0.30 to 0.50 range are moderate and might be considered appropriate. Levels above 0.50 are probably too high in many situations and may impede diffusion, performance, or collective action, among other outcomes. Even this suggestion must be tentative however since even the most basic measure, network density, has been posited to be advantageous for spreading new information but deleterious when accepting new information from the outside [11]. And in one intervention, increased density was found to inhibit adoption of evidence-based programs [50]. In addition, there is rarely just one network that evolves; rather, there are often multiple networks, at multiple layers, with multiple perspectives that need to be coordinated. The second function of network analysis at this stage is to determine if there are individuals, groups, or organizations that are isolated or marginally connected to existing networks. In Fig 1, nodes 18 and 27 are isolated from the network, and this isolation makes it difficult for them to receive information, resources, or services that may be necessary to maintain their health or healthy lifestyles. This may be an area for the implementation of strategies that foster expanded network linkages. Note also in Fig 1, persons 10 and 12 are connected to each other but disconnected from the main group, which means they also may be in a disadvantaged network position; persons 9 and 19 have few connections to the group and may be outside the normal flow of information. The third function of network analysis in the needs assessment phase is to use the network information to identify individuals or groups that can and should be solicited to help identify community needs, barriers to change, and positive motivations for change. Typically, this would entail identifying individuals who are central in the network and thus can be thought of as representative of the community's needs. Care must be exercised however to insure that the network analysis has not simply identified the power brokers in a community but indeed has identified those that everyone thinks is important. Although there are many measures that identify central nodes in a network [33,51], only some are appropriate for use in the exploration phase. We wish to identify nodes that are prominent in the network, whom many people consider important or whom they trust and/or would go to for advice. At the same time, we want these individuals to be broadly representative of the community so they span the network most efficiently. In other words, we want to avoid identifying a set of individuals that are interconnected and represent the same subgroup in the network. Thirdly, we would like to identify individuals whose first-hand experiences with the program are positive or are positively inclined towards using evidence-based programs (e.g., scoring high on an instrument such as the EBPAS [52] or demonstrating skills required for implementation). Borgatti [53] developed the Key Player program, which provides precisely the measure we seek, identifying important nodes that represent different subgroups in the network. Key Player is a tool that researchers can use to identify important nodes, while ensuring that these nodes are located in different sub-sections of the network so the voices from various factions are heard and incorporated into the needs assessment. Combining EBPAS with network measures would be a logical set of tools for this exploratory phase. The fourth function of network analysis within the exploratory phase is to determine if there are subgroups in the network that should become inter-connected or that need to be addressed separately. There are many community detection algorithms developed to determine the extent to which a network can be described as consisting of subgroups [54,55]. The algorithms provide a measure of modularity, called Q, which indicates how much the network is clustered within sub-groups or communities identified by the analysis. Q values greater than 0. Finally, it is important to determine if individual attributes are associated with network connections and structure. In Fig 1, the nodes are colored by the departments within the organization for each person. It is apparent that many of the nodes connect solely or primarily to others in their department. The yellow cluster is almost entirely contained within itself, and there is a tendency for the grays, reds, and blues to be inter-connected. (Indeed, in a statistical model, network members in the same department were more likely to be connected to one another.) The influence of individual attributes is important to consider since the attributes may be strongly associated with the behavior under investigation and they may provide clues for how to stimulate change in the network. For the network in Fig 1, it is clear that links across departments may need to be created to allow more rapid flow of information and influence in the network. In many cases, individual sex, ethnicity, education, age, religious affiliation, or economic As in all network research, deciding exactly which networks to measure and how to measure them can be challenging. For the needs assessment phase, the network analysis should identify individuals "who can motivate the community to be healthy." In other words, ask network questions that will identify individuals who would be best suited to helping with identifying community problems and solutions. In many cases, a full census of the community may not be possible and so a snowball approach in which key informants are recruited and they name others that should be consulted [56]. Some sample of the people named in the first wave should then be interviewed, typically those that receive some threshold level of nominations; and potentially some named in the second wave. Many studies follow protocols that recruit individuals from different positions of leadership such as criminal justice, education, government, media, healthcare services and so on [57]. Unfortunately, many studies fail to document these procedures and so research is needed on how community leaders and partners are recruited to participate in health promotion interventions. --- Program design Once satisfied that a successful needs assessment has been conducted, the implementation team can turn to the task of designing the intervention either by adapting an existing evidencebased one so that it meets the needs of the community, or designing a new program. Research has shown that who delivers the program, and the social context of its receipt has a strong impact on program appeal and impact [22,58]. Program recipients do not interpret program materials in a vacuum; rather they interpret the content within their social and environmental contexts either explicitly or implicitly. Social network analysis can be used at this stage in at least five ways: (1) identifying opinion leaders to act as change agents, (2) using community members as recruiting agents, (3) consideration of other network interventions methods, (4) consideration of the social context of program delivery, and (5) attending to social media and other communication needs. One of the most frequently used network intervention tactics has been the recruitment and training of peer identified opinion leaders to implement behavior change programs [19,20,22]. Network-identified peer opinion leaders have been used in dozens of studies to accelerate behavior change. Most of these studies have been conducted among physicians or other service providers in health care settings. These studies have relied on using network in-degree, a count of the number of times a person is named in response to a network question, to identify the leaders or champions for change. Opinion leader interventions are effective because they solicit community input to determine who should implement the program, and because the leaders come from the community they often have considerable trust and social capital. Moreover, since the leaders know the community, they can help suggest necessary program adaptations during implementation. Finally, because the leaders come from the community and are embedded within the community, the interventions have the potential to be more sustainable [59]. However, despite their leadership role within the community, it is important to ensure that the opinion leaders are well trained in the intervention. Many interventions have used community members as recruitment agents. There has been a long history of successful HIV prevention interventions using peer networks to reach individuals at risk such as injection drug users (IDUs) [60]. The most common intervention has been the respondent driven sampling (RDS) approach in which IDUs who have received a behavioral intervention are then given coupons to distribute within their social networks so these peers can then also receive the intervention [61]. The initial indexes and their alters are both incentivized to participate. These RDS interventions have been shown to be quite effective at reducing risk behavior [62]. Many other network interventions are possible, and Valente [22] classified them into four broad strategies, each comprised of several tactical alternatives with each tactic having from one to many operational choices. For example, opinion leader interventions fall under the strategy of selecting individual change agents, but tactically for some programs one might choose to identify bridges, or marginal, or low threshold adopters. Operationally, if one selects leaders, there are dozens of specific mathematical algorithms that identify the most central nodes. Other strategies include network segmentation, induction, and alteration. Space prohibits review of all of these and interested reader is directed to consult [22]. Selecting a network intervention from this wide array of strategic, tactical and operational choices depends on the behavior of study, the goals and objectives of the intervention, the characteristics of the population, and the context in which the intervention will be implemented. The fourth use of network analysis for program design is consideration of the social context of program receipt. Many interventions are designed to be delivered in group settings in part to increase their cost-effectiveness and in part to harness the power of group dynamics to facilitate behavior change [63]. Alcoholics and Narcotics Anonymous interventions use group meetings to encourage commitment to behavior change; and even media campaigns would be wise to consider the social context of message receipt [64]. In short, designing interventions that encourage social networks to reinforce behavior change messages are likely to be more effective than those that ignore it. Finally, the increasing importance of social media use among all populations means that public health interventions are likely to re-imagined and re-interpreted with a social media lens [65]. Program participants will tweet and post their reactions to intervention materials and activities. If positive, these actions may increase program participation and effects and can in some cases take on a life of their own. Designers are thus encouraged to consider how their interventions may be adapted to online platforms, how to create content amenable to social media, and how to encourage participants to use social media to increase program effects. In the program design stage researchers usually measure networks of expertise and trust operationalized as "who do you go to for advice" and "who do you discuss problems with" [66,67]. The advice question identifies individuals who are knowledgeable about the issue at hand and have the necessary technical know-how to persuade non-users. The discussion question identifies individuals who are trustworthy and who are "people people" and good communicators. When the barriers to change are technical, advice networks are important; when the barriers to change are cultural, then discussion ones are important. In other words, when people report they do not know how to do the behavior, advice or expertise is needed, but when people report they do not know what behaviors to do, discussion or trust is needed. There is strong evidence that opinion leader and RDS interventions are effective behavior change approaches [20,68]. There is some evidence that identifying leaders and forming groups around them [58]; or forming groups and identifying leaders within them are also effective approaches [69]. What is lacking is comparison between network intervention methods. For example, do opinion leader approaches work better than snowball (RDS) ones; and to what extent does the kind of network measured influence network intervention effectiveness (i.e., are advice networks more effective than discussion ones). --- Implementation Most interventions are not one shot "single-time, single dose," but instead are implemented over many months and involve many sessions. For example, school-based interventions typically occur over 10-12 weeks with a single lesson each week during one class period. Similarly, many community-based interventions are delivered in group settings with multiple sessions, often facilitated by a leader or interventionist who in some cases may also vary. There are many network tools and statistics that can be used to facilitate program implementation and enable mid-course program adjustments. Gesell and colleagues [70] introduced the concept of network diagnostics in which network data are collected and analyzed during program implementation in order to provide diagnostic information useful to the program's implementation during its delivery. These investigators worked with a 12-week program designed to improve lifestyle choices for parents in underserved communities. The intervention taught healthy shopping, cooking, eating, and lifestyle choices in a group format. The theory guiding the intervention was that increased social cohesion during the group sessions would improve information seeking and social support among group members, which would translate into better outcomes. At two intervals during the intervention (weeks 3 and 6), network data were collected and the results of the network analysis were given to the interventionist and provided specific advice on activities to use to increase group cohesiveness. This advice was based on the following metrics [70]: (1) isolates-they should specifically be included in activities, (2) asymmetry in connections-this should be minimized, (3) densitywhich should be greater than 15% but less than 50%, (4) subgroups-there should be no distinct subgroups, everyone should be able to reach everyone else in the network either directly or indirectly, (5) centralization-this should be minimized so no one individual or group dominates the network, and ( 6) transitivity and cohesion-these group characteristics should be encouraged but limited. The recommendations regarding appropriate metrics and thresholds for these metrics were based primarily on experiences with one pilot study, and much work remains to be done in this area. However, this example demonstrates the usefulness of the network analysis approach in the implementation stage of an intervention. During the implementation stage, researchers should measure the same networks as the ones measured in the design stage. The most salient measures will be to determine the centrality of those who have changed their behavior in response to the intervention. In addition, it is important to identify any marginal individuals or groups who may not be responding to the intervention and who thus need to be recruited into participating more fully. A further consideration at this stage is to make an assessment of whether the intervention is changing the networks: Are intervention agents becoming more or less central in the network. We know of only one study, Gesell and others [70], that has used network data during the implementation stage so clearly more experience is needed to determine whether networks can be used to assist program implementation. --- Sustainment and Monitoring After the program has been implemented, the hope of most interventionists is that the program will continue to be implemented and the impact of the program on outcomes sustained over the long term. Many studies have deplored the tendency for programs to lose funding, focus, or purpose once the initial investments are removed [71,72]. And unfortunately many effective programs are discontinued prematurely and many ineffective ones continued when they should be terminated [73]. We expect that following the guides provided here using network analysis to aid in needs assessment, program design, and implementation will result in programs that are more sustainable than ones that do not consider the importance of social networks. Here, we specify procedures for using network data to document program sustainability, and suggest remedies should the evidence indicate a lack of sustainability. The dynamic systems framework proposed by Chambers and others [74] provides a model that is consistent with the network proscriptions provided here, namely that continuation of a program provides added opportunity for learning and adaptation of the program rather than attempting to freeze it in place. There are at least four metrics/network analytic techniques that can be used to aid in the evaluation of sustainability: (1) documentation of the continuation of the network; (2) documentation of the behavior of leaders; (3) estimation of contagion (or learning) effects; and (4) strategies to help ensure that participants hear positive things about the program from their peers. As in the program design phase, one task in the sustainability phase is to be sure there is a network relevant to the program. Asking participants whether they continue to discuss the content of the program and whether lessons learned are incorporated into the day-to-day interactions of the community is important to document as evidence that the program is still relevant. This type of process data should be routinely included in sustainability evaluations. Aside from assessing whether the community in general is aware of and values the program, it is also important to document that the program is still relevant to important community leaders and gatekeepers. Again, it is important to know whether central individuals embrace the program and still see it as relevant and a priority. This type of data collection and monitoring should be included in standard program evaluations so that researchers know the network positions of their informants. Interventionists may often think the program is still being embraced by the community because the people they come into contact with regularly still report that it is important. But these individuals may not be central in the network for which the programs are developed and so the surveillance information available to the interventionist is not a valid reflection of what is happening in the community. Most programs generate their effects either directly or indirectly through interpersonal persuasion and communication [75]. Individuals change their behaviors because they perceive themselves to have the personal supports to change, because they see the new behavior as normative, and because they see how it benefits themselves and those around them [21]. Sustainable programs are ones for which these perceptions are maintained among program participants. Network analyses can document these perceived normative influences and determine if participants maintain a perception that their interpersonal environment supports and endorses the new behaviors. Moreover, truly successful programs turn program participants into advocates, and network analyses can demonstrate whether participants provide positive influences to their peers. For example, with a standard network survey researchers can ask whether participants discussed program components with one another. Finally, related to the last point above, sustainable programs have created "buzz" that makes people excited about the program and its success, especially within various social media platforms. Researchers need to document what people think about the program and what they hear "on the street" about it. While not specific to SNA, they can also measure current perceptions of the program and determine if recent participants' experience of the program is similar to earlier participants. The important consideration is to assess how people discuss the program, and if they do, whom they discuss it with. Often, we need to answer the question: Are program graduates recruiting new people? Chambers and others [74] write (p. 118) "As a consequence, assessment of organizational characteristics (e.g., structure, climate, culture, resources) is seen as an essential component of sustainability, and indeed, the fit between context and the intervention is at the center of a sustainability phase." A central tenet of this network paper is that a key organizational characteristic to assess is the network context of program delivery and receipt. By monitoring the network of delivery agents and organizations, researchers can more adequately determine how the program is changing and how its effects may vary. During the sustainment and monitoring stage the same networks measured for program design should be assessed. Researchers should also ask program recipients whether they have had contact with intervention change agents. For analyses, researchers should test whether there is behavioral contagion within the community and among program recipients to others in the community [76]. Researchers also need to determine if prominent/central network members embrace the behavior being promoted. This is also an opportunity to monitor social media platforms to determine what people say about the intervention and whether it is becoming part of the community discourse. There are many studies showing that networks are important influences on behavior change, but few that directly assess whether intervention agents influence program recipients. Table 1 summarizes the research questions, common network measures used, concepts, and outcomes when applying network analysis methods to these program implementation stages. --- Potential Benefits Using network data to inform all stages of program development and implementation offers considerable potential benefits in a number of areas. First, it will substantially advance theory about how programs work and how best to design and implement them. Human service interventions are designed, delivered, and consumed by people; and for most people, the most important aspect of their lives are their social networks. Interventions can be conceived as programs that enable people to build and maintain better, health-promoting, and supportive networks in addition to individual changes in cognitions, attitudes, and/or behaviors. Existing evidence indicates that network-informed interventions and programs are more effective than non-networked ones. All programs think about networks and most probably acknowledge that networks are important components of program design and delivery. Yet, most programs do not explicitly account for networks in their design, delivery, and evaluation. Network data will enable us to improve our programs without substantially increasing costs or revising existing interventions. Indeed, it is likely that network-based interventions are more cost-effective because of their stronger impacts and because local buy-in and delivery are enhanced. Collecting network data has become easier in today's digital world. Mining and analyzing online networks makes use of accurate data, is cost-effective, and may be less o

This paper introduces the use of social network analysis theory and tools for implementation research. The social network perspective is useful for understanding, monitoring, influencing, or evaluating the implementation process when programs, policies, practices, or principles are designed and scaled up or adapted to different settings. We briefly describe common barriers to implementation success and relate them to the social networks of implementation stakeholders. We introduce a few simple measures commonly used in social network analysis and discuss how these measures can be used in program implementation. Using the four stage model of program implementation (exploration, adoption, implementation, and sustainment) proposed by Aarons and colleagues [1] and our experience in developing multi-sector partnerships involving community leaders, organizations, practitioners, and researchers, we show how network measures can be used at each stage to monitor, intervene, and improve the implementation process. Examples are provided to illustrate these concepts. We conclude with expected benefits and challenges associated with this approach.

development and implementation offers considerable potential benefits in a number of areas. First, it will substantially advance theory about how programs work and how best to design and implement them. Human service interventions are designed, delivered, and consumed by people; and for most people, the most important aspect of their lives are their social networks. Interventions can be conceived as programs that enable people to build and maintain better, health-promoting, and supportive networks in addition to individual changes in cognitions, attitudes, and/or behaviors. Existing evidence indicates that network-informed interventions and programs are more effective than non-networked ones. All programs think about networks and most probably acknowledge that networks are important components of program design and delivery. Yet, most programs do not explicitly account for networks in their design, delivery, and evaluation. Network data will enable us to improve our programs without substantially increasing costs or revising existing interventions. Indeed, it is likely that network-based interventions are more cost-effective because of their stronger impacts and because local buy-in and delivery are enhanced. Collecting network data has become easier in today's digital world. Mining and analyzing online networks makes use of accurate data, is cost-effective, and may be less obtrusive than pen-and-paper surveys. Online data can be used at any point during the different stages of program implementation, whether it is from emails for coordination, or fully integrating social media websites into an intervention. Much of the technology that supports online interactions is inherently designed to connect different persons and groups, and thus is ideal for network exploration. As more methods of communication become digital, researchers need to capitalize on the plethora of network data that can be made available. In addition, experience indicates that collecting network data via surveys, either electronically or with paper, is easy and efficient. Everyone knows who they go to for advice or whom they spend time with. Finally, incorporating network data into program design and intervention may enable us to conceive of new interventions. For example, school-based programs are often created with the intention of delivering the same content to all students. Yet, we know that students cluster into identifiable cliques and groups. Consequently, it makes sense to design separate interventions for different groups or cliques of students, particularly when those groups are distinct. For example, a school-based group of "jocks" may benefit from different programs than a group of "nerds." --- Conclusions This brief introduction to the application of social network analysis for program implementation is not comprehensive and highlights that much work remains to be done. Many of the suggestions provided here are based on limited evidence at this point in time, but the available evidence is promising. Social network analyses can be applied to intervention development and implementation to foster and enhance the implementation process. There is no rigorous comprehensive database of network experiences for this applied work at the present time. There is, however, a wealth of knowledge about how to collect and analyze social network data, and a plethora of measures, both individual-and network-level, that can be used to improve the design, delivery, and assessment of public health interventions. --- Data Availability Statement: All relevant data are within the paper. --- Author Contributions Wrote the paper: TV LP SC KHC CHB.

This paper introduces the use of social network analysis theory and tools for implementation research. The social network perspective is useful for understanding, monitoring, influencing, or evaluating the implementation process when programs, policies, practices, or principles are designed and scaled up or adapted to different settings. We briefly describe common barriers to implementation success and relate them to the social networks of implementation stakeholders. We introduce a few simple measures commonly used in social network analysis and discuss how these measures can be used in program implementation. Using the four stage model of program implementation (exploration, adoption, implementation, and sustainment) proposed by Aarons and colleagues [1] and our experience in developing multi-sector partnerships involving community leaders, organizations, practitioners, and researchers, we show how network measures can be used at each stage to monitor, intervene, and improve the implementation process. Examples are provided to illustrate these concepts. We conclude with expected benefits and challenges associated with this approach.

Introduction Since it started in 2020, the global COVID-19 pandemic has drastically changed people's daily lives and work styles. As an infection prevention measure, it was recommended that people refrain from or limit normal social activities such as eating out and traveling, and respect physical distance from other people. In Japan, telecommuting was recommended because COVID-19 infections mainly occurred in people of working age, and reducing human contact was deemed important for preventing the spread of infection (1). Consequently, whereas 20% of companies in Japan used telecommuting before the COVID-19 pandemic, this figure rose to 57% after the pandemic occurred (2). Loneliness is an emerging public health issue resulting from the COVID-19 pandemic situation (3,4). Recommendations for physical distance and restrictions on social activity have reduced opportunities for people to interact. In particular, young people, people who are separated or divorced, and those with health issues have reported feeling more lonely during the pandemic (4). Loneliness is associated with reduced quality of life, physical health, and mental health (3)(4)(5). From a public health perspective, it is important that nonmarried people, a high-risk group for loneliness, are able to establish new romantic relationships, because spending time with family can alleviate loneliness and help maintain good mental wellbeing (6). However, as physical distance is encouraged and social activities continue to be restricted for months or even longer, we considered that it was increasingly difficult for single people to find romantic partners. It is known that the COVID-19 pandemic has impacted romantic relationships in various ways (7,8). However, little is known about factors and attributes that influence new romantic activities during the pandemic. In April 2020, a state of emergency was declared for the first time in Japan. Subsequently, in winter 2020, the number of COVID-19 infected persons and deaths was much higher than in previous waves, and Japan was placed on maximum alert, called the third wave. Since many of those infected were of working age, many companies followed the government's voluntary request and promoted infection control measures within their companies. Infection control in the workplace, which is where workers spend most of their time and therefore a major place of human contact, has been reported to affect not only infection prevention but also workers' mental wellbeing and performance (9)(10)(11)(12). COVID-19 measures taken by Japanese employees include refraining from going to work when sick, measuring their body temperature, disinfecting the workplace, and teleworking. However, the extent to which the various control measures are implemented varies with the size of the company, the type of work, and other factors (13). Improved infection control measures in the workplace are expected to affect workers' perception of infection risk and influence their daily behavior and quality of life. Indeed, enhanced infection control measures in the workplace were also reported to be associated with individual preventive behaviors such as hand washing (10). We hypothesized that infection control efforts in the workplace may encourage people's interactions, including romantic activities, during the COVID-19 pandemic. Here, we conducted a prospective cohort study to specifically examine whether better infection control in the workplace influences workers' behaviors related to romantic relationships. --- Methods The study was conducted under a prospective cohort design between December 2020 and December 2021. Baseline and followup surveys were conducted using self-administered questionnaires via the Internet. All participants gave informed consent, and the study was approved by the ethics committee of the University of Occupational and Environmental Health, Japan (reference No. R2-079 and R3-006). The protocol for the baseline survey has been described elsewhere (14). The target population included workers aged 20 to 65 years and in employment at the time of the baseline survey. Sampling was conducted taking into account geographic region, occupation, and sex. Regions were divided into five levels, covering 47 prefectures, according to the level of COVID-19 infection. Occupations were divided into office workers and non-office workers. Thus, a total of 20 blocks of five regions, two occupations, and two sexes was created, and each block was sampled in equal numbers. We planned to collect data from 30,000 people overall, aiming to reach at least 1,500 participants in each block. The survey was commissioned to Cross Marketing Inc. (Tokyo, Japan). Of their 4.7 million pre-registered monitors with information such as sex, age, place of residence, and occupation, approximately 600,000 were sent an email request to participate in the survey. Of these, 55,045 participated in the initial screening, with 33,302 satisfying the final inclusion criteria. Of those 33,302 participants, 27,036 were included in the analysis, after excluding those judged as submitting untrustworthy responses. The following criteria were used to determine untrustworthy responses: extremely short response time (<unk>6 min), reporting extremely low body weight (<unk>30 kg), reporting extremely short height (<unk>140 cm), inconsistent answers to similar questions throughout the survey (e.g., inconsistency on questions about marital status and area of residence), and wrong answers to a question used solely to identify unreliable responses ("Choose the third largest number from the following five numbers"). The follow-up survey was conducted in December 2021, 1 year after baseline. A total of 18,560 (68.7%) participated in the follow-up. Finally, 6,486 individuals (2,779 men, 3,707 women) who were not married and who were not in a romantic relationship at baseline were included in the analysis. As the study is an internet-based survey, there are no missing values. --- Evaluation of infection control measures in the workplace at baseline Participants were asked to answer yes or no concerning whether the following measures had been implemented in their workplace: refraining from and restrictions on business trips; refraining from receiving and restrictions on visitors; refraining from meeting or recommending a limit on the number of people at social gatherings and dinners; refraining from or limiting face-to-face internal meetings; wearing masks at all times during working hours; installing partitions and revising the workplace layout; recommending daily temperature checks at home; encouraging telecommuting; prohibiting eating at the work desk; and requesting employees not come to work if feeling unwell. The questionnaire also asked participants to rate their company's infection control measures, using the question: "Do you think your company has taken adequate infection control measures for its employees?" Participants responded on a four-point scale: "strongly agree", "agree", "disagree", "Strongly disagree". --- Assessment of romantic relationship status The subjects were asked about their activities in relation to developing romantic relationships during the period from baseline to follow-up (1 year). Participants who answered yes to either of the two questions: "Have you taken any action in the past year to find a romantic partner?" "Have you engaged in any marriage activities in the past year?" were classified as "taking action in a romantic relationship". Those who answered "yes" to the question: "Have you been in a new romantic relationship in the past year?" were classified as "having a new romantic relationship". --- Other covariates Information on the subject's characteristics were collected at the baseline. Participants answered the following questions about themselves in an online form: age, sex, prefecture of residence, marital status (unmarried, bereaved/divorced), job type (mainly desk work, mainly involving interpersonal communication, and mainly labor), number of employees in the workplace, educational background, income, smoking status, and alcohol consumption (6-7 days a week, 4-5 days a week, 2-3 days a week, <unk>1 day a week, hardly ever). --- Statistical analyses Age-sex adjusted odds ratios (ORs) and multivariate adjusted ORs were estimated using a multilevel logistic model nested in the prefecture of residence to take account of regional variability. The ORs of the number of infection control measures in the workplace, and evaluation of those infection control measures associated with "taking action in a romantic relationship" were estimated. The model adjusted age, sex, marital status, job type, income, education, selfrated health, smoking, alcohol drinking, number of employees in the workplace, and the incidence rate of COVID-19 by prefecture at baseline. In addition, the ORs associated with "having a new romantic relationship" were also estimated. A p <unk> 0.05 was considered statistically significant. All analyses were conducted using Stata (Stata Statistical Software: Release 16; StataCorp LLC, TX, USA). --- Results Table 1 shows the baseline characteristics in relation to the number of infection control measures implemented in the workplace. In workplaces where no infection control measures were taken, 60% of the respondents were men. As the number of infection control measures increased, the percentage of women increased. Compared to workplaces with more infection control measures, workplaces with no infection control measures had lower incomes, more smokers, and were of smaller size. Table 2 shows the ORs for the number of infection control measures and activities related to romantic relationships. Compared to workers in workplaces with no infection control measures, those in workplaces with seven or more measures were more likely to engage in romantic relationship-related activities (OR = 1.84, 95% CI: 1.42-2.47, p <unk> 0.001). The multivariate adjusted model also showed this association (OR = 1.90, 95% CI: 1.45-2.48, p <unk> 0.001). We also found a linear relationship between the number of infection control measures and activities in a romantic relationship (p for trend <unk> 0.001). Similarly, workers who perceived their company's infection control self-perception as adequate compared to those who considered it inadequate were more likely to engage in romantic relationship-related activities (OR = 1.68, 95% CI: 1.24-2.28, p = 0.001). This relationship was similar in the multivariate model (OR = 1.55, 95% CI: 1.13-2.12, p = 0.007). Furthermore, workers' perception of infection control in the workplace showed a linear relationship in the order of strongly disagree, disagree, agree, and strongly disagree (p for trend = 0.003). Table 3 shows the ORs of participants who actually established a romantic relationship during the period between baseline and follow-up. The more infection control measures were implemented in the workplace, the more people started to have new romantic relationships. In the multivariate analysis, the OR for those who reported seven or more measures in the workplace compared to those who reported that no infection control measures existed was 1.79 (95% CI: 1.20-2.66, p = 0.004). There was also a linear relationship: the higher the number of measures, the greater the likelihood of starting a new romantic relationship (p for trend <unk> 0.001). In addition, there was a non-significant tendency (p for trend <unk> 0.081) for those who felt that the level of infection control in the workplace was adequate to establish a new romantic relationship. --- Discussion This study showed that greater implementation of infection control measures in the workplace is associated with more activities toward initiating romantic relationships. Furthermore, workers who expressed satisfaction with their company's infection control measures were more likely to be active in initiating a relationship than those who felt that their company's measures were inadequate. These results support the hypothesis that infection control efforts in the workplace influence workers' romantic activities during the COVID-19 pandemic. Several explanations can be offered for the association between better infection control measures in the workplace and romantic activity. First, infection risk perception, wherein more controls and satisfactory controls may reduce risk and then increase likelihood of spending time with potential partners, then leading to romantic relationship formation. Second, perhaps people with more stringent infection control measures are more motivated to seek out romantic partners to compensate for lesser social contact at work. Third, perhaps workplaces that implemented restrictions on travel or were less opportunities to meet people outside the company due to their infection control measures simply afforded workers more opportunities to for people to interact with coworkers relatively to more lenient workplaces, thereby allowing other predictors of attraction such as similarity and familiarity to take place. Pietromonaco et al. (8) proposed a conceptual framework for the impact of the COVID-19 pandemic on romantic relationships, based on the adaptive process from the perspective of relationship science. The adaptive process involve interaction in which couples respond to external stressors and difficulties by mutually supporting each other and functioning through their problems (8,15). When subjected to an external stressor, COVID-19 being a good example, couples that have existing problems will experience a decline in their relationship. The co-presence of emotional factors such as anxiety and depression also results in a lack of supportive affection when couples need to support each other, resulting in negative interactions (16,17). A positive association between adequate infection control measures in the workplace and workers' mental health and anxiety has been reported (12). Infection control in the workplace may have a moderating effect on workers' feelings of vulnerability by alleviating anxiety, and thereby positively influence within-couple relationships. Adequate infection control measures in the workplace may act as a deterrent to self-regulatory depletion in workers and promote the initiation and construction of romantic relationships. The theory of self-regulatory depletion has been proposed as a mechanism by which external stress can lead to decreased cooperation and satisfaction among partners (18). Because coping with external stress requires individual effort, it depletes self-regulatory capacity, leading to more negative behavior toward partners and inhibiting dyadic relationships (15,18). Couples experiencing increased daily stress may show more criticism of their partner, which has been attributed to self-regulation depletion (18). In the COVID-19 pandemic, various external stresses including economic problems, loneliness, employment instability, physical limitations, and limited social activities can make it difficult for couples to function in a complementary and supportive manner, and diminish romantic relationships. Working under a high perceived risk of infection can lead to negative attitudes toward one's partner or potential partner and a decline in the relationship, as one's self-regulatory capacity is depleted because of the effort required to avoid crisis. Adequate infection control measures in the workplace can alleviate the external stress and anxiety that COVID-19 brings to couples, including the anxiety that infection may be introduced to couples and families; they can help maintain good dyadic adjustment and psychological wellbeing, and thereby the quality of the relationship. We chose socioeconomic factors related to marriage as confounding factors in this study because romantic activities are in some ways similar to those of marriage, albeit that the sociodemographic drivers of romantic activities are not clear. Even after adjusting for factors such as age, income, and health status, we found an association between workplace infection control and romantic activity. This result implies that the mechanisms underlying the association of adequate workplace infection control with romantic activity may depend on factors other than socioeconomic ones. It has been reported that infection control in the workplace is associated with workers' risk perception, individual preventive behavior, mental health, and stress (10,12,19). It seems likely that these factors can either promote or retard the intention to engage in romantic behavior. Nevertheless, this question warrants further examination. Some limitations of this study can be mentioned. First, although we assessed romantic activities using self-reports, there is little reason to doubt participants' recollections about whether or not they started a romantic relationship. Also, as the study involved an anonymous survey via the Internet, there was little motivation for false reporting. Second, our operational definition of romantic activity might be questioned. The presence or absence of a sexual relationship, homosexuality, or multiple partners are unknown. However, we believe that any impact of not distinguishing between these factors was minimal in the context of this study. Third, it is not clear how participants found their romantic partners, whether at parties, or by social networking applications, for example. It is possible that social networking activities have become more prevalent during the COVID-19 pandemic (20)(21)(22), but any impact on this study remains unknown. Fourth, there is no theory or model that explains romantic behavior during an infectious crisis. Romantic behavior is a preliminary step leading to marriage that affects demographics. With social activities restricted by the COVID-19, elucidating the behavioral patterns related to people's romantic activities is an important future issue for public health. Under the COVID-19 pandemic, implementation of infection control measures in the workplace and the degree of satisfaction with these measures promoted romantic relationships among non-married individuals who did not have a romantic partner. We propose that infection control measures at work facilitate romantic activities and the establishment of romantic relationships by alleviating anxiety and supplementing self-regulatory capacity depleted due to the external stress of COVID-19. Single people are at high risk for loneliness, and the latter has intensified during the COVID-19 pandemic. Meeting and establishing a relationship with a romantic partner may be an important factor in maintaining mental wellbeing, for as long as physical distancing continues to be recommended. --- Data availability statement The datasets presented in this article are not readily available because of ethical restrictions. Requests to access the datasets should be directed to YF, [email protected]. --- Ethics statement The studies involving human participants were reviewed and approved by the Ethics Committee of the University of Occupational and Environmental Health, Japan. The patients/participants provided their written informed consent to participate in this study. --- Author contributions YF: conceptualization, methodology, software, formal analysis, investigation, data curation, writing-original draft preparation, project administration, and funding acquisition. AH, KM, and KI: investigation, data curation, and writing-review and editing. TN, ST, MT, and TI: investigation, data curation, writing-review and editing, and funding acquisition. All authors contributed to the article and approved the submitted version. --- Conflict of interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. --- Publisher's note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Objective: Due to the COVID-pandemic, non-married people are at high risk of loneliness. With social interactions restricted, it is important for non-married people to acquire a new romantic partner for their mental health and quality of life. We hypothesized that infection control e orts in the workplace influence people's social interactions, including romantic activities.We conducted an internet-based prospective cohort study from December (baseline) to December , using self-administered questionnaires. Briefly, , workers completed the questionnaires at baseline, and when followed up after year, , ( . %) participated. A total of , non-married individuals with no romantic relationship at baseline were included in the analysis. At baseline they were asked about the implementation of infection control measures in the workplace, and at follow-up they were asked about activities they performed with a view to romantic relationships during the period from baseline to follow-up. Results: Compared to workers in workplaces with no infection control measures, the odds ratio (OR) associated with romance-related activities for those in workplaces with seven or more infection control measures was . ( % CI: . -. , p < . ), and the OR associated with having a new romantic partner was . ( % CI: . -. , p = . ). Discussion: Under the COVID-pandemic, the implementation of infection control measures in the workplace and the expressed satisfaction with those measures promoted romantic relationships among non-married, single individuals.

Introduction Bipolar disorder (BPD) is a major psychiatric illness characterized by fluctuations of mood. It disrupts the patient's personal and social life; and it inflicts a huge economic burden on the family. Its prevalence varies between 0.2% and 6% [1][2][3][4][5] across different countries. In the bipolar spectrum form, its prevalence ranges from 2.6% to 7.8% [6]. A recent study conducted in 11 countries-mainly in the Americas, Europe and Asia-found a lifetime BPD prevalence of 2.4% [7]. BPD is the ninth leading cause of years-of-healthy-life-lost due to premature mortality and disability [8]. In 2004, BPD affected an estimated 29.5 million people worldwide [8]; and an estimated 0.9% of the total global burden of disease was attributed to BPD. In 2013, BPD accounted for 9.9 million disability-adjusted life-years (DALYs), or 0.4% of total DALYs and 1.3% of total years lived with disability [9]. BPD is a heritable illness [10,11]. The first-degree relatives of a patient show a significantly higher rate of mood disorder and social cognitive deficits [12]. BPD prevalence also has a significant positive relationship with hypertension, dyslipidemia and diabetes [13]. Environmental factors influence its severity and clinical course [14]. Stressful life events, both in childhood and in adulthood, and alcohol or substance abuse affect the onset, recurrence and severity of BPD [15,16]. Its incidence is increased by viral infection, substance abuse and trauma [17]. Common mental comorbidities of BPD include anxiety, substance abuse, conduct disorders, eating disorders, abnormal sexual behavior, attention-deficit/hyperactivity, impulse control, autism spectrum disorders, etc. Medical comorbidities are migraine, thyroid illness, obesity, type II diabetes and cardiovascular diseases [18]. BPD starts at age 18-22 years. It is common in both males and females, though the course of the disease in the two gender groups differs [19,20]. Usually, women show a predominance of depression and mixed mania; and they commonly develop it at an older age with one or more physical comorbidities [21]. In Bangladesh, there have been some studies on general mental disorders. A few of these studies also investigated, though in a small scale, the prevalence of clinically diagnosed BPD along with other mental disorders [22,23,24]. Some other studies have been conducted on specific mental disorders such as depression [25,26], schizophrenia [27], anxiety disorder [28], substance abuse disorder [29,30], and obsessive-compulsive disorder [31]. To the best of our knowledge, no study has been done in Bangladesh exclusively on BPD-neither in the general population, nor among married women. BPD causes long-lasting adverse effects on psycho-social functioning of the individual, and it generates negative financial implications causing intense suffering for the diseased individual. Moreover, in Bangladesh, wives and mothers play a major role in doing household works and rearing children. If they become sick, they cannot perform the duties efficiently. Consequently, the family is hurled into an abyss of suffering, and the country faces a great public health concern. In light of above-mentioned adversities, our study was aimed at determining the prevalence of BPD among married women in Rajshahi City, Bangladesh, and at identifying the associated risk factors and quantifying their effects on BPD. --- Materials and methods We conducted a cross-sectional household study. All households in Rajshahi City constituted the population for this survey. We shall explain the sampling design in the next two paragraphs. From each selected household, one married woman was invited to respond to the survey. All selected married women were currently living with their husbands. The respondents' age ranged from 15 to 82 years. --- Sample size determination Rajshahi is one of the four big cities of Bangladesh having an area of 97.18 sq. km. and a total population of 4,48,087 [32]. The city is divided into 30 wards, which are further subdivided into muhallas (neighborhoods), and consist of 99,222 households. The mathematical formula n = N/(1+Nd 2 ) was used to determine the sample size for this study [33], where n = sample size, N = population size and d = margin of error. Choosing d = 0.05, the formula indicated that n = 398 would suffice for this study. Assuming a 90% rate of response from the selected women, initially 450 married women were selected to participate in the study. --- Sampling method The survey participants were selected using a multistage random sampling (Fig 1). In the first stage, three wards were randomly selected out of the 30 wards of Rajshahi City, using a probability proportional to size sampling scheme. In the second stage, three muhallas were selected from each chosen ward by random sampling, again using a probability proportional to size sampling scheme. In the third stage, 50 households were selected from each chosen muhalla, using a simple random sampling. If the chosen household had only one married woman, she was invited to participate in the survey; otherwise, only one of the married women was chosen at random. All information about the number of households within each muhalla was collected from Rajshahi City Corporation Office. The randomizations were implemented by senior researchers. --- Ethical approval Before collecting data, ethical clearance for the study was taken from the Institutional Animal, Medical Ethics, Biosafety and Biosecurity Committee (IAMEBBC) for Experimentation on Animal, Human, Microbes and Living Natural Sources, Institute of Biological Sciences, University of Rajshahi, Bangladesh (Memo No: 120/ 320/ IAMEBBC/ IBSc, dated 11 April, 2019). --- Data collection For data collection, we used a semi-structured questionnaire, which was originally written in English and later translated into Bangla to help participants understand it easily. The first author prepared the first draft of the translation; subsequently, other authors reviewed and improved it. The final version of the questionnaire included the 20 questions (Questions 15-34) of the bipolar spectrum disorder scale (BSDS). Three teams were trained to collect data; and one team was assigned to each of the three selected wards of the city. Each team consisted of one male and one female postgraduate student of the Department of Statistics, University of Rajshahi. The interviewers discussed the details of the research with the participants. A total of 96 women (slightly over 20%) declined to give any information. The remaining 354 agreed to provide information, and their written consent was taken. For respondents under 18 years old, consent of their guardians (family heads) was also taken (incidentally, although the legal minimum age of marriage in Bangladesh is 18 years for females, occurrence of child marriage is still very high [34,35]). Fifty three responding women took the questionnaire with them, and asked the interviewers to collect it another day. Among them, 16 women failed to return their questionnaires. The survey was conducted at the respondents' place of choice during the period May 15 through July 30, 2019. When data were entered in a spread sheet, we detected some (59 married women) responses had one or more missing values; consequently, we excluded these respondents. Finally, complete responses from n = 279 married women were available for analysis. Thus, the achieved sample size fell short of the desired size of 398. Accordingly, the margin of error increased to d = 0.06. --- Outcome variable The outcome variable, BPD in married women, was determined using BSDS, which is an effective tool with sensitivity 0.76 and specificity 0.93 [36]. A study compared the diagnostic accuracy of several screening tools, and found that BSDS had the highest reliability (0.83) [37]. The BSDS was also used in developing countries such as Iran [38]. The score of points of the 20 questions ranged from 0 to 25 [39]. In this study, we classified our sample into three classes such as (i) no bipolar disorder (0-12 points), (ii) probable BPD (13-19 points) and (iii) BPD (20-25 points). However, as the prevalence of probable BPD and BPD were very low, these two classes were merged into one, and the combined class was simply called 'BPD'. Thus, our respondents were classified into two categories, which were used for chi-square tests and binary logistic regression model. --- Independent variables Based on similar studies conducted in the past and keeping in view the objectives of our study, some socio-economic, demographic, anthropometric, familial and psycho-social factors were considered as independent variables in this study. The 23 independent variables were: age group, nutritional status, religion, respondent's and their parents educational level, respondent's occupation, type of family, number of family members, family's monthly income, age at first marriage, duration of present conjugal life, miscarriage/abortions, death of children, number of children alive, number of marriage, comorbid stress/anxiety, relationship with husband, if sick treated immediately, comorbid chronic disease, family members' chronic disease, comorbid mental disorder, blood relative's mental disorder, and death of beloved one/s. --- Statistical analysis A frequency distribution was used to determine BPD prevalence. Chi-square test and binary logistic regression model were used respectively to detect associated significant factors and to measure their effects on BPD among married women in Rajshahi City, Bangladesh. The software SPSS (IBM, version 22) was used to analyze the data. --- Results The frequency distribution revealed that the prevalence of BPD, probable BPD and no BPD among married women in Rajshahi city, Bangladesh were 2.5%, 7.2% and 90.3% respectively (Fig 2 ). Chi-square (<unk> 2 ) tests identified the following ten variables as statistically significant factors associated with BPD among married women (Table 1): respondent's education level, family's monthly income, age at the first marriage, relationship with husband, if sick treated immediately, comorbid chronic disease, family members' chronic disease, comorbid mental disorder, blood relative's mental disorder and death of beloved one/s (Table 1). Only significant factors (detected by chi-square test) were included as independent variables in the logistic model. Table 2 shows the results of the binary logistic regression model, which pronounced six of the ten factors as statistically significant in altering the odds of developing BPD. Women with comorbid mental disorder had eight times higher odds to develop BPD [AOR = 8.323, 95% CI = (1.397, 50.000), p<unk>0.05]. Women having poor relationships with their husbands had twelve times higher odds to have the disorder [AOR = 11.775, 95% CI = (2.070, 66.667), p<unk>0.01]. Women coming from poor families were 16 times more vulnerable to develop BPD than those from rich families [AOR = 16.000, 95% CI = (2.086, 122.709), p<unk>0.01]. Surprisingly, women with higher education had six times higher odds to have BPD than women with only primary education [AOR = 0.177, 95% CI = (0.037, 0.843), p<unk>0.05]. Women who were not treated immediately if sick showed about three times more chance to develop BPD [AOR = 2.941, 95% CI = (1.259, 6.871), p<unk>0.05]. Women whose beloved one/s died were about three times more vulnerable to have BPD than those who did not lose their dear ones [AOR = 2.768, 95% CI = (1.130, 6.777), p<unk>0.05] (Table 2). --- Discussion Our study aimed at determining the prevalence of bipolar disorder and its associated factors among married women in Bangladesh. For this purpose, a survey was conducted in Rajshahi City, Bangladesh. The prevalence of BPD found in this study was 2.5%, compared to 2.4% in 11 countries of the Americas, Europe, and Asia [7], 2.0% in England [40], 2.2% in Canada [41] and 1.2% in Singapore [42]. In our study, the prevalence of probable BPD was found to be 7.2%. Hence, the prevalence of BPD ranged from 2.5% to 9.7%, which is consistent with a global prevalence of 2.6% to 7.8% [6]. In comparison, the prevalence of BPD was estimated to be 8.6% in India [43] and 14.3% in Pakistan [44]. BPD prevalence is usually higher in urban environments than in rural areas [45]. This might also be a cause of the comparatively higher prevalence of BPD in our study, as all of our subjects came from urban areas. Such dissimilar findings necessitates conducting more studies either using the same scale and strategy, or using different scales and strategies. Our study revealed that the women with comorbid mental disorder were eight times more prone to develop BPD, which is comparable to that found in other studies in Europe [3], Family members' chronic disease 15.112 0.001 No, 178 ( United States [46] and the entire globe [47]. In this study, we also found that poor relationship with husband was an important factor affecting BPD among married women. Marital life sometimes becomes stressful and can trigger onset or relapse of mental illness such as BPD; on the other hand, marriage can also protect couples from mental disorders [48]. Hence, relationship with husband is a crucial issue: poor relation can either create or trigger mental disturbances; good relation can prevent mental illness. Nonetheless, we could not compare our finding with other studies as no other study included husband-wife relationship as a study variable. Women from poor families were found more vulnerable to develop BPD than those from rich families, probably because poverty exposed the poor women continually to insecurity, anxiety and stress. A US study agreed that people with low family income were more vulnerable to BPD [49]. Another study observed that adult women of low socioeconomic status had twice the chance of developing mood disorders compared to middle-and high income groups [50]. Surprisingly, our study revealed that women with high education were more likely to have BPD than those with only primary education. A probable explanation for this may be that women with high education suffered from despair for not getting due recognition, power and honor within the family or in society-privilege their high educational status should have earned them. In fact, our observation matches a finding that BPD patients showed a higher likelihood to complete the highest level of education compared to their normal relatives [51]. On the other hand, a Norwegian study found that the association between educational level and BPD prevalence was not statistically significant, although social and occupational functioning was lower among BPD patients compared to healthy ones [52]. Furthermore, in our study, occurrence of BPD among uneducated women was not significantly different from that among women with higher education. Hence, we could not make any conclusive statement regarding the relationship between educational level and BPD occurrence. Our study found that women who were deprived of getting immediate treatment if sick had a three times higher odds of developing BPD. No other study is available to compare this finding. We can say that such a situation probably breeds a sense of insecurity, agitation and irritation in these women. Death of beloved one/s was found to be an important risk factor of BPD among married women. This issue is poorly documented in the literature: A study in Denmark found that parental death, especially maternal, increased the chance of BPD in their offspring [45]. Death of dear ones imprints on the human mind a long-lasting psychological effect; and that may be an explanation behind our finding. Our study determined, for the first time in Bangladesh, the prevalence of bipolar disorder among married women; and it successfully identified some associated significant factors. However, this study also had some limitations. The self-reported responses to the BSDS questionnaire, being dependent predominantly on the respondents' perceptions, may have been affected by differing levels of understanding and bias. Moreover, the cross-sectional observational study could not detect any actual cause-and-effect relationship. Comorbidities (chronic and mental disorders of the subjects and their blood-relatives) could not be accounted for. Also, some other important issues such as environmental factors, lifestyle, familial customs, effects of drugs, treatment options and outcomes, etc. could not be studied. Recognition of these limitations ought to propel the scientific community to implement new, more in-depth and elaborate research strategies. --- Conclusions The current study determined the prevalence of bipolar disorder and detected some associated risk factors of BPD among married women in Rajshahi City, Bangladesh. We found that 2.5% and 7.2% married women were suffering from BPD and probable BPD respectively. Among the significant risk factors were mental disorder, poor relation with husband, poverty, high educational level, lack of immediate treatment if sick and death of beloved one/s. As no other study on BPD has been conducted in Bangladesh, this current study has laid the foundation for further research regarding different aspects of BPD such as familial environment and culture, comorbidities, treatment options, treatment outcomes, biochemical picture, environmental factors, etc. Furthermore, based on our study, we recommend that health professionals focus on comorbidities and family matters while providing treatment and rehabilitation services to BPD patients. Government authorities and concerned nongovernmental and social organizations should take adequate steps to reduce repression on women and work for ensuring their rights and empowerment both at the household level and in society. --- All relevant data is found in the manuscript and supporting information files --- Supporting information --- S1 Data. (SAV) --- Author Contributions

Bipolar disorder (BPD) is a major mental disorder which not only affects the personal and social functioning of an individual, but also inflicts a huge economic burden on the family. Yet, the study of BPD in Bangladesh is rare and poorly documented. Responding to the dire need, we conducted a new study to determine the prevalence of, and detect the associated factors of, BPD among married women in Rajshahi City, Bangladesh.We conducted a cross-sectional study, selecting households in Rajshahi City using a multistage random sampling design. The data consisted of 279 married women, who were screened for BPD using the bipolar spectrum diagnostic scale (BSDS). Frequency distribution, chi-square test and binary logistic regression model were used respectively to determine the prevalence, identify the associated factors and quantify their effects on BPD.The prevalence of BPD among married women in Rajshahi City was 2.5%, with an additional 7.2% classified as probable BPD. A binary logistic regression analysis established the following six main factors of BPD: (1) comorbid mental disorder [AOR = 8.232, 95% CI = (1.397, 50.000), p<0.05]; (2) poor relationship with husband [AOR = 11.775, 95% CI = (2.070, 66.667), p<0.01]; (3) poverty [AOR = 1.600, 95% CI = (2.086, 122.709), p<0.01]; (4) high educational level [AOR = 0.177, 95% CI = (0.037, 0.843), p<0.05]; (5) lack of immediate treatment if sick [AOR = 2.941, 95% CI = (1.259, 6.871), p<0.05]; and (6) death of beloved one/s [AOR = 2.768, 95% CI = (1.130, 6.777), p<0.05].

Background Although work-related issues are major causes of mental distress among Japanese people of working age [1][2][3], personal concerns also have a significant effect on the mental health of this population. There are a variety of concerns in life, such as the negative consequences of marriage and personal relationships, as well as financial difficulties and illness. Relatively few studies have examined the variety of personal issues that affect psychological distress among Japanese workers, especially those which have been adjusted for work-related factors [4]. Ogami and colleagues [5], for example, found that financial matters, divorce, and illness were associated with depressive symptoms among Japanese discretionary workers who are working under the premise of the deemed working hours system. Other studies have further described how unemployment, low income, and divorce are associated with mental distress, and even with suicide, among Japanese adults, especially men [6,7]. Depression among workers represents a major occupational health concern in Japan from the viewpoint of reduced performance and substantial economic loss [8], for which various measures have been taken to minimize work stress and prevent depression among workers [9][10][11]. However, personal problems have not been well addressed by occupational health service because the effects of personal problems on psychological distress are not well delineated, or because healthcare workers in a clinical setting may find it difficult to address personal problems that are not health-related in a physical sense. The Ministry of Health, Labour and Welfare (MHLW) of Japan regularly conducts comprehensive surveys of living conditions that provide a large database on individual problems and concerns, including information on a variety of personal problems, work condition issues and psychological distress [12]. The current study examined the most concerning personal problems which have the greatest impact on psychological distress among working-age Japanese men, rather than issues relating to work tasks or duties; which have often been investigated in previous research. --- Methods --- Data collection The Ministry of Health, Labour and Welfare has conducted a comprehensive survey of living conditions in Japan every year since 1986 to help evaluate living conditions, welfare, health, and income; to assist with planning, management and policy implementation [12]. Conducted every three years, this survey covers approximately 289,363 households nation-wide in randomly selected areas. Participants are randomly selected from 5,150 municipalities of the National Census. In designated municipalities, instructors at public health centers in prefectures conduct training for persons who are in charge of the survey and conduct interviews with selected households. To conduct the present study, we requested data from the 2010 database according to procedures governing the use of official MHLW statistics. The file included de-identified data on all family members from each of the 228,664 households who had agreed to participate in the survey during 2010. We analyzed data from all men aged 20-59 years who were at the time, employed and the head of a household whose earnings maintained family. --- Questionnaire Questions in the MHLW survey included basic demographic information, such as age, sex, occupation, and employment status. There were 13 categories of occupation: 1) management, 2) professional and technical work including teachers, health care workers and researchers (professional), 3) clerk, 4) sales, 5) service, 6) security, 7) agriculture and fishery (agriculture), 8) manufacturing, 9) transportation or machine operator (transport), 10) construction, 11) cleaning, packing operators (cleaning), 12) others, and 13) unknown. Employment status consisted of self-employed, employers of companies, regular workers, non-regular workers, and others [13]. The first question regarding the problems on present personal problems causing annoyance and stress in life was: "Do you have any problems that make you annoyed or feel stress in daily life at present (Yes/No)?" The people who answered "Yes" were asked to select all possible causes of annoyance or stress from 21 categories, and then to select the most concerning present personal problems: 1) personal relationships with family, 2) personal relationships with anyone except family, 3) love or sex, 4) marriage, 5) divorce, 6) bullying or harassment, 7) Concerned about one's purpose in life, 8) no time for myself, 9) financial difficulties, 10) own diseases, 11) disease and care of a family member, 12) pregnancy of their wife, 13) child-rearing, 14) household chores, 15) Education for himself, 16) education of his children, 17) own tasks or duties of work, 18) family's working conditions, 19) housing and environmental conditions, 20) others, and 21) do not know. Some respondents did not choose any items from this list even though they had answered "yes" for the first question asking if they had any problems that made them annoyed or stressed in life. The questionnaire also included the Japanese version of the Kessler 6 (K6) which is comparable with the center for Epidemiologic Studies Depression Scale (CES-D) [14,15]. The cut-off of the K6 for determining psychological distress was over 5, based on a previously validity study of a Japanese version of the K6 with 100% sensitivity, 68.7% of specificity based on the Youden index which provided the same optimal cut-off point for the scale [15]. --- Statistical analysis We first conducted chi-square analyses to determine the associations between psychological distress and each variable of the most concerning present personal problem. Logistic regression was used to examine potential associations of the most concerning present personal problems with psychological distress, adjusting for occupational variables (occupation and employment status) and age, using problems from own tasks or duties of work as the reference [16,17]. Logistic regression calculated the odds ratios (OR) and confidence intervals (CI). All analyses were performed using IBM SPSS Statistics 20, with the level of statistical significance set at p <unk> 0.05. All ORs were adjusted using Zhang's correction formula for common outcomes, given that the prevalence of psychological distress was relatively high [18]. --- Ethics statement This study involved a retrospective analysis of data that had already been obtained during a national survey. As we did not use any personally identifiable information, based on regulations existing in Japan, ethical approval was not required. --- Results The demographic characteristics of participants are shown in Table 1. The data from a total of 76,950 survey participants were analyzed in this study. The most frequent occupation was professional (28.1%) and the most frequent employment status was regular (71.7%). Nearly half (47.7%) of the sample answered that there were no annoying or stressful problems in their daily life at present, while 27.7% of the participants reporting having psychological distress. Table 2 shows the associations between the most concerning present personal problems with psychological distress. A relatively high proportion of psychological distress was associated with bullying and harassment, divorce, and a lack of purpose in life, compared with work tasks and duties. Table 3 shows the results of multiple logistic regression analysis. This analysis revealed that psychological distress was associated with bullying and harassment (OR: 2.05, 95% CI: 1.50-2.56), divorce (OR: 1.90, 95% CI: 1.53-2.26), concern about one's purpose in life (OR: 1.73, 95% CI: 1.59-1.88), personal relationships with family (OR: 1.49, 95% CI: 1.37-1.60), personal relationship with anyone except family (OR: 1.38, 95% CI: 1.29-1.48), own diseases (OR: 1.24, 95% CI: 1.15-1.33), and financial difficulties (OR: 1.16, 95% CI: 1.12-1.20), compared with work-related problems. We also found a negative association of psychological distress with wife's pregnancy (OR: 0.51, 95% CI: 0.31-0.82), child-rearing (OR: 0.63, 95% CI: 0.48-0.81), children's education (OR: 0.75, 95% CI: 0.66-0.84), housing and environment (OR: 0.79, 95% CI: 0.70-0.90), and family's working condition (OR: 0.84, 95% CI:0.71-0.99); when compared with own work tasks or duties. --- Discussion In this study we identified several personal problems that might have a greatest impact on mental health among working-age males in Japan, perhaps more so than issues related to work tasks or duties. While such problems might be expected to be solved individually, they are usually difficult to solve by themselves. The number of individuals who were concerned about personal problems that could have a significant effect on mental health was not large; therefore, efficient screening to identify those who need support for their problems and their mental health needs to be considered. Problems caused by interpersonal relationships are known to influence mental health outcomes, especially, their most negative consequences, such as bullying and harassment. There is ample evidence that bullying and harassment have a strong association with poor mental health outcomes [19][20][21]. In this study, we did not classify whether bullying and harassment occurred in the workplace or in personal relationships. Personal relationships with family and others also can be associated with psychological distress. Educational and skills-based interventions could be applicable to preventing the negative consequences of disruptive relationships [22]; however, there is insufficient evidence on the effectiveness of these interventions in working-age populations. Multidisciplinary collaboration between communities and workplaces are therefore necessary to help individuals improve their interpersonal relationships and move towards increased mental resilience. Divorce, a generally negative outcome of marriage, is a major factor that affects the mental health of Japanese men, even increasing their risk of suicide [5,6]. Statistics on the reasons for divorce in Japan indicate that the most common reason is mismatch of characteristics (64% for male and 44% for female), followed by domestic violence, and psychological conflicts between wives and husbands [23]. These reasons might be difficult to address using interventions at the workplace, even though communitybased interventions might not solve these problems either. Since the effect of divorce is so large, interventions for preventing divorce should be researched further. Having a purpose in life, which the Japanese call ikigai, is a great motivation for life and has been shown to reduce the risk of mortality among elderly men [24][25][26]. It is often viewed as a dichotomous concept. A previous study [27], for example, reported that almost half (48.7%) of men simply answered "no" to the question "Do you have ikigai in your life?" However, only 0.9% of participants in the current study chose this lack of purpose in life as the problem about which they were most concerned. The exact meaning of a lack of ikigai, or a not having a life worth living, may differ between those of working age and the elderly. As such, further studies should now be undertaken to establish the more specific components of ikigai, especially the intergenerational aspects. One's own diseases are often recognized as risk factors for mental distress or as somatic manifestations of depression [5,28]. A variety of diseases have been suggested which may result in presenteeism and absenteeism among individuals of working age [8], including chronic pain, mental disorders, and other life-threatening diseases, such as cancer. In the current study, we did not identify what exact diseases or symptoms the participants had, although different diseases and symptoms could conceivably have significant effects on psychological distress. Therefore, further research should now be undertaken to better identify and address specific diseases and symptoms that may require the most support. Financial difficulties were frequently listed as problems of most concern in the current study, followed by working conditions. This result is somewhat expected as financial difficulties have often been suggested as a risk factor in psychological distress during other studies [5,7,29]. In the current investigation, the presence of financial difficulties was based on each participant's self-perception, rather than a quantitative assessment of income or indebtedness. Mortgage delinquency on housing loans is another common issue (17% of household) for the working population of Japan [30], and is known to cause anxiety about future plans in an era of economic downturn [17]. With regard to indebtedness, given that health services may find it difficult to help in this particular situation, increased collaboration between other governmental departments and service agencies might be necessary in future. There are a few limitations of the current study. Firstly, as a cross-sectional investigation it was not possible to determine causality. Future, longitudinal research studies will therefore be needed to address these issues. Secondly, the classification of personal problems used in the original MHLW survey has not been validated in previous research. In our current study, almost half the participants did not select any items describing annoying or stressful factors in their daily life. A Survey on the State of Employees' Health conducted in Japan during 2013 for example, also reported that 40% of men responded "no" to any factors causing strong anxiety, or annoyance and stress in their work and life [31]. There may have been some participants who suffered from mental health issues or stressors issues that were not addressed in the current study. Furthermore, the validity of questions used to identify personal problems that were annoying or stressful in life is not known, and indeed, some questions in the official MHLW survey were somewhat imprecise. --- Conclusion This study suggests that there are several personal factors which have a larger influence on the mental health status of working-age Japanese men; rather than work tasks or duties which have often been the focus of previous research. Asking workers what personal problems most concern them could help identify those in most need of psychological support. Multidisciplinary interventions for life concerns may be necessary for helping to reduce psychological distress in Japan, as elsewhere. --- Competing interests The authors declare that they have no competing interests. Authors' contributions KW conceived and designed the study. KW and DY analyzed the data for the study. KW, HE, JO, and DS contributed to the interpretation of the results and the writing of this paper. All authors approved the final version of the article. Author details 1 International Health Cooperation, National Center for Global Health and Medicine, 1-21-1 Toyama, Shinjuku-ku, Tokyo 162-8655, Japan. 2 Department of Public Health, Kitasato University School of Medicine, 1-15-1 Kitasato, Minami-ku, Sagamihara, Kanagawa 252-0374, Japan. 3 Department of Statistical Science, School of Advanced Sciences, the Graduate University for Advanced Studies, Kamiyamaguchi Midori-Cho 10-3, Tachikawa-Shi, Tokyo 190-8562, Japan. 4 Employee Health Office, Oriental Consultants Co. Ltd, 3-12-1 Honmachi, Shibuya-ku, Tokyo 151-0071, Japan. 5 School of Health Sciences, Faculty of Health and Medicine, University of Newcastle, Brush Road, Ourimbah, New South Wales 2258, Australia.

Background: Personal problems are known to influence mental health among workers. The current study investigated the most concerning present personal problems which have the greatest impact on psychological distress among working-age Japanese men, rather than issues relating to work tasks or duties. Methods: We obtained data from the 2010 Comprehensive Survey of Living Conditions conducted by the Ministry of Health, Labour and Welfare. The original survey interviewed 289,363 households in 5,150 randomly selected municipalities of Japan, from which 228,664 households agreed to participate. We analyzed the data pertaining to men who were 20 to 59 years of age and the head of a family. The questionnaire included occupation, employment status, the most concerning present personal problems, and a measure of psychological distress (the Kessler 6 scale). Multiple logistic regression analysis was conducted to delineate the association between present personal problems and psychological distress. Results: A total of 76,950 males were included in the analysis, 27.7% of whom reported some type of psychological distress. Statistical analysis revealed that psychological distress was associated with bullying and harassment (Odds

Introduction Ujamaa policy or model practiced in Tanzania as a social development (Developmental social welfare or the developmental approach) with a broad approach to the promotion of holistic well-being of a nation, through aligning the human and other aspects of development i.e., political, social, economic, health etc. Ujamaa adopted with the major components of African history, values, and ecology (Ubuntu). Ujamaa as a theory, political ideology and model was developed by the Late Dr. Julius Kambarage Nyerere to restore African development in Tanzania following the socioeconomic problems faced by newly Independent Tanzania as most of the problems were caused by colonial destructions. Ujamaa policy represented developmental social work in African context that embraced African Ubuntu values where manifestation of change process was realised through a planned social change designed to promote the well-being of the population as a whole within the context of a dynamic multifaceted development process. Despite of various challenges it faced yet Ujamaa succeeded to achieve social and economic justice through strengthening people and their communities' livelihood capabilities and made a legacy that developmental social work was rooted in Africa and was widely applied to ameliorate social and economic problems facing Africa before the emergence of modern social welfare and social work practice. This developmental thinking and practice was implemented by using ecology and collective perspective where the families, kinship, neighborhood to member groups through community participation and empowerment to achieve collective goals. Lastly Ujamaa while collectively increased production yet society emphasized on the realization of universal human rights and development in the human face where collectively and holistic social protection for all through social welfare provision. Ujamaa was not a new practice but rather a new name that restored Indigenous African (Ubuntu) human development model. He thought Ubuntu model was relevant for post-colonial Tanzania and he embraced Ubuntu values through Ujamaa Policy from the dawn of Independence in 1961 and subsequently. --- Background Social development (also known as developmental social welfare or the developmental approach) is a broad approach to the promotion of the well-being of a nation, through aligning the human and economic aspects of development (Patel, 2005& Van Breda 2015). According to Midgley (2014: 13) has most recently defined social development as "a process of planned social change designed to promote the well-being of the population as a whole within the context of a dynamic multifaceted development process". Decolonization process left post-colonial Africa countries into dilemma to opt the right path of social, economic and political development (Nyerere, 2011). Colonial legacy influenced African countries and their leaders since there was poor hand over of power and unsmooth transition process from colonialists to African rulers. In Tanzania during the dawn of independence, Ujamaa model was a form of developmental approach that was developed and adopted with the major components of African history, values, and ecology (Ubuntu). Ujamaa as a theory, political ideology and model was developed by the Late Dr. Julius Kambarage Nyerere to restore African development in Tanzania following the socio-economic problems faced by newly Independent Tanzania as most of the problems were caused by colonial destructions (Lembuka, 2023). Postcolonial Tanzania was left poor and most of the country was still rural underdeveloped by colonial legacy (Nyerere, 2011). On the process to seek for the right path of development for Tanzania that could be relevant to African context Dr. Nyerere being a son of Chief Burito Nyerere from interlacustrine region, he combined his indigenous knowledge and experience to restore developmental policy and model relevant for African context. African socialism "Ujamaa" the Swahili word for extended family, was a social, political, and economic policy developed and implemented in Tanzania and later was adopted in other African countries. The policy based on African history, culture, values, and ecology in relation to human development. After the independence of Tanzania, the late Dr. Julius Nyerere being the First President of Tanzania he envisioned to restore holistic African human aspects of life through the establishment of Ujamaa Policy. With the Arusha Declaration of 1967, President Dr. Julius Nyerere set a path for Tanzania towards Ujamaa, with villagisation as an important component (Nyerere 1967& Ndiege 2019). Ujamaa was a means of restoring African model of life for relevant and sustainable human development in African setting where Ubuntu values could be realized. Ubuntu is a collection of values and practices that black people of Africa or of African origin view as making people authentic human beings. While the nuances of these values and practices vary across different ethnic groups, they all point to one thing an authentic individual human being is part of a larger and more significant relational, communal, societal, environmental and spiritual world (AJSW, 2020). For broader understanding, Ubuntu model was practiced in sub-Sahara Africa expressed with different terms but entailed the similar meaning and goals. For example in Angola, gimuntu, Botswana (muthu), Burkina Faso (maaya), Burundi (ubuntu), Cameroon (bato), Congo (bantu), Congo Democratic Republic (bomoto/bantu), Cote d'Ivoire (maaya), Equatorial Guinea (maaya), Guinea (maaya), Gambia (maaya), Ghana (biako ye), Kenya (utu/munto/mondo), Liberia (maaya), Malawi (umunthu), Mali (maaya/hadama de ya), Mozambique (vumuntu), Namibia (omundu), Nigeria (mutunchi/iwa/agwa), Rwanda (bantu), Sierra Leonne (maaya), South Africa (ubuntu/botho), Tanzania (utu/obuntu/bumuntu,ujamaa), Uganda (obuntu), Zambia (umunthu/ubuntu) and Zimbabwe hunhu/unhu/botho/ubuntu) etc. (Mugumbate, 2019). Ujamaa as a value and model of African Ubuntu, were shaped by many sources, most notably, history, particularly, the historical events which shaped Tanzania and Africa; the cultural traditions, ecology and customs of Africans in sub-Sahara Africa, the political traditions of liberalism and communism, and by the socio-political and economic situation which prevailed in the country (Cornel, 2012). This provides the direct link between Ujamaa and Ubuntu when Africa was decolonized herself from colonial legacy restoration of Ubuntu model in post-colonial Africa was inevitable across the continent. Ujamaa policy embraced African cultural values and ecology to form a model based on the idea of collective efforts in all aspects of human life such as collective farming, cooperative unions, sharing of resources, respects of human dignity and free provision of social welfare services to all, as a result Ujamaa increased level of self-reliance at both an individual and national level (Lembuka, 2023). Villagisation was seen as a way to improve productivity of farmers and a way to facilitate the delivery of social welfare services, such as health care and education, with which people living in scattered homesteads had been harder to reach. Villagisation often had a political agenda, ranging from an ideological framework to the need for control or counter-insurgency and it also tended to have a modernising agenda (Ndiege, 2019). Not only Ujamaa policy covered development in human face bur also it emphasized on preventive social welfare approach from micro, mezzo, and macro level. With Ubuntu values, Ujamaa policy highlighted collective interventions and shared responsibilities of individuals, families, extended families, groups, ethnic leaders, cooperative societies, party, and government officials in realization universal human rights, human dignity, and social welfare in the process of development. Ujamaa policy was a kind of policy that united all Tanzanians and overcame tribalism compared to other African countries as it embraced on African culture, respect of human dignity and development was meant for Africans not infrastructures (Lembuka, 2023). --- Ujamaa policy The late Dr. Julius Kambarage Nyerere, African philosopher, Ubuntu champion, anti-colonial leader, first president of Tanzania, and globally respected international statesman who developed Ujamaa Policy for newly independent Tanzania that covers all the aspects of human life (Sanga, 2020). Ujamaa based on the nature of extended families in traditional Africa and the practices which fed into Ujamaa, namely, participation and inclusion, sharing and cooperation, and communal ownership of major means of production (Jaimungal, 2019). The emphasize of Ubuntu was made by Dr. Nyerere where Karioki (1973) is quoted that "In Africa of the past, Nyerere argued, social interaction was based on the spirit of the extended family, and in that setting, the individual worked for the community, and in turn, the community took care of him. Exploitation, laziness and parasitism were forbidden; conflict between individual and society were unimaginable. That, according to Nyerere, was an ideal society." It was this sense of community and social values that he wanted Tanzania to return to and attempted to do so with Ujamaa. According to Lal (2010), national development in Tanzania was heavily anchored to ideas of self-reliance and security. Additionally, between 1964 and1975, development politics in Tanzania came to be organized around a version of Ujamaa (Lal, 2010& Jaimungal, 2019). Lastly Ujamaa while collectively increased production yet society emphasized on the realization of universal human rights and development in the human face where collectively and holistic social protection for all through social welfare provision. Ujamaa was not a new practice but rather a new name that restored Indigenous African (Ubuntu) human development model. He thought Ubuntu model was relevant for post-colonial Tanzania and he embraced Ubuntu values through Ujamaa Policy from the dawn of Independence in 1961 and subsequently (Lembuka, 2023). --- Ubuntu and social work Ubuntu has several tenets that support achievement of social case work in African settings where social welfare service was a shared responsibility to each member of society, family member, extended family or local community. In pre-and post-colonial African societies, the responsibility of every community member to look after people who are in needy within their natural ecology with no institutionalization. Following the colonialism in Africa that emphasized institutionalization of children, disabled persons, old people and offenders yet it has not been very successful in the African setting. These institutions never existed in the Africa setting (Mugumbate, et al, 2013). Social work with groups utilises the group as a strategy to solve social problems where groups are a key feature in African society. What binds groups together is the ubuntu spirit. However, group work as it stands in social work is meant to bring people from varying backgrounds together to solve their problems through group processes. This looks alien in the African context because in most cases problems are solved within the family and not with aliens (Mugumbate et al, 2013). Ubuntu has greater relevance in the application of social work with communities. Unhu inters that the community lives as a collective. In community work, cooperation and collaboration are key principles. The concept of zunderamambo in Zimbabwe, Ujamaa in Tanzania and perhaps other parts of Africa is based on ubuntu values (Mugumbate et al, 2013) where the concept involves community members working together to grow food and relevant activities for the welfare of community guided by vision of the king or chief who later distribute the share to the wider community basing on the needs of existing groups and clans. Evidence based research has always existed even in the African setting. However, the research paradigms differed. In the African setting, mythology and lived experiences especially from elders were key elements in the research process. Sources of knowledge in the African setting are elders and svikiros (spirit mediums). Current research, based on positivism, does not recognise these sources of knowledge. In most African settings, when researching on sampled individuals where the belief stems from the concept of dare, a practice whereby ideas are generated and verified by a group of people (Mugumbate et al, 2013). --- Developmental social work As cited by Manyama (2018), that developmental social work can be defined as a process of planned social change designed to promote the well-being of the population as a whole within the context of a dynamic multifaceted development process (Mingle, 2013). It aims to achieve social and economic justice through strengthening people and their communities' livelihood capabilities (Patel, 2005). Developmental social work was rooted in Africa (Midgley, 1995) and was widely applied to ameliorate social and economic problems facing Africa before the emergence of modern social welfare and social work practice (Mupedziswa, 2005;Mwansa, 2012). This developmental thinking and practice was implemented by the families, kinship and neighborhood to member groups through community participation and empowerment (Lombard & Wairire, 2010). Developmental social work is defined as an integrated, holistic approach to social work that recognises and responds to the interconnections between the person and the environment; links micro and macro practice; and utilises strength-based and non-discriminatory models, approaches and interventions, and partnerships to promote social and economic inclusion and well-being (cf. Patel and Hochfeld, 2008;Lombard, 2007;Patel, 2005;Mayadas & Elliott, 2001;Gray, 2006;Midgley, 1995). Developmental social work affirms the commitment of the social work profession to social justice and human rights and to the eradication of poverty and inequality. It is practiced within a developmental social welfare system (Gray, 2006) and positions social work as a role player and partner in social development in the social welfare sector (Lombard, 2007). Being closely associated with the poor and vulnerable in society, social workers are key social partners and change agents in development and should therefore play an important role in the national effort to reduce inequality and eliminate poverty (Lombard, 2008a& Wairire, 2010). --- Reflection Table 1: Reflections of Ubuntu, Ujamaa and Developmental Social Work In Figure 1.0 I have presented the reflections of Ubuntu, developmental social work and Ujamaa model respectively, where they do closely relate to each other from goals to the African values. One can see the how these aspects bound and originate in traditional practice of African continent and human was a key for any form of development (Spitzer, 2019). I conquer with Dr. Nyerere who argued on the three basic assumptions made a traditional Ujamaa life-style possible through mutual respect, communal living with social equality based upon the common control of property, and the obligation for those able to work to do so (Abraham et al, 1974& Nyerere, 2011). One can agree with me that realization of human rights was a milestone for Ubuntu, Ujamaa and developmental social work. We know in African context classes is not our goal and our mission is first attain through respect human dignity before you share the resources and fruits of the society, all individuals are cared for in a mode of solidarity emanating from the value of respect. I have distinguished how Dr. Nyerere (2011) emphasised on sharing all the basic goods in the community are held in common under Ujamaa however the same trend I have observed in Ubuntu and developmental social work. Ubuntu being the core of African values, I pragmatic how it influenced Ujamaa and developmental social work, first in redesigning existing social welfare infrastructures and detach them from the negative aspects of their colonial heritage, second in reinforcing traditional social support networks of • Ujamaa used villagization approach where people were free to inherit land and protect their ancestors' land. Natural resources were protected for the future generations basing on African models. • The developmental approach ensures that economic development is to the benefit of the majority of the nation, notably reducing "mass poverty and inequality. • Collective rights to benefit from economic growth and to live in the family kinship groups and the community (Hall, 1996). My view on current World is open to revisit Ujamaa in decolonial perspective and if we can capture the strengths from within or not (Lembuka, 2023). Yet Ubuntu, Ujamaa and developmental social work are holistic they composed social, political, environmental, agricultural, militaristic, administrative, or a mix of several of these. --- Methodology The study applied desk research method in collecting and analyzing information from available secondary sources, such as documents, reports, academic publications and other materials available online or in libraries. Desk research helped to gain a broader perspective on the influence of Ujamaa policy in developmental social work in Tanzania, as well as to supplement or confirm knowledge on the topic. Desk research was thought to be relevant and useful for the study since it entails historical events or processes and theoretical studies (Rwegoshora, 2014). The Desk review method considered inclusion aspect by considered various sources of literatures from physical and online libraries including African Journals Online Database to enrich the study based on what most writers provide in their articles. Then intention was to measure and report the size (amount) of literature available on the influence of Ujamaa policy in developmental social work in Tanzania. The review was guided by one of the Ubuntu theories known as Ujamaa theory, the theory was developed by the late Dr. Julius Kambarage Nyerere who practical believed and attained to restore African model that embraces African history, culture, values and ecology in the process of human development. Ujamaa theory envisions development in human face and any process or stage of development should focus first on human beings and not things or infrastructures (Nyerere, 2011). Ujamaa theory was based on Ubuntu values where Nyerere expanded on the objective of building Ujamaa in Tanzania based on the traditional family values of mutual respect, sharing of basic goods and social welfare services held in common, and the obligation of everybody to work and protect vulnerable populations such as disabled, children, elderly and poor (Kombi 1995). His famous concept of ujamaa (which means 'familyhood' in Swahili) refers to'socialism in the villages'. The villages would be rural economic and social communities where people live together and work together for the good of all (Nyerere, 2011). --- Findings Post-colonial Africa faced multiple problems but I the uniqueness of Tanzania's problems was grounded by even the type of colonial administration that took place of indirect rule. These problems included the high rate of unemployment and poverty (Sanga, 2020). Tanzania was very poor and economically underdeveloped, and the economy depended mainly on underdeveloped agriculture. The period after independence, before Ujamaa policy, it has been portrayed by some historians as a difficult time in people's lives (Crouch: 1986) Establishment of Ujamaa as developmental model was to restore and emphasize unity and cooperation in the country. People started to live and work together; and the result of their joint labor belonged to the family (Sanga, 2020). As already pointed out by Spitzer ( 2014), in Tanzania, developmental social work method was unconsciously employed when Tanzania decided to adopt socialism and self-reliance (Ujamaa na kujitegemea), as a governing political ideology. The primary intention of this method was to pool together the community efforts and resources and facilitate the provision of social welfare services while realizing sustainable development in African context (Ibhawoh & Dibua, 2003). Ujamaa represented developmental social work as it invested in holistic and collective process of socio-economic development. The developmental social work method under Dr. Nyerere was meant to enhance the spirit of togetherness, trying to up-hold the African ways of helping each other in their communities and that the state will provide support when people are in groups (Sanga, 2020). To put this ideology into practice, villagization campaign was employed, where many social workers were engaged in assisting the communities to identify their resources, work and develop together for their betterment (Nyerere, 197, Okoko & Kimse, 1987). To overcome the national problems Dr. Nyerere advocated and succeeded to decolonize colonial mentality among Tanzanians through self-reliance practice. The development of the country and the building of Ujamaa must be dictated by the needs of Tanzanians (Sanga, 2020). Ujamaa was against the usage of money as the basis for development and related foreign aid dependency syndromes (Sanga, 2020). Thus, Ujamaa succeeded to deliver interdependence and cooperation in the country (Nyerere, 2011). With Ubuntu values in mind, Ujamaa sought to learn from African traditions to build a modern Ujamaa state and a kind of modern state that overcame some of the negative aspects that traditional African communities embraced from colonial legacy (Lembuka, 2023). The negative aspects, according to Ujamaa, are mainly two: (1) the inequalities women had to endure, and (2) that production, overall was in levels of quality and quantity very insufficient for a modern state (Nyerere, 2000). During Ujamaa Tanzania's education system was changed to meet the needs of the kind of society and Ujamaa replaced colonial education with Education for Self-reliance. This new system of education capacitated learners a sense of commitment to the community to instill self-reliance and dignity among the individuals and encourage their involvement in the building of the country (Nyerere, 2000). The life of the community and that of school was integrated and each school should have its own farm or workshop for training purposes and to encourage self-reliance. Education was practical and evidence based to equip Tanzanians to master their environment and human development (Yona, 2008). Under ubuntu values, Ujamaa policy went beyond caring other Africans beyond Tanzanian border, Ujamaa insisted that Tanzania was not full independent until other African countries were liberated where the country risked itself in the values of Ubuntu by providing massive land to the freedom fighters and hosted a number of immigrants from Africa. Ujamaa succeeded to demonstrate the moto of Ujamaa that "Africa is one and we are brothers" (Nyerere, 2011). Evidently of Ujamaa contributed to the continent's liberation struggle from Zambia, Zimbabwe, Mozambique, South Africa etc. For example, Ujamaa hosted in the country ANC's Umkotho we Sizwe as far back as 1962 and issued Tanzanian passports to liberation fighters like the late Madiba Mandela and Thabo Mbeki so they could travel and agitate for their people's freedom (The citizen, 2021). --- Discussion The main goal of Ujamaa was achieved as the primary mission was people's development and not things. This critical about some Ujamaa policies, she is particularly impressed by the policy on Education for self-reliance. According to Crouch (1986) who portrayed that Ujamaa policy as having made an impact on the development of people and its endeavour to reduce the rate of illiteracy in the country and she traced the problem of the high rate of illiteracy back to the colonial era (Yona, 2008). The perception of benefits was sometimes generational, with young people having a greater appreciation for the social and educational opportunities offered by the villages yet elders were reached through adult education all over the country. Also, proximity to roads and transport afforded greater mobility (Yona 2008& Sang, 2020). Universal provision of social welfare services was observed such as free Universal primary education (UPE), universal health coverage and improved socio-economic infrastructures in rural areas. Also, the sick, the disabled, the old, the orphans, widows, unmarried people and other vulnerable populations were looked after by the community as a whole. No member of a village would suffer and remain poor while other members of the village were in a better economic situation. This was emulated from traditional African society through Ubuntu values (Sanga, 2020). Freehold (1979) portrayed that Ujamaa was a policy solution to the economic problems of Tanzania and also something that brought the principle of human equality (Utu) in different aspects of life. Under this policy people were able to share equally the resources of the country and also the equal distribution of production. According to Freyhold (1979), everybody had a right to be respected, an obligation to work, and the duty to assure the welfare of the whole community (Sanga, 2020). Through Ujamaa people were then encouraged to stay in the countryside to form the cooperatives and the Ujamaa villages where the living conditions of the people in the rural areas improved. This resulted to an increase in life expectancy from 37 to 41 years. The infant mortality was much better than 1957, it was now between 160 and 165 per thousand live births (Yeager, 198234 & Yona, 2008). Some African countries are currently experiencing the unfortunate problems of xenophobia against the foreigners (Jaimungal, 2019). Tanzania experienced an ethnicity problem and dealt with foreigners peacefully during Ujamaa (few remained colonialists and some African who came during colonial time) which is somehow similar to xenophobia and there is a commonality between these two as they are both driven by hatred (Yona, 2008). Ujamaa became anti tribalism and anti-racism tool that successfully dealt with ethnicity and foreigners, so this approach stood the test of time. African countries can learn many things from Nyerere and the Ujamaa period including the strategies Dr. Nyerere used when dealing with the problem of ethnicity and foreigners. The Ujamaa policy caused the emergence of internal conflicts between two groups; these groups emerged to oppose the application of the Ujamaa policy (Shivji, 2012). The conflicts were inside the governing class association among two intermediate classes which included large initiatives and bureaucracy (Resnick, 1981). The presence of conflicts by the shareholders of the Ujamaa policy hindered the development of the policy, because those who are responsible for implementing and supervising the Ujamaa policy only focused on their interests. During the 1970s Tanzania faced the problem of low agricultural production this problem was due to the lack of reliable rainfall which affected nearly all of the country. The shortage of rainfall led to a severe drought which led many plantations to fail in production (Shivji, 2012). Thus, the drought continued to attack agricultural activities especially the growing season between 1973 and 1975 as a result of food shortage. After failure of harvest for two seasons, people began to complain loudly in shops for famine relief (Freyhold, 1979). Women were the critical part in the development of the Ujamaa policy; they worked for 12 or 14 hours a day and also, they worked even Sundays and holidays (Nyerere, 2011). But in other ways, women were severely oppressed during the Ujamaa policy. In this case, neither Nyerere nor TANU who made any effort to stop the inhuman actions against women; and thus, it was normal for a woman to be slapped by her supervisor in the presence of male Tanzanians (Shivji, 2012). When the government used military power to force people to settle in Ujamaa villages resulted in many Tanzanians to lose their lives and their properties were badly destroyed (Jennings, 2017). This could mean that the equality was not popular with everybody during the application of Ujamaa policy (Hunter, 2015). --- Implications Ujamaa is both a policy and model that proved workable in African context and elsewhere in the World that social workers can (re)learn and apply it in the contemporary World to address existing problems (Lembuka, 2023). As the key stakeholders of SDGs by 2023, social workers are called upon applying the principles of Ujamaa such as interdependence, communality, cooperation, respect of human dignity and cultural diversity, shared responsibility and resources for collective development in realizing SDGs (Nyerere, 2011). African Ubuntu, Ujamaa and developmental social work have similar vision and common goal of collective and holistic development with a human face (Manyama, 2018). With reference to Ujamaa policy, a comparative study on developmental social work in African countries are needed and can provide the baseline data to obtain a comprehensive profile of the status and recognition of social work in Africa (Lembuka, 2023). To act collectively as a region, it is critical that African countries need to identify key areas, themes, and common and specific priorities for developmental social work through Ubuntu perspectives with experience of Ujamaa as a benchmark case study for understanding African development (Hyden, 1980). --- Recommendation • With applicability of Ujamaa in Tanzania proved that social development has its origins in Africa (Midgley, 1995). It is therefore appropriate for the social workers in the region to demonstrate how developmental social work could contribute to building a better Africa and a better world for all the people of the African region.

The study applied desk research method to analyse literatures to gain a broader perspective on the influence of Ujamaa policy in developmental social work; it was developed by the late Dr. Nyerere's in post-colonial Tanzania in the process of decolonization and restore development approach relevant to African context. Review shows that Ujamaa became the social development model (Developmental social welfare or the developmental approach) with a broad approach that promoted collective and holistic well-being of a nation. Ujamaa embraced values that entail African history, cultural values and ecology, applicability of Ujamaa policy marked the essence and applicability of developmental social work in Africa. Despite of various challenges, Ujamaa succeeded to achieve collective development through strengthening people and their communities' livelihood capabilities, ameliorate social and economic problems. The model integrated families, kinships and groups into community participation and empowerment to achieve collective development in human face. Ujamaa's principle of universal welfare, communalism, interdependence, shared resources, human dignity and participatory development, is widely studied in the world today. Based on the success of Ujamaa model, in October 2009 the UN General Assembly named Dr. Nyerere 'a World hero of social justice' and August 2014 he was awarded as Ubuntu Champion by the National Heritage Council of South Africa for his practical contribution in African Ubuntu. Conclusively, application of Ujamaa model in realization of SDGs by 2030 is vital. It's recommended for social workers to demonstrate how developmental social work could contribute to building a better Africa and a better world.

INTRODUCTION Hope College is a 4-year residential institution located within the city of Holland, MI (pop. 30,300). The college has a current enrollment of 3,061 undergraduates served by 328 instructional faculty. Several months before the start of the fall 2020 semester, Hope College President, Matthew Scogin, committed to students and families to do everything possible "to provide an in-person experience for all our students, which includes in-person classes and on-campus living" (1). On May 20, 2020, a framework was shared to re-open the campus for safe, in-person living, and learning. It was our intent that this framework would lead to a safer and more equitable learning environment for all students. A recent Gallup study across higher education suggests it may also be a better one, finding that students who transitioned from an in-person learning environment to online learning said the quality of their education experience declined (2). Anecdotally, but in agreement with the Gallup study, Hope College professors reported that when the college was online during the Spring 2020 semester students could be seen taking exams in cars outside the local library due to a lack of reliable internet access at home. Thus, our desire to provide an in-person college experience was driven by the best interest of our students. Our framework for a return to an in-person college experience for the 2020-2021 academic year included starting classes 2 weeks earlier than normal and reducing break days to complete the semester before Thanksgiving, adapting instructional spaces, implementing safeguards and health screens with accountability, and frequently communicating with students, families, and employees. Our plans also included strategies for student testing, contact tracing, and isolation. The outcomes and lessons learned from these strategies are outlined here with the hope that others can learn from our work to provide safe, in-person learning experiences of their own (Figure 1). --- COMPREHENSIVE TEAM APPROACH As we prepared for the fall semester, it became clear that we would need to work together across multiple areas of the organization if our framework were to succeed. In July, a team was formed that consisted of two biology faculty members, two IT staff members, a residential life staff member, the head athletic trainer responsible for contact tracing, a vice president responsible for COVID-19 response and public affairs, the director (registered nurse) of the campus health center, and a team lead assigned from another area of college administration. Meeting at least three times a week throughout the semester, this group made sure that the testing, contact tracing, and quarantine and isolation aspects of our framework acted as a single process. Having a diverse cross-functional team played a critical role in sharing and interpreting data from multiple sources in order to take decisive actions when needed. --- TESTING Between July 29 and November 24, 2020, we conducted 10,700 tests at no cost to students and employees. Our testing plan was tailored for our community and informed by the expertise of public health officials and our faculty and staff. We recognized that challenges could arise in both the supply chain for testing equipment and the turnaround time for test results. For this reason, it was critical to take a multi-faceted approach. Our testing plan was meant to supplement and monitor, not replace, all of the other protections put in place as part of our broad pandemic mitigation strategy. --- Baseline Testing Our first goal was to start the academic year with zero cases of COVID-19 on campus. To achieve a baseline of zero, students and employees were tested with an at-home kit sent directly to them (3). The kit featured a saliva-based test that was medicallysupervised via Zoom video conferencing. Partnering with Vault Health (NY), students and employees were tested 8-10 days before arrival. Because this test captured the result for only that particular moment in time, students and employees were expected to do everything possible to minimize their risk of exposure to the virus as they prepared to arrive on campus. A relatively small number of students, faculty, and staff were unable to complete the saliva-based test prior to arrival on campus. These individuals were required to get a rapid antigen test performed on-site and were not allowed to check into oncampus housing (access card not activated) until they received a negative test. Results of the tests were communicated only to the student and members of the test team in accordance with HIPAA guidelines. The housing office was provided a list that indicated if a student was cleared to move in or needed testing upon arrival. Between the at-home and on-arrival testing, a total of 3,878 tests were administered as part of baseline testing. Thirty-five students tested positive with the at home test and were not allowed to travel to campus. Three students tested positive upon arrival and were immediately placed into isolation housing. No faculty or staff tested positive with pre-arrival testing. On the first day of classes (August 17, 2020), our 0.98% positive test rate was lower than the national positive rate of 6.1% and state positive rate of 2.5% according to the Johns Hopkins Coronavirus Resource Center (4) (Figure 2). --- Surveillance (Asymptomatic) Screening Our surveillance testing plan was developed to monitor the campus for the presence of SARS-CoV2, catching as many asymptomatic cases as possible, and containing the spread of the virus. As testing strategies were being developed, several mathematical models were published to predict the effectiveness of viral transmission mitigation strategies (5,6). In these models, the greatest reduction in case number was achieved by many of our strategies (i.e., mask wearing, physical distancing, etc.). Therefore, we determined that a lower rate (1% of campus population daily) of surveillance screening would be sufficient to detect outbreaks in campus housing. This is a much lower testing frequency than advocated in some other models (7), but was determined to be our best strategy based on our constraints. We were able to contract with a service provider to enable rapid (15 min) testing for surveillance that allowed for a maximum of 60 tests to be run per day. As new testing methods were introduced and reagents produced nationwide, our campus health team was able to begin administering rapid (15 min) antigen tests. Thus, our daily surveillance testing averaged 73 tests per day for the semester. Once the contracted service was complete (2 weeks before the end of the semester), all rapid antigen tests were run by the Hope College Health Center nursing team. When needed, we were able to perform large-scale testing events. Early in the semester, saliva-based PCR testing was used to conduct one of these testing events. As the semester progressed, testing events used the rapid antigen tests. For surveillance testing, students were selected at random and the testing took place at the campus Test Center every weekday. Notifications and reminders were sent from a dedicated Test Center email account to the student's college-provided email account. In most cases, students had up to 48 h to complete their test. Out of 5,696 requests to test, only 24 failed or refused to be tested for a response rate of 99.6%. If a student failed to show up for a test, then they would receive two email reminders on subsequent days. Students were then sent to the Student Development office for follow up with the judicial process if the individual still did not comply. --- Wastewater Surveillance Testing Between August 27 and September 6, the wastewater surveillance testing program was ramped up with coverage of approximately 55% of the entire student population, including 70% of those in college-owned housing. This led us to switch to a more targeted testing strategy based on wastewater data for the covered student population. Those living outside of the wastewater testing zones continued to be screened through randomized surveillance testing. Wastewater coming from nine specific residential zones on campus was collected using dedicated autosamplers each weekday, yielding 24 h composite samples. Each zone had between 100 and 250 residents. Wastewater samples were collected and analyzed for the presence of viral genomic material by quantitative PCR. This method allowed for same-day results regarding the presence or absence of viral genetic material in the campus wastewater zones. Thus, we were able to react quickly in response to infected individuals within campus housing even if those individuals were not showing symptoms. Follow-up testing of individuals based on wastewater samples was conducted on 29 different occasions between August 31 and November 16. On multiple occasions, as testing capacity allowed, entire residential halls were asked to test within 24 h. Including both the 1% sample and individual wastewater follow up testing, 5,696 surveillance tests were conducted during the semester, resulting in 57 positive cases (a 1% positive rate) (Figure 2). It is important to note that these were asymptomatic cases (or more likely mildly symptomatic and not likely to seek testing) identified and isolated, with contact tracing leading to additional quarantined students. --- Subset Testing There were some groups of individuals, including residential life staff and student-athletes, that were tested regularly. While the Michigan Intercollegiate Athletic Association postponed conference competition, our athletic teams continued team activities and conducted additional testing regularly. Testing cadences for each team were executed from the NCAA recommendations. Recommendations were determined by risk of transmission while participating in sport. This was completed by athletic training staff in partnership with the overall testing process. An additional testing subset at the end of the semester included students who indicated they would be returning to housing during the time between semesters and would unavoidably be in close contact with an immunocompromised family member. For these situations, students were provided "peace of mind" testing prior to leaving campus. --- Symptomatic Testing Symptomatic testing was reserved for students who experienced and reported symptoms of COVID-19. These tests were conducted by registered nurses from the Hope College Health Center. Students were asked to self-quarantine until they received a test and result. Employees experiencing symptoms were asked to test through their healthcare provider. During the semester, 960 symptomatic tests were conducted, resulting in 124 positive cases. The positive rate for symptomatic tests was 12.9%. --- CONTACT TRACING Students were also required to participate in the contact tracing process. Each positive case prompted a contact tracing investigation to determine close contacts. Close contacts were defined as individuals who had been within 6 feet of a positive case for a cumulative total of 15 min or more. As we began the semester, trained advocates, often college staff from other areas, helped students begin the process of identifying close contacts so that they were prepared to work with the health department. While the Ottawa County Department of Public Health remained a close partner all semester long, these investigations shifted increasingly to college staff as contact tracing resources in the surrounding community became unavailable. A contact tracing team of 7 individuals conducted over 150 investigations throughout the semester leading to 670 close contacts in quarantine. While our semester average of close contacts per positive case was between 4 and 5, we identified a trend of more close contacts per positive case in the latter part of the semester as, presumably, student adherence to safeguards lessened. --- ISOLATION AND QUARANTINE All students that tested positive for SARS-CoV2 were required to isolate. Isolation separated infected individuals from others and lasted 10 days from the first date of symptoms (or test date for asymptomatic cases). Quarantine separated and restricted the movement of people who were close-contacts with a known infected individual and lasted 14 days from last known contact. In general, isolation and quarantine resulted in students leaving their campus housing and moving into designated housing. However, in some cases (apartments, cottages) where all roommates were considered close contacts students had the option to remain in their original housing. Students were also allowed to go home to their permanent residence to quarantine or isolate unless instructed otherwise by the local health department. At the start of the semester, the college reserved 126 rooms for isolation and quarantine purposes. This increased to 176 rooms by the end of the semester. The peak of students in isolation and quarantine, including those in isolation or quarantine at home, was 369 on November 11. All courses made accommodations for remote learners (i.e., students in isolation or quarantine). Typically, this included streaming or recording of lectures and online submission of materials or exams. A team of trained advocates and healthcare professionals supported students who were in isolation or quarantine to make their experience as comfortable as possible. These advocates helped with moving, informed students of resources, checked in on them, facilitated communication with faculty, and helped them understand their role in the contact tracing process. While in isolation or quarantine, students participated in classes remotely. Faculty were prepared to engage students in their courses using online tools. Dining services created a special menu and delivery service. Symptomatic students were asked to take their temperature and asymptomatic students were asked to monitor for symptoms. The process, including the availability of housing and advocates, applied to all students, whether they resided on or off-campus. We used a symptom-based and time-based strategy, not a test-based strategy, to determine a return date for individuals diagnosed with confirmed or suspected COVID-19. This strategy took into account the time since the diagnosis and the time since recovery as well as the presence or absence of symptoms. The decision to end isolation or quarantine and return to campus was made in consultation with healthcare providers and the local health department. We observed 21 cases where a close contact student that went into isolation became symptomatic and tested positive. Thus, quarantining close contacts likely reduced the number of infections on campus. Our experience serves as an example of how contact tracing and quarantine procedures worked to mitigate spread. --- OTHER MITIGATION MEASURES --- Masks As required by the Michigan Department of Health and Human Services order, students and employees were required to wear masks in all public spaces, including classrooms. --- Physical Distancing Prior to the academic year, faculty and staff determined capacity limits for classroom and laboratory spaces that would provide the recommended physical distancing (6 feet) between individuals. The registrar then determined which classes, based on enrollment, could fit within designated classroom spaces. If a large enough classroom could not be found, then the instructional modality was changed. For Fall 2020, 58% of courses were conducted in the traditional (in-person) manner compared to 100% in all previous years. Eighteen percent of courses were taught in a completely online manner and the rest were taught in a hybrid manner (combination of in-person and online course environment). --- Screening Form Students who would be on campus were asked to complete a screening form that asked if they were experiencing any COVID-like symptoms. Students were asked, but not required, to complete this form. Participation in this screening form dropped from 1,882 submissions on the first day of classes to 316 on the final day of classes. This measure was deemed to not be effective and was not continued for the Spring semester. Employees who were going to be on campus were required to complete a daily screening form for COVID-19 symptoms as directed by the Michigan Department of Occupational Health and Safety. If students were experiencing symptoms they were directed to the Campus Health Center while employees were directed to see their physician. --- Travel As mentioned in the Introduction, the academic calendar was modified to discourage opportunities for long distance travel. Breaks were limited to single days separated from weekends, such that there were never more than two consecutive days without courses in the fall semester. All faculty, staff, and students were also encouraged to minimize travel, though no formal restrictions were put into place. Hope College is tightly interwoven with the city of Holland, MI, which has a vibrant downtown area. Thus, it would have been impossible to restrict students from leaving campus. --- DISCUSSION As we prepared to continue in-person learning during the spring of 2021, we considered the following lessons and offer them to others preparing for similar situations: • Information must be available and actionable. The teams involved in the various stages of the process (from wastewater results to quarantine capacity) need a consistent and accurate way to share information, interpret available data, and make data-driven decisions. • Diverse perspectives lead to better decision-making. Having representatives from each area regularly meet to share information, have honest and difficult discussions, and make recommendations to decision-makers helps make sure we make the right decisions at the right time. • Constraints must be acknowledged and managed. We try to start with the ideal approach and work backward based on identified constraints. These could include testing capacity, staffing, housing capacity, regulations, or finances. • Talented teams make difficult work possible. Including the COVID-19 Steering Committee and sub-teams around wastewater, testing operations, contact tracing, academics, safety operations, and housing, it is estimated that 150 employees, approximately 20% of our workforce, have had at least part of their job realigned to respond to COVID-19. During a meeting on October 30, amid a local, state, and national outbreak, local health officials confirmed that our students were likely safer on campus within our framework than they were elsewhere. We note that our student population differs from the overall population of Michigan and the country with regard to age and underlying health conditions making such comparisons difficult. However, the Ottawa County Department of Public Health shared that while our campus situation reflected the reality of the broader West Michigan region where viral spread was picking up rapidly, because it is a highly controlled environment Hope was actually better positioned than our surrounding communities. We were able to quickly identify areas of viral spread, schedule tests, isolate and quarantine individuals, complete contact tracing, and notify close contacts very effectively and efficiently. This stands in contrast to what may have been experienced in communities with large universities (>20,000 students) where the incidence of infection was not sufficiently contained (8). Therefore, we recommend that all higher education institutions seek to implement a comprehensive framework similar to the one outlined here and implemented at Hope College. --- DATA AVAILABILITY STATEMENT The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. --- AUTHOR CONTRIBUTIONS ST wrote the article. AB, KB, ND, JF, PF, TK, BGK, BPK, JP, MP, CS, and AS edited the article. AB, BGK, BPK, and MP performed wastewater testing. PF, AS, and ST analyzed non-wastewater related data. All authors contributed to the article and approved the submitted version. --- Conflict of Interest: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Due to the COVID-19 pandemic, higher education institutions were forced to make difficult decisions regarding the 2020-2021 academic year. Many institutions decided to have courses in an online remote format, others decided to attempt an in-person experience, while still others took a hybrid approach. Hope College (Holland, MI) decided that an in-person semester would be safer and more equitable for students. To achieve this at a residential college required broad collaboration across multiple stakeholders. Here, we share lessons learned and detail Hope College's model, including wastewater surveillance, comprehensive testing, contact tracing, and isolation procedures that allowed us to deliver on our commitment of an in-person, residential college experience.

Introduction Socioeconomic status (SES) indicators are essential factors that the population and individuals' health. Several studies have shown that people with higher SES have lower chronic diseases, better mental health, and longer life expectancy [1,2]; however, this effect is mediated by perceived discrimination [3]. However, the effect of SES on health outcomes has been shown to be mediated by psychological distress [4], social support [5], and perceived discrimination [3]. Marginalized populations, including immigrants, experience healthcare inequities and health disparities when compared to their native-born counterparts. In this study, we focused on whether immigration status moderates the effect of educational level on perceived discrimination. Various theoretical definitions have been proposed for discrimination in sociological literature. Oxford Sociology Bibliographies describes discrimination as an "action or practice that excludes, disadvantages, or merely differentiates between individuals or groups of individuals on the basis of some ascribed or perceived trait" [6]. Furthermore, in the Oxford Dictionary of Sociology, Scott suggests that discrimination can be broadly considered 'treating unfairly' [7]. Early sociologists, such as William G. Sumner and Franklin H. Giddings, have viewed discrimination as an expression of ethnocentrism, which defines discrimination as a cultural phenomenon of 'dislike of the unlike'. This interpretation is different from more recent views that defines discrimination as a pattern of dominance and oppression, viewed as expressions of a struggle for power and privilege [7]. More relevant to our study, Krieger 2014 [8] defined discrimination as follows: "the process by which a member, or members, of a socially defined group is, or are, treated differently (especially unfairly) because of his/her/their membership of that group" [9], extended this definition with notions from the Concise Oxford Dictionary of Sociology that discrimination involves not only "socially derived beliefs each [group] holds about the other" but also "patterns of dominance and oppression, viewed as expressions of a struggle for power and privilege" [10]. Finally, Krieger concludes that discrimination is "a socially structured and sanctioned phenomenon, justified by ideology and expressed in interactions between individuals and institutions, that preserves privileges for dominant groups at the cost of deprivation for others" [8]. We adopted this last statement as our working definition in our study. The effect of perceived discrimination on health has been documented in a longitudinal study conducted by Fuller-Rowell et al., which showed that about 14% of the longitudinal association between SES and self-rated health was explained by perceived discrimination [3]. Perceived discrimination may be related to intersecting identities around race/ethnicity, indigenous status, age, gender, sexuality, disability, and immigration status [8,[11][12][13][14][15][16]. A study based on the data of 4733 White American citizens demonstrated that higher educational level was positively associated with higher perceived discrimination [17]. Another study by Zhang et al. found higher levels of perceived discrimination in Asian-American populations with a college degree or above than those with lower educational levels [18]. Educational level promotes the health status of people. Grosse and Auffrey, in 1989, concluded that literacy is a crucial tool for adapting to different conditions [19]. This effect, however, is tenuous among people who belong to minority groups [20,21], a phenomenon called marginalization-related diminished returns (MDRs) [11,12]. In a study of 25,659 US adults, Assari et al. demonstrated that people with high educational level had a protective effect against developing cardiac disease; however, this effect was weaker in Hispanic and African American individuals than non-Hispanic White people [22]. Multiple studies replicated MDRs, defined as the attenuated positive impact of educational level on the health status of marginalized groups in various settings [23][24][25][26][27][28][29][30][31][32][33][34]. Hence, people from marginalized groups have limited access to education due to structural racism and racial discrimination. They may take less advantage of educational attainment partly because they feel discriminated against [3,8,35,36]. The differential return of higher levels of education is different from lower access to education. Historically, most researchers have argued that minorities have fewer opportunities to attain advanced education due to systematic racism/discrimination, particularly for lower-income individuals who work and have more difficulty investing time into higher education. Previous research demonstrated that immigrants experience high levels of discrimination in Europe [37][38][39], which even extends to their next generations [40]. Immigrants in Europe face economic and social challenges [41,42] that can affect their health and economic outcomes [43]. Research in the field of Marginalized-related Diminished Returns (MDRs) previously showed that immigration can be considered a moderator for the association of SES resources and health and economic outcomes in the US [44][45][46][47]. Within the European context, immigration was shown to moderate the association between education and self-rated health [48]. These diminished returns have been attributed to structural barriers that prevent marginalized minority groups from gaining equal benefits from their SES resources [49,50]. Therefore, in this study, we aimed to assess whether and to what extent immigration can modulate the protective effect of educational level on perceived discrimination in European countries. Multiple studies have been conducted to test the moderating effects of immigration on positive health results from educational level [30,[51][52][53][54]. To the best of our knowledge, no study has been identified that examines the role of immigration as a moderator on the association between educational level and perceived discrimination in Europe. Although we know that the association between educational level and perceived discrimination exists [17,18,55,56], no previous European studies have examined the role of immigration as a moderator in the association between educational level and perceived discrimination. Related research was conducted by Bakhtiary, who used data from 30 European countries and suggested that SES affects health status. However, this effect depends on immigration status. Based on this study, future research is needed to evaluate the moderating role of immigration on the phenomenon of MDRs [57]. To address this gap in the literature, we aimed to study whether an inverse association exists between educational level and perceived discrimination in European countries and if immigration status moderates the association between educational level and perceived discrimination. First, we hypothesized an inverse association between educational level and perceived discrimination. Second, we hypothesized that this association would be weaker for immigrants. --- Methods This study used data from the 10th round of the cross-sectional European Social Survey (ESS) [57], conducted from September 2020 to the end of January 2022 in ten European countries, including Bulgaria, Czechia, Estonia, Finland, France, Croatia, Hungary, Lithuania, Slovenia, and Slovakia. ESS is a cross-national survey founded in 2001 to study whether the MDRs pattern applies to other countries with different contexts from the US. Unlike previous rounds, in which data were gathered through an hour-long face-to-face interview, a self-administered questionnaire was used in the 10th round due to the COVID-19 pandemic. Additionally, some countries used video interviews as a backup plan [57]. Countries that did not report data on the participants' gender, educational level, perceived discrimination, and immigration status were excluded from the analysis. --- Study Population The total sample was 17,596 individuals aged 15-90 who lived in one of the countries included in the 10th round of this survey, either native residents (16,632; 94.5%) or immigrant residents (964; 5.5%). --- Measures Evaluating the association between educational level and perceived discrimination, we selected the years of formal education as an independent and the latter as the outcome variables. Although there are differences between educational systems of various countries participated in this study, this issue was addressed in the design of this study by harmonizing the educational levels between different educational systems using International Standard Classification of Education (ISCED). Hence, the educational level in our study is a universal variable that applies to all these countries included. Other variables, including age, gender, and country of residency, are considered confounding factors. Immigration status was the moderator. Dependent Variable: The outcome was measured by asking a question from the respondents about discrimination. The question was, "Do you describe yourself as a member of a group that is discriminated against in this country?". Here, "yes" and "No" answers were recorded as "1" and "0", respectively. Independent Variable: Educational level: The International Standard Classification of Education (ISCED) was used to classify the participants' educational levels. ISCED has nine categories, coding from "1" to "9". Code "1" was used for participants who did not complete primary education, "2" for those who completed primary education or less than two years of vocational studies, "3" for lower secondary education, "4" for lower-tier upper secondary education, "5" for upper-tier upper secondary education, "6" for advanced vocational sub-degrees, "7" for those with bachelor or equivalent degrees, "8" for those who had a master degree or equivalent, and "9" for those had a doctoral or equivalent degree. Moderator: Immigration status was the moderator variable determined through the following question. "Were you born in [country]?". "No" answers, equivalent to Immigrant status, were recorded as 1, and "Yes" answers, equal to native-born participants were coded as 0. Confounders: In this study, age and sex were considered confounders. Age (years) was a continuous variable ranging from 15 to 90, and sex was a dichotomous variable, coded 1 for men and 0 for women. Other potential confounders, such as income or behavioral traits, can be considered for this study, which we decided not to include because of lacking data or risk of overadjustment [58][59][60]. --- Data Analysis Statistical Package for Social Sciences (SPSS) version 23 (IBM Inc., NY, USA) was used for data analysis. Descriptive statistics were reported as frequency (%) and mean (<unk> standard deviation) for ordinal and numerical variables. The logistic regression models reported adjusted unstandardized regression coefficients (b) and corresponding 95% confidence intervals (CIs), as well as p values. Four multivariable logistic regression models with perceived discrimination as the primary outcome were used. Model 1 and Model 2 were performed in the pooled sample. Model 1 did not include any interaction term; however, Model 2 included immigration status by educational level as an interaction term. Model 3 and Model 4 were specified in the native-born participants and immigrants, respectively. We tested our central hypothesis in Model 2 with the significance of the interaction term between immigration status and educational level. This approach tests population differences in correlates of factors such as educational level or perceived discrimination [11,35,[61][62][63][64][65][66][67][68][69][70][71]. We applied a statistical method, which is a standard procedure to test diminished return and has been widely applied in previous research [72][73][74]. We tested the main effects of educational level and immigration on perceived discrimination in model 1. It is necessary to test the main effect in a separate model (model 1) before testing interactions in model 2. One cannot test statistical interactions without priory testing the main effect [75]. Because life conditions are different for various groups and the levels of exposure and vulnerability of groups toward risk and resilient factors can be different, one need to do the stratified model and include all covariates model 3 and 4. By doing this, a researcher can observe the differential roles of covariates across social groups. Another function of this stratified model is to confirm the finding of interaction observed in model 2 [76,77]. The main effect shows the overall effect in the population, assuming no heterogeneity in the effects across groups [78]. However, in the presence of any hypothesis about innate difference between groups, the main model should be followed or expanded by interaction models. The interaction model is an extension of the previous model, which evaluates the significance of the difference (moderating effect) between groups. Robustness check. A sensitivity analysis was also conducted using model tests with the following modifications. The educational level variable was recoded from 0 to 3, with the following groups. Code "0" was used for participants in education category "1", who did not complete primary education, as well as category "2" for those who completed primary education or less than two years of vocational studies, and education category "3" for lower secondary education. Code 1 was used for upper secondary education, comprising education category "4" for lower-tier upper secondary and category "5" for upper-tier upper secondary education. Code 3 was for advanced education, composed of categories "6" for advanced vocational sub-degrees, category "7" for those with bachelor's or equivalent degrees, education category "8" for those who had a master's degree or equivalent, and education category "9" for those had a doctoral or equivalent degree. To analyze age, we ran the models with the following categorical age groups: 15-24, 25-49, 50-65, and 65+. We also tested models with age or age and age square. We also ran models with dummy variables for countries as covariates. As the results did not change for our interaction terms that reflected the MDRs, only one set of results was reported. --- Results Table 1 shows the demographic characteristics of all participants based on immigration status. The frequency of men and women participating in this study was similar between native-born and immigrant respondents. However, more immigrants reported perceived discrimination compared to their native-born counterparts (13.4% vs. 7.9%). Regarding the country of origin of participants in each country, Bulgaria and Estonia reported the lowest and the highest rate of immigrant participants, i.e., 0.7% and 13.7%, respectively. Table A1 in the Appendix A shows the number of participants from each country. Data were reported as frequency (%) or Mean (<unk>SD) as indicated. * p <unk> 0.05 for comparison of immigrant and native-born groups. For education, 1 = did not complete primary education, 2 = completed primary education or less than two years of vocational studies, 3 = lower secondary education, 4 = lower tier upper secondary, 5 = upper tier upper secondary, 6 = advanced vocational, sub-degree, 7 = bachelor or equivalent degrees, 8 = master degree or equivalent, and 9 = doctoral or equivalent degree. --- Logistic Regression in the Overall Sample Table 2 summarizes the results of two logistic regressions in the pooled sample, with perceived discrimination as the primary outcome. Model 1, which did not include any interaction term, shows the main effect of educational level; in contrast, Model 2 considered the interaction between immigration status and educational level. For education, 1 = did not complete primary education, 2 = completed primary education or less than two years of vocational studies, 3 = lower secondary education, 4 = lower tier upper secondary, 5 = upper tier upper secondary, 6 = advanced vocational, sub-degree, 7 = bachelor or equivalent degrees, 8 = master degree or equivalent, and 9 = doctoral or equivalent degree. According to the information revealed by Model 1, the pooled sample, a higher educational level was correlated with lower odds of perceived discrimination. Model 2 demonstrates a statistically significant interaction between educational level and immigration status on the reported perceived discrimination. Therefore, the protective effect of educational level against perceived discrimination is more robust in native-born participants than in their immigrant counterparts. --- Stratified Logistic Regressions Table 3 shows the logistic regression results in native-born participants (Model 3) and the logistic regression in immigrants (Model 4). In these models, educational level was the independent variable, and perceived discrimination was the outcome (dependent variable). Model 3 shows that in native-born participants, the higher the educational level they reported, the lower the discrimination they perceived. However, Model 4 showed that educational level did not protect immigrant respondents against perceived discrimination. For education, 1 = did not complete primary education, 2 = completed primary education or less than two years of vocational studies, 3 = lower secondary education, 4 = lower tier upper secondary, 5 = upper tier upper secondary, 6 = advanced vocational, sub-degree, 7 = bachelor or equivalent degrees, 8 = master degree or equivalent, and 9 = doctoral or equivalent degree. --- Discussion This analysis included data from 17,596 European residents aged 15-90 who participated in the 10th round of the cross-sectional European Social Survey (ESS), and participants were predominantly native-born residents (16,632; 94.5%). Results revealed that as the educational level rises, perceived discrimination falls. Interestingly, our findings are inconsistent with the results of studies by Das, who reported that white US citizens with a higher educational level faced more discrimination [17]. Some other studies suggest that perceived discrimination generally increases with higher educational level among ethnic minorities in the U.S. For example, Zhang et al. found that the Asian-American population with a college degree or above reported a higher level of perceived discrimination than those with a lower educational level [18]. Higher educational level among immigrant groups in the US may be associated with an increased level of acculturation. In another US-based study examining the relationship between perceived discrimination and mental illness, higher levels of acculturation led to a significant increase in discrimination's association with mental illness [79]. We found that immigrants also benefited from the protective effect of educational level against discrimination; however, the effect is less robust than the native-born population. A study conducted on immigrants in the US showed that, similar to other marginalized groups, the protective effects of higher educational levels is less remarkable in decreasing the risk of psychological distress and chronic diseases and improving subjective health status for immigrant populations [51]. A study by Steinmann on Turkish and Polish immigrants residing in Germany showed that the frequency of reporting discrimination among participants with higher educational levels depended on the "bright boundaries" [80] they faced; the brighter the boundaries they felt, the harder they found it to assimilate. For instance, Turkish immigrants reported discrimination more frequently than their Polish immigrant counterparts because Turkish immigrants thought of themselves as a different community [81]. On the other hand, a study conducted among undergraduate students of Ethiopian origin in Israel showed that perceived discrimination motivated students of Ethiopian origin to pursue higher education. Thus, the association observed in crosssectional studies may have unexpected consequences longitudinally [82]. There is extensive evidence of discrimination against immigrants in European countries [83]. Borgonovi and Pokropek, using data from four rounds of the European Social Survey conducted in several European countries between 2010 and 2016, indicated that individuals with a higher education level are associated with lower levels of opposition to migrants [84]. Noticeably, feelings of threat mediated around 60% of the effect of the educational level on opposition to migrants. Additionally, a larger foreign-born population was significantly associated with a steeper education gradient for feeling threatened. Such attitudes do not necessarily equal greater opposition to migration, because feelings of threat are less associated with support for restrictive migration measures in countries with high numbers of foreign-born residents [84]. In studies of discrimination in European countries, it is important to identify relevant populations of immigrants, as various ethnic-racial groups, such as European immigrants and visible or non-European immigrants, do not experience it to the same level [83]. Heath and Cheung showed that non-European minorities experience ethnic penalties in accessing the job market. These disadvantages extend beyond the next generations despite the educational progress of children of immigrants [85]. Nuances such as phenotypic features, including skin color [86,87], religion [88-90], and accent [91,92], may have a role in the differential treatment of European and non-European immigrants and their subsequent perceived discrimination. Unfortunately, information about the characteristics of immigrants, such as region or country of origin, was not available in our dataset. Immigration status should not be considered a confounder or covariate but a complex contextual variable. Immigration status alone may be a proxy of diverse personal experiences, including an individual's cultural background and perceived marginalization [51]. However, immigration status as a variable may not be enough to capture more nuanced factors, including length of stay, citizenship status, generation status, or even context of migration (forced displacement, including asylum or refugee status). Additionally, complex associations may be observed for perceived discrimination across sub-groups of immi-grants. For example, skin color and accent may explain why some groups of immigrants experience discrimination. Halanych et al. analyzed data from 1800 African American (45%) and White participants (55%) and showed a direct positive association between educational level and perceived discrimination [56]. While these associations were more prevalent among African American participants, sub-group differences such as gender, age, and income levels exist for African American and White Americans. Another study based on data from 2606 European adults found that perceived discrimination is associated with poor health outcomes among first-generation immigrants from low-income countries who live in European countries but not among their descendants [93]. These studies indicate the complexity of experiences of perceived discrimination among immigrant groups and the heterogeneity of perceived discrimination within and across immigrant groups that should be considered. While studies have established the association between perceived discrimination and health, there is a need to examine structural discrimination on health outcomes beyond interpersonal discrimination [8]. There is a need to investigate the social and political influences that give rise to perceived discrimination. While restrictive immigration policies have been found to be associated with higher levels of discrimination among immigrants [94], there is a need to examine how neighborhood factors such as ethnic enclaves and sanctuary cities may be protective against perceived discrimination. Moreover, there is a need to incorporate unique dimensions of the immigration experience, including citizenship and generation status. --- Limitations This study was cross-sectional. Perceived discrimination may be a driver for pursuing higher educational level [82]. Moreover, the data included no information regarding immigrants' length of stay. It is shown that new immigrant responders report perceived discrimination more frequently [81]. Additionally, we analyzed data from 10 European countries with different policies regarding immigrants. There is a need to study how much immigrants feel assimilated into the newly adopted culture [81]. In addition, the response rate in this study was slightly below 50%, which may cause sampling bias. Additionally, there may be several potential confounders that were not controlled in this study. Furthermore, no information was available regarding the country where they got their educational degree. According to a US study by Zhang et al., immigrants who had completed their education outside the US reported less discrimination than those who studied in the US [18]. Further studies are needed to evaluate if there are any differences between native-born and immigrants regarding the positive impact of educational level on health outcomes. --- Conclusions This study found that educational level was a protective factor against perceived discrimination among European adults. Despite the overall protective effect of educational level against discrimination, this effect was more robust in the native-born participants than in their immigrant counterparts. Future studies are needed to evaluate if there are any differences between native-born and immigrants in European countries regarding the positive impact of educational level on other health outcomes. --- Author Contributions: N.M., A.R., B.N. and H.S. designed the study, analyzed the data, and wrote the paper. N.M., A.R. and H.S. designed the conceptual design and revised the paper. H.Z. supervised the work, helped with the analysis, and revised the paper. All authors have read and agreed to the published version of the manuscript.

Background: Multiple studies have been conducted to test the moderating effect of immigration on the positive health results yielded through educational attainment. However, no study has been conducted to examine the role of immigration as a moderator in the association between educational level and perceived discrimination in Europe. Aim: We aimed to study whether an inverse association exists between educational level and perceived discrimination in European countries and whether immigration status moderates the association between educational level and perceived discrimination. Methods: Data from the 10th round of the cross-sectional European Social Survey (ESS) were used in this cross-sectional study. A total of 17,596 participants between 15-90 years old who lived in European countries were included. The independent variable was educational level, a categorical variable, and the dependent variable was perceived discrimination. Immigration status was the moderator, and age and sex were confounders. Results: Of 17,596 participants, 16,632 (94.5%) were native-born and 964 were immigrants (5.5%). We found that higher levels of educational level were protective against perceived discrimination, which was also found in immigrant participants; however, the effect was weaker. Conclusions: This study found that educational level was a protective factor against perceived discrimination. This effect, however, was more robust in the native-born participants than in their immigrant counterparts.

Introduction Afghanistan, in addition to the decades of armed conflict and the severity of the humanitarian crisis, is frequently at the top news headlines or global agenda for opium production, drug trafficking, and substance use. 1,2 Although there are local routine reports on opium production and drug trafficking, they discount the information on its long-term bio-psycho -social sequelae, such as adverse health outcomes and social problems, which in turn may lead to ill health and even social malfunctioning. 2 While it is well-known that substance use leads to the emergence of bio-psycho-social imperfections in a large number of cases, these disorders are largely under-reported. Depending upon substance use vulnerability factors, such as socioeconomic conditions, substance availability, peer behavior, and access to therapeutic and preventive measures, the prevalence of substance use disorders (SUD) is significantly variable. [3][4][5] Worldwide, over 2% of the population has alcohol or other substance use. 6 The prevalence of these disorders is high as 5% in the USA and Eastern Europe. 6 In 2015, the Afghanistan National Drug Use Survey (ANDUS) found that the prevalence of substance use was 11% and higher among male adults. 7,8 This high prevalence is attributed, in part, to the ongoing decades of conflict, unlimited availability of cheap narcotics, and limited access to treatment and preventive services. 5,7 Opioids (4.9%), cannabis (2.2%), and prescription sedatives (about 1%) are the most common substances used in Afghanistan. 7,8 The biological impact of substance use has been explored in the relevant literature. 9,10 According to the literature, substance use, including alcohol, heroin, marijuana, and cocaine, is independently associated with an increased risk for chronic pain, cancer, and heart disease that compromise the normal functioning of people with substance use disorders. 3,9 In addition to chronic disorders, there has been an increase in the risk of infectious disease transmission, including Human Immunodeficiency Virus (HIV) and hepatitis C virus (HCV). 9,10 However, in the developing world, especially Afghanistan, its impact is under-reported. The psychological consequences of substance use and its effects on family members are substantial. 9,11 Depression, anxiety, stress, and sleeplessness are frequently reported. 9 Long-term impacts include paranoia, psychosocial dependence and a higher inclination to chronic mental disorders, suicidal ideation, and self-harm intentions. [9][10][11] Pertinent literature reported that patients with mental health disorders are prone to substance use and vice versa. 10,11 Additionally, substance use and its related bio-psycho-social challenges are likely to harm the families of the substance users involved. 12 The nearly four decadeslong persistent warfare in Afghanistan and its inevitable multitude of consequences have most certainly contributed immensely to the increased psychological burden, particularly among Afghans with substance use disorders. In light of the aforementioned global data from peaceful and resourceful countries, the extent of the impact of this public health concern in Afghanistan, a country torn by nearly half a century-long war that continues to this day, is almost neglected. Considering the above gap in knowledge, this study aimed to address this lacuna by assessing the bio-psychosocial profile of people with substance use disorders treated at locally assigned treatment facilities in Kandahar, Afghanistan. This assessment will probably prove worthwhile in reflecting the necessity for healthcare services, especially mental health services, in these settings and providing implicit insight to relevant humanitarian assistance providers. --- Materials and Methods --- Study Settings and Design We conducted this facility-based cross-sectional study among people with substance use disorders receiving care at three locally assigned treatment facilities in Kandahar, Afghanistan, from November to December 2022. The locally assigned treatment facilities were a public rehabilitation center (22 clients), private rehabilitation center (13 clients), and correctional institution (586 clients). At the time of inclusion, the aforementioned centers were locally assigned to treat people with substance use disorders and were supported by governmental and non-governmental organizations. People with substance use disorders come to these centers in multiple ways; about 80% of them are referred by local authorities, and the remaining are self-referred. Patients admitted to these centers can receive primary health screening at the on-site health clinics. This initial visit also serves as an opportunity to test for sexually transmitted diseases, including HIV and HCV. Within the first week of arrival, the substance users at these centers visit the treatment facilities to develop a treatment plan. Psychiatric services are also routinely provided on-site; additional referrals to Mirwais Regional Hospital are arranged (if required) for more specialized care. The services provided in these three centers are largely similar to those described in our paper, with predominately several advantages (better treatment) in private rehabilitation centers. --- Study Population All people with substance use disorders living in the southwest of Afghanistan during the data collection period were the source population. The study population consisted of all substance users treated at locally assigned treatment facilities in Kandahar, Afghanistan. We excluded substance users with severe illness, those with compromised consciousness (unable to understand/utter a normal verbal response or not fully awake/alert), and those who refused to participate in the study. We approached 685 people with substance used disorders treated at locally assigned treatment facilities. The response rate was 92.2% (632). The final analysis consists of 621 participants with their complete data sets. --- Study Measures The socio-demographical characteristics of our participants encompassed age, gender, marital status, educational background, occupation, and documented travel history or living abroad. In this study, we employed the term substance use disorder to "any problematic pattern of using alcohol or another substance that results in impairment in daily life or noticeable distress" based on DSM-5. 13 Additionally, the term "psychological" refers to the symptoms of depression, anxiety, and stress. Substance use-related variables were the type and frequency of drugs used and the age at first use. Other variables of interest included a prior attempt to quit and their reasons for failure. The presence of concomitant medical conditions and hospitalization were health-related characteristics of the people with substance use disorders. This study employed the Depression, Anxiety, and Stress (DASS-21) scale to measure the presence and severity of psychological symptoms during the last two weeks and assess the prevalence of probable depression, anxiety and stress. 14,15 DASS-21 is a 4-point scale from 0 (not at all) to 4 (extremely) that yields a total score from 0 to 84. The questionnaire is a globally credible instrument with good psychometric properties for assessing depression, anxiety, and stress based on DSM-IV criteria. 15 The internal consistency (Cronbach's alpha) value for the Pashtu version in the present study, respectively, was 0.75, 0.77, 0.70, and 0.88 for depression, anxiety, stress, and DAS-21 total scales. --- Data Collection The study questionnaire was initially drafted in English and subsequently translated into local language (Supplementary Figure 1) for the convenience of interviewing the respondents. After the translation process, a pilot test was conducted to assess the clarity and simplicity of the questionnaire among a small group of participants who were similar to the study population in terms of age, sex, and education level (Kandahar Teaching Hospital). The pilot test results were used to refine the questionnaire and ensure that it was culturally appropriate and linguistically accurate for the target population. The data were collected by four male nurses and one supervisor (Master of Public Health Program trainee). They were provided with one-day classroom training on sampling, interviewing, filling out questionnaires, and addressing potential ethical concerns that may emerge during the study, followed by one day of practical field training. The interviewer screened and interviewed every consecutive client (non-probability sampling technique) treated in one of the locally assigned treatment facilities. Before their invitation to voluntarily participate, the recruiters offered the potential participants a written synopsis of the study objectives and information on the course of the research. If the client would like to participate, they could willfully date and sign the consent form. Only then the survey team interviewed the participants. Each interview took about 15-20 minutes. Daily, we checked the questionnaires for completion. We kept the completed questionnaires in locked cabinets and stored the electronic data in password-protected computers to which only the principal investigators had access. --- Statistical Analysis We employed the Statistical Package for Social Sciences (SPSS) version 22 for data analysis. 16 We conducted descriptive statistics to understand participants' bio-psycho-social characteristics. The total score of DAS-21-total, depression, anxiety, and stress was calculated as the sum of the component scores and multiplied by two. 14 Since the three subscales of depression, anxiety, and stress were strongly associated with each other (depression-anxiety: Pearson r= 0.7, P <unk>0.001; anxiety-stress: Pearson r= 0.7, P <unk>0.001; depression-stress: Pearson r= 0.8, P <unk>0.001). Therefore, we used the DASS-21 total score in a linear regression model to assess factors associated with psychological symptoms in the sample. 17 The alpha level was set at P <unk> 0.05. The following characteristics of the people with substance use disorders were entered into the model: Model: Educational attainment, employment status, marital status, pre-existing medical condition, level of social support, and history of living abroad in the past ten years. --- Results From the three locally assigned treatment facilities, a total of 621 male substance users were included in the analyses. The mean age of the participants was 34.41 (<unk> 10.10 SD) years with a range of 17-70 years, and a majority (475; 76.5%) of them were younger than 40 years. Regarding their marital status, the majority (528, 85.0%) of them were married. About 11.1% (69) of the participants have lived abroad in the past ten years. Data on participants' educational attainment and employment status are summarized in Table 1. The mean age at first substance use was 23.01 (<unk> 5.89 SD) years, with 82.4% (512) reporting hashish as an initiation to substance use. About half (331, 53.3%) of the substance users in our sample reported daily substance use prior to their treatment. The most common reasons for substance use were friend offers (303, 48.8%) and family disputes (112, 18%). Almost all (99%, 615) participants had made unsuccessful attempts to quit substance use in the past. The top three reasons for relapse to substance use were peer pressure (58.8%), economic problems (18.4%), and lack of social support (14.0%). Table 2 portrays the detailed characteristics of the participants about their substance use behaviors. --- 93 The majority of the subjects (81.2%) perceived their physical health as either very good or fair. More than two-thirds (73.4%) rated their social support as high. Overall, 46 (7.4%) and 32 (5.2%) of the study enrollees, respectively, had a physical injury or illness and were hospitalized in the past 30 days (Table 3). Of all participants, 541 (87.1%) had symptoms of depression, 569 (91.6%) of anxiety, and 442 (71.2%) of stress. The prevalence of severe depression, anxiety, and stress was 34.8%, 65.8%, and 27.3%, respectively (Figure 1). --- Factors Associated with Psychological Symptoms in People with Substance Use Disorders The results in Table 3 portray the multiple linear regressions. The multiple linear regression showed that several attributes of people with substance use disorders [ie, having a low level of education (<unk>=0.12, p=<unk>0.001), being unemployed (<unk>=0.31, p=<unk>0.001), having a low level of social support (<unk>=-0.35, p=<unk>0.001), had a pre-existing medical condition (<unk>=-0.28, p=<unk>0.001), and having lived abroad in the past ten years (<unk>=0.10, p=0.001)] were significantly associated with higher DASS-21 total scores (Table 4). --- Discussion This study reveals bio-psycho-social profile of people with substance use disorders treated at locally assigned treatment facilities in Kandahar, Afghanistan. Most of our subjects (81.2%) perceived their physical health as very good or fair. Notably, the majority of the participants rated their social support as high. This study portrays a high prevalence of depression, anxiety, and stress in our subjects, reflecting the necessity for mental health services. We observed that several attributes of people with substance use disorders (ie, their level of education, employment status, level of social support, history of a pre-existing medical condition, and history of living abroad in the past ten years) were highly associated with psychological symptoms. The majority of substance users (81.2%) in this study rated their physical health either as very good or fair. Parallel studies conducted on substance users in diverse developing countries show mixed results on the physical health of people --- DovePress with substance use disorders. 11,13 We argue that our participants may have adjusted their responses and made the responses appropriate for the interview setting. From a more realistic standpoint, however, a lack of bio-psycho-social support programs, financial restraints, and an increased risk for sexually transmittable infections, including HIV and HCV are hypothesized to be the crucial underlying factors that may have led to a poor physical health in our sample. 13 Our findings indicate that hashish (82.4%) is commonly used as a starting point for individuals who later develop substance use disorders. Moreover, our results support the findings of pertinent literature that social factors, such as peer influence and family conflict, play a significant role in the development of substance use disorders. 18,19 In the present study, the DASS-21 scale has been used to screen the existence of depression, anxiety, and stress in our subjects. The high levels of depression, anxiety, and stress are illustrated in Figure 1, with notably higher rates of psychological symptoms than that reported in other studies from developing countries. 11,13,20 A limited number of rehabilitation facilities in terms of infrastructure and personnel, the exceptional severity and lengthy duration of warfare, social discrimination and marginalization, unemployment, poverty, and illiteracy in this geography are some of the factors that may have brought about such a staggering level of psychological symptoms. Therefore, the provision of mental health services for this population is of paramount importance. Our results epitomize the difference in psychological symptoms based on the participants' educational attainment. A significantly higher number (357; 57.1%) of people with substance use disorders with a low level of education were more susceptible to psychological symptoms than their educated counterparts. This finding, reported in the literature, may somehow signify the difficulty in accessing care, poor resilience capability, limited access to health awareness programs, and other sociocultural impediments in populations with a low level of education. 11,21,22 Consistent with the pertinent literature, our findings indicate that psychological symptoms were higher in respondents who were unemployed than those who were employed. This result is supported by other studies revealing that unemployed individuals are more likely to report psychological symptoms. 11,23 This finding may elucidate additional stressors, such as severe financial hardship and poverty, debt, boredom, homelessness, and the loss of confidence and self-esteem that people with substance use disorders may have to endure. 23 Thus, government and health authorities should develop appropriate intervention strategies and policy guidelines to protect and preserve the psychological health of those susceptible to psychological symptoms. A great deal of evidence has shown that low social support markedly increased the rate of psychological symptoms. 11,23,24 Notably, we have found that people with substance use disorders with low social support were more likely to report higher rates of psychological symptoms than their counterparts with high social support. The higher levels of depression, anxiety, and stress in people with substance use disorders having low social support have been attributed to the impeded social interactions on the family and societal levels that may have been quite stressful. Considering the high rates of the psychological burden from low social support in people with substance use disorders, families and friends should provide social support and help them in lessening the bio-psycho-social problems linked to substance use. Expectedly, and as established in previous studies, we found that the presence of a pre-existing medical condition was highly associated with psychological symptoms, including depression, anxiety, and stress. 18,23 A significant body of epidemiologic studies highlights that the solo presence of medical conditions, such as cancer, Diabetes Mellitus (DM), and HIV, is associated with poor mental health. 25,26 Hence, people with substance use disorders having pre-existing medical condition should be identified and treated at the earliest to circumvent additional medical conditions and their psychosocial sequelae. Substance use is common among Afghans settled in refugee camps in neighboring countries. 1,27 Additionally, we observed that people with substance use disorders having a history of living abroad in the last ten years were more likely to develop psychological symptoms than their counterparts. The ongoing multidimensional conflict and subsequent displacement crises have resulted in severe psychological trauma for many Afghans on multiple levels. --- Limitations Several limitations of this study should be considered. Firstly, this is not a detailed bio-psycho-social profile of people with substance use disorders, as we were more inclined to specific aspects of biological, social, and mental health problems. Therefore, our findings might be biased, not representing a comprehensive profile of bio-psycho-social diagnosis in our sample. Secondly, we only included people with substance use disorders who were treated for substance use. It is likely that individuals not seeking care could have substantially more severe conditions. Thirdly, this study was done in one area (Kandahar province); generalization should be made with caution. Finally, the study is also limited by a relative lack of access to reliable HIV and HCV data, which were deemed confidential and could not be disclosed. --- Implications for Future Research Future research may investigate chronic medical conditions in people with substance use disorders as well as their associations with the time span since initiation to substance use. Another venue to explore may be related to the effective role of the rehabilitation programs on their bio-psycho-social health. Furthermore, the bio-psycho-social health of people with substance use disorders not enrolled in rehabilitation programs will be an area for future research. --- Conclusion We concluded that people with substance use disorders were vulnerable to a high magnitude of bio-psycho-social problems, necessitating multilevel bio-psycho-social interventions to alleviate the burden of substance use. Moreover, with an additional emphasis on those with a low level of education, those who have lived abroad in the past ten years, are unemployed, have a low level of social support, and those with a pre-existing medical condition. --- Data Sharing Statement The primary data used to support the findings of this study are available from the corresponding author upon request. --- Ethical Approval The study was approved by the Research and Ethics Committee (REC) of Kandahar University (Letter No. 76; Dated: 15/ 09/2022). We followed the ethical principles outlined in the Declaration of Helsinki. Additionally, the interviewer provided every potential participant with information concerning the study objectives, benefits/risks of participation, and their right to retract from the study at any time they desire. We obtained written informed consent from every subject. --- Author Contributions Both authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work. --- Disclosure The authors have no conflicts of interest in this work. --- Substance Abuse and Rehabilitation --- Dovepress --- Publish your work in this journal Substance Abuse and Rehabilitation is an international, peer-reviewed, open access journal publishing original research, case reports, editorials, reviews and commentaries on all areas of addiction and substance abuse and options for treatment and rehabilitation. The manuscript management system is completely online and includes a very quick and fair peer-review system. Visit http://www.dovepress.com/testimonials.php to read real quotes from published authors. Submit your manuscript here: http://www.dovepress.com/substance-abuse-and-rehabilitation-journal --- DovePress Substance Abuse and Rehabilitation 2023:14

Background: Substance use and its associated bio-psycho-social problems are public health concerns with harmful individual and social consequences. Objective: This study assessed the bio-psycho-social profile of people with substance use disorders (SUD) treated at locally assigned treatment facilities in Kandahar, Afghanistan. Methods: We conducted this facility-based cross-sectional study among 621 substance users receiving care at three locally assigned treatment facilities in November-December 2022. We performed multiple linear regression to determine factors associated with psychological symptoms. Results: The mean age of the participants was 34.41 (± 10.10 SD) years. The majority of our subjects (81.2%) perceived their physical health as very good or fair. More than two-thirds (73.4%) rated their social support as high. Of all participants, 541 (87.1%) had symptoms of depression, 569 (91.6%) of anxiety, and 442 (71.2%) of stress. The prevalence of severe depression, anxiety, and stress was 34.8%, 65.8%, and 27.3%, respectively. The multiple linear regression showed that several attributes of people with substance use disorders [ie, having a low level of education (β=0.12, p=<0.001), being unemployed (β=0.31, p=<0.001), having a low level of social support (β=-0.35, p=<0.001), had a pre-existing medical condition (β=-0.28, p=<0.001), and having lived abroad in the past ten years (β=0.10, p=0.001)] were significantly associated with higher DASS-21 total scores.This study highlights the importance of providing bio-psycho-social support programs and implementing therapeutic interventions to help people with substance use disorders, particularly those who are most susceptible to higher levels of bio-psychosocial problems.

Background Skewed distribution of the health workforce is a global phenomenon [1] that intensifies the well-known crisis of scarce personnel in many countries' health sectors. This greatly affects low-and middle-income countries (LMIC), although sub-Saharan Africa (SSA) remains the most affected region [2,3]. The shortage of health personnel in absolute numbers which has received a considerable attention among researchers is one of the health systems' pressing problems in developing countries [4,5] including Tanzania [6]. This has been acknowledged as a serious threat towards meeting the Millennium Development Goals (MDG) [7][8][9]. According to Barden-O'Fallon et al. [10], accurate knowledge of characteristics of the health workforce that affect health care production is of critical importance to health planners and policymakers. A global picture shows that skewed distribution of the health workforce geographically, and by specific characteristics, poses significant challenges to quality health care delivery [3]. Some forms of the skewness include skill mix, over specialization and gender [1], the latter being subtly discussed. So far, gender has been much less considered in health workforce-related matters both in the formal and informal systems of health care production [11,12]. However, a clear gender component reportedly exists in several ways. In formal health care systems, for example, women are less likely to be in senior, managerial and policy making roles than their male counterparts [13,14], whereas non-institutional care for the sick is often carried out by women [15]. Since poor health care utilization for some individuals due to the absence of a provider of a particular gender has been reported [16], gender balance in the cadres of health care providers may be an imperative response. The United Republic of Tanzania's health workforce is reportedly small, both by international and national standards [2]. In 2006, the Ministry of Health and Social Welfare (MoHSW) estimated a massive shortage in the health workforce by 65% of the staffing requirement [17]. This situation was declared a threat to effective delivery of quality health services [18]. The shortage is aggravated by, among other factors, population expansion and attrition as well as increasing disease burden due to HIV/AIDS, tuberculosis (TB) and malaria [19]. It is noted that the health workforce challenges in the country are significantly related to poor working conditions, a situation that drives some staff, especially highly trained ones, to seek employment outside the country (brain drain); those who remain working in the country become greatly demoralized [20]. A cross-sectional study conducted by Leshabari and colleagues [21] at the Muhimbili National Hospital (MNH) to assess health workers' service and care delivery motivation found that nearly 50% of doctors and nurses were not satisfied with their jobs. Low salaries, persistent unavailability of the necessary equipment for service delivery, inadequate performance evaluation and feedback, poor communication between workers and management and lack of participation in decision-making were among major reasons reported for the dissatisfaction. Moreover, previous evaluations have established that the health workforce in the United Republic of Tanzania is inequitably distributed [3]. There are more health personnel in urban-based health facilities than their rural counterparts [1]. xEven between districts, disparities in staffing levels exist. This is partly due to the fact that some districts host regional or tertiary hospitals, thus requiring more health personnel than the ordinary district hospitals [3]. The problem is more serious in cases where new districts were formed after dividing the original districts. There is also geographical skewness in the skill mix distribution among the existing cadres, whereby the most skilled and specialized personnel are less likely to work in rural facilities [6]. The most qualified health personnel are concentrated in a few centralized locations mainly in urban or peri-urban centres where they can access basic social services and desirable infrastructural facilities, while severe understaffing reigns at dispensary levels, especially those located in rural and peripheral settings [22]. Apart from these challenges, most of which are basically distributional, there is a general lack of empirical analysis of the status of the gender-based distribution of the health workforce in the United Republic of Tanzania. The slim evidence available shows a link between gender and geographical imbalance in the distribution of the United Republic of Tanzania's health workforce. Reportedly, while rural facilities are severely understaffed, they are also less likely to be served by female providers [10]. Evidence from other countries shows the presence of more women in lower-status health occupations usually performed by personnel of low education on the one hand and fewer women than men among highly trained professional staff for direct health service delivery and management positions on the other [23]. The distributional skewness favours women among nurses, but women are poorly represented among doctors, dentists, pharmacists, clinical officers and managers [23]. Poor representation of women in higher-status health cadres may lead to poorer understanding of problems that are specific to women [24]. It has been further reported that female general practitioners practice differently from their male counterparts. They are capable of managing a variety of medical conditions with some differences due to patient mix and patient selectivity. On the other hand, research studies reveal that some women in more traditional areas will not seek health care for themselves or even for their children unless they have access to a female provider [25]. It has also been established that women cannot be seen by male doctors in some parts of the world [16]. While gender imbalance in the distribution of the health workforce may be affecting service production, delivery and utilization in the United Republic of Tanzania, there is no evidence available to reveal the status of gender in the distribution of HWs. The gender-based distribution of the health workforce remains barely addressed, and no literature so far clearly documents the situation in the United Republic of Tanzania. Even the few attempts in other countries that have discussed gender in relation to the distribution of the health workforce, lack details as they did not focus solely on gender. As a result, they have simply reported frequencies, without considering such robust techniques of analysis as regressions, where gender skewness could be assessed while other characteristics, such as location, age, education and so on, were controlled for. This paper, thus, has two objectives: (1) to examine the composition of the United Republic of Tanzania's health workforce cadres by gender and (2) to assess the predictive effect of gender on each of the United Republic of Tanzanian health cadres surveyed, namely, maternal and child health aide or medical attendant (MCHA/MA), nurse, midwife, clinical officer (CO) and medical doctor (MD), using multivariate logistic regression. --- Methods --- Study design, sampling, area and population This is a secondary analysis of data collected in 2008 in a cross-sectional health facility survey in the United Republic of Tanzania. The survey was conducted by the Ifakara Health Institute (IHI), United Republic of Tanzania, in collaboration with Columbia University, USA, as part of the implementation of the Health Systems Strengthening for Equity (HSSE) project. Based on the eight United Republic of Tanzanian zones, one region from each zone was selected randomly through a multi-stage sampling technique which brought up eight regions: Dodoma, Pwani, Mwanza, Tanga, Mbeya, Iringa, Tabora and Mtwara. From each district in these regions, two health facilities (one hospital and/or one health centre) providing emergency obstetric care (EmOC) were selected. This made a total of 88 health facilities from which 825 health workers (HWs) participated in the survey. Of these HWs, 815 (98.8%) with non-missing data on gender and cadre were extracted from the parent database for the current analysis. --- Data collection tool Selected HWs for the primary study responded to a selfadministered provider questionnaire which comprised mostly closed-ended questions and a few open-ended ones. Broadly, the questions pertained to the HWs' background and employment, pre-or in-service training programs attended, feelings of job satisfaction, and a discrete choice experiment which aimed at understanding factors that affect employment preferences. Following interviewer training, the tool was pre-tested in facilities similar to those actually surveyed to check for relevance and answerability of the survey questions. --- Data management and statistical analyses Original cadres were operationally regrouped by merging those that were closely related because the numbers of respondents for some cadres, such as specialists, were very small. Maternal and child health aide (MCHA), medical attendant (MA) and nursing assistant were joined to form a single category, "MCHA/MA"; registered public health nurse (PHN), enrolled public health nurse (EPHN), registered nurse (RN) and enrolled nurse (EN) were combined into a single category and referred to as "nurse"; registered midwife and enrolled midwife were grouped together as "midwife"; clinical officer (CO) remained unchanged; and assistant medical officer (AMO), medical officer (MO) and specialist were combined and referred to as "medical doctor (MD)". Gender-specific proportions of the HW in various categories of socio-demographic characteristics were calculated. The degree of association between gender and sociodemographic characteristics was tested using Pearson's Chi-square (<unk> 2 ) and Student's t-tests for categorical and continuous variables, respectively. Further analyses were performed using multivariate logistic regression to assess the relationship between gender and each of the cadres, controlling for potential confounders. Each of these cadres was assessed as a separate dependent variable with two (binary) categories that classified a HW as either an MD or not an MD, CO or not a CO and so on. As coding of the outcome variable in logistic regression requires, a code of '1' was assigned if a HW belonged to a particular cadre and '0' if not. Therefore, the probability that a HW was an MD for example was expressed in a multivariate logistic regression model as: p 1<unk>4 e b 0 <unk>b 1 X 1 <unk>b 2 X 2 <unk>...<unk>b k X k <unk> <unk> 1 <unk> e b 0 <unk>b 1 X 1 <unk>b 2 X 2 <unk>...<unk>b k X k <unk> <unk> where p is the expected probability that a HW is an MD; X 1 through X k are k distinct independent variables; and b 0 through b k are the regression coefficients. The model was then re-written with the outcome expressed as the expected natural logarithm of the odds that a HW is an MD as: In p 1 <unk> p 1<unk>4 b 0 <unk> b 1 X 1 <unk> b 2 X 2 <unk>... <unk> b k X k The main independent variable was gender (female = 1, male = 0). This variable was taken along with other several independent variables including age, educational attainment, region, health facility ownership, and health facility type. Educational attainment was included because of evidence from other countries showing that some women do not prefer courses that take a long time to graduate, resulting in fewer women in specialized professional jobs [26]. Facility type was included in the analysis as an indicator of facility location. In the United Republic of Tanzanian health system context, a hospital is the highest level of care that serves either a region (regional hospital) or a district (district hospital) and is usually located in the headquarters of regions or districts. Therefore, it may be appropriate to consider hospital locations in the United Republic of Tanzania as urban. Health centres on the other hand, which are the second highest level of care in the United Republic of Tanzania, exist mostly in rural and sometimes in urban settings. Therefore, this variable to a larger extent reflects the rural-urban distribution of HWs in the country. Data analysis was performed using STATA (Version 11) statistical software (Stata Corp, Texas, USA). --- Ethical approval Ethical approval to conduct the main survey from which this paper stems was granted by the Medical Research Coordinating Committee (MRCC) of the National Institute for Medical Research (NIMR) in the United Republic of Tanzania. Participation in the study was voluntary with all consenting individuals having to sign an informed consent form first. To ensure integrity and confidentiality, the database was anonymous with no information (e.g. names) that could identify the participant. Storage of completed questionnaires and consent forms was carefully managed, and access to the data was restricted to a few experts. --- Results Of the 815 HWs analyzed, 75% were women. Their mean age was 39.7 <unk> 9.0 years, ranging from 15 to 63. The proportions of female staff in different cadres were as follows: MCHA/MA, 86%; nurse, 86%; midwife, 91%; CO, 28%; and MD, 21% (Figure 1). Table 1 presents the distribution of female and male HWs by their background characteristics. The contribution of each cadre to the total number of HWs was: MCHA/MA, 21.4%; nurse, 26.9%; midwife, 31.5%; CO, 11.0%; and MD, 9.2%. These proportions were gender maldistributional (P <unk> 0.001), with female HWs being the majority among MCHA/MA, nurses and midwives but the minority among COs and MDs. With respect to the highest level of education attained, 59.6% (70.2% female and 27.4% male) were certificate holders; 28.7% (24.2% female and 42.1% male) held an ordinary diploma while 11.8% (5.6% female and 30.5% male) held at least an advanced diploma. These differences in the highest level of education by gender were statistically significant (P <unk> 0.001). With regard to the regions where the HWs worked, 13.9% (13.8% female and 13.9% male) worked in Dodoma; 14.2% (15.5% female and 10.5% male) worked in Iringa, 16.1% (16.6% female and 14.4% male) worked in Mbeya; 4.8% (3.8% female and 8.0% male) worked in Mtwara. Also, 13.4% (13.8% female and 11.9% male) worked in Mwanza, 12.4% (11.6% female and 14.9% male) worked in Pwani; 12.5% (10.6% female and 18.4% male) worked in Tabora and 12.8% (14.3% female and 8.0% male) worked in Tanga. Gender differences across the regions were statistically significant (P = 0.002). Although the distribution of the female and male HWs by health facility ownership was symmetric (P = 0.995), about 90% of the HWs worked in government facilities. Also, about two-thirds (66.1%) of all HWs worked in hospitals and the rest in health centres with no significant differences by gender (P = 0.125). --- Regression results The relationship between gender and the cadres of the United Republic of Tanzania's health workforce was further assessed in a multivariable style, adjusting for age, education, region, health facility ownership and health facility type (Table 2). The results reveal a massive distributional skewness of the workforce by gender. The odds of an MCHA/MA being a woman was 3.70 times higher than that for their male counterparts (OR = 3.70, 95% CI 2.16-6.63). Similarly, a nurse was 5.61 times more likely to be a woman than to be a man (OR = 5.61, 95% CI 3.22-9.78). It was also almost three times more likely that a midwife was a woman (OR = 2.74, 95% CI 1.44-5.20). On the other hand, the chance that a CO was a woman was 92% less likely than being a man (OR = 0.08, 95% CI 0.04-0.17) and, likewise, an MD was 96% less likely to be a woman than to be a man (OR = 0.04, 95% CI 0.02-0.09). Apart from gender, there were other factors with a significant association with the cadres. A one year increase in age was associated with a 6% increased likelihood of a health worker being an MCHA/MA (OR = 1.06, 95% CI 1.04-1.08); 7% less likely to be a midwife (OR = 0.93, 95% CI 0.91-0.96); and 4% more likely to be an MD (OR = 1.04, 95% CI 1.01-1.07). In terms of educational attainment, nurses were almost 15 times more likely to be holding an ordinary diploma than a certificate (OR = 15.07, 95% CI 9.45-24.03), whereas midwives were 98% and 92% less likely to be holders of an ordinary diploma (OR = 0.02, 95% CI 0.01-0.05) and advanced diploma or higher (OR = 0.08, 95% CI 0.03-0.21), respectively, than a certificate. COs were about 46 times more likely to hold an ordinary diploma than a certificate (OR = 46.45, 95%. There was also a geographical imbalance in the distribution of some cadres. With Dodoma being a baseline or a reference category, there was a higher likelihood of more MCHA/MAs in Mwanza than Dodoma (OR = 2.47, 95% CI 1.22-4.99). The presence of nurses in Tanga was twice as likely as in Dodoma (OR = 2.04, 94% CI 1.01-4.12). Also, midwives were about four times more likely in Mtwara than in Dodoma (OR = 3.88, 95% CI 1. 24-12.19). While the presence of COs in Iringa was 84% less likely compared to Dodoma (OR = 0.16, 95% CI 0.03-0.86), it was 3.35 times more likely that they were present in Pwani than in Dodoma (OR = 3.35, 95% CI 1.02-11.01). Furthermore, the availability of MDs was 96% less likely in Iringa (OR = 0.04, 95% CI 0.01-0.36), 66% less likely in Pwani (OR = 0.34, 95% CI 0.12-0.97) and 80% less likely in Tabora (OR = 0.20, 95% CI 0.07-0.57) compared to the Dodoma region. With respect to health facility ownership, MCHA/ MAs were about three times more likely to be working in private facilities than in government facilities (OR = 2.73, 95% CI 1.52-4.89), whereas midwives were 67% less likely to be working in private facilities compared to government facilities (OR = 0.33, 95% CI 0.16-0.66). In terms of facility type, MCHA/MAs were 53% less likely to be working in hospitals than health centres. Similarly, COs were 86% less likely to be working in hospitals than health centres. In contrast, midwives and MDs were 2.72 (OR = 2.72, 95% CI 1.71-4.32) and 3.98 (OR = 3.98, 95% CI 1.92-8.25) times more likely to be working in hospitals than in health centres, respectively. --- Discussion This paper sought to examine the gender-based distribution of the United Republic of Tanzania's health workforce cadres. This question has so far received limited attention in previous analyses [11,12]. Our findings demonstrate that the distribution of the health cadres in the United Republic of Tanzania is dramatically gender skewed, probably reflecting gender inequality in health career choices. Nursing and midwifery cadres are generally large and female-dominant, while CO and MD cadres are small in absolute numbers and male-dominant. The possibility of this happening may be associated with cultural, familial or systemic structures underlying values and specialty preferences in choosing careers. Our findings are consistent with evidence from other studies in which female health personnel are the minority among the highly trained cadres [23], but the majority in other health specializations [27]. The overall picture of three-quarters of the HWs being women in this study area is similar to that reported in Congo where women are the majority [28]. The gender-based distributional skewness in the United Republic of Tanzanian health workforce cadres persisted and remained highly significant even after controlling for other variables in a multivariate logistic regression model. There is a possibility that care-related jobs, such as nursing and midwifery, attract more women than men in the United Republic of Tanzania. There is also a chance that the services delivered by people from the MCHA/MA, nursing and midwifery cadres tend to be perceived as more suitable for women than men. This, however, needs further elucidation. Previous studies have shown that the dominance of women among nurses is a historical phenomenon [29] and generally apparent in lower cadres [30]. That is why it is important to examine why this has been so and whether achieving a balance can in any way reduce the HW shortage. We observed the presence of more female-than maleholders of certificates, while the proportion of women declined rapidly among those with higher qualifications. Although the higher the qualification the lower was the proportion of both women and men, the data show that the proportion of men outnumbered their female counterparts for diploma, advanced diploma and higher qualifications. This trend somewhat prevailed in terms of the leaving level for basic education, such that the proportion of women with primary education was 1.5 times as high as that for men, whereas the proportion of those holding an ordinary secondary education was similar between women and men. However, the proportion of women with advanced secondary education which is a key entry point into colleges and universities for higher education was five times as small as that of men. Leon and Kolstad [31] report a lack of primary interest in medicine among medical school entrants. The authors further report skewed recruitment and the absence of a rural-related clinical curriculum as facilitators of inequitable distribution of doctors in the United Republic of Tanzania. Although the study did not clarify the observed effect by gender, it may be that the overall effect of these factors is gender-sensitive. Further research is, therefore, needed to provide explanations. Moreover, the demand for more specialized cadres such as MDs may be higher for hospitals than health centres because the former is a higher level of care with more specialized services than the latter. Hospitals also constitute the receiving end of the referral pathway in the health service delivery framework of the United Republic of Tanzania. That is why facility type was a significant explanatory variable of the cadres' distribution, with hospitals being more likely than health centres to have MDs, for example. Additionally, most hospitals are in urban settings where MDs prefer to work as already noted [22]. Our findings also indicate that MCHA/MAs and COs are less likely to work in hospitals, implying somewhat that these are the likely cadres in health centres and most rural facilities where MDs in the United Republic of Tanzania are rarely present. Generally, the gender skewness observed in this study may be linked to gender differences in entering and leaving various academic junctures. In a study on gender differences in performance in science subjects in the United Republic of Tanzania, Mushi [32] found that more girls than boys fail in science subjects in the Advanced Certificate of Secondary Education Examination (ACSEE). Also, very few girls enroll in science subjects at the universities, implying that fewer women eventually proceed to highly trained medical cadres. We agree with previous authors that a strong health system requires among other things a strong, adequate and balanced workforce to ensure effective and efficient delivery of quality health services [17]. Therefore, we additionally emphasize that that gender should be considered important in human resources planning to ensure a balanced health workforce. This is because of the existing evidence showing that many women prefer female providers in certain kinds of health services such as those involving intimate physical examination [33,34]. Moreover, it has been reported that some people, especially women, fail to seek health care when they have no access to a provider of their gender [25]. While all this may be true in the United Republic of Tanzania, there currently exists no evidence showing whether or not this is so; thus, there is a need for researchers to provide clarifications. In this case, it may be good to explore whether or not men would prefer particular health services for their wives, such as those associated with pregnancy and childbirth, to be delivered by female or male providers. Therefore, it is important that the United Republic of Tanzanian school admission policies, recruitment and on-the-job training modalities are re-examined in order to ensure a good balance between female and male HWs, especially in highly trained cadres. This will not only respond to the shortage of HWs in the United Republic of Tanzania where records show a very low doctor-to-population ratio [35], but also cater for gender-specific roles and needs. --- Limitations and strengths Although the present paper is limited to HWs in health facilities providing EmOC, the findings and their interpretation respond to the existing gender-related gaps in the health workforce distribution in the United Republic of Tanzania. Factors such as marital status and spouse's place of work which may also influence one's choice of career specialization were not available. Future research should examine more cadres at all levels of the United Republic of Tanzanian health care delivery framework. Despite these limitations, this study makes an important contribution to the human resources for health literature by deeply considering the gender dimension in the distribution of the health workforce in the United Republic of Tanzania. Unlike some studies in the field [6,16], our study goes beyond descriptive analysis to apply multivariate analysis to control for confounding factors and see clearly the role of gender in the distribution of HWs in the United Republic of Tanzania. Gender is an aspect often overlooked in medical and general public health studies that address questions related to human resources for health in various countries [26]. --- Conclusions The gender-based distribution of the United Republic of Tanzania's health workforce cadres is significantly skewed, with MCHA/MA, nursing and midwifery cadres being large and female-dominant, whereas CO and MD cadres are slim in absolute numbers and male-dominant. It is important that gender be considered a dimension of crucial importance in human resources planning in the United Republic of Tanzania as this will not only contribute to uplifting the United Republic of Tanzania's doctor-to-population ratio, but will also respond to possible health service users' preferences regarding the gender of the service providers. As a result, this may enhance health seeking behaviour and health care utilization. --- Competing interests All authors declare that they have no competing interests. Authors' contributions AE conceptualized the problem, designed the study, participated in field data collection, performed data analysis and drafted the manuscript. AML participated in designing the study, data analysis and critical review of the manuscript. NW participated in drafting the manuscript and critical review of it. GM oversaw the study design as well as critical review of the manuscript. SP and GMM contributed to interpretations and overall review of the manuscript. All authors read and approved the final draft.

Background: While severe shortages, inadequate skills and a geographical imbalance of health personnel have been consistently documented over the years as long term critical challenges in the health sector of the United Republic of Tanzania, there is limited evidence on the gender-based distribution of the health workforce and its likely implications. Extant evidence shows that some people may not seek healthcare unless they have access to a provider of their gender. This paper, therefore, assesses the gender-based distribution of the United Republic of Tanzania's health workforce cadres. Methods: This is a secondary analysis of data collected in a cross-sectional health facility survey on health system strengthening in the United Republic of Tanzania in 2008. During the survey, 88 health facilities, selected randomly from 8 regions, yielded 815 health workers (HWs) eligible for the current analysis. While Chi-square was used for testing associations in the bivariate analysis, multivariate analysis was conducted using logistic regression to assess the relationship between gender and each of the cadres involved in the analysis. Results: The mean age of the HWs was 39.7, ranging from 15 to 63 years. Overall, 75% of the HWs were women. The proportion of women among maternal and child health aides or medical attendants (MCHA/MA), nurses and midwives was 86%, 86% and 91%, respectively, while their proportion among clinical officers (COs) and medical doctors (MDs) was 28% and 21%, respectively. Multivariate analysis revealed that the odds ratio (OR) and 95% confidence interval (CI) that a HW was a female (baseline category is "male") for each cadre was: MCHA/MA, OR = 3.70, 95% CI 2.16-6.33; nurse, OR = 5.61, 95% CI 3.22-9.78; midwife, OR = 2.74, 95% CI 1.44-5.20; CO, OR = 0.08, 95% CI 0.04-0.17 and MD, OR = 0.04, 95% CI 0.02-0.09.The distribution of the United Republic of Tanzania's health cadres is dramatically gender-skewed, a reflection of gender inequality in health career choices. MCHA/MA, nursing and midwifery cadres are large and female-dominant, whereas COs and MDs are fewer in absolute numbers and male-dominant. While a need for more staff is necessary for an effective delivery of quality health services, adequate representation of women in highly trained cadres is imperative to enhance responses to some gender-specific roles and needs.

Introduction The realm of gender in sport is a complex, ever-changing area that is closely linked to participation levels, opportunities and the treatment of individuals based on gender differences. This theme not only crosses gender lines between sports, but also delves into enduring social ideas surrounding gender roles and abilities. Over time, men's sports have received increased attention, financial support, and media coverage, increasing the visibility and prospects of male athletes. This stark contrast perpetuates a long-standing gender gap in sport, affecting important areas such as pay, sponsorship, access to facilities and resources available for training and competition (Ariyo, E. S., 2022). The surge of women's soccer onto the global stage in the 1990s marked an era of immense growth and promise within the sports landscape. Despite this, the realm of professional soccer remains predominantly male-centric, creating a stark imbalance in the treatment and compensation of female players. A glaring discrepancy exists between the salaries, resources, and investment afforded to male and female soccer players, perpetuating a systemic inequality that hampers the progress and livelihoods of women in the sport (Murray, C., 2019). The call for equal pay for equal work has reverberated strongly within the domain of women's soccer, spurred by the principles of feminism advocating for equitable treatment across all spheres. This demand has gained substantial traction, especially as the U.S. professional women's soccer team has made remarkable strides in international tournaments, drawing attention to the glaring disparities in remuneration and resources between male and female athletes. Motivated by the pressing need to address these disparities and advocate for fair treatment, this research endeavors to delve into the multifaceted aspects of the demand for equal pay in professional women's soccer. Drawing upon sociological and feminist perspectives, this study aims to dissect the controversy surrounding this issue, dissecting the implications and significance of equal pay not just for women in soccer but also for the broader pursuit of gender equality within the global sporting arena. This examination will illuminate the path towards establishing parity between male and female athletes, forging a vision of inclusivity and equality within the realm of sports worldwide. The impact of the Civil Rights Movement on sports cannot be underestimated, sparking important conversations and actions to address inherent inequalities. The disparity between women's and men's soccer in the United States is a stark example of this ongoing inequality. Despite the U.S. Women's National Soccer Team's remarkable global success, players still face significant disparities in pay, working conditions and resources compared to the men's team. Their pursuit of fair treatment and fair compensation is a powerful example of the obstacles female athletes continue to face. The deeply entrenched gender bias in sports, portraying football as a predominantly masculine domain, perpetuated this discrimination against female players. The patriarchal grip marginalized women in a sport they excelled in, robbing them of deserved rights and recognition. --- Efforts for In the 1970s, FIFA began recognizing and supporting women's football, but it was not until the early 1990s that a dedicated women's committee was established (Scelles, N., 2021). The inaugural Women's World Cup victory by the U.S. women's team in 1991 set the stage for eight subsequent tournaments, with the U.S. team clinching four titles. Despite this success, pay inequality plagued the U.S. women's soccer team, consistently earning lower wages compared to their male counterparts. From 2008 to 2015, data from U.S. Soccer revealed a staggering $200,000 disparity between the highest-paid men and women during the same period. This discrepancy extended beyond salaries, affecting bonuses, training, medical care, and venue resources for female players. The substantial investment required with minimal returns made it arduous for many female athletes to sustain their careers and aspirations. Inadequate training environments and equipment, coupled with limited medical resources, increased their vulnerability to various health issues. In a significant stride towards rectifying this inequality, in March 2019, 28 women's soccer players filed a gender discrimination lawsuit against U.S. Soccer. They highlighted that between 2013 and 2016, the women's national team earned less than one-third of what the men's team received. While arguments had been made justifying financial disparities based on audience size and revenue generation, these arguments were debunked when the 2015 World Cup-winning U.S. women's team generated over $23 million in revenue compared to the U.S. men's team's $4.6 million profit (Andrew Das, 2016). After relentless advocacy and struggle, a historic collective labor agreement was finally signed by the U.S. Women's National Team, the U.S. Men's National Team, and the United States Soccer Federation (USSF). This groundbreaking agreement ensures equal pay for equal work, marking a significant milestone in the battle for gender pay parity. Notably, the deal mandates that 90 percent of World Cup prize money be evenly split between players from the 2022 Men's World Cup roster and the 2023 Women's World Cup roster, a groundbreaking move unique to the United States among top soccer nations (Andrew Das, 2022). This agreement not only values labor equally but also champions fairness, prevents discrimination, and fosters the advancement of women's soccer in the United States. --- The Contributions of Various Feminist Perspectives to Advancing Women's Rights in the Realm of Football Feminism plays a crucial role in championing parity in remuneration for comparable work within women's football. Liberal feminism underscores the values of parity, autonomy, and human rights (Wade & Ferree, 2019). Specifically, within the realm of women's football, this ideology advocates for equitability in opportunities, resources, and treatment for female athletes. Notably, the enactment of Title IX in 1972 stands as a significant milestone, ensuring parity in educational programs and sports for both genders (Famuditimi-Bello, K., 2020). This pivotal legislation mandated educational institutions to furnish equivalent resources and scholarships to female athletes, sparking a transformation in women's involvement in sports, particularly football. Consequently, it paved the way for amplified investments in women's sports, fostering broader participation and the benefits stemming from endeavors like women's football. On the other hand, radical feminism identifies patriarchy as the root cause of women's subjugation (Wade & Ferree, 2019). Within sports like football, this patriarchal influence manifests through gender-based segregation, historically relegating women's football to a marginalized position with far less support compared to men's football. The perception of women's football as inferior mirrors entrenched patriarchal beliefs predominated by men. Financial disparities and inadequate sponsorship further underscore this disparity. Additionally, societal expectations concerning childbirth and familial duties can disrupt the careers of female athletes, contributing to their economic vulnerability. The goal of radical feminism is to deconstruct these patriarchal frameworks that impede the progression of women in sports. Meanwhile, feminist postmodernism challenges fixed classifications and underscores the diverse experiences of women predicated on factors such as race, class, nationality, ability, and age. This perspective underscores that women engaged in football, or any other domain, hail from varied cultural backgrounds and possess distinct aspirations (Wade & Ferree, 2019). It scrutinizes the concept of a universal experience, proposing that localized and individualized theories offer a more accurate comprehension of the intricacies surrounding women's involvement in football. It prompts the consideration of the distinctive hurdles encountered by women from diverse backgrounds in the realm of sports. Collectively, these feminist viewpoints call for persistent advocacy for equitable resources, the disruption of patriarchal norms and structures, recognition of the diverse experiences of female athletes, and the advocacy of inclusive policies and support mechanisms tailored to address the unique challenges faced by women in football. --- Comparative Analysis of Chinese and American Women's Football Players --- Comparison of the Situation Between the US Women's Football Team and the Chinese Women's Football Team In the field of intersectional studies involving sports, gender, and ethnicity, the U.S. women's soccer team and the Chinese women's soccer team encounter distinct circumstances. Within the realm of sports, examining the intersection of gender and ethnicity reveals contrasting scenarios for the U.S. women's soccer team and the Chinese women's soccer team. China's women's soccer team boasts nine Asian Cup victories and holds the 19th spot in FIFA World Rankings, a stark contrast to the 74th position held by China's men's soccer team. Despite these achievements, there exists a considerable disparity in their respective salaries (Jenny Feng, 2022). A 2018 report from the Chinese Football Association indicates that fewer than 20% of players in the top two women's leagues in China, the Chinese Women's Super League and League One, earn over 10,000 yuan ($1,572) monthly. Notably, the highest-earning female players in these leagues receive merely about 20% of the wages earned by their male counterparts (Jenny Feng, 2022). In comparison, the women's soccer league within U.S. college campuses and the robust soccer structure established through socialized semi-professional and professional leagues reflect a different scenario (Nesseler, C., Gomez-Gonzalez, C., & Gasparetto, T., 2021). While the Chinese women's soccer team has showcased commendable performances, the league struggles with poor marketing and negligible attention, hampering its commercial development and corporate support. Consequently, this situation results in low incomes for Chinese women's soccer players and inadequate post-retirement security. Moreover, attracting top-tier talents and ensuring equitable treatment proves challenging when juxtaposed with the income of Chinese men's soccer players. This disparity in systems and circumstances complicates the pursuit of equal pay within the Chinese soccer landscape. Nevertheless, with the escalating success of women's soccer garnering increased attention, there have been notable advancements. Earlier this year, the Chinese payment platform Alipay, which pledged a ten-year, 1 billion yuan ($157.2 million) commitment in 2019 to bolster the growth of women's soccer in China, announced a reward of 13 million yuan ($2.04 million) for the players following their triumphs. This development potentially opens avenues for augmenting salaries in Chinese women's soccer (Jenny Feng, 2022). --- Employing Feminist Theory to Assess Disparities in the Treatment of Female Athletes in China and the United States The disparities between China and the United States in the treatment of female soccer players can be better comprehended through the lens of postmodern feminism. This approach highlights the intricate nature of gender treatment and acknowledges the multifaceted influences of social, cultural, and economic factors. Postmodern feminists scrutinize power dynamics within the football industry, emphasizing the impact of media portrayal, commercialization, and global trends. They question the constructed narratives around women's soccer in both countries, examining how these narratives mold public opinions and affect the treatment of female players. In China, conventional gender roles and societal expectations significantly shape perceptions of women's roles in sports. Despite the achievements of China's women's soccer team, the wider cultural narrative might not accord the same recognition and status to female athletes as seen in the United States. These societal attitudes might result in reduced attention and investment in female players, ultimately leading to lower wages. Radical feminists in each country focus on distinct aspects due to their unique social environments. In China, they may stress how ingrained gender roles and societal expectations hamper recognition and support for female athletes. Conversely, in the United States, their emphasis lies in critiquing the historical marginalization of women in sports and its ongoing impact on female football players. Furthermore, substantial disparities exist in the infrastructure and developmental trajectories of women's soccer between the two nations. The U.S. college system serves as a robust platform for nurturing talent and enhancing visibility for female athletes (Nesseler, C., Gomez-Gonzalez, C., & Gasparetto, T., 2021). In contrast, the Chinese Women's Football League encounters hurdles in marketing, visibility, and corporate backing, directly affecting players' earnings and career prospects. Investment levels and institutional support for women's soccer vary significantly. The United States boasts a more established structure with well-funded professional leagues and a track record of investing in women's sports. Although China has shown considerable progress recently, it may still lag behind in terms of financial backing and sustained support for women's soccer. Liberal feminists in both countries advocate for more equitable policies within football associations and sports bodies to address inequality. However, changes in policies and societal perceptions evolve slowly and might not immediately translate into fair treatment and wages for female players. --- Suggestion Female athletes wield significant power in catalyzing substantial changes within the sports industry, extending beyond just football and toward achieving parity in compensation. Firstly, establishing transparency within salary structures and bonuses is pivotal. Athletes must champion systems that are clear and equitable, linking pay directly to performance and acknowledging achievements comparable to their male counterparts. Simultaneously, ensuring a balance in sponsorship and endorsement opportunities is crucial. Encouraging brands to invest equally in female athletes, aligning with their skill sets and market appeal, can help bridge the existing financial support gap. Addressing disparities in media coverage stands as another essential aspect. By advocating for increased visibility of women's sports and equal representation across media platforms, female athletes can enhance their prominence and acknowledgment. This endeavor is instrumental in reshaping public perceptions and garnering backing for equitable compensation. Furthermore, active involvement in collective bargaining agreements holds significance. By pushing for clauses in contracts that guarantee parity in pay for equal work, female athletes can secure tangible commitments from sports organizations. Beyond individual endeavors, advocating for legislative and governance reforms proves crucial. Pursuing reforms mandating fair pay practices for female athletes within the sports industry is pivotal for fostering enduring gender equality. These collective strategies empower female athletes to challenge existing norms and disparities, fostering an environment where equal compensation becomes an intrinsic facet of their careers. --- Conclusion In conclusion, this paper delves into the wide gaps in treatment, wages, and resources between male and female soccer players, notably in the U.S., emphasizing the ongoing battle against gender bias. Feminist viewpoints have significantly shaped discussions on gender parity in sports, advocating for equity, challenging entrenched norms, and acknowledging diverse experiences in soccer. These perspectives stress the importance of continuous advocacy, inclusive policies, and targeted support systems to address the specific hurdles faced by women in sports. Examining the status of female soccer players in China and the U.S. underscores the complexity of disparities shaped by intersecting factors. While both countries encounter obstacles in achieving fair pay, the nuances differ due to distinct societal, cultural, and structural elements. Suggestions to remedy these gaps spotlight the pivotal role of female athletes in instigating change. Ultimately, the quest for equal pay in professional women's soccer mirrors the broader struggle for gender equality in sports. It demands collective action, policy shifts, and a change in societal perspectives to establish an environment where female athletes receive fair treatment, acknowledgment, and rightful compensation for their contributions to the sporting realm. However, this study focus primarily revolves around the disparities between male and female soccer players in the U.S. While this provides valuable insight, it might overlook nuanced challenges faced by female athletes in other regions or sports, limiting the breadth of the analysis. A potential avenue for future study could involve a more comprehensive global examination of gender disparities across various sports, taking into account regional, cultural, and economic influences.

This article delves into the multifaceted realm of gender disparities in sports, focusing on the enduring challenges faced by female soccer players, notably within the United States. Highlighting the intertwined complexities of gender biases, treatment discrepancies, and resource inequalities, it underlines the pervasive gender gap prevalent in various facets of sports, encompassing pay differentials, sponsorship limitations, and unequal access to training facilities. This research aims to explore the multifaceted demand for equal pay in professional women's soccer through sociological and feminist lenses. It scrutinizes implications not only for female soccer players but also for broader gender equality pursuits in global sports. This investigation endeavors to illuminate pathways toward achieving parity between male and female athletes, envisioning inclusivity and equity in the sporting world. The study concludes with a series of recommendations to achieve equal pay for women in sport, emphasizing the importance of transparency, sponsorship, recognition, media representation and advocacy in collective bargaining agreements and governance. This research contributes to the ongoing debate on gender equality in sport and provides insights into potential solutions to the pay gap.

INTRODUCTION In recent years, the number of women in the labor market has steadily increased (e.g., for Belgium1, 62.6% at the beginning of 2017 versus 67% at the beginning of 2019). This influx of women marks a change in mentality, giving rise to legal provisions and policies intended to guarantee them equal opportunities with men (Institut pour l'égalité des femmes et des hommes, 2013). beliefs and stereotypes describing and prescribing social roles for male and female managers, in addition to interpersonal and situational issues, also increase women managers' perception of being treated differently and therefore their perception of a glass ceiling. Moreover, while Elacqua et al. (2009) focused their study on the causes of the glass ceiling, we want to extend the understanding of the phenomenon by also investigating its consequences for women managers' organizational attitudes and well-being at work. Although some studies have demonstrated the deleterious effects of the glass ceiling on different concepts (e.g., intention to quit, Stewart et al., 2011;lower self-esteem, Tran, 2014; and a reduction in capability to build networks and support structures for one's own career, Freeman, 1990), there remain some gaps in this domain, notably in terms of work-family interface. Indeed, to the best of our knowledge, no previous study has investigated the perception of work-tofamily conflict (WFC) in association with the model proposed by Elacqua et al. Therefore, we investigated the consequences of women managers' perception of a glass ceiling in terms of workfamily interface and well-being (i.e., job strain, job engagement, job satisfaction, and intention to quit). By investigating these issues, our study allows us to respond to specific recommendations and also fill some gaps in the literature on the glass ceiling. Therefore, through this research, we will try to better understand this phenomenon by considering both its antecedents and its possible consequences for well-being at work. --- LITERATURE REVIEW The Glass Ceiling Glass ceiling refers to the fact that a qualified person whishing to advance within the hierarchy of his/her organization is stopped at a lower level due to a discrimination most often based on sexism or racism. The glass ceiling refers thus to vertical discrimination most frequently against women in companies. The difficulty inherent in this theme is the diversity of definitions and approaches describing the glass ceiling. There are also no objective and easily observable criteria that would make it possible to establish with certainty the real existence of a glass ceiling in a company. Nevertheless, based on several studies, the glass ceiling can be defined as subtle but persistent barriers/obstacles, underpinned by discriminatory, conscious and unconscious practices, and attitudes that hinder access to top/senior management positions for qualified women (e.g., Jackson and O'Callaghan, 2009;Bendl and Schmidt, 2010;Zeng, 2011). Glass ceiling refers thus to discriminatory barriers that prevent women from rising to positions of power or responsibility and advancing to higher positions within an organization simply because they are women (Li and Leung, 2001). This phenomenon of the glass ceiling is based on several assumptions. Indeed, compared with other forms of discrimination and inequality, the glass ceiling is a particular and specific form of inequality due to several criteria (Cotter et al., 2001). First, the essence of the glass ceiling is the discrimination against women in management. The glass ceiling would therefore affect women despite their level of education, experience and skills. The glass ceiling is observed diachronically, i.e., it is advancements in women's careers, promotions to managerial positions, that need to be taken into account, rather than the number of women in those positions at a specific time (Cotter et al., 2001). The glass ceiling also refers to the growing inequalities between men and women as they evolve in their professional careers within the company. Second, this bias is difficult to observe given that current equal opportunity policies prohibit open discrimination against these populations. Moreover, this includes norms/stereotypes revealed through practices, actions, facts, procedures, or attitudes that are frequently not directly observable. Finally, the existence of invisible barriers hinders hierarchical ascension (Cotter et al., 2001). Indeed, this definition focuses on top/senior management, with the assumption that the glass ceiling occurs more frequently at this level than at middle and lower grades. These obstacles to advancement are more present as one approaches the top of the hierarchy (Cotter et al., 2001). Indeed, most researchers admit that the singularity of this phenomenon lies in its predominance at higher levels of management (Baxter and Wright, 2000;Albrecht et al., 2003;Elliott and Smith, 2004;Prokos and Padavic, 2005;Zeng, 2011;Dambrin and Lambert, 2012;Lupu, 2012). The Antecedent of the Glass Ceiling: The Model of Elacqua et al. (2009) A model for understanding the phenomenon of the glass ceiling is that developed by Elacqua et al. Through their study, these authors investigated why women managers rarely reach the highest levels of their organization. Among 685 managers at a large Midwestern insurance company, they proposed a model in which beliefs about organizational variables of an interpersonal and situational nature were positively related to the perceptions of differential treatment between men and women, which, in turn, was positively related to the perception of a glass ceiling. Therefore, these authors suggest that perceptions of differential treatment mediate the relationships between both these organizational factors (i.e., interpersonal and situational issues) and perceptions of a glass ceiling. --- Interpersonal Factors According to these authors, some interpersonal relationships can influence how women and their male counterparts are treated differently within the company. In their study, Elacqua et al. examined in particular: (a) mentoring, (b) the existence of an informal network of senior managers, and (c) the friendly relationships with company decision-makers, as these concepts are all related to career advancement. Indeed, regarding the first of these three aspects, research has highlighted that a lack of high-level organizational mentors is deleterious to women's career progression, especially since mentoring is an important source of information (e.g., Ibarra et al., 2010). In addition, mentor-supported individuals perceived more opportunities for promotion (Allen et al., 2004), and the likelihood that they will actually be promoted is greater (Dreher and Ash, 1990;Allen et al., 2004). These people would also be more satisfied with their careers and work (Kammeyer-Mueller and Judge, 2008). Thus, mentoring would be a significant enabler in women's career advancement and even business development (Elkin, 2006). As mentioned by Elacqua et al. (2009, p. 286), "employees whose supervisors act as their mentors are more likely to feel that they are not excluded from important information and opportunities, and, therefore, assume that is true for others as well." As a result, they perceive less differential treatment among employees in their company (Raabe and Beehr, 2003). The second aspect of interpersonal factors investigated by the authors is the existence of an informal social network of senior men within the firm. Networks refer to the development and use of career-relevant contacts in which members exchange valuable strategic information (concerning new positions, ongoing projects, and managerial decisions, etc.), contacts and recommendations (Burke, 1984). Several studies have shown that women are being assigned positions with lower visibility, limiting their opportunities to connect with high-ranking individuals and develop social networks (e.g., Ragins et al., 1998). In companies where there is an informal social network of senior men, women managers may not be treated in the same way as men because of lack of visibility (Elacqua et al., 2009). Limited access to such a network would reduce the chances of promotion and therefore, lead to a perception of a glass ceiling (Brass, 1985). The third aspect of interpersonal factors refers to the friendly relationships with the company's decision-makers. Individuals often like to form friendships with people of the same sex who have had similar experiences. Women managers might then face an additional difficulty: the "queen bee syndrome" (Keeton, 1996). This syndrome describes the fact that some women managers who have managed to get to the top feel that they have had to work hard to get where they are. They think that other women should work as hard to succeed. Indeed, according to Cech and Blair-Loy (2010), women breaking the glass ceiling tend to attribute their success to merit rather than to overcoming the structural barriers that senior women are able to do something about. In their study of a large U.S. sample of female accountants, Cohen et al. (2020) highlighted the importance of these interpersonal factors in the occurrence of the glass ceiling. Indeed, they found that a lack of mentoring opportunities, networking opportunities, social support from male organizational leaders and high-profile job assignments have a strong positive influence on female accounting professionals' glass ceiling perceptions. --- Situational Factors Regarding these second kind of factors, Elacqua et al. considered two aspects that would influence the perception of a glass ceiling through the perception of differential treatment. The first aspect is the existence of objective criteria for procedures established within the company (e.g., hiring, and promotion). Lyness and Heilman (2006) have shown that the promotion criteria are more severe and more related to job performance for female line managers compared to their male colleagues. Women are particularly sensitive to and supportive of being promoted on the basis of their performance (Beehr et al., 2004). Employees who believe that their company uses objective criteria related to skills and performance for the promotion process rate the process as fairer and do not perceive differential treatment among employees (Beehr et al., 2004). In their study, Cohen et al. (2020, p. 22) also found that "female accounting professionals who believe that female employees are not treated the same as male accounting professionals with regard to performance evaluations will be more likely to report a glass ceiling within their organizations." The second aspect refers to "the number of women managers who have been in managerial positions long enough to be considered serious candidates for advancement to higher levels" (Elacqua et al., 2009, p. 287). According to Elacqua et al., one would think that a woman manager is a serious candidate for a promotion if she already occupies a managerial position and participates in the development activities of the company. However, as mentioned by Elacqua et al. (2009, p. 287), "women experience these two situations less frequently than men." Therefore, there will not be as many women in position for promotion to higher-level management positions (Elacqua et al., 2009). "If managers believe that this is happening in their organization, they are likely to perceive differential treatment of sexes and consequently the existence of a glass ceiling" (Elacqua et al., 2009, p. 287). --- Perception of Differential Treatment of Women Men and women are often treated differently in the world of work (Blau and Kahn, 2007;Kochan, 2007). These disparities arise when "personnel decisions are based on gender, an ascribed characteristic, rather than on an individual's qualifications or job performance (Gutek et al., 1996;Ngo et al., 2002)" (cited by Foley et al., 2005, p. 423). According to Elacqua et al., perceiving differences in a company's treatment of women would lead workers to believe that there is a glass ceiling in the company. While gender differences may be relatively small in terms of, for example, promotion at each level of the hierarchy, they add up to form a gap between the number of men and women occupying the highest positions in the company (Agars, 2004). --- Organizational Culture in Relation to Gender In their study, Elacqua et al. suggested further investigation of other factors that may influence the perception of differential treatment of men and women and of a glass ceiling. Indeed, according to these authors, "glass ceilings and perceptions of it are especially likely to be found in cultures that encourage differential views of and treatment of men and women" (Elacqua et al., 2009, p. 293). Another factor influencing the emergence of the glass ceiling through the perception of differential treatment could be the culture implemented within organizations, in other words the organizational culture, referring to organizationally shared values and beliefs that reflect workers' judgments of how things should be and really are (Lord and Maher, 1991). The literature has shown that two main aspects of organizational culture in relation to gender are important barriers to women's progress (e.g., Eagly and Johnson, 1990;Valian, 1998), i.e., the "male-oriented" organizational culture and beliefs about the incompatibility of the roles of mother, wife and manager. Both these aspects refer to the set of beliefs and stereotypes describing and prescribing social roles for male and female managers that are conveyed by the organization and some of its members. Senior managers, overwhelmingly men, define a "gendered" culture that excludes and marginalizes women. This culture consists of a series of norms and organizational practices that promote and define values, stereotypes, behaviors, and a vision of management and leadership that are "masculine" (van Vianen and Fisher, 2002;Broadbridge and Hearn, 2008;Koenig et al., 2011). In line with the social role theory (Eagly, 1987), the image of the manager is often associated with that of a man with so-called "masculine" qualities, such as authority, independence, competitiveness, and aggressiveness (e.g., Weyer, 2007). Therefore, women, who have been associated for centuries with diametrically opposing qualities (e.g., collaboration, listening, sensitivity, and sympathy), would be less committed to their careers and unable to manage (e.g., Weyer, 2007). These gender stereotypes about women have an adverse impact on their assessments and judgments (Lyness and Thompson, 1997;van Vianen and Fisher, 2002). As a result, early in their careers, women managers are assigned different responsibilities from those of men. Women managers are then faced with a twofold constraint: (a) if they do not conform to male norms, they risk being judged and evaluated negatively; and (b) if they adopt a "masculine" attitude, they get hurt by their colleagues (Oakley, 2000;Eagly and Karau, 2002;Mavin, 2008;Kumra and Vinnicombe, 2010). Other beliefs about women managers can lead their superiors to fail to consider them as serious candidates for top management positions. Among these beliefs are the incompatibility of the roles of mother, wife and manager; being a manager requiring too much investment, flexibility and travel. In line with this idea, Hoobler et al. (2008) have shown that managers think that work-life conflict is greater for women than for their male counterparts. However, this belief has implications since it will reduce the organization's perception of women's adequacy in the workplace, decreasing the probability of female promotion (Rudman and Phelan, 2008). Therefore, based on the above, we postulate that perceptions of interpersonal issues, situational issues and the organizational gender culture will influence perceptions of differential treatment, which in turn, will affect the perception of a glass ceiling: Hypothesis 1a: Perceptions of differential treatment will mediate the relationships between interpersonal issues and the perception of a glass ceiling. Hypothesis 1b: Perceptions of differential treatment will mediate the relationships between situational issues and the perception of a glass ceiling. Hypothesis 1c: Perceptions of differential treatment will mediate the relationships between the organizational gender culture and the perception of a glass ceiling. --- Consequences of the Glass Ceiling Although there are many articles on the glass ceiling, very few, to our knowledge, and investigate its effects on workers. That is the reason why one of the objectives of our study is to extend the understanding of this phenomenon by also investigating its consequences for organizational attitudes and women managers' well-being. Better assessing the consequences of the glass ceiling seems essential to sensitize companies to this issue. Through our study, we want to investigate the links between the perception of the glass ceiling and five concepts, i.e., WFC, job strain, intention to quit, job engagement, and job satisfaction. Specifically, we intend to investigate the mediating role of WFC in the relationship between the glass ceiling and the other four outcomes. We have included job strain and intention to quit in this study because they are the strongest outcomes related to WFC among work-related outcomes (Allen et al., 2000;Amstad et al., 2011), and job satisfaction because it is the workrelated outcome that has attracted the most research attention (Allen et al., 2000;Amstad et al., 2011). Considering that little is known about the effects of WFC on positive indicators of wellbeing (Peeters et al., 2013), we have considered job engagement in relation with WFC because this concept is a well-known indicator of well-being (Bakker andDemerouti, 2008, 2017). --- Relationship Between the Glass Ceiling and WFC Individuals must daily take on several social roles (i.e., parent, spouse, and employee, etc.). Contradictory demands can arise from these multiple social roles in which individuals have to perform. The force to comply with these contradictory demands can produce incompatibilities between professional and family roles (Duxbury and Higgins, 1991). Indeed, given that individuals have limited resources notably in terms of time and energy, they can not satisfy all of these demands (scarcity hypothesis, Sieber, 1974). Therefore, when individuals have to compose with too many demands arising from work, they may experience WFC defined as "a form of inter-role conflict in which the general demands of time devoted to, and strain created by the job interfere with performing family-related responsibilities" (Netemeyer et al., 1996, p. 401). To the best of our knowledge, no previous study has investigated the link between the glass ceiling and WFC. However, we can argue that the glass ceiling increases the perception of WFC. One theory for postulating this link is the conservation of resources theory (COR theory, Hobfoll, 1989Hobfoll,, 2001Hobfoll,, 2011)). According to this theory, individuals and groups are threatened by the potential or actual loss of that which they value highly, namely resources. Therefore, people are motivated to obtain, acquire, retain, preserve, protect, foster, and expand valued resources for anticipated future needs (Hobfoll, 1989). Indeed, as mentioned by Hobfoll (2011, p. 117), people "employ key resources in order to conduct the regulation of the self, their operation of social relations, and how they organize, behave, and fit in to the greater context of organizations and culture itself." Within this theory, resources are crucial and refer to "those objects, personal characteristics, conditions, or energies that are valued by the individual or that serve as a means for attainment of these objects, personal characteristics, conditions, or energies" (Hobfoll, 1989, p. 516). We can reasonably think that, in perceiving a glass ceiling, women managers have access to fewer resources at work (e.g., little access to information, advice, social support, supervisory coaching, opportunity to develop knowledge, performance feedback, and promotion). The glass ceiling also undermines personal characteristics (e.g., self-esteem, self-efficacy, and optimism) and energies (e.g., knowledge, and money). Indeed, women facing such a form of discrimination are disadvantaged with regard to choice of job, salary, and prestige (Baxter and Wright, 2000). Perceptions of a glass ceiling also inhibit women from seeking and obtaining promotions (Powell and Butterfield, 1994), and reduce their self-esteem (Tran, 2014) and their capabilities to build networks and support structures for their own careers (Freeman, 1990). As mentioned by Downes et al. (2014, p. 133), "women also internalize negative evaluations and stereotypes by those in the majority to the point where they limit themselves and turn down opportunities for advancement due to the fear that they will not succeed (Ilgen and Youtz, 1986)." It is also possible that women managers, in order to fight or override barriers/obstacles related to the glass ceiling, invest resources to try to progress in the hierarchy but that this investment is not effective. Indeed, using resources to cope with a situation of resource loss is also stressful because this may deplete an individual's stock of resources (Hobfoll, 1989). As mentioned by Hobfoll, in situations where "resources expended in coping outstrip the resultant benefits, the outcome of coping is likely to be negative" (Hobfoll, 1989, p. 518). As a result, when there is a loss of resources and when no action is taken or when resource investment is unsuccessful, a spiral of loss of resources appears in which more and more resources are lost. In other words, initial loss of resources produces further loss (Hobfoll, 2001). This spiral contributes to poor psychological and/or physical health. These negative consequences (e.g., negative emotions, impaired psychological well-being, and ultimately impaired mental and physical health) spill over and negatively affect individuals' functioning at home (Edwards and Rothbard, 2000), increasing the perception of WFC. --- Relationship Between WFC and Job Strain The imbalance in professional and family roles is an important stressor, influencing work, and private domains and affecting health and well-being in general (Frone, 2003). According to the COR theory, stress is defined as "a reaction to the environment in which there is (a) the threat of a net loss of resources, (b) the net loss of resources, or (c) a lack of resource gain following the investment of resources. Both perceived and actual loss or lack of gain are envisaged as sufficient for producing stress" (Hobfoll, 1989, p. 516). Losing resources, or the threat of such a loss, may cause stress and strain. Applied to work-family conflict, Grandey and Cropanzano (1999, p. 352) mentioned that "inter-role conflict leads to stress because resources are lost in the process of juggling both work and family role." Supporting this theoretical view, considerable empirical research has shown that the WFC is positively related to job strain (e.g., meta-analyses of Allen et al., 2000 andAmstad et al., 2011). For example, in their study on firefighters in Taiwan, Wong et al. (2014) found a positive relationship between WFC and job strain. In studies of employees of a Belgian hospital (Babic et al., 2015) and of employees of a Belgian company (Babic et al., 2019), WFC was also found to be positively related to job strain. --- Relationship Between WFC and Intention to Quit According to the COR theory, "because resource loss is stressful and because people must invest resources to offset further resource loss, once initial losses occur, people become increasingly vulnerable to ongoing loss" (Hobfoll, 2001, p. 355). Moreover, as mentioned by Vinokur-Kaplan (2009, p. 231), "because people have fewer resources as they lose resources, they are decreasingly capable of withstanding further threats to resource loss." This situation ultimately results in a loss spiral in which more and more resources are lost. Facing such situations of resources loss, in order to protect the remaining resources, people perceiving WFC want to flee this situation (Hobfoll, 1989) and leave their organization (e.g., Greenhaus et al., 1997). Empirical research has supported this theoretical view. The metaanalyses conducted by Allen et al. (2000) and Amstad et al. (2011) showed that intention to leave the organization significantly related to WFC. Through a study on US workers, Anderson et al. (2002) found that perceiving WFC increased workers' turnover intention. Hammer et al. (2009) also found a positive relationship between these constructs. On 197 Maori employees working in 13 New Zealand organizations in diverse geographical locations, Haar et al. (2012) also found that WFC was significantly related to turnover intentions. Through studies of 509 employees of a Belgian hospital, Babic et al. (2015) found a positive relationship between WFC and intention to quit. --- Relationship Between WFC and Job Engagement As previously mentioned, workers losing resources due to WFC (Hobfoll, 1989) want to protect their remaining resources (Hobfoll, 2002). Another way for workers to protect their resources is to reduce their level of engagement in their work. Moreover, in line with the source attribution perspective of WFC (Shockley and Singla, 2011), workers perceiving that their work interferes negatively with their family sphere blame their professional domain given that it is the source of the perceived conflict. When facing such situations of resource loss or dissatisfaction, workers react or cope by adjusting their attitudes (e.g., reducing their engagement in their work). Supporting these theoretical views, empirical research has shown that the WFC employees have to deal with can affect their job engagement (e.g., Wilczek-Ruzyczka et al., 2012;Opie and Henn, 2013;Fiksenbaum, 2014). Through their study on 267 South Africans working, Opie and Henn (2013) found that employees experiencing WFC engaged less in their work. In a sample composed of 978 workers from a Belgian Federal Public Service, Babic et al. (2017) found that WFC had a negative impact on work engagement. Through their study of 98 nurses from southern Poland, Wilczek-Ruzyczka et al. (2012) found that, compared to nurses perceiving lower WFC, those perceiving higher WFC engaged less in their work (lower level of vigor and dedication). In a sample of employees of a Belgian Federal Public Service, Babic et al. (2020) also found that WFC was negatively associated with vigor. --- Relationship Between WFC and Job Satisfaction According to the source attribution approach (Shockley and Singla, 2011), in perceiving WFC, individuals are dissatisfied with the work domain because they psychologically blame the domain that is the source of conflict (i.e., work). Therefore, when work causes difficulties in fulfilling family responsibilities, individuals obtain less satisfaction from work (e.g., Höge, 2009). Empirical research has supported this theoretical view. Through their study on US workers, Anderson et al. (2002) found that WFC was negatively related to job satisfaction. Through their study on 360 employees from 12 stores of a grocery store chain in the Midwestern United States, Hammer et al. (2009) found a similar result. Carlson et al. (2010) also found that perceiving WFC leads workers to have lower level of job satisfaction. In a study of 509 employees of a Belgian hospital, Babic et al. (2015) found a negative relationship between WFC and job satisfaction. This negative relationship was also highlighted by several metaanalyses (e.g., Kossek and Ozeki, 1998;Allen et al., 2000;Amstad et al., 2011;Shockley and Singla, 2011). Therefore, based on the above, we postulate that the perception of a glass ceiling will increase the perception of WFC, which, in turn, will increase job strain and intention to quit, and decrease job engagement and job satisfaction: Hypothesis 2a: WFC will mediate the relationships between the glass ceiling and job strain. Hypothesis 2b: WFC will mediate the relationships between the glass ceiling and intention to quit. Hypothesis 2c: WFC will mediate the relationships between the glass ceiling and job engagement. Hypothesis 2d: WFC will mediate the relationships between the glass ceiling and job satisfaction. --- MATERIALS AND METHODS --- Sample The organization in which our study was conducted is a global healthcare company. In terms of culture and value, this company is careful to treat its employees with respect and in an equal manner. This company aims to create an inclusive, engaging/stimulating work environment that encourages employees to get involved to help achieve company goals. To do so, the company offers all of its employees a range of learning opportunities, training and professional development programs. On-the-job learning is encouraged through projectbased work, while providing support through more traditional training programs, mentoring and coaching. In terms of learning opportunities, the organization, for example, sets up feedback from the line manager and peers. Regular meetings with managers are also set up in order to plan development objectives. These meetings focus on the development aspirations, skills, experience, and needs of workers. This company is attentive to the well-being of its workers and to their possibility to balance their professional and private life. For example, the organization offers benefits such as access to health care and wellness programs or joining a retirement plan. The company invests heavily in the development of its employees at all stages of their careers, from young graduates to senior managers. In terms of inclusion and diversity, the company designs workshops to sensitize and empower employees to create a workplace free from discrimination. Different campaigns have been put in place to recognize and manage unconscious prejudices with regard to gender, age, sexual orientation, gender identity, ethnicity, or even religion. In order to test our hypotheses, a self-reported questionnaire was administered to women occupying managerial positions within this global healthcare company. One thousand two hundred thirty women managers were invited to participate in this research. We received 320 questionnaires in return, corresponding to a response rate of about 26%. Twenty-six percent of respondents were between 36 and 40 years old. Forty percent had been employed by their company for between 2 and 5 years. Eighty-one percent were either married or living with a partner. Seventy percent had one child or more at home. Fifty-six percent led a team. Sixty-six percent occupied a first line management position, thirty-one percent occupied a middle management position, and only three percent were part of senior management. Using the full partial covariate effects (Little, 2013), three socio-demographic variables (age, grade, and presence of children at home) were significantly related with the constructs of our model. Consequently, we included these three sociodemographic variables as covariates to control for their effects in our analyses. --- Procedure This study was carried out in accordance with the recommendations of the APA Ethical Principles of Psychologists and Code of Conduct with written informed consent from all subjects. All subjects participated in a free and informed manner and gave written informed consent in accordance with the Declaration of Helsinki. Data were collected through an online survey. Participants received an e-mail explaining the purpose of the study and emphasizing the confidentiality of the responses (anonymous participation). The informed consent of each participant was obtained. People were given 1 month to complete the survey that was anonymous and confidential. As participants speak French and English, questionnaires were written in these two languages. Lacking all the scales in these two languages, we translated them following a translation back-translation procedure (Brislin, 1980). For all scales, there was no major discrepancy between the original and translated versions, so the translation process was considered appropriate. --- Measures The glass ceiling and its antecedents were assessed using the questionnaires developed by Elacqua et al. (2009). To develop their questionnaire, Elacqua and her colleagues had established a task force composed of employees of the company in which they carried out their study and an industrial/organizational psychologist with expertise in survey research. This task force developed a series of items relating to the different variables the authors wanted to measure. These items reflected the company's concerns about gender equality at work. However, these items were developed in reference to an American culture. Therefore, in order to have items more in line with European culture, we added to the items of Elacqua et al. some questions created in collaboration with the company in which this study was carried out (see Supplementary Appendix 1 for the overall questionnaire). All items (original and new) were higher than 0.50, which is the recommended cut-off score for factor loadings (Kline, 2011). All were measured using a 4-point Likert-type rating scale ranging from "strongly disagree" to "strongly agree." For all of the scales, a high score indicates discrimination against women managers. Interpersonal issues were assessed with the four original items developed by Elacqua et al. (e.g., "Having a personal friendship with the decision-makers determines whether an employee will be considered for a promotion at my company.") and with four new items (e.g., "Few women managers have access to strategic information coming from the senior management"). Cronbach's alpha was 0.75. Situational issues were assessed with the three original items developed by Elacqua et al. (e.g., "There are few women at the top rung of management at my company because they haven't been in the management ranks long enough") and with three new items (e.g., "In my company, promotion procedure is based on objective criteria"). Cronbach's alpha was 0.72. Organizational gender culture was evaluated with four new items (e.g., "My colleagues and superiors estimate that women managers are less mobile and flexible than their male counterparts"). Cronbach's alpha was 0.70. Differential treatment was evaluated with the five original items developed by Elacqua et al. (e.g., "At my company, there are differences in salaries and titles for men and women in the same positions that are not explained by differences in performance, education, experience") and with two new items (e.g., "In my company, performance criteria are different for men and women managers"). Cronbach's alpha was 0.71. Perceptions of a glass ceiling were evaluated with the three original items developed by Elacqua et al. (e.g., "Female managers/supervisors at my company generally progress to a certain level, then go no further") and with three new items (e.g., "In my company, with equal experience and expertise, men have access to higher positions in the hierarchy than women"). Cronbach's alpha was 0.84. We performed an exploratory factor analysis and found that both new items and those developed by Elacqua et al. loaded adequately on their respective factors. The full measurement model concerning these 31 items are presented in the Supplementary Appendix 2. Work-to-family conflict was assessed using the appropriate SWING subscale (Geurts et al., 2005). This nine-item subscale evaluates the negative impact of the professional situation on family life (e.g., "I'm irritable at home because my work is demanding"). People responded on a 4-point Likert-type scale (0: never to 3: always). Cronbach's alpha was 0.87. Job strain was measured with the Negative Occupational State Inventory subscale developed by Barbier et al. (2012). This scale is comprised of nine items (e.g., "My work stresses me"). People responded on a 4-point Likert-type scale (1: never to 4: always). Cronbach's alpha was 0.78. Job engagement was measured with the Positive Occupational State Inventory subscale developed by Barbie

Despite significant promotion of diversity in companies, as well as legislation for equal opportunities for women and men, it must be noted that women still remain largely in the minority in decision-making positions. This observation reflects the phenomenon of the glass ceiling that constitutes vertical discrimination within companies against women. Although the glass ceiling has generated research interest, some authors have pointed out that theoretical models have made little attempt to develop an understanding of this phenomenon and its implications. Therefore, our study aims to fill this gap and to better understand the phenomenon of the glass ceiling by considering both its antecedents and its possible consequences. More precisely, we extend the model developed by Elacqua et al. (2009), proposing a more comprehensive model including organizational gender culture as a third factor (in addition to situational and interpersonal issues) in the emergence of the glass ceiling through the perception of differential treatment. We also investigated the glass ceiling's consequences for organizational attitudes and well-being at work by considering work-to-family conflict (WFC) as a possible mediator. We surveyed 320 women in managerial positions in a Belgian organization. Our study highlights the importance of all three factors in the emergence of the perception of differential treatment and, ultimately, the perception that a glass ceiling exists. Moreover, our results show that WFC fully mediates the effects of the glass ceiling on job strain and job engagement, and partially mediates the effects of the glass ceiling on job satisfaction and intention to quit.

Cronbach's alpha was 0.84. We performed an exploratory factor analysis and found that both new items and those developed by Elacqua et al. loaded adequately on their respective factors. The full measurement model concerning these 31 items are presented in the Supplementary Appendix 2. Work-to-family conflict was assessed using the appropriate SWING subscale (Geurts et al., 2005). This nine-item subscale evaluates the negative impact of the professional situation on family life (e.g., "I'm irritable at home because my work is demanding"). People responded on a 4-point Likert-type scale (0: never to 3: always). Cronbach's alpha was 0.87. Job strain was measured with the Negative Occupational State Inventory subscale developed by Barbier et al. (2012). This scale is comprised of nine items (e.g., "My work stresses me"). People responded on a 4-point Likert-type scale (1: never to 4: always). Cronbach's alpha was 0.78. Job engagement was measured with the Positive Occupational State Inventory subscale developed by Barbier et al. (2012). This scale included eight items (e.g., "When I'm working I forget my tiredness"). People responded on a 4-point Likert-type scale (1: never to 4: always). Cronbach's alpha was 0.81. Job satisfaction was measured with the scale used by Eisenberger et al. (1997). This scale comprised four items (e.g., "All in all, I am very satisfied with my current job"). People responded on a 5-point Likert-type scale (1: strongly disagree to 5: strongly agree). Cronbach's alpha was 0.87. Intention to quit was estimated using Hom and Griffeth's (1991) scale which comprises three items (e.g., "I often think about quitting my organization"). People responded on a 5point Likert-type scale (1: strongly disagree to 5: strongly agree). Cronbach's alpha was 0.92. --- Data Analyses Structural equation modeling analyses were performed using Lisrel 8.80 (Jöreskog and Sörbom, 2006). Data were analyzed following a two-stage process suggested by Anderson and Gerbing (1988). First, we assessed the measurement model through a series of confirmatory factor analyses to evaluate the independence of constructs examined in our study. Second, we proceeded with the assessment of the hypothesized structural relationships among latent variables. For this second stage, in order to limit the number of parameters to be estimated, we reduced the number of items per factor by combining them to create a limited number of indicators per construct (Landis et al., 2000). Using the balancing technique, we generated aggregate indicators by averaging items with high and low loadings. We used this technique in order to have parcels equally balanced in terms of their discrimination (Little et al., 2002). We thus reduced the number of items to three for each of our constructs. We decided to use parceling strategy for different reasons. Considering psychometric characteristics, parcels have higher reliability, greater communality, a higher ratio of common to unique factor variance, a lower likelihood of distributional violations, and more-equal intervals compared to item-level analyses (Little et al., 2013). Moreover, considering model estimation and fit characteristics, parcels have fewer parameter estimates, lower indicator-to-sample size ratio, a lower likelihood of correlated residuals and dual factors loading and reduces sources of sampling errors (Little et al., 2013). --- RESULTS --- Discriminant Validity We tested the distinctiveness between the variables included in our study by comparing several nested models (Bentler and Bonnett, 1980). First, we examined the fit of our hypothesized ten-factor model (i.e., interpersonal issues, situational issues, organizational gender culture, differential treatment, the glass ceiling, WFC, job strain, job engagement, job satisfaction, and intention to quit). The results indicate that this hypothesized measurement model fit the data reasonably well [<unk> 2 (360) = 628.79, p <unk> 0.001, RMSEA = 0.05, NNFI = 0.97, and CFI = 0.98]. Moreover, loadings of all items, including new items created to assess the glass ceiling and its antecedents, were higher than 0.50, which is the recommended cut-off score for factor loadings (Kline, 2011). Starting from this ten-factor model, we tested a series of more constrained measurement models. In particular, we tested ten nine-factor models, one eight-factor model, one seven-factor model, and one six-factor model obtained by combining the glass ceiling and its antecedents. Chi-square difference tests were used to compare the fit of these nested models with that of the tenfactor model (Bentler and Bonnett, 1980). Results indicated that the ten-factor model was significantly superior to all alternative models. Consequently, we used this ten-factor model to test our hypotheses. Table 1 displays fit indices for some of these alternative models. --- Relationships Among Variables Means, standard deviations, Cronbach's alphas and correlations among variables are presented in Table 2. Internal consistency reliabilities ranged from 0.70 to 0.92. Based on the results of the confirmatory factor analyses, we examined the structural relationships among latent variables through a series of alternative models (Models 2 to 8). Table 3 presents the fit indices for these alternative models. Model 1 (i.e., the hypothesized model) fit the data reasonably well [<unk> 2 (468) = 912.36, p <unk> 0.001, RMSEA = 0.06, NNFI = 0.96, and CFI = 0.96]. To evaluate whether this model offered the best depiction of our data, we successively added paths from each of the three antecedents of differential treatment to the glass ceiling (Models 2 to 4). However, these three latter models did not have a significantly better fit than Model 1. Starting with Model 1, we successively added paths from the glass ceiling to job strain (Model 5) and to job engagement (Model 6) but these latter two models did not have a significantly better fit than Model 1 [ <unk> 2 (1) = 0.01, p > 0.05]. We also added a path from the glass ceiling to job satisfaction (Model 7). This Model 7 presented a fit that was superior to Model 1 [ <unk> 2 (1) = 4.11, p <unk> 0.05]. However, starting with this Model 7, we added a path from the glass ceiling to intention to quit (Model 8), and this latter model had a significantly better fit than Model 7 [ <unk> 2 (1) = 36.05, p <unk> 0.001]. Standardized parameter estimates for Model 8 are shown in Figure 2. For ease of presentation, we show the structural model rather than the full measurement model. Regarding our first hypothesis, situational issues, interpersonal issues and organizational gender culture were positively related to differential treatment. This latter concept was in turn positively associated with the glass ceiling. We used the bootstrapping technique to estimate indirect effects (Preacher and Hayes, 2008). As shown in Table 4, the indirect effects of situational issues, interpersonal issues and organizational gender culture on the glass ceiling through differential treatment were all significant. Thus, differential treatment totally mediates these relationships. These findings support Hypotheses 1a, 1b, and 1c. Concerning our second hypothesis, results showed that the glass ceiling was positively associated with WFC, which in turn, was negatively related to job engagement and job satisfaction and positively related to job strain and intention to quit. The glass ceiling was also directly and positively related to intention to quit and negatively related to job satisfaction. As shown in Table 4, the indirect effects of the glass ceiling on outcomes through WFC were significant for all four outcomes (i.e., job strain, job engagement, job satisfaction, and intention to quit). Thus, WFC totally mediates the effects of the glass ceiling on job strain and job engagement, and partially mediates the effects of the glass ceiling on job satisfaction and intention to quit. These results totally support Hypotheses 2a and 2c, and partially support Hypotheses 2b and 2d. Ancillary analyses were conducted in order to investigate whether differential treatment, the glass ceiling and WFC sequentially mediate the relationships. As indicated in Table 5, the indirect effects of the situational issues, interpersonal issues and organizational gender culture on the outcomes (i.e., job strain, job engagement, job satisfaction, and intention to quit) through differential treatment, the glass ceiling and WFC were all statistically different from zero. In sum, there is a triple mediation in sequence for all of these outcomes. --- DISCUSSION In this study, we try to better understand the phenomenon of the glass ceiling by considering both its antecedents and its possible consequences for well-being at work. More precisely, the present study has two objectives, allowing it to respond to a specific recommendation and to fill some gaps in the literature concerning this phenomenon of vertical discrimination against women in companies. Firstly, following the recommendation of Elacqua et al. (2009), we considered the organizational culture in relation to gender as a third factor (in addition to situational and interpersonal issues) in the emergence of the glass ceiling through the perception of differential treatment. (i.e., Hypothesis 1c). Secondly, we investigated the glass ceiling's consequences for women managers' organizational attitudes and well-being at work (i.e., Hypothesis 2). We did that to fill gaps, in particular with regard to relationships between the glass ceiling and workfamily interface (i.e., WFC), and also because, to the best of our knowledge, no previous study has investigated the perception of WFC in association with the model proposed by Elacqua et al. These hypotheses were tested by asking women occupying managerial positions in a global healthcare company to selfreport their perceptions. Through our first Hypothesis, we postulated that perceptions of differential treatment will mediate the relationships between interpersonal issues, situational issues, organizational gender culture and the perception of a glass ceiling. Results indicated that interpersonal, situational issues and an organizational gender culture were all positively related to differential treatment, which was in turn positively related to the glass ceiling. As a reminder, we found no direct path from these three antecedents of differential treatment to the glass ceiling, suggesting that differential treatment fully mediates these relationships. Results concerning interpersonal and situational issues are in line with previous findings (e.g., Elacqua et al., 2009;Cohen et al., 2020). On one hand, the lack of access to a mentor leads women to feel that they are excluded from communication and important opportunities for promotion (Allen et al., 2004). Moreover, as women are assigned positions with lower visibility, they have fewer opportunities to connect with high-ranking individuals and develop their social networks with senior managers (e.g., Ragins et al., 1998), reducing therefore their chances to exchange valuable strategic information, contacts and recommendations and, consequently, limiting their opportunities for promotion (Cohen et al., 2020). Lastly, by having fewer opportunities to develop friendly relationships with decision-makers, women managers have less chance of being promoted. These three interpersonal issues lead women mangers to perceive more differential treatment between them and their male colleagues, increasing therefore their perception of a glass ceiling within their organization. On the other hand, two situational issues also influence the perception of differential treatment and ultimately the perception of a glass ceiling. Indeed, women considering that their company does not use objective criteria related to skills and performance for the promotion process rate the process as being unfair (Beehr et al., 2004), thus increasing their perception of a glass ceiling within their organization. Moreover, in observing that there are few women who have been in managerial positions long enough to be considered serious candidates for advancement to higher levels or that there are not as many women in a situation for promotion to higher-level management positions, women managers are more likely to conclude that they are treated differently than their male colleagues (Elacqua et al., 2009). Not feeling they are treated the same as their male counterparts, women managers will be more likely to perceive a glass ceiling within their organizations (Cohen et al., 2020). However, in comparison to the study conducted by Elacqua et al. (2009), the added value of our research lies in the inclusion of a third factor in the emergence of differential treatment and ultimately the glass ceiling phenomenon. Indeed, the organizational culture in relation to gender within an organization can also increase women managers' perceptions of being treated differently from their male counterparts. Working in an organization wherein women are excluded or/and marginalized, wherein gender equality policies are not applied or promoted, wherein gender stereotypes are common, women managers perceive a difference in the way they are treated and considered by the organization. When woman managers observe that their organization and/or their colleagues convey and encourage a "masculine" image of management and express, not necessarily consciously, negative stereotypes about their flexibility and mobility (e.g., van Vianen and Fisher, 2002;Broadbridge and Hearn, 2008;Hoobler et al., 2008;Kumra and Vinnicombe, 2010), they tend to consider that women and men are treated differently and that women managers are victims of a glass ceiling. Concerning differential treatment, thinking that the performance of men and women is evaluated differently and that the distribution of rewards depends on gender (e.g., Ngo et al., 2002;Elacqua et al., 2009) increases women's perceptions that some barriers inhibit their progression/advancement to higher positions in their organization. Through our second Hypothesis, we postulated that WFC will mediate the relationships between the glass ceiling and job strain, intention to quit, job engagement, and job satisfaction. Results indicated that WFC was positively related to the glass ceiling, which was in turn positively related to job strain and intention to quit, and negatively related to job engagement and job satisfaction. We also found a direct positive relationship between the glass ceiling and intention to quit, and a negative one between the glass ceiling and job satisfaction. Therefore, these results suggest that WFC fully mediates the effects of the glass ceiling on job strain and job engagement, and partially mediates the effects of the glass ceiling on job satisfaction and intention to quit. A glass ceiling reduces access to job resources (e.g., mentoring, important information/advice, social networks, opportunities to develop knowledge, and performance feedback) and undermines personal characteristics (e.g., self-esteem, self-efficacy, and optimism) and energies (e.g., knowledge, money) of women managers (e.g., Freeman, 1990;Powell and Butterfield, 1994;Baxter and Wright, 2000;Tran, 2014). In order to progress in the hierarchy despite obstacles related to the glass ceiling, women managers invest their limited resources (Hobfoll, 2011). However, in their attempts to override these almost insurmountable barriers, they deplete their resources. This situation ultimately results in a loss spiral in which more and more resources are lost (Hobfoll, 2001). The poor psychological and/or physical health emerging from this situation of a loss of resources negatively affects individuals' functioning at home (Edwards and Rothbard, 2000), increasing women managers' perceptions of WFC. Women managers perceiving WFC are more strained (e.g., Babic et al., 2015Babic et al.,, 2019) ) due to the loss of resources in the process of juggling work and family roles (Hobfoll, 1989). In a such situation of WFC, they also are more likely to plan to leave their organization (e.g., Babic et al., 2015) and tend to reduce their level of engagement in their work (e.g., Babic et al., 2020) in order to preserve and protect their remaining resources (Hobfoll, 1989). Women managers psychologically attribute blame to the work domain and are dissatisfied with this domain given that it is the source of the conflict they perceive (Shockley and Singla, 2011). Therefore, when women managers perceive that their work interferes strongly and negatively with their family domain and causes difficulties in fulfilling family responsibilities, they are less satisfied with their job (e.g., Babic et al., 2015). Our results also indicated that the glass ceiling was directly related to job satisfaction and intention to quit. Women managers who perceive a glass ceiling report being less satisfied with their job and having more intention to quit. These effects could be explained by the lack of career advancement opportunities they have/perceive when facing such a situation of vertical discrimination (e.g., Nelson et al., 1990;Stroh et al., 1996;Foley et al., 2002). --- Limitations, Strengths, and Future Perspectives Our study has some limitations leading to interpreted findings with caution. The major limit of our study concerns the crosssectional design of our research. Indeed, such a design precludes any inference of causality among the variables. Therefore, the present study should be replicated by using longitudinal designs with repeated measures to investigate the direction of causality between the variables included in our model. Another limitation refers to the fact that data are based on self-reported measures. Indeed, our study focuses on women managers' perceptions. However, using self-reported data may have reduced the validity of our results. Indeed, this kind of data can produce two important biases. The first one concerns the social desirability influence bias and the second refers to the common method bias (Podsakoff et al., 2012). Nevertheless, in order to counter such bias, we followed several recommendations and took several precautions at both methodological and statistical levels. On one hand, at the methodological level, participants were assured that their responses were anonymous and confidential and that there were no right or wrong answers to the questions. We also used largely validated questionnaires to assess the glass ceiling's consequences for organizational attitudes and well-being, and performed confirmatory factor analyses to demonstrate their validity. On the other hand, at the statistical level, we performed the Harman's single-factor test (Podsakoff et al., 2012). Results indicated that the common method bias was not a major threat to our results. Although we included age, grade and presence of children at home as covariates, our model could be influenced by other socio-demographic variables. Indeed, the age of the youngest child, for example, influences the perception of WFC (e.g., Byron, 2005). It is therefore impossible for us to guarantee that the relationships we studied have been isolated from spurious influences (Bollen, 1989). As our data were collected in one specific organization, it is also difficult to generalize our results. Moreover, the type of sector (i.e., public or private) seems to influence the perception of the glass ceiling. Indeed, Sever (2016) found that those who worked in private sectors claimed that they felt a glass ceiling effect more than those who worked in the public sector. As we conducted our study within a private company, it would be interesting to replicate it with other samples coming from the public sector. Although we are aware of these limitations, our study also has several strengths. The first is that we tend to better understand the phenomenon of the glass ceiling by considering both its antecedents and its possible consequences for wellbeing at work. We extend the model developed by Elacqua et al. (2009) by proposing a more comprehensive model including the culture in relation to gender within the organization. We also investigated the impact of the glass ceiling on women managers' organizational attitudes and well-being at work by considering WFC as a possible mediator. In doing so, our study contributes to the literature considering that, to the best of our knowledge, no previous research has investigated the relationship between the glass ceiling and WFC through the model developed by Elacqua et al. (2009). Our findings are consistent with theories/models that are largely recognized in the literature (e.g., the COR model, Hobfoll, 1989Hobfoll,, 2001Hobfoll,, 2011; and the spillover theory, Edwards and Rothbard, 2000). Having said that, our study may have gone one step further. Through the present study, we focused only on the negative side of the work-family interface (i.e., WFC). However, over the last years, there has been a move to understand and examine the potential benefits of occupying multiple roles. Researchers refer to this positive side of the work-family interface by often employing the term of work-family enrichment defined as the "extent to which experiences in one role improve the quality of life in the other role" (Greenhaus and Powell, 2006, p. 73). Therefore, it would be interesting to include the positive side of work-family interface in the present study in order to have a comprehensive understanding of the links between glass ceiling and work-family interface. Indeed, considering that conflict and enrichment coexist (e.g., Grzywacz and Butler, 2005;Rantanen et al., 2013), focusing only on one of the two sides limits our understanding of the underlying processes (Boz et al., 2016). Furthermore, as developed earlier, resources seem to play an important role in the relationship between the process of the glass ceiling, WFC and well-being, as evidenced by, for example, the conservation of resources theory detailed throughout this research. Therefore, to understand the underlying mechanism more fully, future research should investigate the role of resources or affects as mediators in these relationships. --- Practical Implications Our results suggest that the glass ceiling may have deleterious effect on work-family conciliation, on well-being at work (i.e., job strain and job engagement) and on organizational attitudes (i.e., intention to quit and job satisfaction). Our study also highlights the importance of all three factors (i.e., the interpersonal, situational issues, and organizational gender culture) in the emergence of the perception of differential treatment by gender within an organization and ultimately to the perception that a glass ceiling exists. Therefore, one way to break down women managers' perceptions of being differently treated from their male colleagues is to act on these three factors. Indeed, organizations/employers could adopt some strategies or practices to eliminate (or at least reduce) the perception of the glass ceiling. Of course, women themselves are also essential players in breaking down barriers related to this vertical discrimination and its consequences for WFC, well-being and organizational attitudes. From the interpersonal issues perspective, it will be important to give women managers access to a mentor in order to include them in important communication and to exchange valuable strategic information, contacts and recommendations. Indeed, mentoring programs are one of the most effective ways to avoid barriers related to the glass ceiling and move into top management positions (Dworkin et al., 2012). Thus, organizations have to provide a strong mentoring program. Failing that, women have to be proactive in finding a senior level (female) mentor (within the organization or not) who has faced and successfully overcome similar obstacles and challenges (e.g., Ragins and Cotton, 1999). Moreover, organizations can also give women managers a more visible position by allowing them to develop friendly relationships with decision-makers and connect with high-ranking individuals and improve their social networks with senior managers. In other words, organizations have to provide sufficient support for women's career progression and to implement/promote strategies creating interpersonal communication and networks (Ohemeng and Adusah-Karikari, 2015). From the situational issues perspective, organizations have to use objective criteria related to skills and performance in the promotion process. It is also important that employers place women managers in important developmental positions, hire more women for high-level management positions and give them opportunities to develop their skills and competences to move more readily to the top of the hierarchy. Organizations have to invest in training and support specifically designed for women, including career and leadership development programs (Kassotakis, 2017;Broadbridge and Fielden, 2018). In benefiting from such programs, women have the possibilities to develop their leadership competences in particular. Human resource policies have to encourage women to advance or be promoted just as men are. In so doing, women are more likely to be considered serious candidates for advancement to higher levels within their organization. Lastly, from the organizational gender culture perspective, organizations have to apply/promote gender equality policies, especially in order to avoid women feeling excluded or/and marginalized. It is also crucial to combat gender stereotypes by, for example, maintaining some strict rules and regulations against adverse unprofessional acts (Bell et al., 2002). Organizations could also implement sensitization programs about the glass ceiling in order that all employees learn and adopt the values of respect for all, social knowledge and expected good behaviours (Haynes and Ghosh, 2011). Given all of these (non-exhaustive) strategies or practices, breaking through the glass ceiling is complex and requires action on several fronts but it is not an insurmountable task. However, as mentioned by du Plessis et al. (2015, p. 45), "practices that facilitate gender equality and the removal of the glass ceiling might increase the cost for organizations, but in the long run the pros will definitely outweigh the cons. The better the employees are treated, the higher profits and productivity that organizations have in return as job satisfaction is one of the most crucial factors to improve employees' performance and organizations' performance as a whole." --- DATA AVAILABILITY STATEMENT The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. --- ETHICS STATEMENT Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. The patients/participants provided their written informed consent to participate in this study. --- AUTHOR CONTRIBUTIONS AB was the main project leader and IH was the supervisor of the project. Both authors contributed to the article and approved the submitted version. --- SUPPLEMENTARY MATERIAL The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpsyg. 2021.618250/full#supplementary-material Conflict of Interest: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Despite significant promotion of diversity in companies, as well as legislation for equal opportunities for women and men, it must be noted that women still remain largely in the minority in decision-making positions. This observation reflects the phenomenon of the glass ceiling that constitutes vertical discrimination within companies against women. Although the glass ceiling has generated research interest, some authors have pointed out that theoretical models have made little attempt to develop an understanding of this phenomenon and its implications. Therefore, our study aims to fill this gap and to better understand the phenomenon of the glass ceiling by considering both its antecedents and its possible consequences. More precisely, we extend the model developed by Elacqua et al. (2009), proposing a more comprehensive model including organizational gender culture as a third factor (in addition to situational and interpersonal issues) in the emergence of the glass ceiling through the perception of differential treatment. We also investigated the glass ceiling's consequences for organizational attitudes and well-being at work by considering work-to-family conflict (WFC) as a possible mediator. We surveyed 320 women in managerial positions in a Belgian organization. Our study highlights the importance of all three factors in the emergence of the perception of differential treatment and, ultimately, the perception that a glass ceiling exists. Moreover, our results show that WFC fully mediates the effects of the glass ceiling on job strain and job engagement, and partially mediates the effects of the glass ceiling on job satisfaction and intention to quit.

Introduction Generation Z is located at the point of convergence between the quest of substantial interpersonal relationships and the pursuit of continual connectivity within the context of the digital era, which is a contextual environment that is constantly shifting and evolving. Because they are the first generation to enter adulthood in a culture that is primarily impacted by social media, members of Generation Z are confronted with the challenges and opportunities that come along with an unprecedented level of virtual participation. This study's major purpose is to analyse the complex association between the extensive use of social media by Generation Z and the evident drop in in-person encounters. Specifically, the relationship between the two is of particular interest. The objective is to get a full understanding of the many implications that these events have on the social dynamics, overall wellbeing, and interpersonal capabilities of the individuals in question (Sheikh, D. H., 2019). The purpose of this study is to analyse the complex relationship between the tendency of Generation Z to engage in heavy use of social media and the observed drop in the number of interactions that take place in person. The purpose of this study is to provide insight into the larger consequences for society norms, personal wellbeing, and the cultivation of important interpersonal qualities. This will be accomplished through an investigation of the patterns, motivations, and outcomes of individuals' digital interactions. While we are in the process of doing this investigation, it is of the utmost importance to determine the delicate equilibrium that exists between the advantages that are offered by social media and the potential drawbacks that may result from a decreased emphasis on face-to-face encounters. Our objective is to make important contributions to the ongoing academic discourse concerning the societal structure of Generation Z and its evolving engagement with technology. This will be accomplished by gaining a full understanding of these complexities. In the present day and age, members of Generation Z are confronted with the task of navigating the contrasting realms of a wide digital social environment and the subtle dynamics of interactions that take place in person. The pervasive allure of social media, which is distinguished by its speedy and convenient means of contact as well as the careful curating of self-representation, has evolved into an essential component of the routines that the current generation engages in on a daily basis (Cheeran., et.al., 2019). Concurrently, there has been a growing concern regarding the potential repercussions that could result from this extensive involvement in digital technology, particularly in regards to the influence that it could have on traditional face-to-face social relationships. --- Literature review In a study conducted by Ahn (2014), the author delved into the complexities surrounding the manner in which adolescents navigate the digital realm and interact with various social media platforms. In her work, boyd critically examines prevailing preconceptions and apprehensions regarding teenagers' use with technology, underscoring the multifaceted and intricate nature of their online experiences. The book is based on an extensive decade-long research endeavour, which involved conducting interviews with youngsters hailing from diverse socio-economic situations. In her work, boyd delves into various topics like privacy concerns, the development of identity, the dynamics of online connections, and the influence of technology on the social lives of adolescents. The author stated that whereas adults frequently view social media as a detriment to privacy and in-person communication, adolescents employ these platforms as a means to navigate their social sphere and sustain interpersonal relationships. Ahn (2014), created the term "networked publics" as a conceptual framework to delineate the virtual environments in which adolescents engage in online interactions. The author emphasises that adolescents are not merely passive recipients of technology, but rather play an active role in building their digital landscapes. The book advocated for readers to transcend prejudices and actively participate in more knowledgeable discussions around the convergence of technology, society, and young culture. Ahn (2014) conducted a comprehensive examination of the social interactions of adolescents in the era of digital technology, presenting a thoughtprovoking analysis that challenges prevailing misunderstandings and presents a multifaceted comprehension of teenagers' utilisation and interpretation of social media. The research conducted by Primack et al. (2017) examines the correlation between the utilisation of various social media platforms and the occurrence of symptoms linked to depression and anxiety among young adults residing in the United States. The study is designed to have a nationally-representative sample, with the objective of offering insights into the potential influence of varied social media involvement on mental well-being. The authors utilise a comprehensive methodology, taking into account many social media platforms and evaluating both the frequency and intensity of usage. The study is to examine the relationship between social media usage patterns and self-reported symptoms of depression and anxiety across a diverse sample of young adults in the United States. The results indicate a positive association between the utilisation of numerous social media platforms and a heightened probability of encountering symptoms associated with depression and anxiety. This study provides significant insights into the ongoing scholarly conversation regarding the mental health consequences associated with diverse forms of social media involvement. It underscores the necessity for additional investigations and a comprehensive comprehension of the intricate dynamics between social media usage and the psychological welfare of young individuals. This study aims to examine the effects of multitasking, namely the concurrent utilisation of several electronic devices or engagement in numerous online activities, on the academic performance of students. The study done by Junco and Cotten (2012) aimed to investigate the impact of multitasking on students' concentration levels during academic assignments. In their study, Junco and Cotten (2012) utilised surveys and data analytic techniques to investigate the multitasking behaviours of students and establish potential associations with their academic performance. The research conducted revealed a significant correlation between elevated levels of multitasking and diminished academic achievement. The findings indicate that individuals who often partake in multitasking, particularly during educational tasks, may encounter challenges in sustaining concentration and attaining maximum educational achievements. The research made a valuable contribution to the continuing scholarly conversation regarding the influence of technology utilisation on academic achievement. It specifically highlighted the possible adverse effects of multitasking on students' capacity to excel in their educational pursuits. The results indicated the significance of formulating efficient approaches to handle diversions and cultivate concentrated learning environments in order to augment academic performance. --- Poushter et al. (2018) conducted a comprehensive study that examined global patterns in social media usage, with a particular emphasis on differentiating between developing and industrialised nations. The researchers examined the evolution of social media adoption in various locations across time, utilising data gathered by the Pew Research Centre. The results of the study revealed that there is a sustained increase in the use of social media platforms in emerging nations. The use of social media platforms in these regions is experiencing a notable increase among users, who utilise them for a range of objectives such as communication, information dissemination, and social networking. Nevertheless, the study underscored that social media utilisation in industrialised nations has reached a state of saturation, characterised by a stabilisation in growth rates. The study provides valuable insights into the various patterns of global social media adoption, highlighting the continuous development of technology usage in many socio-economic environments. The presence of a plateau in wealthy nations indicates a state of maturity or stability in the utilisation patterns of social media. Conversely, the ongoing increase in emerging nations emphasises the ever-evolving nature of technology adoption within these areas. 2018) conducted a study that examined the worldwide patterns of social media usage, highlighting noticeable disparities between developing and industrialised countries. The results of this study have made a valuable contribution to the greater comprehension of the evolutionary patterns of technology behaviours within diverse socio-economic contexts. --- Poushter et al. ( In a study conducted by Primack et al. (2017), the objective was to investigate the correlation between interaction with social media and the perceived degree of social isolation among individuals in the young adult age group. The study encompassed a thorough examination of the social media behaviours and self-reported experiences of social isolation among a cohort of young adults residing in the United States. The researchers employed questionnaires as a means of gathering data pertaining to participants' patterns of social media usage, encompassing metrics such as the frequency and duration of their interactions across different platforms. The researchers also evaluated the subjects' reported social isolation by employing known metrics. The study's primary outcomes indicate a substantial correlation between extensive utilisation of social media platforms and an elevated probability of experiencing social isolation among individuals in the young adult demographic. According to a study conducted by Primack et al. (2017), it was shown that although social media platforms appear to enhance connectedness, excessive usage or specific engagement patterns may lead to feelings of isolation. This research provides significant contributions to understanding the intricate relationship between the utilisation of social media and the subjective perception of social isolation among individuals in the young adult demographic. This statement promotes the investigation of the psychological and social effects resulting from prolonged involvement with social media platforms. It underscores the importance of developing a comprehensive comprehension of how digital communication influences individuals' perception of being socially connected. the study revealed that specific parenting strategies, such as maintaining open lines of communication with teenagers regarding media content, are linked to higher levels of proactive monitoring. Furthermore, the inclination of adolescents to reveal their media engagements to their parents has been recognised as a crucial determinant impacting parental monitoring practises. This study elucidates the significance of communication and transparency within the parent-adolescent dynamic with regards to media consumption. The comprehension of these processes might provide insights into approaches for fostering positive media behaviours and conscientious media usage among adolescents, with an emphasis on the significance of parental engagement and proficient communication in moulding practises related to monitoring media. Twenge (2017) explored the distinct characteristics and obstacles encountered by the cohort that has come of age during the era of smartphones, social media, and uninterrupted connectivity. Twenge (year) examines the impact of technology on the cognitive framework, interpersonal engagements, and psychological well-being of iGen, hence altering their perspectives on employment, social connections, and the transition into adulthood. In her study, Twenge (2017) examined several significant themes, such as the diminishing prevalence of rebellious conduct, the increasing acceptance of other perspectives, the influence of social media on psychological welfare, and the probable lack of readiness among iGen individuals for the obligations associated with maturity. Twenge extensively utilised a substantial body of research data, encompassing surveys and interviews, to substantiate her observations and analysis. The suggested title posits that iGen, albeit displaying higher levels of tolerance and connectivity, raises worries regarding their emotional well-being and preparedness for the demands of maturity. Twenge urges readers to contemplate the wider societal ramifications of these patterns and provides perspectives on the possible effects on education, employment, and interpersonal connections. Twenge (2017) conducted a comprehensive analysis of the attributes and actions exhibited by the cohort that has come of age in the era of digital technology. This study not only elucidated the favourable and arduous parts of their encounters but also presented broader societal consequences. --- Research methodology The present study goal to do a factor analysis of generation Z's youth perception about use of social media & & decreased face-to-face encounters. A study with a quantitative approach focus was done to collect primary data from a sample of 90 respondents. The participants were linked to the standardised questionnaire. A total of 120 replies were collected, of which 110 were subjected to examination, while 10 were excluded due to the presence of distorted results. Descriptive study has used with the support of convenience random sampling. Youths gathered from college campus randomly. --- Research objectives 1. To study the perception about excessive use of social media & decreased face-to-face encounters 2. To analyse the factors of generation Z's youth for excessive social media use & decreased face-to-face encounters --- Results & Discussion Objective 1 "To study the perception about excessive use of social media & decreased face-to-face encounters". This implies a focused investigation into how individuals perceive two interrelated phenomena: the excessive use of social media and a decline in face-to-face encounters. --- Statement Generation Z's (Youth) Perception --- Results --- Excessive Use of Social Media Measurement This study aims to comprehend and measure individuals' subjective perceptions of their own or others' use of social media platforms (K, A. D., 2019). --- Attitudes and Beliefs This inquiry aims to examine the prevailing attitudes and opinions held by individuals concerning the parameters that define excessive utilisation of social media platforms. --- Impact on Daily Life Examine the manner in which individuals understand the function of social media within their everyday activities, encompassing both possible advantages and disadvantages. --- Decreased Face-to-Face Encounters: --- Quantification Evaluate the degree to which individuals perceive a reduction in in-person social connections within their personal spheres. --- Attribution Examine the participants' perspectives of the factors contributing to the reported decline in face-to-face interactions. Impact on Relationships Examine the impact of diminished face-to-face encounters on the overall quality of personal and professional relationships as perceived by individuals (Srivastava, A., & Agarwal, K., 2014). --- Interpretation: The objective of this study is to get a deeper understanding of individuals' subjective experiences, attitudes, and beliefs pertaining to the usage of social media and its potential effects on face-to-face relationships. This particular study has the potential to yield significant insights on the social ramifications of digital communication and its impact on conventional in-person communication habits. The sufficiency of the sampling was assessed using the Kaiser Meyer Olkin measure, which yielded a value of 0.874 in Table 3. This result surpasses the acceptable threshold of 0.5, indicating that the sample size was sufficient for the analysis. Therefore, it is apparent that there is a significant degree of information overlap among the variables. Therefore, it is reasonable to carry out factor analysis. According to Barlett's Test of Sphericity, the chi-square value of 456.871 at a p-value of 0.000 suggests that it is appropriate to utilise component analysis. --- Interpretation: The objective of this analysis is to offer a thorough comprehension of the elements that impact the social media usage habits of Generation Z's youth and the decrease in face-to-face interactions. This will be achieved by examining and discussing these components in detail. This knowledge possesses significant value for individuals in the field of education, policymakers, and researchers who are interested in fostering good communication practices and comprehending the intricacies of contemporary social interaction. --- Findings of the study • People of all ages use social media extensively. --- • The way we communicate, consume information, and engage with one another has all been profoundly affected by the pervasiveness of technology in modern life. --- • You can connect with people, build your network, and share information easily through social media sites. --- • Because of the integration of technology, the broadcast of information has become much easier and faster. --- • The ways in which different generations utilise social media and adapt to new technologies are noticeable. --- • People of a more senior age may have various expectations and preferences when it comes to using technology. --- • When it comes to integrating new technologies into their daily lives, younger generations tend to be more open and enthusiastic. --- • Thanks to technology, we can now communicate instantly, regardless of our physical location. --- • Online conversations can enhance in-person meetings, but they won't be able to totally supplant them. --- • Because of our dependency on digital communication, some people may feel lonely or lose the ability to interact with others in person. --- • The use of social media can have both beneficial and bad impacts on one's mental health. --- • Potentially beneficial outcomes may encompass enhanced social connectivity, freedom of expression, and accessibility to networks that provide assistance. --- • As a result of constantly comparing oneself to others and seeking approval from others online, poor selfperceptions, anxiety, and depression are possible side effects. --- Conclusion The comprehension of generational disparities is crucial in facilitating good communication and fostering teamwork. The older cohorts can derive advantages from the favourable attributes of technology, whereas the younger cohorts can acquire knowledge and insight from the experiences and sagacity of their forebears. In summary, the amalgamation of social media and technology has revolutionized the manner in which individuals establish connections, exerting an influence on their psychological state and interpersonal bonds. Achieving equilibrium between digital and offline contacts is of paramount importance, and acknowledging intergenerational disparities facilitates the establishment of a more comprehensive and interconnected societal framework. The advancement of technology has significantly contributed to the enhancement of global connectedness. However, in order to foster genuine personal relationships, it is crucial to strike a balance between encounters that occur online and those that take place offline. The establishment and cultivation of interpersonal connections through face-to-face interactions continue to be crucial for one's holistic welfare. The psychological ramifications of social media are multifaceted, exhibiting considerable variability in individual experiences. It is imperative to strike a balance between fostering positive connections and maintaining an awareness of the potential negative outcomes that may arise. The implementation of strategies aimed at fostering digital well-being and cultivating mindfulness in the use of technology has the potential to reduce the negative psychological consequences associated with its use. The integration of social media and technology has become indistinguishable from contemporary existence, exerting influence over the manner in which individuals establish connections, disseminate information, and participate in interactive discourse. The advantages encompass improved communication and the ability to access a wide range of information. Nevertheless, there are legitimate concerns that arise in relation to matters such as privacy, cyberbullying, and addiction, hence underscoring the imperative for ethical utilization of technology.

Generation Z, commonly referred to as individuals born approximately between the mid-1990s and early 2010s, exhibit a notable trait of being digitally native due to their upbringing in a period heavily influenced by technology and the prevalence of social media platforms. Numerous studies constantly demonstrate a notable prevalence of active participation in social media platforms among individuals belonging to Generation Z. Excessive use is commonly defined in academic literature as the cumulative amount of time individuals spend online, the frequency with which they engage in online activities, and the emotional consequences that social media usage has on their overall well-being. The utilisation of social media among individuals belonging to Generation Z is characterised by its comprehensive nature, embracing several dimensions such as psychological, educational, cultural, and ethical issues. Ongoing investigation is crucial for comprehending the changing patterns of social media use within this demographic and the potential consequences on their welfare and interpersonal relationships. The main aim of gthe study is to study & analyse the factors of generation Z's youth for excessive social media use & decreased face-to-face encounters.

Introduction Economic inequality, mastery of the means of production among the have, results in a deep economic gap between the rich and the poor. Inequality in the distribution of wealth is caused by the greed of an economic system that favors the owners of capital and makes society miserable (Nikmah, 2021). As a result, economic tyranny is increasingly widespread. In the end, a handful of conglomerates' power can control most Indonesian people's economy. Awareness of Indonesia's economic and political conditions led to building Indonesia with philanthropy based on zakat, infaq, alms, and waqf funds. Philanthropic funds can contribute positively to Indonesia's economic development. Even though philanthropic institutions have not significantly contributed to poverty alleviation, at least these philanthropic fund-based programs have been felt by the poor and vulnerable communities (Bahjatulloh, 2016). The broader definition of philanthropy is the awareness to give and help to overcome difficulties and improve people's welfare in the long term. Muslims' philanthropic potential is realized through zakat, infaq, alms, endowments, grants, and other kindnesses (Sholikhah, 2021). Philanthropy is a concept that already exists in Islam, which aims for good (al-birr). Considering the conditions of different social and economic levels of society, the idea or concept of philanthropy is an alternative for a community group to reduce social inequalities among people (Linge, 2015). Philanthropy is love for fellow human beings in general, doing good to them sincerely, and efforts or tendencies to improve good human life and happiness, such as through generosity, charity, or donations (Pramesti, 2020). As a movement or idea that has developed in Indonesia, philanthropy has made quite a contribution to the development of Indonesian Islamic society since the Dutch colonial era to the current reformation period, both in the form of materials and services (Saripudin, 2016). The history of Islamic philanthropy in Indonesia is deeply rooted. It has complex dynamics and twists and turns, where the religiosity of many Indonesian people adheres to Islam, providing its role and enthusiasm in the development of Islamic philanthropy. Nahdatul Ulama (NU) and Muhammadiyah are the two largest Islamic organizations in Indonesia which are concrete facts in the development of Islamic philanthropy in Indonesia. These Islamic organizations have special institutions to deal with social and economic issues such as waqf, zakat, and infaq (Sulkifili, 2018). According to James O. Midgley, philanthropy is one of three approaches to promote welfare, including poverty alleviation efforts, namely social service (social administration), social work, and philanthropy approaches. Philanthropy as a social capital has been integrated into a communal culture (tradition) that has been rooted for a long time, especially in rural communities. Cultural facts show that philanthropic traditions are preserved through giving alms to less fortunate friends, family, and neighbors. The importance of innovation in maximizing philanthropy in modern times is a must for realizing justice (Aiz, 2020). According to Ajay and Goodstein in Yoestni and Eva (2007), if you want to influence someone, the best way is to learn what they think. Differences in life experiences experienced indicate that each generation has a different lifestyle, understanding, expectations, values, and attitudes (Wijayanti et al., 2021). It becomes interesting to know how each generation's perception or assessment of the importance of philanthropy an integral part of Islamic social finance is. The generation classification broadly follows the concept of several experts with the following generation characteristics. The baby boomer generation born between 1946-1964 tended to be competitive. From the next baby boomer generation, there is generation X born between 1965-1980, a generation born in the early years of the development of technology and information. In that year, many new technologies were discovered that were beneficial to human life, such as the use of PCs, video games, cable TV, and the internet. Because generation X was born at the beginning of the development of technology and information, they can adapt and easily accept new things if there is a change in the era/next generation. Generation X will tend to be resourceful, logical, good problem solvers (Yustisia, 2016). Generation X (born 1965-1980): Able to adapt; Able to accept change well and is called a tough generation; Have independent character and loyal (loyal); Highly concerned with image, fame, and money; hardworking type; The drawback is always calculating the contribution that the company has made to its work. Generation Y (born 1981-1994): The characteristics of each individual are different, depending on where he grew up, economic strata, and social family; The pattern of communication is very open compared to previous generations; Fanatical social media users whose lives are heavily influenced by technological developments; More open to political and economic views, so they are seen to be very reactive to changes in the environment that are happening around them; Have more concern for 'wealth' or wealth. Generation Z (born 1995-2010): This digital generation is proficient and passionate about information technology and various computer applications. They will quickly and easily access information needed for educational or personal purposes; they like and often communicate with people from all walks of life, especially through social networks such as Facebook, Twitter, or SMS. Through this media, they are freer to express what they feel and think spontaneously, tend to be tolerant of cultural differences and care deeply about the environment, and are accustomed to various activities at the same time. For example, reading, talking, watching, and listening to music simultaneously. This is because they want everything fast-paced, not long-winded, and convoluted; they tend to be lacking in verbal communication, egocentric and individualist, want everything instant, impatient, and do not appreciate the process (Hafifah & Widjayatri, 2022). Istifhamah's research concludes that zakat, as Islamic philanthropy, can be distributed for Muslims to empower the people's economy in Indonesia (Istifhamah, 2015). This then corresponds with research conducted by Abrori and Kharis which concluded that implementing Islamic philanthropy relies on solutions to address the problems of poverty and injustice in Indonesia (Abrori & Kharis, 2022). Meanwhile, Zanil, Akbar, and Masykur's research concluded that philanthropy discussed in the Qur'an is not only seen as a charity movement with religious motives, but it is a form of humanity to care for one another (Akbar et al., 2020). This study aims to describe the importance of philanthropy in today's difficult times according to the classification of generations based on Beresford Research (Baby Boomers, Generation X, Generation Y, Generation Z, and Generation Z Alpha). While the research questions to be answered include Islamic philanthropy as a part of Islamic social finance important in today's difficult conditions; is there a relationship between the grouping of generations according to Beresford Research with the perspective of the importance of philanthropy in today's difficult times. --- Literature Review The word 'philanthropy' (English: philanthropy) was a term unknown in the early days of Islam, although recently, several Arabic terms have been used as equivalents. Philanthropy is sometimes called al-'ata' al-ijtima'i (social giving) and sometimes called al-takaful al-insani (human solidarity) or 'ata khayri (giving for good). However, terms such as al-birr (good deed) or as-sadaqah (alms) are also used (Linge, 2015). According to Dawam Rahardjo, philanthropic practices actually existed before Islam, bearing in mind that the discourse on social justice had also developed. Meanwhile, Warren Weaver, director of the Rockefeller Foundation (United States of America), emphasized that philanthropy is not a new tradition known in modern times because a person's concern for fellow human beings was also found in ancient times (Encyclopedia). Plato, for example, is said to have given his productive land as a waqf for the academy he founded. Within Christianity, the early followers of this religion also heavily emphasized the philanthropic tradition. Among adherents of Zoroastrianism, philanthropy has become one of their important commitments in life (Irham, 2019). Today, philanthropy has a number of goals that are not solely religious but also social and political. For example, some philanthropic institutions target only social services, believing that by providing services, the burden of community poverty can be reduced or even eliminated (Nur & Bakir, 2020). Meanwhile, philanthropic institutions also engage in social change, making social justice their main goal. In other words, these two philanthropic models want a better social life by smoothing the way for its realization through many economic, political, legal, and so on empowerment (Murti, 2016). Philanthropy is very closely related to the teachings of Islam. In Islamic teachings, Muslims have many instructions to contribute to social life and other people (Arifin, 2021). Here are some of the teachings of the Islamic religion: 1. Give alms secretly. Verily, Allah SWT loves alms that are given secretly. Bakr's answer was more astonishing. Umar said, "By Allah, I cannot surpass you after today, O Abu Bakr." This is the third time Abu Bakr has donated all of his wealth after his early conversion to Islam and when he migrated with Rasulullah SAW. Abu Bakar believes that the world and everything in it is temporary. Thus, there is not the slightest love for wealth, so it defeats his love for Allah SWT and His messenger. He is willing wholeheartedly and does not feel sad about living a simple life (As-Suyuthi, 2010). --- Hypotheses Testing the hypothesis on signs in this study is to answer whether philanthropy as a part of Islamic social finance is important in today's difficult conditions. Previous research supporting this hypothesis was based on Bahjatulloh's research which concluded that awareness of Indonesia's economic and political conditions led to the idea of developing Indonesia with philanthropy based on zakat, infaq, alms, and waqf funds. Philanthropic funds can contribute positively to Indonesia's economic development. So, philanthropy becomes important in today's difficult circumstances (Bahjatulloh, 2016). This research is in line with Madjakusumah and Saripudin's research that ZISWAF funds can be maximized to develop the people's economy so that it is an essential instrument in difficult economic conditions (Madjakusumah & Saripudin, 2020). The null hypothesis essentially states that philanthropy as a part of Islamic social finance is unimportant in today's difficult conditions. While the alternative hypothesis states that philanthropy as a part of Islamic social finance is important in today's difficult conditions. For this reason, in this case, study the null hypothesis and the alternative hypothesis are symbolically formulated as follows: H0: PImportant Philanthropy = PNot Important Philanthropy H1: PImportant Philanthropy > PNot Important Philanthropy While the independence test was carried out to answer whether there is a relationship between the grouping of generations, according to Beresford Research, with the perspective of the importance of philanthropy in today's difficult times. This hypothesis is based on Udin Saripudin's research on how the link between Islamic philanthropy (zakat, donation, and charity) and poverty reduction that occurs in a country, so the role of zakat, donation, and charity as an instrument of distributive justice can be felt by the poor (Saripudin, 2016). This research is also in sync with Abdiansyah Linge's research, which concludes that philanthropy's effectiveness in reducing social inequality is inseparable from the role of philanthropic institutions that manage these activities. The potential sourced from philanthropy is very large by distributing capital to those who cannot afford it, so that they can play a role in economic activity as producers in increasing income (Linge, 2015). In this case study, the null hypothesis and the alternative hypothesis are formulated as follows: H0: There is no influence between the grouping of generations according to Beresford Research and the perspective of the importance of philanthropy as a part of Islamic social finance in difficult times. H1: There is an influence between the grouping of generations according to Beresford Research and the perspective on the importance of philanthropy as a part of Islamic social finance in difficult times. --- Research Methods The approach in this study uses a mix methods research type. Previous research used a mix of methods with almost relevant topics applied in Muhammad Ali Mustofa's research (Mustofa, 2021) and Muh. Ihsan (2021). The research sample included 400 respondents who traveled with their final destination in Pemalang Regency with loci at the Pemalang bus terminal and station from December to January 2023. Respondents were put into a generational classification, according to Beresford's research. The sampling technique used random sampling. While the instrument used in this research is a questionnaire. The data analysis technique uses hypothesis testing on signs and independence hypothesis testing through the Khai-Square distribution. Testing the hypothesis on signs in this study is to answer whether philanthropy as a part of Islamic social finance is important in today's difficult conditions. Determination of the null hypothesis status in the sense of whether it is accepted or rejected is based on a comparison between the calculated khai-squared value and the khai-squared value in the distribution table and then aligned with the applicable testing criteria. The magnitude of the calculated khai-squared value can be found by applying the formula. X 2 = [(n 1 -n 2 ) -1] 2 n 1 + n 2 Where X 2 is the khai-squared value, n1 is the number of positive differences and n2 is the number of negative ones. In general, the hypothesis testing step through the sign testing method is carried out through the following processes: a) Formulate null hypothesis and alternative hypothesis. b) Determine a certain level of significance. c) Formulate test criteria: 1. If the valid hypothesis test is a two-tailed test, the null hypothesis is accepted if -X (<unk> 2d <unk>.f.1) 2 <unk> X 2 <unk> X (<unk> 2d <unk>.f.1) 2 While the null hypothesis is rejected if While the null hypothesis will be rejected if X 2 <unk> X (<unk>d.f.1) 2 d) Calculating the khai-squared value. At this stage, the khai-squared value is calculated by applying the formula described earlier. e) Formulate conclusions. The conclusion on this hypothesis testing method is formulated by comparing the khai-squared value in the distribution table with the calculated khai-squared value then aligned with the applicable testing criteria. X 2 > X (<unk> 2d <unk>.f.1) 2 atau X 2 <unk> -X (<unk> 2d <unk>.f.1) 2 2. If Independence testing is carried out to test the hypothesis of whether or not there is a relationship between one variable and another. While the number of individuals in a population is categorized into two variables or categories, each consisting of several alternatives. In this study, a test was conducted to determine whether there is a relationship between the grouping of generations according to Beresford Research with the perspective of the importance of philanthropy in today's difficult times. Independence testing is carried out with the help of a contingency table. --- Results and Discussion According to Beresford Research, in general the grouping of generations is as follows: 1. Generation Alpha, namely those born between 2010-2011 until now. Generational differences often cause problems because each generation is attached to certain ideal values that may differ from one another. The following is a description of the classification of respondents to a sample of 400 respondents who are in two locations, namely, Pemalang bus terminal and Pemalang station. --- Figure 1. Description of Respondents Of the 400 respondents, the most 30%, including 120 respondents, were Generation X, 29% of respondents as many as 115 belonged to BabyBoomers, 17% of respondents namely 70 respondents belonged to Generation Y, followed successively 13%, and 11% were Generation Z of 50 respondents and Generation Alpha as many as 45 respondents. 400 respondents then answered the questionnaire instrument given for testing the hypothesis on the sign and testing the independence hypothesis with a description of the respondents' answers as follows: The description of the answers of respondents who answered that philanthropy as part of Islamic social finance is important at present described in the following three thoughts: First, 40 percent of respondents who answered that philanthropy is important argued that the nature of human nature as an individual and social being is a fact and cannot be denied. Humans as individuals have the meaning of something that is whole and cannot be divided anymore. Therefore, it is natural for humans in their lives to fulfill their individual needs. Humans must recognize their interests. On the contrary, he is obliged to fulfill it seriously. However, in fulfilling his individual needs, one thing must be guarded and cared for: he (a human) must not exceed his limits and ignore other people's interests. if humans in fulfilling their needs exceed the limits, ignoring other people's interests, then humans are individualistic. In principle, in meeting their individual needs, humans should not lead to individualism. Because the other side of the individual is a group, fulfilling individual needs must also pay attention to the group's needs. Everyone must realize that individual interests are only a part of group interests. Individuals can only develop with a group which is the container for individual work. Therefore, humans must be fair to themselves, namely fulfilling the needs that are their obligations. Whole human beings are those who are able to know themselves and who are able to care for themselves properly. Because only yourself knows the problems that befall him and only, yourself knows the problems that befall him, and only yourself feels all the consequences that arise because of his negligence. Second, 35 percent of respondents answered that humans are social creatures. Social nature is innate in human nature because it must be fulfilled as well as possible. This social nature cannot be eliminated. Ignoring these qualities is self-defeating. What needs attention is how best to fulfill these characteristics. If someone ignores his interests, he will become an altruist. Humans who are altruists only pay attention to the interests of others and completely ignore their interests. When this happens, there will be confusion within. Whole human beings can realize the existence of this social nature and act to fulfill it within reasonable limits. What is meant by a reasonable limit is not violating his nature as an individual being. Humans must realize that social nature is only a part of the whole nature possessed. In normal situations, it is not good if it fulfills its social needs excessively. Nevertheless, prioritizing social interests above individual interests is a commendable attitude of a human being. Third, 25 percent of respondents answered that every person (human being) is in line with the uniqueness and uniqueness as a person who also has the freedom and should not just dissolve in existing views. However, we must also be aware that existence as a person exists in togetherness and exists in a certain form of society. Because of that, however, we also have to be in tune with the basis of individual and group relationships chosen by the community so that the common goals that have been set can be achieved properly. Broadly speaking, zakat has implications for social welfare, so the intent of the Shari'a (maqashid sharia) is realized. In line with that, Yusuf Qardawi believes that zakat is a maliyah ijtima'iyyah worship related to wealth and social development (Nikmah, 2021). From the description above, a quantitative approach is then carried out by testing the hypothesis on the sign. Several steps to test the hypothesis that must be carried out include: Formulating the null hypothesis and alternative hypotheses. The null hypothesis essentially states that philanthropy as a part of Islamic social finance is unimportant in today's difficult conditions. While the alternative hypothesis states that philanthropy as a part of Islamic social finance is important in today's difficult conditions. For this reason, in this case, study the null hypothesis and the alternative hypothesis are symbolically formulated as follows: H0: PImportant Philanthropy = PNot Important Philanthropy H1: PImportant Philanthropy > PNot Important Philanthropy After we look at the data display in the graph, it can be seen that the number of positive signs is 315, which is greater than the number of negative signs, which is 85. Thus, given that the number of positive differences is greater, the applied hypothesis testing is the right-hand side test. In this case, the significance level applied is 5% or 0.05. In the khai-squared table, the khaisquared value for degrees of freedom is 1 and a significance level of 0.05 is 3.841. As for the criteria for testing this case, the applied hypothesis testing is the left-hand side test. So, the testing criterion applied in this case is that the null hypothesis is accepted if x 2 <unk> 3,841 While the null hypothesis is rejected if x 2 > 3,841 So, based on the formula for calculating the khai-squared value that has been displayed, the khai-squared value is equal to, Based on the calculation results above, the khai-squared value is 129.96. The khai-squared value is greater than the khai-squared value in the table of 3.841. Thus, the null hypothesis is rejected, and the alternative hypothesis is accepted. The description of the answers of respondents who gave very important answers is simplified as follows: 54 respondents answered that why philanthropy is very important is because in essence, humans are individuals, individual creatures, but it cannot be denied that they are also related to other creatures, including other humans. He does not live and live alone. On the contrary, he is always with and associated with other beings and people. Humans, as a reality, are also social beings. Sociality (relationships between humans) and personality encourage humans to socialize or live together. Based on the respondents' answers, several experts also explained that humans could not be separated from society. Society is a collection of people who have relationships and need each other. Robert M. Maclver, in The Web of Government says that society is a system of ordered relations (society means a system of ordered relations). Then Harold J. Laski in The State in Theory and Practice, said that society is a group of human beings who live together and work together to fulfill their mutual desires (A society is a group of human beings living together and working together for the satisfaction of their mutual wants). Based on this definition, society includes all relationships and groups within an area. Meanwhile, Major Polak in "Sociology" said that society is a group of people who have relationships and need each other. That relationship is a social relationship between individuals and individuals, between individuals and groups, between groups and groups, formal and material, and static and dynamic. Meanwhile, Murtadha Mutahhari in "Society and History" (Society and History) said that a society consists of groups of people who are related to each other by systems, customs, rites and special laws, and who live together. Common life is when human groups live together in a certain area and share the same climate and food. While 28 percent of respondents who gave answers strongly agreed in giving the argument that philanthropy cannot be separated from our lives because human life in togetherness is natural, humans were indeed created as creatures that need each other and must help each other in meeting their needs and solving life's problems. Togetherness is also open, that is, a good relationship will be realized if mutual understanding, mutual respect, and mutual awareness that each subject has the same dignity. Humans will only succeed in realizing a life together in harmony in an atmosphere that loves and cares for each other. The meeting of a harmonious life together will only occur in a certain space and time which is called society as a social unit. The largest organizational unit of society is the state and the smallest unit is the family. Meanwhile, 18 percent of respondents think philanthropy is a form of solidarity. Solidarity is a principle that describes the attitude of caring for each individual (or family) to contribute to a group (society). The contribution is responsibility for common (general) welfare, such as a sense of belonging to a group, a sense of obligation to participate in it, and a willingness to defend society's honor. In other words, humans only become themselves as long as they are in correlation with other/other humans. In this correlation, each individual must respect the humanity of others to become fully human. Every human being is responsible for each other. Solidarity is defined as social justice. Solidarity expresses the basic assumption that we want to unite in one community and a country that shares the same destiny. We are willing to sacrifice for society, nation and state, and feel responsible for others. Based on the arguments from the respondents who gave answers that strongly agreed, it can be extracted that it cannot be denied that human relations are constitutive in human existence. Existence is existence with others. The existence with others presumably shows something deeper than just being together physically/physically. That existence also attains its highest manifestation in love. Because of that, the relationship between humans also reaches perfection in a relationship marked by love. But in perfect human relations such as that rarely found in reality. The relationship between humans is not infrequently only a certain function or something functional. Relationships between people require a foundation of awareness about both and relationships between people more comprehensively. It is not enough if humans only come to be aware that themselves are an autonomous subject. He must also be aware of the existence of something "not me", which has the same dignity and degree as "me". In the relationship between "I" and "another me" or between "me" and "you", humans meet each other on the same level, the level of brothers. The relationship between humans or sociality is one of the elements that distinguishes humans from other creatures. Philanthropy is one of three approaches to promote welfare, including efforts to eradicate poverty: social service (social administration), social work, and philanthropy (Akbar et al., 2020). Philanthropy as a social capital has been integrated into a communal culture (tradition) that has been rooted for a long time, especially in rural communities (Arifin, 2021). Cultural facts show that philanthropic traditions are preserved through giving alms to less fortunate friends, family, and neighbors. The community's demands show another feature to prioritize alleviating the burden on low-income people, whose numbers increased from 1 to 48% during the economic crisis that hit Indonesia in 1997 (Sulkifli, 2018). Based on the case description above, a series of testing steps to prove the validity of the hypothesis must be carried out. The steps that must be taken are as follows: In this case study, the null hypothesis and the alternative hypothesis are formulated as follows: H0 : There is no influence between the grouping of generations according to Beresford Research and the perspective of the importance of philanthropy as a part of Islamic social finance in difficult times. H1 : There is an influence between the grouping of generations according to Beresford Research and the perspective on the importance of philanthropy as a part of Islamic social finance in difficult times. Meanwhile, in this study, the significance level was determined at 5% or 0.05. From the description of the data in this case, the number of rows is five and the number of columns is 5. Thus, the degrees of freedom are row 4 (5 -1) and column 4 (5 -1). Thus, the chi-squared value in the table for a significance level of 0.05 and 16 degrees of freedom is 26.296. From the chisquare values in the table, the test criteria are formulated. The null hypothesis, in this case, is declared accepted if X 2 <unk> 26,296 While the null hypothesis is rejected if X 2 > 26,296 To ensure the status of accepting the null hypothesis in the above case, the calculation of the chi-square value is carried out. Previously, the expected frequency value in each cell had to be determined. After that, the chi-square value is calculated using the following calculation. The value is The conclusion is formulated by comparing the chi-squared value in the table with the calculated chi-squared value and then adjusting it to the test criteria. Based on the calculation results, the chi-square value is 22.199. That is, the value is smaller than the chi-square value in the table. thus, the null hypothesis is accepted, and the alternative hypothesis is rejected. So that there is indeed no influence between the grouping of generations, according to Beresford's research, with the perspective of the importance of philanthropy as a part of Islamic social finance in difficult times. The independence test proves that each generation's perspective is not different, in the sense that all generations have the perspective that philanthropy as part of Islamic social finance is important in these difficult times. If their thinking is simplified, they tend to see the existence of philanthropy as a form of socialism. Societal in humans is the interaction between one person and another. The interaction has its pattern because it also has its style. It is a separate whole from the forms of social groups mentioned above. Society establishes the roles of its members. It has its own meaning, expression, and purpose. Furthermore, society determines in advance how that role must be carried out. This happens by conveying his hopes to those who have that role. However, the implementation of that role is not something that is outwardly but also something that forms personality in a very deep way. Personal identity is only achieved through or in social interaction. People also choose their identity and development by choosing a particular form of society. The substance of modern philanthropy is clearly seen in its orientation, in institutional and systematic change (Aiz, 2020). In modern philanthropy, the resources collected are aimed at activities that lead to social change with the main methods of community organizing, advocacy, and public education. This orientation is in line with the orientation of social movement organizations (Social Movement Organizations), which are generally represented by civil society organizations (Abrori & Kharis, 2022). The basic characteristics of human sociality, according to Sudiarja in his book "Social Philosophy," can be explained as follows: First, human sociality or relations between humans have very broad dimensions. Indeed, in his life experience, a human becomes a human only when he associates and allies with other humans. Humans cannot live alone; no man is an island, says a proverb. Aristotle called humans zoon politicon, social beings. Meanwhile, existentialist philosophers reaffirm human existence in a new way as Mitsein (Heidegger) or Co-existence (Gabriel Marcel). This is the true essence of human sociality. Humans cannot be called human except because of their social life and togetherness with others. Sociality is an essential characteristic that cannot be denied, not a feature added to humans or conditions determined from the outside, but something that is attached to them from birth. Second, sociality related to human nature leads to a wider, fuller, and more perfect humanity. Human sociality is open and prospective and can develop in a good direction as long as community members are aware of and responsible for the prospects. Even so, social relations are very complex and include different levels. Third, social relations occur for two reasons, namely: (1) social relations occur because of close ties. It could be due to similarities in class, ethnicity, religion, or other cultures. This social relationship occurs more because of primordial instincts. Therefore this social relationship is more emotional; the bond that occurs is "from within" the social group members. (2) social relations occur because of a mutual need for one another. These social relations are more rational and produce social divisions into regular functions, a bond that occurs "from the outside." Fourth, human social nature should not be interpreted as simple or instinctive as a swarm of bees or ants. Human social nature as a reality of togetherness must still be viewed within the framework of human "autonomy and freedom," which allows various forms of social relations (competition, alliance, hostility, friendship, and so on), depending on what social basis is emphasized. --- Conclusion Based on the hypothesis testing steps above, it is stated that for testing the hypothesis on the sign it is concluded that philanthropy as a part of Islamic social finance is something that is important in today's difficult conditions. As for the independence test, it was concluded that there was no influence between the grouping of generations, according to Beresford Research with the perspective of the importance of philanthropy as a part of Islamic social finance under challenging times. In the sense that each generation's perspective is not different, all generations have a perspective that philanthropy as part of Islamic social finance is important in these difficult times. The existence of society is associated with philanthropy, not solely due to deficiencies and needs in life, human imperfections, but stems from the roots of human nature. Humans are essentially social creatures. Besides that, fundamentally, the goal of human society everywhere is the same, namely a prosperous life even though the supporting conditions that are inherent and owned by each person-individuals are not the same as one another. We must always uphold human dignity in every thought and action we have in society. Social interaction can be understood in two forms: interaction between humans and vertical interaction. In the interaction between humans, all roles need and influence each other because it is called horizontal interaction. Meanwhile, vertical interactions take place between two ends or two poles, specifically between the lower end/pole and the upper end/pole. The lower end/pole is the special interactivity of society contained in its members, while the upper end/pole is the ideals and values between social "provisions" and personal "freedom", between social facticity and the goals aspired to by society. From horizontal interaction, we arrive at vertical interaction. That is, the members of a society not only communicate with each other, but they also communicate in the stage of shared ideals and values. They have a common goal. This goal has yet to be realized and still needs to be realized. Society also aspires to the same values. These values give a special form to society. Therefore, one social form is different from other social forms. In the overall sense, every form of sociality has its own uniqueness or peculiarities

Purpose -This study aims to describe the perception of the importance of philanthropy in today's difficult times according to the classification of generations based on Beresford.The approach in this study uses a mixed methods research type. The research sample included 400 respondents who traveled with their destination in Pemalang Regency from December to January 2023. The sampling technique used random sampling. While the instrument used in this research is a questionnaire. The data analysis technique uses hypothesis testing on signs and independence hypothesis testing through the Khai-Square distribution. Findings -Based on the hypothesis testing steps above, it is stated that for testing the hypothesis on the sign it is concluded that philanthropy as a part of Islamic social finance is important in today's difficult conditions. As for the independence test, it was concluded that there was no influence between the grouping of generations, according to Beresford Research on the importance of philanthropy as a part of Islamic social finance in difficult times. Implications -Philanthropy is a form of programmed social generosity and is aimed at alleviating social problems (such as poverty) in the long term. Originality -The fundamental difference between this study and previous research lies in using data analysis techniques using non-parametric statistical tests. So, it has a more scientific level of testing results.

same as one another. We must always uphold human dignity in every thought and action we have in society. Social interaction can be understood in two forms: interaction between humans and vertical interaction. In the interaction between humans, all roles need and influence each other because it is called horizontal interaction. Meanwhile, vertical interactions take place between two ends or two poles, specifically between the lower end/pole and the upper end/pole. The lower end/pole is the special interactivity of society contained in its members, while the upper end/pole is the ideals and values between social "provisions" and personal "freedom", between social facticity and the goals aspired to by society. From horizontal interaction, we arrive at vertical interaction. That is, the members of a society not only communicate with each other, but they also communicate in the stage of shared ideals and values. They have a common goal. This goal has yet to be realized and still needs to be realized. Society also aspires to the same values. These values give a special form to society. Therefore, one social form is different from other social forms. In the overall sense, every form of sociality has its own uniqueness or peculiarities as human beings as individuals or persons in togetherness. The research implies that philanthropy is a form of programmed social generosity to alleviate social problems (such as poverty) in the long term. Philanthropy is also defined as personal or institutional voluntary actions that are encouraged to uphold the public good or voluntary actions for the public good. Zakat, infaq, alms, and waqf are configurations of philanthropy in the Qur'an and have a significant role in realizing social welfare. This study recommends that the philanthropy collected is aimed at activities that lead to social change with the main methods of community organizing, advocacy, and public education so that, in the long run, it can overcome social problems such as poverty. This research is limited to only answering the importance or not of the existence of philanthropy as part of Islamic finance in difficult times or times. It has yet to be able to answer how much the importance of philanthropy is quantitative. So it is open for other researchers to answer the shortcomings of this study with different methods or data analysis tools.

Purpose -This study aims to describe the perception of the importance of philanthropy in today's difficult times according to the classification of generations based on Beresford.The approach in this study uses a mixed methods research type. The research sample included 400 respondents who traveled with their destination in Pemalang Regency from December to January 2023. The sampling technique used random sampling. While the instrument used in this research is a questionnaire. The data analysis technique uses hypothesis testing on signs and independence hypothesis testing through the Khai-Square distribution. Findings -Based on the hypothesis testing steps above, it is stated that for testing the hypothesis on the sign it is concluded that philanthropy as a part of Islamic social finance is important in today's difficult conditions. As for the independence test, it was concluded that there was no influence between the grouping of generations, according to Beresford Research on the importance of philanthropy as a part of Islamic social finance in difficult times. Implications -Philanthropy is a form of programmed social generosity and is aimed at alleviating social problems (such as poverty) in the long term. Originality -The fundamental difference between this study and previous research lies in using data analysis techniques using non-parametric statistical tests. So, it has a more scientific level of testing results.

A s the United States experiences a historic degree of change in its demographic and linguistic landscape (Betti, 2018), Catholic education has the opportunity to serve an increasingly diverse range of families, students, and communities. Historically, the U.S. Catholic Church has been enriched by inclusive representation of cultures and languages that reflect the wider demographic shifts in the country (McLaughlin et al., 1996). Today, over 40% of U.S. Catholics identify themselves as Latino, with this percentage expected to continue growing in decades to come (Ospino, 2015). While the majority of Catholic Latinos were born in the United States, the Church remains at the central part of religious and cultural contact for newly arrived immigrants from Latin America. Both U.S.-born and immigrant Latino families navigate the cultural landscape of the United States, often while maintaining robust cultural and familial relationships across national boundaries (Oliveira, 2018;Skerrett, 2015). In this article we examine how a group of principals in a Network of Catholic Bilingual Schools (NCBS) serve Latino/a families. We do that by analyzing the results of a survey distributed to 16 principals who are part of the NCBS as well as one in-person observational visit at one of the schools in the network. NCBS has been rooted in promoting Catholic education within the Latino population, and this study highlights a need to examine how bilingual Catholic schools perform outreach to Latino families in their communities. Specific attention was paid to the principals' narratives in order to understand how schools bridge their services toward the Latino community. With the decrease of student enrollment in Catholic schools (Kim, 2016;Ospino & Weitzel-O'Neill, 2016), NCBS was launched in order to cater for the educational needs of Latino populations (Ospino & Weitzel-O'Neill, 2016). Bilingual education has been known to promote better test scores, better job possibilities, and flexibility in thinking (Bialystok et al., 2012;Grosjean, 2010) while raising awareness toward Catholic identity that embraces human dignity and public good (Scanlan & Zehrbach, 2010). Also, the school network system is expected to have cultivated an innovative leadership structure, promotion of students' academic excellence through professional development, and enhancement in Catholic identity (Huchting et al., 2017). With this background, this study addresses the following research questions: To what extent, according to principals in a nation-wide network, do Catholic bilingual schools in the NCBS serve Latino populations through family engagement? How might these schools' services impact the experience of teachers and families in the program? The frameworks that inform this study are approaches in family engagement as a construct toward the understanding of Latino/a family's experiences with bilingual/Catholic schools. We situate this framework within the anthropology and sociology of family engagement in the context of Latino populations in the United States. How do family, school and Church contribute to understandings of why school-based family engagement is a crucial practice within Latino populations? Family and community engagement with/in (with or in) schools are promoted in the literature as essential to the social and academic flourishing of students (Driscoll & Goldring, 2005;Epstein & Sanders, 2006;Henderson et al., 2007;Lareau, 1996;Nieto, 2004). When there is rupture between schools and families and there is no support for the learning and development of students, student success in school is hampered (Driscoll & Goldring, 2005). Vera et al. (2017) in their study of English Language Learners (ELL)' parental involvement in Catholic schools revealed the important role of teachers in mediating parental involvement of English Language Learners in schools. Their findings indicated that, "...Latino parental involvement may increase when teachers of ELLs are well-prepared and equipped to hold students to high expectations, invest students in educational practices, support student achievement, and create learning environments that welcome and celebrate both students and families..." (p. 16). In the same vein, our study further probes the role of principals in bilingual/Catholic schools in encouraging and reaching out to Latino families in their community. --- Literature Review --- Catholic Education and Latino/a populations In the United States, over 40% of Catholics are Hispanic. Over half of the school-age Catholics in the United States are Hispanic, but only 4% of Hispanic school-age Catholics attend a Catholic school. In the 2013-2014 school year, only 15% of students in Catholic schools were Hispanic. Due to this, it is important that Catholic schools do more to increase Latino involvement and enrollment (Ospino & Weitzel-O'Neill, 2016). Catholic elementary schools have been found to underserve Latino families in general, especially those with limited English proficiency (Scanlan & Zehrbach, 2010). With the growing Catholic Latino population in the United States, it is important that they are represented in Catholic schools (Fraga, 2016). Despite lower enrollment in Catholic schools among Latino families, there is a strong connection between Latino family values and the values of Catholic schools. Many Latino cultures place a strong emphasis on family life, which is in line with the Catholic education's mission of caring for the whole person and its strong emphasis on relationships and the familial bond (Ospino & Weitzel-O'Neill, 2016). Latino parents often want their children to remain Catholic and grow their faith, which strengthens parents' desire to send their children to Catholic schools (Crea et al., 2015;Suhy, 2012). However, one challenge to Latino families sending their children to Catholic schools is the high cost of tuition (Fraga, 2016;Scanlan & Zehrbach, 2010;Suhy, 2012). In one study done in Dallas, TX, Latino parents expressed that they felt Catholic schools were not affordable, although they did want their children to attend. All participants in this study said that if Catholic schools were free, they would send their children to them (Suhy, 2012). --- Family Engagement in Bilingual and Catholic Schools In order to make Latino families feel more welcome in Catholic schools, it is important that teachers and staff reach out to form connections and establish communication, as Latino parents are more likely to be involved in the school when they feel welcomed by teachers. Some Catholic schools even use community parishes to reach out to parents through social events after Mass where teachers and parents can communicate with one another (Vera et al., 2017). However, there are some barriers to Latino family engagement in Catholic schools. One main challenge is the language barrier that may exist between families and schools (Crea et al., 2015). Another challenge is the perception by Latino parents that Catholic schools are not welcoming to Latinos, especially to lower income Latino families (Scanlan & Zehrbach, 2010;Suhy, 2012). Latino families may not feel welcome due to schools not having bilingual programming, having a limited number of Latino teachers, or by not reflecting a multicultural community (Scanlan & Zehrbach, 2010). However, bilingual Catholic schools can help offset some of these challenges that Latino families face. Two-way Immersion (TWI) programs within Catholic schools can help increase Latino enrollment as well as family engagement within a school. TWI programs provide an inclusive way for Catholic schools to reach Latino students, can help attract more Latino families to the Catholic school, and can help eliminate the language barrier that may exist in the school (Scanlan & Zehrbach, 2010). One study found that TWI programs helped increase Latino access to Catholic schools, and helped parents see Catholic schools as providing their children with unique learning opportunities (Fraga, 2016). It has also been found that TWI Catholic schools in Spanish and English have high numbers of bilingual and bicultural faculty and have higher levels of Latino family engagement than Catholic schools without TWI programs (Ospino & Weitzel-O'Neill, 2016). It is also important to note that even when a Catholic school is not bilingual, having resources and outreach in Spanish helps increase Latino enrollment and family engagement within a school. For example, Latino enrollment increased in one school due to marketing materials in Spanish, tuition assistance, and Spanish language outreach (Fraga, 2016). It is also helpful for schools to have a parent coordinator or a bilingual-bicultural liaison between Latino families and the school to help support Latino families and facilitate communication between parents and the school staff (Crea et al., 2015;Fraga, 2016). --- The Role of School Leadership in Family Engagement in Bilingual Catholic Schools Principals and school leaders play an important role in Latino family engagement in Catholic and Catholic bilingual schools. In order to increase Latino family engagement, it is important that principals make this a priority. For example, in one TWI Catholic school, the principal sought to create mixed groupings of parents through informal coffee groups to support each other and encourage them to support their children at home (Scanlan & Zehrbach, 2010). To increase family engagement, it is important that principals work with families to build trust and acceptance, while fostering a sense of community (Marasco, 2016). Additionally, in order to effectively communicate with Latino families and increase Latino enrollment and family engagement in Catholic schools, principals must receive training in cultural competency in relation to Latino communities. It has been found that principals of Catholic schools who speak Spanish or have participated in a cultural competency training are more likely to ensure that their schools are welcoming to Latino families (Ospino & Weitzel-O'Neill, 2016). In one Catholic school, a new principal who attended a cultural competency training was able to triple enrollment in three years. This was done partly by speaking about the school in English and Spanish at Masses, and by organizing community events sponsored by the school. Another principal of a Catholic school recognized that the school must be more culturally welcoming and that Latino parents should be on the school board to help give this population a stronger voice in the school (Ospino & Weitzel-O'Neill, 2016). Additionally, some principals have even participated in doorto-door visits with families in order to immerse themselves in the culture, connect with families, and offer them support (Fraga, 2016;Marasco, 2016). In Catholic schools, the most effective school leaders accept the diversity of the community and recognize the need to participate in the community (Marasco, 2016). --- Methodology --- Research Site The data for this study were drawn from a larger one-year exploratory case study in collaboration with NCBS. There are 19 total member schools of NCBS located in 10 different US states. Of the member schools, 16 participated in this study, a response rate of 84.2. Fifteen schools had Spanish-English bilingual programs, while one school had a Mandarin-English bilingual program. Eleven schools served Pre-Kindergarten to 8th grade (PK-8), two schools served Kindergarten through 8th grade (K-8), and one school served Transitional Kindergarten to 6th grade (TK-6). Two schools did not report the grade levels they serve. The locations of the schools were evenly dispersed: four schools were in the Midwest; four schools were in the Southwest; three schools were on the West coast; two schools were in the Northeast and one school was in the Southeast. The average percentage estimate of Catholic population by state is approximately 22%, ranging from 17% to 31% (Pew Research Center, 2014). The Latino student population in the schools who are part of this network (NCBS) is quite high at 74%, which is significantly higher than the average percentage estimate of Latino students enrolled in public schools in the same area (16%) (National Center of Education Statistics, 2017). In addition, there is a large gap between the percentage of Latino students in NCBS schools compared to other schools in the same district, as shown in Table 1. This implies that Latino families chose to come to a school that is part of NCBS. Regarding the usage of Spanish as the home language, all the schools who answered exhibited a higher percentage than the one of its district, apart from the two schools, implying that the parents of the NCBS tend to value using Spanish at home. --- Catholic Bilingual School Principal Survey The data for this study were collected using the Catholic Bilingual School Principal Survey (CBSP), a tool developed by Gabrielle Oliveira's team at Boston College to examine how Catholic bilingual schools facilitate family engagement by identifying the types and magnitude of practices used to support family engagement in the context of school. The CBSP was built from the previous literature that contributed to family engagement in school (Lowenhaupt, 2014;Povey et al., 2016). Lowenhaupt (2014) sorted parental engagement into the two categories of access (school-led services that were intended to increase parents' accessibility to existing practices) and participation (the activities that parents actively led, or partnered with schools in planning and organizing). We augmented Lowenhaupt's (2014) analytical tool of access and participation by involving three axes: timeline, degree of involvement, and Catholic values. Given that parents, especially Latino/a immigrant parents have different stages of adjustment to school systems compared to non-immigrant parents (Han & Love, 2015), our analytical framework consisted of three axes (Figure 1): (1) Stages of the engagement process: enrollment and onward engagement; (2) Degree of involvement: access and participation; (3) Catholic values: equity and a sense of community. --- Sample The Catholic Bilingual School Principal Survey (CBSP) employed a total population design targeting the 19 member schools of the Network of Catholic Bilingual Schools (NCBS) in 10 states in the US. During the spring of 2018, 16 administrators answered the survey (84%): 15 principals and one assistant principal. Among the administrators who identified their demographic information, 86.7% were female (n=13) and 13.1% were male (n=2). Roughly one third were in their 30s (n=5, 33.3%), followed by 40s (n=4, 26.7%), 60s (n=3, 6.7%), 50s (n=2, 13.3%) and 20s (n=1, 6.7%). Most of them had more than 10 years of work experience in an education sector, except one principal having less than 10 years work experience: 10-19 years (n=5, 33.3%), 20-29 years (n=3, 20.0%), 30-39 years (n=1, 13%), 40-49 years (n=1, 6.7%). Nearly 87% identified their race as White (n=13) and the remaining two principals identified themselves as American Indian or Alaska Native (n=2, 13.3%). Regarding ethnicity, eight principals self-identified as Latino (53.3%). Nearly 86% (n=12) identified they were born in the US, whereas 14% (n=2) migrated from Latin America. Apart from one principal, all participants spoke Spanish (n=14, 92.7%) and 10 participants perceived their Spanish proficiency as proficient or fluent. --- Analysis The survey data were analyzed to answer the research question about how the school made institutionalized efforts to increase access and participation of the Latino parents in the process of enrollment and onward engagement. First, the political and societal context-related items were analyzed to understand how these contexts impact the way that each school supports parental engagement. Second, the two open-ended items related to the successful activities for enrollment and engagement were quantified based on the frequency and types of the activity reported as successful. The data formation was conducted from the qualitative data (open-ended answer) to the quantitative data. This process allows for forming the central support that the school provided for Latino parents by timeline of enrollment and onward engagement: Enrollment (religious connection; Linguistic inclusion; Outreach and Cultural Understanding) and Onward Engagement (One Stop Shop Approach; Parent-Teacher Organizations; Event Planning by parents. Volunteering.) --- Observational Data Collection and Analysis For in-depth analysis of parental engagement, two authors visited one selected site, School ID 1007, Texas. This is a Catholic dual language, co-educational PK-8th grade school. This school was selected based on the survey response of its principal and consultation with the funder of the Network of the Bilingual Catholic School with the goal of identifying the site that provided variation in terms of geography, student population, and relative success with immigrant family engagement. The school is located in the city near the Southern border where Latino population accounts for 79% of the total population (n=323,048) which makes sense of the percentage of the Latino students of this school (85%) (Kids Count Center, 2020). Twenty four percent of the school students use Spanish as their home language, and its percentage is lower than the one of the school district (52%). This school strives for its mission to engage students in dynamic academic experience preparing them for college and heaven. Dual language programs are one of the tools for this mission. The school sees benefits of dual language with the five points (5 C): Cognitive, Career, College, Cultural and Catholic. --- Findings Catholic bilingual schools offered various activities to engage families pre-and post-enrollment. The activities schools reported as successful for family engagement were similar across Latino and Non-Latino subgroups. However, schools reported Latino parents were more likely to be engaged in activities that were either cultural or religious in nature. When schools were trying to recruit new parents and students, they promoted the accessibility of the schools' services to Latino/a populations by translating materials and hiring bilingual staff. In order to recruit Latino/a families, schools contacted Catholic churches and organizations where Latino parents would get information about schooling. Latino parents took an active role in inviting more students to their school through the schools' parent organization. Once parents enrolled their children in school, schools emphasized active participation of both sub-groups of parents such as leadership roles in parent meetings to integrate and empower families. In this regard, our data demonstrated that the rate of engagement is similar between ethnic subgroups, but the nature of the activities in which they engaged differed. In the following section, we begin with a sketch of the context of schools, specifically, how schools have changed under the influence of the federal government, state, local entity, and diocese. --- Grounding the Context The social contexts have appeared to influence how schools organize parental engagement. Table 2 shows the results of the items in which the principals were asked about the impact of the federal government, state, local entity and diocese on their school's abilities to serve immigrant families. The results allow us to compare the impact of four different exterior levels that the respondents perceive. From the table, it is clear that the "no influence" is the most frequent answer across the four different levels of agencies. This implies that Catholic bilingual school principals perceive that their schools are quite insular from outside influences. With this result, we can assume that each individual school has its own established system to deal with expected and unexpected influences to some degree. Despite the results that Catholic bilingual schools are seemingly insular, when taking a closer look at the open-ended answers, some patterns are shown allowing for deeper understanding of how these contexts influenced serving immigrant students. First, the schools have reported that federal-level anti-immigrant rhetoric has negatively impacted their school's abilities to serve immigrant families (n=6, 43%). The schools addressed anti-immigrant tenor fueled by the current administration as the current challenges they faced in fostering the school enrollment of Latino population. One principal commented that "With the current federal political tone being as it is, we have experienced our Latino families pulling back and not wanting to be as engaged" (School ID 1011, Nebraska). The federal government's impact on school differs from where a school is located. The schools near the US-Mexico South border tend to take this situation seriously. The principals reported that they are worried about their school parents' job loss, parental detention and deportation of the families which may disrupt their children's learning and living. The principal of the school at the border summarized the federal impact by saying "SB1070, The election of the President, DACA" (School ID 1005, Arizona). Compared to the schools at the border, the schools in other regions such as the Midwest or Northeast are more likely to mention financial difficulties caused by federal title funds as the negative impact at a federal-level: "Changes to how federal title funds can be spent have created difficulties in providing the services that students need" (School ID 1013, Minnesota). For the state-level impact, nearly 43% of the principals reported that there was no influence from the state on their abilities to serve immigrant populations. However, two principals addressed the state's financial support for managing their schools: "State provides funding for public school students to attend private schools" (School ID 1010, Maryland), "There is a wonderful, state scholarship fund that keeps us afloat (but is threatened to end each year)" (School ID 2002, State not identified). However, for local-level impact, there was no additional description of how schools get impacted. In contrast to the results for federal-level impact, results for the diocese level show that the principals felt a positive impact from the diocese (35%, n=5). The three schools reported specific support that they have received from archdiocese in low income populations which included immigrant families. One school principal contrasted how dioceses support immigrant families as opposed to federal agency: "I have not noticed a change in helping to serve immigrant families based on policy changes in Washington or nationally. Our [school] has created initiatives that have helped us to better serve immigrant families" (School ID 1003, Washington). According to one school (School ID 1009, New York), the diocese' ways to support schools varied. For example, the diocese can designate a school as an "Embassy" school to support lower income families. The principal explained, "also the diocese pays support funds with a certain period of agreement (e.g. a 3-5 year financial commitment to our school)". Other schools (School ID 1010, Maryland) provided a different example for funding such as money to be used for additional staffing. --- Enrollment The principals were asked to report the successful activities that they organized to improve enrollment of Latino populations (Table 3). The principals reported various outreach strategies and content used for advertisement of the school. With regards to the outreach strategies, the schoolorganized advertisement event was reported as successful by nearly 31% of the respondents (n=8). The schools held a school tour, open house, information session, and a social event. The schools were also linked to various preschools or Catholic organizations (n=3, 13%). Some schools issued newsletters and flyers and spread the enrollment plan through SNS, TV, newspaper, and their schools' websites (n=2, 8%). The content that has been reported as successful for increasing the enrollment of Latino families are: Catholic values of school (n=5, 19%), linguistic inclusion (n=3, 12%) and outreach with cultural understanding (n=3, 12%). In other words, principals reported that Latino parents want to send Note. As some schools specified more than one ac vity, the total number of ac vi es reported is above the total number of principals who responded to the survey. their children to Catholic bilingual school and participate in parental activities at school because the school emphasizes Catholic values, provides various language support, and encourages some parents to take a lead for recruitment activities in respecting those parents' cultural experiences. The following section provides more detailed information about how these three contents, religious connection for Latino families, linguistic inclusion, and outreach and cultural understanding, helped to increase Latino families' enrollment in school. --- Religious Connection for Latino Families As mentioned above, the Catholic bilingual school principals viewed their religious connection as one of the factors which increased enrollment of Latino population. The schools reported to what extent school activities are organized based on Catholic faith. They also reported to what extent schools incorporate Latino families' faith values and customs. In all items, mean ratings are between 4 (Often) and 5 (Always), which demonstrates Catholic bilingual high school's strong commitment to religious connection for Latino families. Most of the principals always focused on the faith development of students (M=4.6, SD= 0.6) and they made school-based activity decisions rooted in the Catholic faith (M=4.6, SD=0.5). Considering the school's emphasis on Catholic values, it seems reasonable for principals to report that the students in their schools are likely to know the value of attending Mass or other prayer services as part of Catholic/Christian school (M=4.4, SD=0.8). The Catholic bilingual schools also reported supporting parents for the spiritual mission of the school (M=4.2, 0.7) Moreover, the schools had specific considerations for Latino families' cultural and spiritual experiences. The schools reported that their values and goals are aligned with those of Latino families and the Catholic church (M=4.4, SD=0.8). They shed light on faith customs of different Latino groups in religious celebrations, which may fulfill the specific demands of different Latino groups for schools (Me=4.2, SD=1.0). That the Catholic bilingual schools accommodate cultural and spiritual experiences of Latino families was also observed in the answers to the open-ended question that asked how schools engaged in practices to include Spanish-speaking families. Some of the schools held bilingual Masses and religious possessions popular in Latino families. In this regard, there was an evidence showing the Catholic school's commitment to foster Latino families' enrollment by accepting their cultural and spiritual experiences: Our school is very strong on cultural, social events which are part of different Latin cultures. These events bring our community together to celebrate, and the non-Spanish speaking parents also embrace them. Since we are next door to a very large Latino church, we also participate in religious celebrations. (School ID 2001, State not identified) --- Linguistic Inclusion The Catholic school principals cited their linguistic inclusion for Latino families as a successful factor for enrollment of Latino families in school (n=3, 12%). The schools provided assistance to help the enrollment of Latino families who prefer to speak in Spanish. Out of the 16 schools, 10 schools (63%) reported that they always provide assistance for enrollment in Spanish, followed by 3 schools (32%) that answered they sometimes provide assistance for enrollment in Spanish. The schools have a sufficient number of teachers who speak Spanish and can assist the Latino population's school enrollment. Table 4 shows that 11 out of 14 schools (83%) reported that more than half the teacher populations in their schools speak Spanish fluently. With regards to ESL certification, 80% of the schools have equal to or less than 50% of the teacher population who are certified in ESL. This means that in the remaining two schools, more than 50% of teachers are certified in ESL. The Catholic bilingual schools also have teachers who are certified in bilingual education. 67% of the schools (n=10) reported that they have less than 25% of the teacher population who are certified in bilingual education. Interestingly, not only teachers but also staff members working for the schools are reported to speak Spanish fluently. 10 schools (67%) reported that more than half of their school staff can speak Spanish fluently. These results show that the schools have a Spanish-friendly environment that may enable Latino families to access a school in which there is less convenience conducive to language barriers. --- Outreach and Cultural Understanding Some of the principals regarded additional linguistic support as successful, for example, the employment of bilingual staff members and Latino outreach coordinators as well as the translation of materials and flyers. Five schools (34%) had a staff member whose task is dedicated to parental outreach. One of the principals provided a detailed explanation about why they hired a Latino enrollment coordinator and what benefits that staff brought to the school in relation to the enrollment of Latino families: The Archdiocese hired a Latino Enrollment Coordinator about 4-5 years [ago] to address the decline of Latino enrollment in Catholic Schools. Since then we have been able to increase our ability to welcome and retain families with immigrant backgrounds. Our coordinator knows our families and their needs. However, due to Federal influence, families are scared and unsure which has contributed to a decline in the last two years. (School ID 1012, Nebraska) The Latino coordinator has knowledge about Latino family needs and guides them to receive the appropriate support given the current anti-Latino climate. Not only did the schools adapt their existing services for Latino parents (access), they also encouraged Latino parents to engage in outreach (participation). For instance, a program called Madrinas paired up Latino parents with prospective parents as a mentor-mentee relationship to help recruit other parents. One principal stated that this program has been effective, and another principal concurred, stating, Madrina program component of pairing up the parents with another returning Spanish parent as a mentor has been really helpful, as well as having them in the Madrina leadership role. (School ID 1004, Texas) Forty-four percent of the schools (n=7) reported that Latino families often or always want to do outreach activities, followed by sometimes (n=4, 25%), rarely (n=2, 19%), and never (n=2, 13%). The degree of willingness to do outreach activities quite varies according to the principal, but some of the schools provided meaningful activities organized by Latino parents as shown above. Overall, in the stage of enrollment, Catholic bilingual schools emphasized the importance of providing services regardless of parents' ethnic groups, yet, in some schools, Latino parents proactively contributed to recruitment for prospective family enrollment with their linguistic and religious assets. --- Onward Engagement Once families enrolled their children in the school, schools provided a wide range of services including linguistic inclusion and academic support as well as medical and legal needs (access). Concurrently, schools encouraged parents to participate in parent-school organizations and to volunteer at school events (participation). Among the aforementioned activities, all the principals reported that the activities with higher degrees of parent participation were more successful (See Table 5). For example, parent-school organizations and meetings got the highest success ratings by the principals. Note. As some schools specified more than one ac vity, the total number of ac vi es reported is above the total number of principals who responded to the survey. --- "One Stop Shop" Approach The Catholic bilingual schools in the NCBS have a "one-stop shop" model for service provision to their Latino populations. On the items asking the principals about how frequently the Catholic bilingual schools provide each service to Latino families ranging from Never (1) to Always (3), six services of the seven recorded the mean ratings above 2 (Sometimes). It presents that the schools offer or connect to various services inside and outside school: Summer camps/academic support (Mean = 2.6, SD = 0.5); Services interpreting letters or other legal documents (Mean = 2.5, SD = 0.7); Services outside of the school (Mean = 2.5, SD = 0.7); After-school/tutoring programs (Mean = 2.4, SD = 0.7); ESL classes (Mean = 2.1, SD = 0.8); Training resources (Mean = 2.1, SD = 0.7); and services studying for citizenship exams (Mean = 1.7, SD = 0.8). With these variety of services, the parents are more likely to have opportunities to communicate with school personnel. The principals were also asked to illustrate additional services that they provided for Latino families if any. The results showed that the schools at the US-Mexico border provided more proactive immigration-related services than others. The schools near the border identified health care and education, legal services for immigration, financial services, including tax preparation, and services provided by Catholic Charities as the main needs for the Latino families in their schools (School ID 1004, Texas;School ID 1005, Arizona;School ID 1006, Texas). --- Parent-Teacher Organizations Sixty-four percent of the principals who responded to the survey stated that their schools have an organization where parents and teachers participate together (n=9, 64%). All the schools have at least one parent-teacher organization meeting per semester: 1 time (n=2), 2 times (n=2), 3 times (n=2), 8 times (n=1) and every other month (n=1). This result presents that a school has an organized event where parents can meet a teacher through formal procedures. Principals evaluated that the schools' parent organizations and meetings were successful for Latino families. The participation rate of Latino families in parent-teacher organizations varied across the schools ranging from 10% to 75%. An open-ended question probed the positive and negative feedback principals have received about their school's Parent Teacher Organization. Only five principals out of the 16 principals responded. One principal stated that "Parents are very active and involved with planning events and fundraisers. The parents on the Board are great planners and implementers, and do their best to engage parents, teachers, and students in the events. (School ID 1012, Nebraska). The schools provided the Spanish translation at all the parent and teacher meetings. However, the three school principals said some Latino parents are reluctant to be in charge of school events and communicate with other parents due to lack of English proficiency. The principal of School ID 1004 (Texas) stated, that the Spanish-dominant parents at their school "don't like to plan [meetings] and don't like agendas" and noted that they support Spanish-speaking parents "with planning/communication when they are in charge of events." Similarly, providing translation was not enough to overcome barriers to participation for Spanish-dominant Latino parents at School ID 1012, where the principal noted "We do translate but it is still difficult to bring in more Latino families to help plan or attend meetings." (School ID 1012, Nebraska) To reduce barriers to parental participation, one principal reported their school was testing a plan to change the time of parent meetings from evening to morning. However, like other participants, they reported challenges related to the format of the meetings: "The agenda-driven format preferred by our president is off-putting to our families, whereas the "open-forum" becomes unwieldy. We need to strike a better balance." (School ID 2002, State not identified) --- Event Planning by Parents One notable point from the findings was that Latino parents and Non-Latino parents had a similar mean value of participation rate in event planning (Latino parents: 26%, Non-Latino parents: 28%). However, Latino parents were more likely to participate in activities related to their culture and Catholic identity. Eight principals also reported that Latino parents took a lead in event planning in collaboration with school staff. One principal mentioned, "The Ambassador Mothers Program has been hugely successful. It is overseen by our Latino Outreach Coordinator." They contribute their gifts and talents in so many ways to our various activities throughout the year. They lead our D<unk>a de los Muertos celebration, creating the altar, volunteering that day to help with Pan de Muerto, etc. Last year we had a D<unk>a de los Muertos celebration typical of Mexico, celebrating children. It was the idea of the Latino moms, they organized it and it was fantastic. It turned out to be a fun festival celebrating our children. (School ID 1001, California) These results may seem contradictory to the answers from the three principals saying that Latino parents feel uncomfortable to plan a parent-teacher meeting. However, when the nature of the event is related to Latino cultural experiences and the target of the events is Latino family, Latino parents are more likely to lead organizing meetings. The principal of School ID 1003 said, We have an Ambassador Mothers group of Hispanic mothers who create and execute events that build sociocultural competence and help recruit other Hispanic families. The Ambassador Mothers (overseen by our Latino Outreach Coordinator) have helped build sociocultural competence through school-wide and classroom events. (School ID 1003, Washington) --- Volunteering The principals valued Latino parents' active engagement, which is contradicted by Lopez's study (2001), which focuses on the lack of Latino parents' participation in schooling. In our findings, one principal stated "Latino parents are very involved by the nature of our school, our location, the demand for Spanish-English dual language education in the area." The principals further reported that they encourage Latino families to volunteer in school events (Mean = 4.8, SD = 0.4) and to volunteer with the community (Mean = 3.9, SD = 1.0) on the scale from 1 (Never) to 5 (Always). Given that the means of encouragement of volunteering for school events is higher than the one for the community, we can

In this article, we examine how a Network of Catholic Bilingual Schools (NCBS) serves Latino populations by examining the leadership's narratives of the school services and outreach. By employing a survey with 16 principals in the NCBS, we argue that the rate of engagement is similar between Latino and non-Latino parents, although the nature of the activities varied. Despite the long-held belief that Latino parents are less likely to participate in schooling than non-Latino parents, we found that Latino parents committed their participation in schools across cultural, linguistic, or religious activities. We highlight how principals acknowledge and describe Latino families' involvement as highly invested in their children's education and their aspirations toward proficiency in both languages, Spanish and English as well a deep-rooted commitment to Catholic values.

Latino parents are more likely to lead organizing meetings. The principal of School ID 1003 said, We have an Ambassador Mothers group of Hispanic mothers who create and execute events that build sociocultural competence and help recruit other Hispanic families. The Ambassador Mothers (overseen by our Latino Outreach Coordinator) have helped build sociocultural competence through school-wide and classroom events. (School ID 1003, Washington) --- Volunteering The principals valued Latino parents' active engagement, which is contradicted by Lopez's study (2001), which focuses on the lack of Latino parents' participation in schooling. In our findings, one principal stated "Latino parents are very involved by the nature of our school, our location, the demand for Spanish-English dual language education in the area." The principals further reported that they encourage Latino families to volunteer in school events (Mean = 4.8, SD = 0.4) and to volunteer with the community (Mean = 3.9, SD = 1.0) on the scale from 1 (Never) to 5 (Always). Given that the means of encouragement of volunteering for school events is higher than the one for the community, we can conclude that overall, Latino families are more likely to start their volunteering at the school, and then enlarge their volunteering to the community. Principals also perceived that Latino families want to build community in the school given the high mean value of 4.1 (SD = 0.7). --- Conclusion Latino parental engagement in the Catholic bilingual school is an important factor to flourish students' learning as well as to revive a declining enrollment in the Catholic school community. The schools' institutionalized efforts play a critical role in fostering Latino parental engagement (Marasco, 2016;Scanlan & Zehrbach, 2010). Within this study, a nationwide network of Catholic bilingual schools was demonstrated to provide an array of support aimed at facilitating school enrollment of the Latino population and engaging the Latino population in diverse school activities. Our goal in this study was to ask how family, school, and church contribute to understandings of why school-based family engagement is a crucial practice within the Latino population. We found that various approaches of the Catholic bilingual schools were aligned with what Latino families were looking for from schools, and some of the approaches actively facilitated Latino parents' active engagement and empowerment in the schools We analyzed the specific successful approaches that Catholic bilingual schools implemented in the timeline. At the school enrollment stage, the schools were able to successfully increase enrollment of Latino families through their strong commitment to forming a religious connection with this community. The various language supports, including bilingual recruitment material and bilingual staffing, also worked, as shown in previous studies (Crea et al., 2015;Fraga, 2016;Ospino & Weitzel-O'Neill, 2016). Moreover, Latino parents' take on the recruitment activities also contributes to an increase in school enrollment. Once Latinos started enrolling in the schools, the schools spurred to engage them in various school activities by providing a wide range of services as if they are a one-stop-shop. Through the schools' support, Latino families were able to actively participate in a parent-teacher organization, school event-planning, and volunteering for the school. Across these findings, we observed that Catholic bilingual schools encouraged Latino parents to engage in school activities and take a leadership role. Thus, Latino families contributed to the school community in many ways. These findings are notable as they contradict much of the literature focusing on what Latino families lack instead of what they can contribute to school (Lopez, 2001). Unlike the long-held belief that Latino parents are reluctant to participate in school activities or take a passive role at school, this study sheds light on the active roles of Latino parents evidenced by principals' narratives. These findings evoke in-depth discussions about how Latino families were able to participate in school activities particularly in Catholic bilingual schools and why. The schools' visions and practical strategies that they identified as successful would be valuable resources to other schools that strive to facilitate Latino family engagement. The reiterative theory of action that the Catholic bilingual schools implemented and identified as successful in this study stem from perceiving Latino parents' cultural, and linguistic experiences as funds of knowledge. For example, the schools had specific considerations for Latino families' cultural and spiritual experiences. They employed bilingual staff members and Latino outreach coordinators to bring in Latino families (Ospino & Weitzel-O'Neill, 2016). Latino parents organized a club for recruitment and planned various school events based on their cultural and social capital. These findings go beyond the evidence provided in previous research that TWI programs can increase Latino populations' access to Catholic schools (Fraga, 2016). Not only access, but also active participation was consistently observed in the findings (Lowenhaupt, 2014). Notably, Latino parents showed interest and strength in organizing activities relevant to their Catholic identity and culture in nature. These findings suggest that schools need to view parents' religious, cultural, and linguistic experiences as funds of knowledge and provide opportunities for parents to take initiatives based on their assets. Within our study, the close relationship of family, school, and church exerted a positive influence in Latino family engagement. For example, the schools recruited the Latino population in connection with the church, and the parents expected their children to grow in faith through school activities. The role of Catholic churches and charities seems more crucial in the current political rhetoric against Latino populations. The schools reported that proactive support from the diocese in many ways was helpful to provide adequate services to Latino populations who feel "scared and unsure" in this era. Thus, the goal of facilitating Latino family engagement will involve seeking continuous collaboration of family, church, and school. Latino parent engagement in the Catholic bilingual school will invigorate school communities by being aligned with the core of Catholic values on equity and community engagement (Ospino & Weitzel-O'Neill, 2016).

In this article, we examine how a Network of Catholic Bilingual Schools (NCBS) serves Latino populations by examining the leadership's narratives of the school services and outreach. By employing a survey with 16 principals in the NCBS, we argue that the rate of engagement is similar between Latino and non-Latino parents, although the nature of the activities varied. Despite the long-held belief that Latino parents are less likely to participate in schooling than non-Latino parents, we found that Latino parents committed their participation in schools across cultural, linguistic, or religious activities. We highlight how principals acknowledge and describe Latino families' involvement as highly invested in their children's education and their aspirations toward proficiency in both languages, Spanish and English as well a deep-rooted commitment to Catholic values.

Background The achievement of equity in health and access to health care have been key policy priorities for the Kenyan Government since the country achieved independence. Progress accelerated after 1994 with the development of the Health Policy Framework and the National Health Sector Strategic Plans and with the recognition of health sector improvement as a critical priority in the Vision 2030 long-term development plan. In resonance with the Constitution of Kenya (2010) and the aim to realize fundamental human rights for the entire population, the Kenya Health Policy 2012-2030 set as its core goal the creation of "equitable, affordable and quality health and related services at the highest attainable standards for all Kenyans" [1]. A key priority in this context is to address persistent socio-economic inequities in access to care by ensuring health services are available to all those who need them and that no undue and prohibitive financial burden is placed on those seeking care [2,3]. Despite considerable health gains over the last decades, marked geographical and socio-economic inequalities in health have persisted in Kenya [4][5][6][7], reflecting a widespread situation in sub-Saharan African countries [8][9][10]. Not only do poorer individuals have worse health than wealthier population groups, but they are less able to access needed care services. While comparative studies tracking inequalities in general outpatient and inpatient care use in the region are scarce, there is convincing evidence of inequalities in use of maternal and child health care service [10][11][12][13], access to HIVspecific care and treatment [14] and preventive care [15]. Furthermore, evidence from country specific studies converges on the conclusion, that impoverished or lower socio-economic status individuals face larger barriers to accessing needed care services [16][17][18]. Within this regional context, socio-economic inequalities in care utilization in Kenya are well documented for a host of care services including reproductive, maternal and child care [19,20], preventive care and immunization [20,21], urgent care [22], inpatient and outpatient care utilization [23,24]. Poverty levels are strongly associated with demand and availability of high quality, formal health care services [25]. Individuals from poorer households show lower propensity to seek care in health facilities (as opposed to relying on traditional healers or self-treating with medicines bought directly from pharmacies) when facing health problems and illness [26] and the quality of service providers is lower in poorer areas [27]. Further socio-economic inequalities in care use can be observed between health care sectors in Kenya. The private sector is more heterogeneous in terms of types of care providers and primarily serves wealthier individuals, whereas those from poorer households more commonly rely on public care providers or use lower standard, often unlicensed, private care facilities [28,29]. These differences can be traced back to spatial factors (e.g. distance to any health care provider and to high quality health care facilities) and to variations in care costs and out-of-pocket payments between public and private care providers [30,31]. While services in private clinics and in private and public hospitals are still subject to considerable user fees, care is free in public health centres and dispensaries (i.e. level 2 and level 3 facilities). Issues with care accessibility are exacerbated by low coverage of health insurance across Kenya (under 20% in 2018, although following a constant growth from only 10% in 2007) and its unequal distribution, disproportionately favouring individuals who are wealthier, formally employed and have higher educational achievement [24,32]. As attempts to address the high levels of inequality in access to care through the establishment of contributory and voluntary insurance schemes have yielded limited results [33], the Kenyan government has launched a concerted effort to ensure all Kenyans and particularly those from disadvantaged groups and regions can access needed care through a national, universal health coverage scheme [34]. In 2017, universal health coverage (UHC) was identified as one of the four pillars of the Kenyan plan for socio-economic growth, with the aspirations of achieving full population coverage, subsidizing all costs for essential health services and cutting medical out-of-pocket expenses for Kenyan households in half by 2022. Following on this pledge, in December 2018 the Kenyan government launched the first phase of an implementation strategy for UHC. Four counties with different population health profiles and facing diverse challenges were included in the pilot phase (expected to reach over 3 million individuals): Kisumu (high rates of communicable diseases), Nyeri (high rates of non-communicable diseases), Machakos (high incidence of road traffic accidents) and Isiolo (one of the poorest Kenyan counties with high rates of maternal mortality). The pilot consists in removing user fees at level 4 and 5 hospitals while compensating facilities for the revenue lost and strengthening availability of pharmaceutical and nonpharmaceutical supplies at all levels of care. Based on the lessons learned from this pilot, the government plans a national rollout of UHC to all 47 counties by 2022. As Kenya embarks on this ambitious policy project, monitoring and evaluation efforts are crucial to assess policy impact. Namely, it is important to establish whether the rollout of UHC will lead to the expected gains in equity in health and access to care and social inclusion, among other objectives. In this context, our study draws on nationally representative data collected shortly before the launch of the UHC pilot to establish a baseline assessment of socio-economic inequality and inequity in health care utilization in Kenya. By setting a basis for comparison, both at national and at county level, our estimates provide an opportunity to track progress towards eliminating inequality and inequity in care use and improving accessibility and availability of care alongside the gradual expansion of UHC throughout the country. To account for structural differences between care sectors, variations in accessibility and organization of diverse care services and the potential of differential policy impact, we independently estimate levels of inequality and inequity in care use by care type (i.e. outpatient, preventive/promotive and inpatient services) and by provider ownership (i.e. private, faith-based and public facilities). To the best of our knowledge, our study contributes the most comprehensive analysis of inequality in health services use in Kenya to date, by accounting for spatial differences (county level analysis), sector differences (provider ownership) and type of care services provided. Whereas previous studies have focused on specific types of care services and care providers or relied on local samples, we use nationally representative data to generate national and county level estimates across care and provider types. Furthermore, to understand to what extent socio-economic inequality in care use (i.e. different levels of care utilization between poorer and richer individuals) is driven by differences in need for care rather than by ability to access needed care, we also run analyses of inequity, which expand the current knowledge base. Through decomposition analysis, we separate variation determined by differences in care needs from inequality that can be traced back to demographic, socioeconomic and regional characteristics of individuals and households. In their entirety, our results provide a detailed overview of socio-economic inequality and inequity in care use in Kenya and of the individual and household level factors that drive these dynamics, therefore providing crucial insights for policy intervention. --- Data and methods --- Data and measurement All analyses presented in this study are based on microdata collected as part of the 2018 Kenya Household Health Expenditure and Utilization Survey (KHHEUS) in April and May 2018 in all 47 Kenyan counties. KHHEUS explores health spending, utilization of health services and health insurance coverage for a representative sample of households at national and county level using a computer assisted personal interview technique [24]. We have maintained for analysis data on all individuals who provided valid responses for health care utilization and household expenditure. The final analysis sample consists of 141,035 individuals from 31,636 households. We consider three main health care utilization variables: outpatient care use (binary variable indicating whether the respondent has used outpatient care in the previous 4 weeks), preventive care use (binary variable for preventive care use during the last 4 weeks) and inpatient care use (binary variable indicating whether the respondent has been admitted to hospital during the past 12 months). Outpatient care was further categorized as public ( We used total household expenditure (a proxy for household income), calculated as the sum of household food and non-food expenditure and consumption, as the ranking variable in all analyses. Values are equivalized using the Anzagi-Bernard scale, assigning a weight of 1 to all household members aged 15 and older, a 0.65 weight to children aged 5 to 14 and a 0.24 weight for all children younger than 4. Poor self-reported health (binary indicators of less than good health), the presence of chronic conditions (including hypertension, other cardiac disorders, diabetes, asthma, tuberculosis, other respiratory disorders, HIV/AIDS, cancer) and the presence of medical conditions are used as indicators of individual health status. Educational achievement (primary, secondary and tertiary education with 'less than primary education' as reference category) and employment status (no employment, informal employment with 'formal employment' as reference category) are defined both for each respondent and for the household head. Finally, we consider a set of household characteristics reflecting living standards, including the size of the household, urban residence and the availability of finished floors, finished walls, electricity, piped water and a flush toilet. --- Analytical approach Following a well-established literature in the field, we use concentration curves and concentration indexes to measure socio-economic inequality in care use and the horizontal inequity index as a measure of inequity in care utilization [35][36][37][38]. Throughout the analysis, we subscribe to an understanding of inequality as the condition of being different or unequal. This pertains to any observed variation in health status or dissimilarities in the utilization of health resources between different groups [39]. We use the term inequity when referring to those inequalities that can be considered unfair or unjust, and that are unnecessary and avoidable [39,40]. In the following, we deem as fair and necessary all those differences in care utilization between groups that are determined by differences in health status, while all remaining inequality after accounting for variation in care needs is considered inequitable. This interpretation is grounded in the "right of every human being to the enjoyment of the highest attainable standard of physical and mental health, without distinction as to race, religion, political belief, economic or social condition" [41]. and the same approach is reflected in universal health coverage policy, which aims to ensure all people have access to needed care, without discrimination and without being exposed to undue financial hardship. Starting from a definition of equality as the state where each individual uses the same amount of care resources as all other members of a population, irrespective of their socio-economic and health status, we use concentration curves to establish departures from this baseline. Concentration curves plot the cumulative percentage of a given health utilization variable against the cumulative population proportion, ranked by socioeconomic status (running from the poorest to the richest population groups). Inequality increases as concentration curves diverge from the equality line. A simple and synthetic measure of total inequality in care use is offered by the concentration index (CI), which measures the area between the concentration curve and the line of equality. The CI varies in the (-1, 1) interval and takes negative values when the care use variable is disproportionately concentrated among the poor, whereas a positive value indicates that inequality favours the wealthier. As the CI approaches zero, lower levels of inequality are present. We calculate concentration indices for the probability to use each of three types of care services: outpatient care, inpatient care and preventive and promotive care (including family planning, immunization, voluntary counselling and testing, ante/post-natal care). As the care utilization variables are binary, we apply the scale correction proposed by Erreygers [42,43] and resulting in the corrected concentration index (CCI). Significant differences in inequality levels between different types of care are assessed via dominance tests for concentration curves [38]. We then carry out a decomposition analysis to provide further insights into the individual and household characteristics that determine observed inequality. Building on a regression analysis technique for the decomposition of the CI [44], we apply an extension for non-linear models based on a partial effects representation [38,45]. In this specification, the CI is expressed as the sum of the contributions of all considered factors (obtained from the elasticity of care use with respect to each factor and the concentration index of each factor) and an error component (called the generalized concentration index for the error term). This approach allows us to decompose the value of the concentration index into the contributions of several key factors: household expenditure, care needs (health status), age and gender, household composition, educational achievement, employment and region of residence. Estimates of inequity in care use through the horizontal inequity (HI) index embody the principle that individuals with equal needs for care should have equal care utilization. We use the indirect standardization method to derive HI values by calculating differences between actual and need-predicted care utilization [46,47]. By separating the determinants of care use into a group describing care needs and a group reflecting other socioeconomic characteristics (non-need), we can use a logistic regression model to estimate how much care each individual would receive, if they were treated equally to other individuals in the sample with equal care needs (i.e. need-predicted care utilization). Inequity in care use (HI) can then be calculated as the difference between need-predicted care use and actual (observed) care use. All data analyses were carried out in Stata 15 [48]. --- Results Descriptive statistics for dependent and independent variables by total household expenditure quintiles are presented in Table 1. Levels of care utilization disaggregated at county level are available in Additional file 1. Richer individuals in Kenya report better health status than those in poorer households, although the prevalence of chronic and medical conditions is higher in richer population groups, as is the average age. Higher educational achievement (particularly at tertiary level) and formal employment (both at individual and household head level) are exceedingly more likely in higher expenditure quintiles, and extremely low among the poorest groups. Richer households tend to be smaller (4.4 household members in the top expenditure quintile as compared to 6.8 in the lowest), located in urban settings and report considerably higher availability levels of modern amenities. Health care utilization in Kenya increases with socioeconomic status. Larger shares of rich individuals (5th expenditure quintile) as compared to poorer population groups use outpatient care, inpatient care and preventive care. A particularly large gap is observed for preventive --- Inequality and inequity in care use Our results show significant pro-rich inequality in the use of all types of care services in Kenya (Table 2). Concentration indexes for outpatient, inpatient care and preventive and promotive care are all positive and statistically significant, indicating richer individuals use disproportionately more health services than their respective population proportion, while lower socioeconomic status groups access a significantly lower share of care resources, irrespective of service type. The pro-rich distribution is maintained even after controlling for different care needs between socio-economic groups, with values for HI indexes positive and significant for all care types. The estimates are overall robust to changing the socio-economic status indicator used as a ranking variable from total to non-food household expenditure (see Additional file 2). We use tests of dominance between concentration curves to establish whether one type of care is more unequally distributed than another. To do so we investigate whether one curve lies significantly above another at all points in the graph (dominates), overlaps the other at some point across the distribution (non-dominance) or lies significantly below another (is dominated). In line with the results for concentration indexes by care type, we find dominance of the equality line over the concentration curves for all types of care utilization (i.e. the equality line lies significantly above the concentration curves) (Fig. 1). Both preventive and promotive health services and inpatient services are more unequally distributed than outpatient care services (i.e. outpatient care concertation curves dominate), but we find no evidence of a marked difference in inequality in care use between the former two (non-dominance between concentration curves for preventive and inpatient care utilization). The magnitude of the contribution of each factor depends on: (i) how sensitive health care use is to variation in the given factor -i.e., its elasticity with respect to it; and (ii) how equal the distribution of a given factor is with respect to the socio-economic status of a household -i.e., its concentration index. Consequently, the largest contributions to overall inequality are relative to those factors that are both unequally distributed and strongly associated with health care use. --- Decomposition of inequality in care use The main driving factors of socio-economic inequality in health care use in Kenya are total household expenditure, educational achievement, household characteristics and living standards, all disproportionately distributed in favour of richer individuals and better off households. The educational achievement of the individuals using health services and of the household head are particularly relevant for the distribution of preventive health service utilization and relatively less relevant for inpatient and outpatient care, whereas household characteristics explain a large portion of inequality in outpatient care utilization. Health care needs drivers (measured here as selfreported health status, number of chronic conditions and the presence of medical conditions) explain a very The region of residence explains a considerable part of total inequality in preventive care utilization but only minor shares of inequality in outpatient and inpatient care use, suggesting that spatial inequalities are not equally pronounced and relevant for all care types. --- Local variation in inequality in care use We find pronounced differences in levels of inequality in health care utilization between Kenyan counties, for all types of care services analysed (Table 3). Wherever statistically significant, inequality favours richer individuals. Out of the four counties included in the UHC pilot, we find systematic inequality in all three types of care use only in Isiolo. We find no statistically significant inequality in care use in Kisumu and Machakos, while in Nyeri only outpatient care is significantly distributed in favour of richer population groups. --- Inequality in private health service utilization The private health sector in Kenya plays a key role in national health care provision, serving an important share of the population across socio-economic status groups. However, previous analyses have found inequalities in access to care provided in privately owned facilities. Our analysis confirms these results, revealing marked prorich inequality and inequity in private care utilization. Richer Kenyans use privately owned care facilities more frequently than lower socio-economic status individuals do, even after controlling for differences in care needs. Pro-rich inequality in private care use is large and statistically significant in the case of both outpatient and inpatient care services (Table 4). We also find pro-rich inequality in the use of care services provided in nonprofit or faith-based facilities, irrespective of whether services were offered on an outpatient or inpatient basis. Conversely, lower socio-economic status individuals disproportionately use outpatient services offered in publicly owned facilities (negative and significant CI). We find no statistically significant inequality in the utilization of public inpatient care in Kenya. To separate the effect of socio-economic status on the probability to use health services from its association with the type of care provider an individual in need of care will use, we repeated the same analysis on the subsample of care users (Table 5). Measured inequality increases considerably, both in the case of outpatient and inpatient care use, favouring richer individuals for privately owned services. Conversely, we find significant pro-poor inequality in the use of health services offered by publicly owned facilities. In other words, if we consider just the population who seeks care, lower socio-economic status individuals will rely on services provided by publicly owned facilities and non-profits more frequently, while richer individuals will represent the largest proportion of care users of for profit, privately owned facilities. Plots of concertation curves by care type and provider ownership for the full sample and for care users only are provided in Additional file 3. --- Discussion Our study provides an overview of socio-economic inequality and inequity in health care use in Kenya and can act as a baseline assessment for evaluating the impact of the newly implemented universal health coverage policy. The results confirm the persistence of marked pro-rich inequality in the utilization of all health care services analysed. This suggests that the series of health reforms implemented over the last decades have not led to the achievement of an equitable and accessible health care system for all Kenyans and that further, coordinated policy intervention is needed. The remaining gap in accessibility and affordability of services affects all types of care, but it is particularly pronounced for preventive and promotive care and inpatient care utilization. While user fees were removed for dispensaries and health centres (level 2 and level 3 facilities), care in hospitals (level 4 and level 5 facilities) was still subject to burdensome fees, which can discourage poorer individuals to seek needed care. Furthermore, a host of indirect costs (e.g. costs of transportation and loss of working time) compound the burden for individuals from poorer households [50]. Even though affordability of care remains a major concern in Kenya, large inequalities in preventive and promotive care use suggest additional barriers to care contribute to lower levels of care utilization among poorer individuals. Our decomposition analysis results emphasize that inequality in living standards and educational achievement explain the largest part of total inequality in health services utilization. As this is particularly the case for preventive care services, our findings constitute further evidence that efforts to improve affordability of care must be paralleled by concerted public health and health promotion initiatives. The results resonate with previous studies, which have found that costs of access but also lack of information and cultural barriers lead to marked differences in careseeking behaviour by socio-economic status, with poorer individuals less likely to seek care in case of illness but also more likely to choose non-modern health care providers, such as traditional healers [26,[50][51][52]. Investments in reducing poverty, increasing awareness of quality of care across sectors and incentivizing the use of formal medical providers are likely to contribute significantly to the reduction of inequalities in care use. The large contribution of household living standards to inequality in care use both for preventive and for outpatient care utilization, in conjunction with the results from the county level analysis, indicate that limited accessibility of health care facilities in the poorest communities might also affect access to care [50,[53][54][55]. The creation of an equitable health care system in Kenya should therefore also include a focus on improving the geographical coverage of care facilities and their balanced distribution across all counties and localities, emphasizing development of health care delivery capacity in the areas most affected by poverty. Confirming previous results [17,18], we found pro-rich inequality in health service use to be significantly higher for care provided in privately owned facilities, while publicly owned care providers serve a much larger share of individuals from lower socio-economic groups. In fact, outpatient care in publicly owned facilities is disproportionately used by lower income individuals, suggesting the impact of policy reforms on the elimination of user out of pocket fees at lower levels of care has contributed to the reduction of barriers in access to care. However, as a considerable share of total care services in Kenya are provided in privately owned facilities, it is important that future policies target improved accessibility and affordability of care across sectors, to avoid the exacerbation of a two-tiered health system, particularly if differences in quality of care between the two sectors persist. We acknowledge two limitations of our study. Firstly, the results on inequity in care use presented in this paper rely on an estimate of care needs that is limited in its precision by the availability and quality of data. As the survey data we use do not include any information on symptoms, we rely on self-reported measures of chronic and medical conditions, both limited proxies for care needs. This limitation is most relevant in the case of preventive and promotive health services utilization, where the definition of care needs through indicators of health status is particularly problematic. Furthermore, we acknowledge reporting accuracy might be limited in the setting considered by our study, as individuals from lower educational and socioeconomic backgrounds are likely to have lower diagnosis rates and awareness about medical conditions and chronic diseases. Secondly, we caution that estimates at county level can be affected by lower sample sizes and particularly by low rates of health service utilization in certain counties [36]. This is primarily the case for inpatient care utilization and the eastern counties of Kenya (further details on rates of care use by care type and county are provided in Additional file 3). --- Conclusions Despite sustained policy efforts, Kenya still faces significant levels of socio-economic inequality in access to health care services, both in inpatient and in outpatient settings. What is more, richer individuals use disproportionately more care services even though they have, on average, better health status and lower care needs. Addressing these persistent inequities in care use and access to needed services is an essential step towards reducing disparities in health and wellbeing between different population groups and promoting fairer, more inclusive societies. In this context, recent efforts of the Kenyan Government to introduce and rollout universal health coverage at national level by 2022 are both timely and appropriate to ensure care services are available to all those who need them, irrespective of their socioeconomic position. Poorer individuals in Kenya often forego essential care services due to burdensome costs and even individuals from higher income households can experience significant financial hardships after seeking care for long-term conditions or severe illness. As limited access to health insurance and the considerable financial burden it places on their household are key factors limiting care utilization among Kenyans (particularly those from lower socio-economic groups) policies aimed at expanding the insurance base and increasing the affordability of care hold the greatest potential to redress the observed inequities in care use. Their success will likely hinge on the ability of Kenyan policy-makers to increase public financing for the health sector, to strengthen prepayment mechanisms and to expand the pooling of health funds both through taxation and through compulsory and voluntary contributory schemes. While addressing affordability is a necessary condition to promote equity in health and access to care in Kenya, it will likely not be sufficient on its own. Large differences in living standards and educational achievement between income groups have a strong impact on care seeking behaviour and act as powerful drivers of inequalities in care utilization. Policies focused on care affordability must be paralleled by efforts to improve health literacy -the lynchpin to an individual's ability to recognize care needs and address them by accessing appropriate care. A primary concern in this regard is to increase investments in strengthening primary care networks and expanding preventive and promotive care service capacity across Kenyan counties. This can ensure all Kenyans can access, understand and use health information correctly in order to maintain and improve their health status. Finally, it is important to recognize that inequities in health and access to care spring from complex socioeconomic dynamics that span beyond the scope of health policies. Multi-sectoral approaches at the local, regional and national level are necessary to address the root causes of social inequity and reduce poverty and persistent disparities between socio-economic groups. To this end, the universal health coverage policy agenda should be embedded into a larger multi-sectoral collaboration (including both private and public stakeholders) focused on addressing the determinants of health and health equity. --- Availability of data and materials The data that support the findings of this study are available from the Ministry of Health, Government of Kenya. --- Supplementary information Supplementary information accompanies this paper at https://doi.org/10. 1186/s12939-019-1106-z. Additional file 1. Levels of care utilization in Kenyan counties (by care type). Additional file 2. Inequality and inequity in care use -comparison of ranking variables (total and non-food household expenditure). --- Authors' contributions All authors are responsible for the conceptualization and design of the study. SI carried out the data analysis and wrote all drafts. LDG and PS provided critical reviews of the data analysis and all manuscript drafts. JC provided overall guidance to the study and critical reviews of the data analysis and all manuscript drafts. All authors read and approved the final manuscript. Ethics approval and consent to participate Not applicable. --- Consent for publication Not applicable. --- Competing interests The authors declare that they have no competing interests. --- Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Background: Kenya is experiencing persistently high levels of inequity in health and access to care services. In 2018, decades of sustained policy efforts to promote equitable, affordable and quality health services have culminated in the launch of a universal health coverage scheme, initially piloted in four Kenyan counties and planned for national rollout by 2022. Our study aims to contribute to monitoring and evaluation efforts alongside policy implementation, by establishing a detailed, baseline assessment of socio-economic inequality and inequity in health care utilization in Kenya shortly before the policy launch. Methods: We use concentration curves and corrected concentration indexes to measure socio-economic inequality in care use and the horizontal inequity index as a measure of inequity in care utilization for three types of care services: outpatient care, inpatient care and preventive and promotive care. Further insights into the individual and household level characteristics that determine observed inequality are derived through decomposition analysis. Results: We find significant inequality and inequity in the use of all types of care services favouring richer population groups, with particularly pronounced levels for preventive and inpatient care services. These are driven primarily by differences in living standards and educational achievement, while the region of residence is a key driver for inequality in preventive care use only. Pro-rich inequalities are particularly pronounced for care provided in privately owned facilities, while public providers serve a much larger share of individuals from lower socioeconomic groups. Conclusions: Through its focus on increasing affordability of care for all Kenyans, the newly launched universal health coverage scheme represents a crucial step towards reducing disparities in health care utilization. However in order to achieve equity in health and access to care such efforts must be paralleled by multi-sectoral approaches to address all key drivers of inequity: persistent poverty, disparities in living standards and educational achievement, as well as regional differences in availability and accessibility of care.

Background An essential long-term outcome of dissemination and implementation research is sustaining the target intervention within its setting [1][2][3]. Sustainability can be defined as "the continued use of program components and activities for the continued achievement of desirable program and population outcomes" [4]. Other terms used by prior researchers include continuation, confirmation, maintenance, durability, continuance, and institutionalization. Sustainability has received increased research attention in recent years [3,[5][6][7], but data-based literature illustrating multiple types of sustainment outcomes is nearly nonexistent. This paper reports sustainability data from Project HEAL (Health through Early Awareness and Learning) a cancer communication implementation trial about the need for early detection, based in African American churches. We report data from multiple measures of sustainability to illustrate the multiple levels of analysis necessary to fully assess the sustainability of health promotion interventions. Interest and research about the long-term sustainability of health-related interventions is growing but often has not operationalized the term "sustainability" as a set of multiple outcomes [3,5,6,[8][9][10]. Researchers, funders, policy-makers, and community practitioners are concerned whether their interventions continue after an initial research project or grant [1,[11][12][13]. Recent advice for researchers also emphasizes the dynamic nature of these processes, as factors in community organizations and contexts interact over time to produce or inhibit long-term sustained program outcomes [7,14]. The importance of the multiple dimensions of sustainability for long-term improvement of community outcomes should not be underestimated. Organizations delivering the interventions must be able to continue implementing program activities for community-based interventions to affect the lives of those they are intended to serve. Given the considerable health disparities that continue to affect medically underserved populations [15], it is important to sustain evidence-based interventions that are delivered in community settings in particular because many of these settings may be limited in resources. Assessing the full sustainability of interventions in underserved communities thus requires a more detailed approach to measurement and data collection than simply asking "Was the project sustained?" For example, involving a broader population requires that the innovative ideas and strategies are spread to other organizations and implementers. Current research about sustainability has highlighted the importance of factors or processes that increase the likelihood of sustaining an intervention (predictors of sustainability) [6,7,16]. Other researchers discuss sustainability as a set of processes or capacities that take place during the initial implementation of a program [17,18], but this approach does not require longer-term data about actual continued implementation or benefits for consumers. In contrast to the focus on processes to increase sustainability, a more detailed conceptual framework for research about sustainability proposes that multiple types of long-term outcomes should be differentiated when reporting research about sustainability [4]. Few if any publications detail empirical results for diverse sustainability outcomes using multiple measures from an intervention project. This paper illustrates empirical examples of multiple measures of sustainability for a cancer communication intervention in African American churches. Scheirer and Dearing [4] recognized the multidimensional nature of the concept of sustainability and proposed that six types of outcome measures should be differentiated. This typology of sustainability outcomes was derived from its authors' prior research and experience, as well as from discussions in two large-scale workshops involving numerous researchers about sustainability measures. These types of sustainability outcome measures can be summarized as follows: These diverse types of sustainability outcomes require different levels of analysis (e.g., data collected from individual participants in the program, data to assess the continuation of the program activities, broader considerations of organizational or community-level change). It is not possible to address all components of sustainability with any one method of data collection and analysis. This adds complexity and cost to attempts to measure the extent of sustainability for a specific project because data about any one level of sustainability outcome does not necessarily encompass other levels of these outcomes. These complexities are illustrated in this multi-method study of Project HEAL sustainability in African American churches. We were interested not only in whether the intervention was continued as originally implemented but also if it was modified or unfolded differently over 24 months. --- Methods Project HEAL-the intervention and research design African American churches have been identified as an important venue for addressing inequities in cancer and other health conditions [19][20][21][22][23][24]. Project HEAL compared two approaches to training lay peer community health advisors (CHAs) to implement an evidence-based cancer educational curriculum in African American churches: a "Traditional" versus a web-based "Technology" approach. The 15 participating African American churches were randomly assigned to either the traditional or the web-based training approach. Each church recruited one male and one female to receive CHA training and certification, who then led a series of three cancer education workshops for their church members (breast, prostate, and colorectal; see Holt et al. [25] for a full description of the intervention). Results for implementation and efficacy outcomes are reported in other papers (Holt CL, Tagai EK, Santos SLZ, Scheirer MA, Bowie J, Haider M, Slade J, Wang MQ, Whitehead T: Is online comparable to in-person training for community health advisors conducting a church-based intervention? 2017. Forthcoming.) [26]. For the overall study, 15 churches were recruited and randomized, 30 CHAs trained, and 42 workshops conducted, serving 375 member participants from those churches The "Traditional" CHA training approach used inperson classroom training for the CHAs to learn about breast, prostate, and colorectal cancer early detection. CHAs then recruited participants from members of their churches to attend the series of workshops. These workshops provided evidence-based intervention content based on the prior research of the principal investigator for this implementation and sustainability research project [21,27,28]. The content of CHA training and intervention workshops in the "Technology" approach were the same, except the CHAs completed their training using a web-based system and with less technical assistance than was provided to the traditionally trained CHAs [25,29]. Technical assistance could range from providing assistance maneuvering through the webbased training system to answering questions about the content of the training and the workshop format. CHAs from both types of training were then responsible for presenting the three in-person workshops for participant members recruited from their churches. The intervention workshop series was to be held within 2 months of completing CHA training, and the CHAs were encouraged, but not required, to later replicate one or more of the workshops. For both study groups, follow-up workshops were held at each church about 12 and 24 months after the intervention workshop series to collect survey data from participants and CHAs, provide recap sessions about cancer screening, and report back data from the baseline survey. Additional engagement was maintained with CHAs and participants through newsletters (sent quarterly after completion of the workshop series, throughout the 24month study period), mobile phone text messages (sent every other week to CHAs and participants who opted to receive them, throughout the 24-month study period), annual holiday events each December, in-person interviews with pastors shortly after the completion of the workshop series, and interviews with CHAs at 12-and 24-months to encourage their continued involvement. Sustainability of health outcomes reported in this paper were collected from participants and CHAs during the 12-and 24-month workshops (Table 2). Project HEAL is a capacity building intervention. By training CHAs as health educators in the churches starting with a focus on cancer early detection, we recognized that there was community interest in additional areas of chronic disease beyond cancer (e.g., heart disease, diabetes), and encouraged and anticipated that the CHAs would naturally expand their educational activities to these areas of interest once the initial Project HEAL three-workshop cancer series was completed. While not required, we believe that this is a strength of the CHA intervention approach and that it would foster this expression of sustainability. Project staff did not do anything specifically to foster sustainability other than encourage the CHAs to continue their health promotion activities and provide technical assistance to them upon request. Study staff did ask the CHAs and study participants about their interest in additional health activities at the 12-month follow-up workshop. Churches, CHAs, and participants received monetary incentives as well as articles of appreciation with the HEAL logo, such as tote bags, bookmarks, pens, and a recognition plaque for each church. Community involvement, especially recruitment of churches, was facilitated by input from several meetings of a community advisory group and by a longstanding partnership with Community Ministry of Prince George's County (Community Ministry), a 41-year-old non-profit organization providing convening and collaborative activities among faithbased organizations. --- CHA demographics CHAs tended to be active volunteer leaders in their churches (two CHAs were also the pastors) and were ages 25 to over 70 (mean age of 51). About two-thirds cited a bachelor's degree or higher as their educational level. Seventeen (61%) of the initial 28 CHAs who provided background data were currently employed, 7 of them in health-related professions or jobs (another 6 were retired, 1 on disability, and 4 did not state their employment status). Facilitating cancer awareness education may have been personally relevant to many CHAs, as 16 of the initial 28 (57%) had experienced cancer, either themselves or through family members. --- Workshop participant demographics Individuals who enrolled in Project HEAL (N = 375) were mostly female (68%) with an average age of 55.28 (SD = 9.28) (Holt CL, Tagai EK, Santos SLZ, Scheirer MA, Bowie J, Haider M, Slade JL: Web-based vs. in-person methods for training lay community health advisors to implement health promotion workshops: Participant outcomes from Project HEAL. 2016. Forthcoming.) [26]. These participants were middle class with a median education including some college and a median family income in the $50-60,000/year bracket. Almost half (47.68%) were married and more than half reported working full-time (53.53%). Most (93.07%) had health insurance coverage. --- Overall project implementation In general terms, implementation was successful: although two churches dropped out early in the project and one was replaced (the second church could not be replaced due to project timeline), 14 of the 15 initial churches continued through the initial intervention phase and 93% of the intended workshops for all 15 churches were facilitated by the CHAs [26]. All 14 churches held the follow-up sessions for participants about 12 and 24 months after their first workshop, which were facilitated by the research staff for both training conditions. --- Data collection methods for Project HEAL sustainability research The RE-AIM (Reach Effectiveness Adoption Implementation Maintenance) Framework [30] was used to assess implementation outcomes such as adoption, reach, implementation, efficacy (reported elsewhere (Holt CL, Tagai EK, Santos SLZ, Scheirer MA, Bowie J, Haider M, Slade JL: Web-based vs. inperson methods for training lay community health advisors to implement health promotion workshops: Participant outcomes from Project HEAL. 2016. Forthcoming.) [26]), and sustainability (maintenance) outcomes across a 24-month follow-up period. During the initial series of workshops, data were collected via participant surveys (see Additional files 1, 2, 3, and 4), observations by project staff, implementation checklists at workshops, and interviews with the CHAs (see Additional files 5 and 6) and church pastors. Some missing data at each level resulted in variations in available sample sizes for several measures. In cases where 24-month data is not presented, these indicators were not collected. Table 1 summarizes the dimensions of sustainability, data sources, and example indicators and indicates when the data was collected. Multiple methods were used to collect data to assess sustainability approximately 12 and 24 months following the start of the workshop series in each church. Due to the unavailability of an appropriate standardized sustainability tool, the study team met early in the development of Project HEAL, inventoried all study measures for indicators of sustainability and added some items as needed, using the Scheirer and Dearing [4] framework as a stimulus. These multiple data sources allowed us to assess several conceptual definitions of project sustainability suggested by the Scheirer and Dearing [4] framework. Paper surveys were completed by church member participants in Project HEAL at the 12-and 24-month follow-up workshops. Participants that did not attend the follow-up workshops were mailed the survey, followed by phone calls and emails. A shortened version was sent to initial non-respondents. We obtained responses from 298 of the 375 participants, a 79% participant response rate at 12 to 18 months following the intervention workshops. Three hundred nine participants completed the participant survey approximately 24 months post-workshop series resulting in an 84% response rate (N = 366). Semi-structured interviews with 24 CHAs (80% of the 30 originally trained CHAs) were conducted by Project HEAL staff either in-person or by telephone, 12 and 24 months following the initial workshop series. These interviews were recorded by the interviewer using a structured response option format, with a few open-ended items. The interviewers had been trained previously on interviewing techniques, when they also interviewed church pastors during the first year of the project. Quantitative data analysis was conducted using SPSS and largely focused on descriptive analyses such as frequency distributions. Analysis of the few open-ended questions took a content analysis [31] approach where it was reviewed for common themes (without an a priori codebook). Data were analyzed for any differences by study group in CHA training methods; however, there were no differences and so for parsimony we present data collapsed across study groups. Project staff maintained a tracking database of participant attendance and staff contacts with each church and CHAs over time, which included church characteristics. The community partner and Project HEAL staff provided descriptive examples of other activities related to Project HEAL in the churches, the community, and their own activities. --- Results --- Results for multiple dimensions of sustainability The data collected from these multiple methods illustrate several perspectives about the sustainability of Project HEAL approximately 12 and 24 months after the start-up of its intervention workshops. The results presented below follow the framework of potential sustainability outcomes suggested by Scheirer and Dearing [4]. A. Outcomes for participants (individual-level sustainability). We assessed several sustained outcomes for the church member participants enrolled in the Project HEAL workshops, including whether those participants continued to engage in the program by attending the 12-month follow-up workshops at their churches. Our attendance records show that 39% (149) of the original 379 participants attended a 12-month follow-up workshop and 37% attended the 24-month workshop. In addition, many participants shared cancer education messages with others, which may be an important social influence to spread cancer awareness messages to others. In their responses to the 12-month survey, 92% reported that they had "shared knowledge from Project HEAL workshops" with someone else while 87% indicated doing so at 24 months. In addition, CHAs estimated that about 235 church members participated in new workshops conducted in their churches after Project HEAL (described below). We were not able to collect any outcome data from these new workshop participants. B. Sustained project activities: workshops (CHA and church-level sustainability). Data at the CHA level are presented in Table 2. Of the 24 CHAs interviewed at 12 months, eight (33%) reported that additional cancer-related workshops were conducted in their churches, after the initial three-workshop series for Project HEAL, while 5 (24%) reported doing so at 24 months. None of the CHAs reported that they replicated the original series of three Project HEAL workshops. Other health-related activities (church-level sustainability). Seventeen of the 24 CHAs (71% percent) reported at 12 months that additional health-related activities had been sponsored by their churches since the initial Project HEAL intervention and at 24 months this number had reduced to 52%. Eight CHAs were involved in conducting these other activities, in addition to 16 other church leaders and some outside agencies. Of the 17 CHAs that reported additional health activities in their churches at 12 months, 6 (35%) indicated that these activities were stimulated by Project HEAL. The CHAs estimated that nearly 500 church and community members had participated in these additional activities. Unfortunately, we cannot precisely report the extent of expansion of these other health activities as we did not explicitly record at the start of the project what health-related activities were already being conducted. It is very likely, however, that Project HEAL contributed to the churches' capacity to provide health-related programs and activities, when church leaders became more aware of the importance of physical health in addition to the spiritual health of their members. Several CHAs commented in their 12 month interviews about the roles of the HEAL project in stimulating other health activities: "HEAL was an awareness alert, a wake-up call informing people." -CHA 1 "The CHA spearhead it -she wanted to reach more people about health issues. Since Project HEAL, the Pastor also talks about health issues Few churches had formal wellness policies at the time of the 12-month follow-up interviews: only one CHA reported having a formal church wellness policy (two more at 24 months), but seven others were not sure if there was a policy (three were not sure at 24 months). However, 11 CHAs (46%) reported at 12 months that there had been changes in church policies involving health. All of these CHAs described serving healthier foods at church (although for several, this was not viewed as a change in policy) and eight reported their church had "no smoking" policies. The community partner confirmed that several participating churches have started serving only nutritious food at church, to teach congregants the value of preparing and serving healthier food. E. Sustained attention to health issues (CHA or churchlevel sustainability). The information detailed above about additional activities and changes in church policies provides evidence of sustained attention to health-related issues in these churches. The diverse types of new health-related activities in churches, including health ministries, new activities, newsletters, and one with a website, will help sustain their members' attention about these issues. In addition, many of the individual CHAs continued learning about health topics 12 months after their training for Project HEAL: 13 of the 24 (54%) had engaged in continued education about cancer, while 16 (67%) had learned about other health topics. For a few, this continuing education was part of keeping up with their requirements for employment, while at least nine obtained health information from the internet for their personal interest. --- Discussion These examples about the multiple dimensions of sustainability for Project HEAL reinforce the diverse meanings of sustainability concepts and demonstrate the utility of the Scheirer and Dearing [4] framework. Overall, this evaluation of sustainability showed diverse results, depending on how sustainability was operationalized at different levels of analysis. By looking overall at the participant and CHA-level data, sustainability slightly attenuated from 12 to 24 months. However, sustainability was still fairly robust given that we relied on volunteer lay CHAs, many of whom are already over-extended in multiple volunteer roles in their churches, to conduct additional health promotion activities beyond the initial Project HEAL intervention and with no additional resources. We assessed many dimensions of sustainability at different levels of analysis: the church level, the CHA level, and the participant level. These data for Project HEAL sustainability ranged from 0% sustainability using a conservative definition-whether the churches replicated the Project HEAL three-workshop series-to 92% of participants sharing Project HEAL intervention content with other people in their lives. The CHAs reported many cancer-relevant health promotion interventions in their churches subsequent to the initial Project HEAL three-workshop series. Importantly, the CHAs also reported subsequent additional health promotion activities in other health chronic disease areas of interest to their congregations. The Scheirer and Dearing [4] framework was useful in encouraging the team to operationalize sustainability by using not just a single measure but with multiple indicators. However, the framework did not provide standardized measures of sustainability, and there is limited guidance on interpreting the relative "weights" of the multiple indicators of sustainability that were collected. Based on the current data one could make radically different conclusions about whether Project HEAL was sustained. Perhaps a reasonable approach would be for program developers and stakeholders to identify a priori dimensions of sustainability that are most important to them at the outset of a project. Collecting data on multiple indicators of any project outcome could be burdensome for projects conducted in low-resource contexts, and some of the data can be difficult to obtain or can be subjective (e.g., sustained attention to the problem). Nevertheless, the framework provides a rich and multifaceted way to consider the important outcome of sustainability. The Scheirer and Dearing [4] framework's conceptualization of sustainability dovetails with that of the RE-AIM Framework [30] to the extent that both frameworks consider both individual-level and setting-level sustainability. Both consider the primary outcome and related outcomes as well as multiple indicators. The RE-AIM Framework [30] explicitly includes program adaptation, which could be more fully integrated into Scheirer and Dearing's framework, particularly in light of the current findings that our CHAs continued many health promotion activities but not a direct replication of the Project HEAL three-workshop series. It is likely that these diverse types of sustainability are interrelated and mutually influence the potential effects of each other. For example, the depth of involvement of the CHAs as shown by the extent of voluntary continuing education is likely to be related to their willingness to plan additional church activities, as well as their influences on the pastors for delivering health-related messages during church services. And as church members participate in some activities or hear messages from their pastor, they are likely to share their health experiences within their families and social networks, and in turn, foster requests for workshops or other information about additional health-related topics. Project HEAL provides an example of how a CHA intervention approach can build capacity in an organization and foster sustainable health promotion. The types of sustainability described as separate concepts in this paper are in reality intertwined in a potential web of factors for growing and sustaining health-related activities in these churches, beyond the specifics of the cancer awareness workshops at the heart of Project HEAL. Similar concepts about the dynamic interrelationships of evolving interventions within changing contexts have been described as the "dynamic sustainability framework" [7]. The dynamic nature of sustainability outcomes also raises questions about the contextual meaning of dissemination of "evidence-based practices" for improving health. In this case, the cancer workshops per se were based on prior research and evidence about the efficacy of similar workshops on breast, prostate, and colorectal cancer [21,27,28]. But the additional activities stimulated by the churches' participation in Project HEAL may or may not have been evidence-based. For example, the CHAs reported that additional workshops about other health topics were conducted in their churches, and many churches participated in health fairs, but we have no information about the evidence base for those additional informational sources or their outcomes. The multi-faceted processes involved in this type of community participatory project means that the evidence behind their sustained delivery cannot be closely controlled. Churches and CHAs could disseminate non-evidence-based messages, as well as research-based information. This reinforces the importance of making evidence-based programs and information accessible to lay audiences and to countering the plethora of non-evidence-based messages in the popular media. The results provided here for the diverse aspects of sustainability for Project HEAL also reinforce the complex nature of change in community contexts. In this case, there was not a linear sequence of implementing Project HEAL, then sustaining its workshops in the same or other churches. Instead, implementing Project HEAL with volunteer CHAs in diverse churches stimulated multiple streams of activities and changes in both the churches and the CHAs, which in turn are likely to foster additional types of activities among the existing networks of people. Pastors and other church leaders also became more attuned to the connections between physical and spiritual health. Throughout these activities, the roles of our community partner were crucial, from initial engagement and recruitment of the churches, to the intervention development, to the CHA training, and to implementation and evaluation. A direct "translation" of evidence-based interventions into community contexts may not be an appropriate metaphor for the dynamic processes of change that promote new health understandings and behaviors among community members. In Project HEAL, it was initially somewhat surprising to some that the CHAs who were trained to conduct cancer educational workshops later began to conduct health promotion activities in other areas of chronic disease. We view this as a success and an indicator of sustainability in that the CHA approach is one of capacity building where CHAs are trained to impart healthrelated knowledge to their peers. That the CHAs started with cancer and drifted into other health areas may be viewed as a natural sustainability process, one that is supported and embraced by the Project HEAL team. However, expansion to other health topics may also be viewed as a negative unintended consequence to the extent that the CHAs reduce their focus on the original topic of interest, in this case, cancer early detection. In the end, we must balance the reality that people are whole bodies and we need to find ways to be responsive to the community's health priorities while staying true to the primary focus of a project and yet not burning people out on a single health topic. Future research should consider thinking more purposively about sustainability and in particular, how to best support lay health educators in these efforts. Participation in Project HEAL is likely to have increased the capacities of the churches to engage in health promotion, as CHAs learned skills and tools for conducting educational activities. In our work with churches, we view organizational capacity as a key factor in understanding outcomes along the implementation continuum, including sustainability. Churches vary greatly in their capacities, and it is likely that churches with greater baseline capacities may have had better sustainability of Project HEAL. However, there are currently limitations in capacity assessment in this type of setting that both hinder this type of analysis and provide an opportunity for future research. These multiple types or meanings of sustainability for Project HEAL also showed diverse extents of sustainability, in measurement terms. While about half the churches sustained at least one cancer workshop, they provided larger numbers of other health-related activities. There were fewer reported examples of sustained partnerships or diffusion to other sites. Therefore, there is not a single measure of sustainability for this project that answers the question, "Was it sustained?" This means that predictors or influences on sustainability are also likely to differ among the different outcomes, further complicating research about factors influencing sustainability in practice. For example, to use the Program Sustainability Assessment Tool developed by Luke and colleagues [6], one would need to specify which specific sustainability outcome is being assessed, as the influencers for individual participant retention, for example, are likely to differ from influencers for church-level activity continuation or for maintenance of partnerships among organizations. In essence, it is not desirable to view sustainability as a single outcome, given the diversity of meanings and measures shown by Project HEAL. A number of recent papers about dissemination and implementation issues have emphasized the need for better measures of the underlying constructs to promote cumulative research and to improve practice [2,3,32,33]. This advice is certainly true for research about sustainability and its influences or predictors. Yet, as this paper has shown, measuring the outcomes of sustainability is not a simple task, and may yield different results depending on the unit of analysis and the data collection measures used. Although one implementation framework included "sustainability" as one outcome within a taxonomy of eight implementation outcomes [2], the data from Project HEAL has shown that sustainability is itself a multifaceted construct. Future research about sustainability should carefully delineate and "unpack" the specific aspects of sustainability that are being measured. However, the methods for Project HEAL have demonstrated that it is feasible to collect multiple sources of data about the diverse levels of sustainability, if this important data is planned for in advance. A priori data collection for sustainability using the Scheirer and Dearing [4] framework should certainly include multiple levels, which in the present case would be the organizational, interventionist, and participant. Continued participant benefit should consider evaluating efficacy outcomes after the intervention or the funding period has ended. Continued program activities may be better documented by talking not only with those tasked with executing those activities but also with program participants, to triangulate those perspectives on the extent to which program activities were sustained. Though we were not initially prepared to assess community-level partnerships or coalitions in Project HEAL, this information could be obtained by interviewing organizational leaders. With regard to organizational practices/procedures/policies, this information is likely best gathered from organizational leaders but again it could be triangulated or verified by asking other organizational members. --- Limitations The current findings are limited by a number of factors. First, as with any intervention implemented in a context, specific findings may be limited in generalizability even though the broader principles are likely applicable to other interventions and contexts. Second, in the absence of baseline data about health activities that the churches or CHAs were conducting prior to Project HEAL, it is difficult to draw conclusions about whether the health promotion activities they conducted in the 24-month follow-up period were directly attributable to project HEAL. Third, Study staff did not attend and document new workshops or additional activities conducted by the CHAs subsequent to the original three workshops. Therefore, we had to rely on reports by the CHAs on these subsequent activities. Fourth, the non-standardized, context-dependent measures of sustainability are study specific and would not be applicable to other studies. These limits on the generalizability of sustainability measures are likely to be true of any research on sustainability, which needs to be intervention and context specific. This is an issue that is not unique to Project HEAL and continues to limit the field of dissemination/implementation/ sustainability science. --- Implications for implementation science The current study operationalizes and provides a datadriven illustration of the Scheirer and Dearing [4] model using Project HEAL as an example of how this framework can be used to evaluate sustainability. The current findings highlight the importance of using multiple indicators to evaluate sustainability including multiple levels, as also recommended by the RE-AIM Framework [30]. We discuss measurement challenges in evaluating sustainability and issues unique to conducting dissemination/implementation research in community settings. --- Conclusions Research about the sustainability of community-based interventions for health promotion has increased in recent years but has not yet led to agreement on conceptual and measurement frameworks that are congruent with the complexity of the underlying processes. This paper has shown that multiple meanings and levels of analysis for sustainability exist for a single intervention project, when its potential outcomes are looked at over time. The data results for the diverse meanings of sustainability vary substantially, for example, from zero churches sustaining all three workshops with fidelity, to nearly three quarters of the CHAs reporting additional health-related activities in their churches. Future research about sustainability should specify which dimensions of sustainability it is addressing, with attention to the likely diversity of influences on those outcomes. It is not possible for research to develop policy suggestions for strengthening the extent of sustainability without first defining which dimension or meaning of sustainability is intended. For example, policies for keeping church members engaged in health promotion activities would be quite different from actions to encourage churches to adopt church wellness policies or to expand their community partnerships. Research-based advice for community-level practitioners is equally complex: there are no easy answers for questions about how to ensure sustainability of their intervention work. Strategies for increasing the extent of sustainability of the specific interventions included in a researchinitiated project (e.g., booster training sessions, targeted technical assistance) may differ substantially from the strategies needed for capacity building to sustain a diversity of health-related activities (e.g., partnership building to provide evidence-based intervention materials across the wide variety of chronic disease interests expressed by community; technical assistance for helping churches build sustainable health ministries and linkages to the healthcare system). Long-term change within community organizations is likely to take a diversity of forms and avenues that reflect the multiple interactions of people, organizations, resources, and interventions. --- Availability of data and materials Intervention materials are available upon request from the principal investigator (CH). Requests for study data will be reviewed on a case-by-case basis by the Principal Investigator (CH) and may be made available upon execution of a data-sharing agreement. --- Additional files Additional file 1: 12-month Men's Participant Survey. This survey was completed by male Project HEAL participants at the 12 Authors' contributions MAS conceptualized and drafted the paper. SLZS assisted with the writing and provided data sources and project management. EKT analyzed the data and assisted with the writing. JB provided a critical review of the paper. JS and RC are community researchers and provided input into all aspects of the study and reviewed the paper. CLH is the principal investigator of Project HEAL and helped with the overall conceptualization and the writing. All authors read and approved the final manuscript. --- Competing interests The authors declare that they have no competing interests. --- Consent for publication Not applicable. --- Ethics approval and consent to participate All human participants were treated in accordance with the Principles of the Ethical Practice of Public Health of APHA and gave their informed consent to participate, and the study was approved by the Institutional Review Board (#10-0691). This trial is registered with clinicaltrials.gov (NCT02076958). --- Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Background: Sustainability of evidence-based health promotion interventions has received increased research attention in recent years. This paper reports sustainability data from Project HEAL (Health through Early Awareness and Learning) a cancer communication implementation trial about early detection, based in African American churches. In this paper, we used a framework by Scheirer and Dearing (Am J Publ Health 101:2059-2067, 2011) to evaluate multiple dimensions of sustainability from Project HEAL. Methods: We examined the following dimensions of sustainability: (a) continued benefits for intervention recipients, (b) continuation of intervention activities, c) maintaining community partnerships, (d) changes in organizational policies or structures, (e) sustained attention to the underlying issues, (f) diffusion to additional sites, or even (g) unplanned consequences of the intervention. Project HEAL provided a three-workshop cancer educational series delivered by trained lay peer community health advisors (CHAs) in their churches. Multiple sources of sustainability were collected at 12 and 24 months after the intervention that reflect several levels of analysis: participant surveys; interviews with CHAs; records from the project's management database; and openended comments from CHAs, staff, and community partners. Results: Outcomes differ for each dimension of sustainability. For continued benefit, 39 and 37% of the initial 375 church members attended the 12-and 24-month follow-up workshops, respectively. Most participants reported sharing the information from Project HEAL with family or friends (92% at 12 months; 87% at 24 months). For continuation of intervention activities, some CHAs reported that the churches held at least one additional cancer educational workshop (33% at 12 months; 24% at 24 months), but many more CHAs reported subsequent health activities in their churches (71% at 12 months; 52% at 24 months). No church replicated the original series of three workshops. Additional data confirm the maintenance of community partnerships, some changes in church health policies, and continued attention to health issues by churches and CHAs. Conclusions: The multiple dimensions of sustainability require different data sources and levels of analysis and show varied sustainability outcomes in this project. The findings reinforce the dynamic nature of evidence-based health interventions in community contexts.

Introduction The general well being of people living with HIV (PLWHIV) is greatly influenced by factors such as mental health and socio-demographic characteristics, in addition to their medical conditions [1]. Good oral health is known to positively influence the quality of life of individuals, this is particularly true in case of PLWHIV [2][3][4]. The inter-relationship between oral health and systemic health is evident especially in patients with chronic diseases, with the signs and symptoms manifested in the oral cavity with varying degree of severity. PLWHIV are more susceptible to oral lesions, the more prevalent being candidiasis, hairy leukoplakia and aphthous ulcers. In addition, they present an increased risk for dental caries, periodontal disease and xerostomia. These oral health issues could affect mastication and swallowing, inadvertently compromising nutrition and possibly adherence to treatment. Furthermore, their immune response could be negatively influenced by poor oral health alluded by the increased levels of immune activation markers found in patients with oral lesions [5]. Though the incidence of HIV-related oral lesions are known to have been reduced from 50% to 10% due to the use of Anti Retroviral Treatment (ART), several studies have shown an increase in the prevalence of dental diseases such as dental caries, periodontitis and tooth wear [6,7]. The Decayed, Missing, Filled Teeth (DMFT) index in PLWHIV ranges from 8.7 to 18.8, depending on the regional, cultural and demographic differences, along with oral health behaviors and preventive practices, nutrition and access to dental services [10]. In India, studies on the oral health status of PLWHIV have reported varied prevalence rates of oral lesions from 68% to 75%, and a poor dental and periodontal status with the DMFT score of 12.83 [11][12][13]. A longer duration of ART, depression and other comorbidities have been associated with an increased risk of dental caries [14,15]. Adding to the burden of dental diseases are the challenges to oral health care mainly due to a lack of oral health awareness, financial concerns and access to care. The findings from the baseline data of a study on oral healthcare in PLWHIV in the US reported that oral health is the persistent unmet need among these individuals, with their oral health deteriorating over time [16]. It is recognized that psychosocial, behavioral and biologic factors play a role in the health of individuals. Behavioral science involves the application of the knowledge of these factors to diagnose and develop interventions for health promotion [1]. Health behaviors and practices arise from and are driven by health beliefs, which in turn influence health outcomes [15]. There are several models that strive to predict and explain health behaviors. The Health Belief Model (HBM) is a psychosocial model and posits that an individual's health behavior depends on his perception of an illness as a threat to his health and well-being, and that the benefit of a certain action would outweigh the negative effects of inaction [17]. The assumptions of this model are based on 6 constructs: perceived severity, perceived susceptibility, perceived benefits, perceived barriers, self efficacy and cues to action. Therefore, this model can be used to understand the behaviors that have lead to a particular disease or could potentially lead to a disease even before the signs and symptoms develop. As PLWHIV are vulnerable to a plethora of oral diseases, the HBM can be applied to explain and predict the oral health beliefs, behaviors and practices that might influence their oral health status. There is a lack of contextual information regarding oral health in PLWHIV derived from the patients' perspective. Furthermore, there is no behavioral research with regard to oral health in the Indian context. Qualitative methods are extensively used in behavioral research to effectively capture the perceptions and experiences of the participants and explain the complexity of behavior. We therefore used qualitative methods along with the theoretical framework of the HBM as a guide to strengthen our research. The findings from this study could aid in finding strategies and developing preventive interventions that focus more on the individual's perception of health for effective oral health outcomes. The data obtained from this research will be used to develop a psychometric instrument to assess the oral health behaviors of PLWHIV. --- Methods This qualitative study is a part of a larger research project, which aims to study the association between oral health behaviors, oral health status and CD4 count in PLWHIV in Mangalore, Karnataka State, India. Our study sample included PLWHIV who were receiving ART at the ART Centre, Wenlock Hospital, which is a government run district hospital in Mangalore. The healthcare personnel at the ART Centre comprise a medical officer, a nurse, a counselor and a pharmacist. PLWHIV visit the Centre on a monthly basis to refill their antiretroviral medication and for a clinical followup. The PLWHIV visiting the centre were approached and requested to participate in the study by authors AR and VH, who administered the informed consent and conducted in-depth interviews. An interview guide developed for the purpose of this study was used for the interviews, and probes were used to elicit additional details or clarify responses. The interviews were conducted in the local language i.e. either Kannada or Tulu. The interviews lasted for about 20 min, were organised in a quiet room and audio recorded with prior permission of the participants. The sociodemographic details of the participants were recorded. Data saturation was achieved with the 14th participant, after which two more interviews were conducted to confirm saturation. The Health Belief model was used to frame the interview guide with key questions based on its constructs. --- Data analysis --- Data was analyzed using content analysis The interviews were translated and transcribed to the English language by a native local language speaker proficient in both the local languages and English. The transcripts were then read by the first and second authors and spot checked for consistency with the recordings. AR then familiarized herself with the transcripts and coded them. The text was broken into smaller meaningful units, or codes, which are the specific and meaningful words and ideas expressed by the participants. All relevant text from the transcripts was grouped under these codes. After coding the first few transcripts, the rest of the transcripts were indexed by applying the identified codes. Additional codes were generated as the analysis progressed. The codes from the different transcripts were then grouped under categories. Discussions between the authors enabled triangulation of the results and the reorganization of the codes and categories to ensure that the interviews were interpreted correctly. The codes and categories were then mapped under the five constructs of the HBM, which provided subthemes and themes for the analysis. --- Ethics approval An approval for the study was obtained from the Karnataka State AIDS Prevention Society, and the Institutional Ethics Committee, A J Institute of Medical Sciences and Research Centre. An informed consent was obtained from the participants prior to enrolment in the study. Verbal and written information regarding the aim, procedure and confidentiality of the study was given to all subjects. --- Findings The constructs of the HBM were used to highlight the factors that influenced the oral health behaviors in PLWHIV (Fig. 1). The demographic details of the participants are shown in Table 1. --- Perceived susceptibility and severity Most participants believed that HIV would not be the cause of any oral problems as they had them even before being infected. They also did not perceive oral health issues as major problems, even though they had past experiences with mouth ulcers, dental decay and extractions. However, they felt that the future is unpredictable. On being asked whether she might be susceptible to oral disease due to HIV, one participant replied: --- "That I don't know. It's with God. (Laughs) it is left to God. It might happen or not. My feeling is that it wont happen. " P_10 (Female) Participants, therefore considered their susceptibility as low, and were even unaware of the oral health implications of HIV. However, on probing deeper, some shared their past experiences with oral problems and how it affected them. Tooth pain and ulcers were commonly reported. The ulcers caused pain and burning that lead to an inability to eat food and also a reduced apatite. Some believed that that the tooth pain was due to beeda/tobacco. In many places in India, it is commonly believed that mouth ulcers are caused due to 'ushna' (increased 'heat') in the body. Therefore, even though a mouth ulcer was the first sign of HIV, it was initially ignored. --- "In the mouth I had some ulcer. They said it was because of 'heat'... I didn't feel like eating at all. This was before testing. Then they said I should get admitted to the hospital.. Then the mouth problems became less. " P_10 (Female) These problems prevented them from going to work, and therefore affected their livelihood, as one participant stated: "I used to earn ten thousand (Rupees) per month, I didn't go to work for three months. I got ulcer (mouth), I thought during work it might have happened. I had severe pain, burning sensation. I was not even able to have a fruit juice. " P_8 (Female) Oral lesions had negatively impacted their physical and psychological health that affected their social life. --- Perceived benefits and self efficacy The participants were aware of the importance of good oral health. Some believed that poor oral health could affect the general systemic health. The benefits included more efficient chewing, especially hard food, better facial appearance and also a reduced illness related burden on the family. Most patients believed that brushing their teeth twice daily with paste and toothbrush was sufficient to maintain good oral health. Some preferred Ayurvedic toothpaste, as they believe it is made of natural ingredients and therefore safer. Tongue cleaning, mouthwashes and salt rinses were also practiced by a few. One participant felt that the practices that were followed by the older generations such as cleaning teeth with salt, charcoal and cashew/mango leaves are extremely effective. "Our fathers they had such strong teeth. They all at that time used salt, cashew leaves.. to scrub. Our teeth fell off, not theirs. " P_12 (Female) Another habit that was practiced was oil pulling. It is an ancient Ayurvedic remedy that is believed to strengthen teeth and improve oral health by swishing a tablespoon of oil for 15-20 min in the mouth. One participant described her family's practice of oil pulling and its benefits: "It (oil pulling) prevents disease as much as possible. All toxins from the mouth will come out. Then we drink water. 1 litre. With that all the toxins will go out. We will get taste with food. We will feel like eating food. As soon as we wake up, we put coconut oil in the mouth. Then for half an hour we do oil pulling. " P_3 (Female) Some participants believed salt has the ability to cure infections. Others felt that chewing paan reduces toothache. When questioned about their diet, most participants reported no dietary changes. However, some avoided sweets. Some avoided non-vegetarian food (meat), as they believed it is bad for teeth, but considered fish healthy. Few could not eat hard food as they had unreplaced missing teeth. There also existed a belief that the body needed to be 'cooled' to counter the 'heat' to prevent problems. "I have been told that body should be made cool. I have tender coconut water... so that body doesn't get 'heat'. They told me to drink juice.. chickoo juice. But I can't drink too cold also. I get cough. So I have ragi and all. These tablets (ART) are 'heat' no?" P_10 (Female) --- Benefits of regular dental visits Tthough participants acknowledged the benefits of good oral health, they believed that regular dental visits and treatment are unnecessary if oral conditions don't affect them. Some have even pulled out the mobile teeth themselves. Few participants have never visited the dentist in the past. --- "No no no... I haven't (visited the dentist). I didn't need to remove teeth or had any pain... I never had any problem before. I only come to this doctor here to get tablets. Only if needed I will go to dentist. Only if there is pain we will go isn't it. (Laughs). " P_5(Male) The benefits of having a full complement of teeth was perceived to be low, as the partially edentulous state did not affect their daily routines. Though they had missing teeth that needed replacement, some participants preferred to remain partially edentulous. Missing posterior teeth did not affect their looks. Mastication was not a problem as they could use their remaining teeth. There was a perception that as age advances teeth replacement is unnecessary. The other reasons were that they have no children, lived alone, and don't socialize, so teeth replacement was not required. Moreover, removable dentures were considered cumbersome and fixed dentures expensive. --- Perceived barriers Psychological barriers like depression and loneliness lead to a lack of motivation and a state of denial of their oral healthcare needs. This, along with a low socioeconomic status and lack of oral health awareness discouraged preventive measures. A female participant said she lived alone, was unaware of where to go, and was therefore dependent on others. She also displayed a low self-confidence and decision-making capability. Lack of time due to busy work schedules and financial constraints were cited as reasons to defer treatment. --- "Yes I feel like cleaning (teeth), but I am not able to go because of work, lot of work... I don't get time. To keep teeth (denture).. I am ready. But its too much of money.. that's why I didn't do. Private and all is very costly.. that's why I didn't do it. " P_9 (Male) Stigma and fear of disclosure were other reasons. There was a fear of a lack of confidentiality that could lead to their stigmatization in the society. To overcome this issue, most participants preferred not to disclose their HIV status to the dentist. They believed that HIV is a personal issue that shouldn't be disclosed. Some have been informed by their peers not to reveal their HIV status. --- "If we tell them (dentists), its like making it public. Next time we will not be able to go there. If we tell we have this problem, won't they tell others or look at us differently. " P_4 (Male) Regarding the possibility of cross infection, they felt that dentists anyway take precautions. Some believed that because they are on medications, they cannot infect dentists. On the other hand, some felt that it is important to disclose their status as medications can affect the dental treatment. Some revealed their HIV status only when asked. --- "When I went to remove my teeth they asked me if I have any problem. I say no. BP, sugar is normal. Then they removed (teeth). They wear gloves isn't it? They wont have a problem. Most doctors wear gloves and work." P_4 (Male) --- Cues to action --- Internal cues to action Many participants had given up deleterious habits like smoking and paan chewing believing that it might adversely affect their general health. The inability to eat food and fear of pain also motivated some of them to maintain good oral health. They didn't want to be a burden on their children. --- "Its is very important, no madam? If there is anything (problem) its difficult to eat. If teeth are not proper, can we eat hard things.. we have to eat soft things, isn't it? And then if we are not careless about it we can live longer. If we keep it clean only we can protect (teeth). If we don't keep clean it will go.. because of decay, cavities..if we eat sweets. It is our duty to protect. We use that toothpaste.. Ayurvedic. If tomorrow tooth pain starts, in the night, because of one tooth, the whole thing.. eyes, head and all will have problem.. because of that we are careful. " P_12 (Female) --- External cues to action Participants revealed that they received little or no advice on oral health, by health care providers, however friends and family did play a role in motivating them to seek dental treatment. Most were ignorant about oral health, and they believed that information by dentists in their clinics, and doctors at the ART Centre could help create awareness on oral health care. However, as it would be difficult to reach out to every patient, patients themselves could pass on this information to their peers. Posters, handouts or even Short Message Services on the mobile phones regarding oral health in the clinics and hospitals, could increase awareness among patients. --- "You can give them information, like a camp (dental) you can do and tell them. It needn't be them only.. people with disease. It can be any normal person who comes.. you can give them information.. Some go house to house to give information. " P_12 (Female) --- Discussion The results of our study indicate that the participants had varied oral health beliefs and behaviors. Overall, the participants failed to interrelate oral and systemic health due to which oral health was not prioritized. The unfavorable behaviors were due to a low perception of susceptibility to oral disease and a low level of oral health awareness. Most of our participants believed that their established daily routine of tooth brushing at least once a day, use of tongue cleaners and mouth rinses with water especially after meals would be sufficient to maintain good oral hygiene. Mouthwashes and interdental cleaning aids were rarely used. However, salt rinses were routinely used by the participants as they are widely believed to improve oral health. Though the evidence is conflicting, it is recommended that salt rinses be used for basic oral health care and maintenance, and cannot replace the conventional toothbrushing and oral hygiene practices [18,19]. In India, people strongly believe in the indigenous or traditional health practices that use herbs and herbal products for the treatment of various diseases. Ayurvedic medicine is quite popular and is considered safe as it is believed to contain natural ingredients. Mango, cashew leaves and chewing sticks obtained from the stem of neem trees are popular alternatives for toothbrushes. They have been shown to possess anti-microbial properties, effective against plaque and caries, and their extracts used in toothpastes and mouthwashes [20,21]. However, chewing sticks have caused increased tooth attrition and gingival recession which has been attributed to their improper use [22]. Though there is some evidence of the effectiveness of these products in the general population, further studies need to be conducted to determine the extent of their effectiveness in PLWHIV. There is also a strong traditional belief that certain drugs and dietary items such as certain vegetables, fruits, meat and spices produce heat as an end result of digestion. This heat is called 'ushna' and is believed to cause mouth ulcers, skin ailments and digestive problems [23]. Participants in our study believed that the antiretroviral drugs increase 'ushna' that causes mouth ulcers, and therefore took measures to counter it by consuming fluids that would 'cool' the body. They however made no other modifications to their diet. The perception of the benefits of regular dental visits was deemed very low, and would be considered only if they experienced pain or discomfort as minor issues could be self-managed. Several studies in the general population have also reported a low frequency of dental visits, mainly attributed to a lack of perceived need, cost of treatment and also fear of dental procedures [24,25]. There were certain barriers to oral health care. Financial barrier has been reported as a major factor for unmet oral health needs even in the general population. Similarly, our participants highlighted financial constraints as a deterrent to oral health care. The participants in our study belonged to the lower socioeconomic class and a few were retired. They lacked health insurance and were unable to meet the expenses of treatment. Personal factors such as loneliness, worry, low level of self-confidence and dependency on others to travel also acted as barriers. Other studies have reported transportation difficulties, cumbersome administrative procedures, long waits at the dental office and problem focused care-seeking behavior as additional barriers [25][26][27]. Few participants in our study were ignorant of the location of dental clinics. They seemed unaware of the presence of a dental clinic in the hospital where the ART Centre is located, suggesting the importance of the role of the healthcare workers at the Centre in educating patients and enabling access to dental care. The fear of disclosure seems to be a universal issue, mainly due to stigma, dentists' behavior and prejudice, and ignorance towards the clinical implications of the disease [28][29][30]. Majority of our participants stated that they would not reveal their HIV status to the dentists due to fear of discrimination, refusal of services and confidentiality. Moreover, they felt that there was no risk of transmission due to the general precautions taken by the dentists. This has clear implications for the dentists as several studies have reported inadequate compliance to cross infection control protocols in private dental clinics and dental laboratories [31][32][33][34]. It is important that dental professionals understand the perceived barriers and take certain measures so that patients are brought into and maintained under their care. A revision in the dental curriculum is necessary, wherein dental students receive training to enhance social awareness, cultural sensitivity and communication techniques, mainly to curb discriminatory practices and to allay fears and anxiety of the patients. Also, cross infection control should be universally employed for all patients in the dental setting to prevent discrimination. Counseling, sedation could be employed by dentists in their practice to mitigate anxiety. The congenial attitudes of the dental staff who value privacy and offer encouragement and respect serve as motivators for healthy oral health practices [35]. Interventions for PLWHIV should be developed to address dental related fears. Studies have shown that educational interventions are extremely effective in improving health literacy and promoting preventive oral health behaviors in the target populations such as adolescents, diabetic patients and pregnant women [36][37][38]. In the current technology driven era, mobile phone applications can serve as attractive means for oral health promotion. A study on the perceptions of an oral health mobile app for oral hygiene routine using animated videos reported that it is a promising tool for patient motivation and education [39]. Further research on development and effectiveness of such technology driven interventions to include behavior modification for PLWHIV, especially from the lower socioeconomic groups need to be explored. Our study sample mainly included patients belonging to the lower socioeconomic strata and had a poor perception of oral health. It was observed that the participants belonging to a higher educational background were better equipped with oral health knowledge and had more favorable oral health beliefs and behaviors as compared to the other participants. There exists a negative association between the socioeconomic level and oral and dental disease as people with a higher income and educational background are more likely to regularly brush teeth, use supplementary dental aids and have regular professional dental visits [40]. Socioeconomic inequalities in oral health are known to be associated with the social hierarchy, with oral health deteriorating towards the lower end of the socioeconomic spectrum [41][42][43]. The findings of our study call for an urgent need for interventions to bring about and sustain positive changes in the beliefs and behaviors towards oral health in PLWHIV. It is recommended that oral health should be integrated with primary health care to facilitate oral health promotion and improve access to oral health services [44]. Health care polices that include attitude and behavior modifications for better oral health should be developed and implemented. Patient education materials in the form of brochures, pamphlets and posters should be made available at ART Centres. Additionally, information could be extended through mobile phones and regular health awareness programs. The professional and lay people involved in providing care to PLWHIV could be instrumental in promoting oral health and should be trained in both general and oral health care, so that they not only create awareness but also refer patients for professional dental care [45]. Initiatives by the government aimed at improving access to oral health care are a prime requisite especially in low and middle income settings, not only for PLWHIV but also for the general population. --- Strengths and limitations Though we only interviewed patients who visited the ART Centre, and belonged to the lower socioeconomic category, the results are likely to reflect the oral health perceptions of PLWHIV at most ART Centres in the State or country, enabling transferability of the results. Further, the context of the study is extensively described in the introduction, enabling the reader judge the transferability of the results. A detailed description of the study procedures supports the dependability of the results. To ensure credibility of the findings, the researchers discussed the interviews as well as the results at different stages of the analysis and resolved conflicting interpretations of the interviews. This also ensured that the results are credible even though the interviews were not transcribed by a professional agency due to limited funds. The first author, understanding her role as a researcher and a dental professional, as well as her awareness of the sociocultural background of the participants, supported the contextualization of participant responses as well as their interpretation. Also, the team of researchers involved are from both the public and private healthcare sectors in the country with experience in HIV treatment, policy development, program implementation, as well as research, which strengthens the interpretation of participant responses. Despite this, we consider that further studies may be required to explore the oral health beliefs and behaviors among PLWHIV with different cultural, regional and socioeconomic backgrounds in the Indian context due to its diversity. --- Conclusion This qualitative study explored the oral health beliefs and behaviors of PLWHIV. The study revealed an alarming lack of oral health awareness and unfavorable oral health behaviors. The findings from this study reveal the utmost necessity of empowering these patients with awareness on the importance of oral health, preventive measures and regular dental visits. There are also implications for health professionals, who need to provide a more holistic approach to patient management as well as develop a rapport to enhance confidence and allay fears and stigma. Also, oral health care should be made more accessible and affordable by its integration into primary health, especially in low and middle income settings. --- Availability of data and materials The data generated from this study are not publicly available, as confidentiality of the participants has to be maintained, but are available from the corresponding author with permission from the Karnataka State AIDS Prevention Society. --- Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12903-021-01549-5. --- Additional file 1. Interview guide based on the constructs of the Health Belief Model. --- Authors' contributions AR and VH were involved in concept, design, data collection, and drafting the manuscript. AVH was involved in data collection. AR and RR were involved in analysis of the data. RR, SGS and DNR were involved in the critical review. All authors read and approved the final manuscript. --- Declarations Ethics approval and consent to participate An approval for the study was obtained from the Karnataka State AIDS Prevention Society (KSAPS/CST/02/2018-19), and the Institutional Ethics Committee, A J Institute of Medical Sciences and Research Centre, (AIEC/REV/210/2018). All methods in the study were performed in accordance with the relevant guidelines and regulations. An informed consent was obtained from the participants prior to enrolment in the study. Verbal and written information regarding the aim, procedure and confidentiality was given to all participants. --- Consent for publication An informed consent for publication has been obtained from the participants. The Institutional consent form was used for this purpose. --- Competing interests The authors declare that they have no competing interests. --- Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Introduction: People living with HIV (PLWHIV) have increased oral healthcare needs due to an increase in the prevalence of oral and dental diseases. Their oral health is influenced by psychosocial, behavioral and biologic factors. The aim of this study was to explore and obtain a deeper understanding of the oral health beliefs and behaviors of PLWHIV that could potentially affect their oral health.We have used the Health Belief Model (HBM) and qualitative methods using in-depth interviews with 16 PLWHIV. Content analysis of the transcribed data was done. The data was grouped under the constructs of the HBM.The perceived susceptibility to oral diseases and awareness on the importance of good oral health was low. Regular tooth brushing and traditional methods for oral hygiene maintenance were considered beneficial. Regular dental visits were not considered important. Psychosocial issues, time and financial constraints were the barriers. Participants believed that information on oral health should be provided by the health providers in hospitals and dental clinics.The findings on the oral health beliefs and behaviors support the need for education on oral health and preventive healthcare practices among PLWHIV. Oral health promotion should include behavioral change as one of its components.

Introduction Over the past few decades, Russian society has undergone drastic changes, which have significantly influenced the role and format of youth leisure. It is important to rethink approaches to the study of leisure practices. Leisure activities are one of the most important areas of selfrealization, self-actualization and an integral component of a person's lifestyle; without leisure the full socialization of the individual is impossible here is much research devoted to the study of leisure, in sociology and in other fields (Best 2009). The study of the leisure time of secondary school students and university students is relevant, because these social groups are more influenced by innovations than others and have a more flexible value system and leisure largely determines their degree of satisfaction with life. --- How Youth Leisure is changing? The dramatic changes that have taken place in Russian society over the past few decades have significantly influenced the place and role of leisure in the socialization of younger generations. This necessitates a rethinking of the approaches to the study of leisure practices, taking into account the modern socio-cultural situation. For most young people, immersion in the media space is an integral part of spending their free time. Young people actively consume content presented on the Internet, are fond of computer games, computer game streams are very popular too. Also youth communicate using social networks (Twitter, TiкTok, Instagram etc.), and also create trends for certain types of recreation, representing their leisure in the digital space. Leading theorists of sociology considered leisure activities in their works. For example, Marx developed a dialectical concept for a more detailed description of the concept of free time: "Free time, which is both leisure and time for more sublime activity, of course, turns the one who possesses it into another subject" (Marx, 1969). Bourdieu emphasized the significant difference in the leisure practices of people belonging to different social strata: the more privileged class is characterized by completely different forms of leisure activities than the provincial bourgeoisie or representatives of the lower classes. He obtained empirical data showing the social structuring of consumption, including leisure forms (Bourdieu, 2005). Kaplan most fully revealed the concept of leisure and developed several models (Kaplan, 1962). Parker considers leisure activity an important condition for the inclusion of the young generation in the sociocultural life of society and identifies such features of leisure activity as choice, flexibility, spontaneity and self-determination (Parker, 1976). Parsons studied the topic of leisure practices in the problem of relations between generations. Sh. Best, presents leisure as a kind of social context within which the self-determination of an individual takes place (Best, 2009). Dumazedier introduced the concept of "the civilization of leisure". His idea is based on leisure not only as a centre of modern cultural practices, but also as a central component of knowledge about a society and its culture (Yaroshenko, 2019;Dumazerdier, 1962). Johns speaks of the free pastime of citizens as one of the important aspects of studying the social life of contemporary Britain (Johns, 1979). Arnold claims there is a close relationship between active leisure and a successful human career (Arnold, 2020). Leisure practices were also the subject of study for many Russian sociologists. Zborovskiy considered leisure (or free time) in close conjunction with culture, which "acts as the content of free time, it's certain "filling" (Zborovskiy, 2006). Lisovskiy, Dobrynina, and Ikonnikova studied the lifestyle and cultural values of modern youth and their influence on the socialization of the younger generation (Lisovskiy, 2000;Dobrynina, 2017). Surtaev examines the leisure practices of young people, and the culture of leisure is presented as "a measure of the realization of the sociocultural potential of an individual in conditions of leisure activity, a measure of the skills she has acquired to regulate leisure time, the readiness of an individual to participate in socially significant types of leisure activity" (Surtaev, 1998). Leisure practices and the impact of their orientation on the professional activities of citizens were studied by Temeneva (2005). Fatov considered the leisure activities of young people in the context of social education; he proposed ways to solve the problem of "the effective use of the educational potential of cultural and leisure activities in the educational system of universities" (Fatov, 2007). Ponukalina and Pogreshaeva studied leisure as a criterion of a person's quality of life and an indicator of his or her psychological state (Ponukalina, 2011;Pogreshaeva, 2012). Drobinskaya studied the problem of the rational use of free time by young people and considered the dependence of personality development on how this time passes (Drobinskaya, 2012). Orlov defined leisure as "a set of human activities in free time, with the help of which immediate physical, mental and spiritual needs are satisfied, mainly of a restorative nature, as well as a specific social method for the further development of human forces" (Orlov, 1983) Currently, there are many different scientific approaches to understanding the essence of leisure, but a single definition of the concept of "leisure" is still missing. In this work, we define leisure as free time, not busy with work or study, activities undertaken for pleasure, entertainment, selfimprovement or the achievement of other goals, and not because of material --- Methods and data The empirical base is a number of studies conducted by the authors on the leisure practices of young people. One of the objectives of the research was to identify the role and place of digital technologies in the structure of the young generation's leisure time. An important part of the research was the annual study (from 2003 to 2009) of the opinion of students of universities in Chelyabinsk Region on various aspects of their life, including leisure practices. The sample types were multistage, serial (university, course and group) and were as follows: September-October cities and districts of Chelyabinsk Region. The methods of collecting primary information were: a questionnaire survey, in-depth interviews, and document analysis. In January-March 2020, the authors conducted indepth interviews with 300 people aged 16 to 30. The informants were young people living in cities of Chelyabinsk Region (Kopeysk, Korkino, Ozersk, Zlatoust and Yuzhnouralsk). The sample for the survey was formed by the method of available cases. The authors also carried out a frame analysis of cases of youth leisure practices. In total, in 2020, 28 cases were analysed on the daily leisure practices of youth in Kopeysk, Korkino, Ozersk, Zlatoust and Yuzhnouralsk. The respondents described their traditional leisure activities, other participants in leisure practices, their attitude to various types of leisure, the possibility of access to certain types of leisure activities. When analysing situations, the methodology of frame analysis made it possible to comprehensively study the specifics of youth leisure in small and medium-sized cities of the region. From December 2019 to March 2020, the authors carried out a content analysis the accounts of social networks of youth in the cities of Kopeysk, Korkino, Yuzhnouralsk and Zlatoust. In total, 150 accounts of young people were analysed. The methodological basis of the author's research was an interdisciplinary approach, which made it possible to consider the problem of leisure practices of all categories of modern Russian youth from various positions, relying on the methods and capabilities of related humanitarian disciplines: sociology, cultural studies, pedagogy, and philosophy. --- The Structure of Cultural and Leisure Activities of Young People Leisure meets the recreational communicative and entertainment needs of a person. The research, conducted over several years, allowed us to identify the features and content of leisure practices of modern students, pupils, working and non-working youth, to trace how their preferences changed and to identify the place and role of the media space in leisure practices. The survey in 2020 identified how much free time young people have. The results of the survey in Kopeysk, Korkino, Ozersk, Zlatoust and Yuzhnouralsk showed that modern youth have a lot of time for leisure. 41.4% of young people aged 16 to 30 had from 2 to 5 hours a day of free time and more than 25% of respondents have from 5 to 8 hours of free time. Only 13% of respondents have less than two hours of leisure time per day. The content of leisure time is perceived by each young person in their own way. Some prefer to spend it with usefully, others exclusively for entertainment. Today's youth are generally satisfied with the conditions of their leisure time. This was confirmed by the studies that we conducted among the student youth of Chelyabinsk Region in the period from 2003 to 2007. Almost 50% of the respondents in different years noted positive changes in this area. The proportion of students who evaluate their leisure time highly increased with each year of the survey (44% in 2003, 47% in 2005 and 49% in 2007). During the study in 2007-2008 (the students were asked the same questions, and the sampling and data collection methods were the same. We were interested in what young people like to do in their free time and what they actually do. The answers to these questions allow us to understand how young people manage their free time. The results clearly demonstrate the desire of modern youth to spend their free time in the media space. The results indicate that most students prefer to listen to music and watch TV. Almost 50% like to play sports. In more detail, the preferences of young people for leisure are presented in Table 1. University students, secondary school students and students of technical schools like to listen to music in their free time a little more than university students. Most of the respondents like sports, but only 25% are actually engaged in them. 36% of university students like sports, which is slightly fewer than among other students. Watching TV is the third most popular leisure practices among secondary school students, students of technical schools and university students. About 65% of students like to study something new. 50% of university students like to visit museums, theatres and cinemas. About 33% of university students like to read (books, newspapers or magazines) and almost the same proportion among There are several more of them among students of technical schools than among other groups of young people. <unk> lot of teenagers like spend free time doing social work, go to Internet cafes, and secondary school students prefer just to hang out. Among other answers included "sitting at the computer", "photo editing, Web design", "playing computer games, network", "programming", "walking and spending time with friends", "I like to write and compose rap, poetry", "talking on the phone", "walking, having fun", "chilling with friends". The opinion of experts on the leisure practices of students and pupils generally coincides with the answers of the youth themselves. Experts note that modern young people most like to listen to music and play sports in their free time (this was indicated by the vast majority of experts); the second preference was to different hobbies, third, watching TV. Experts from the field of higher education more often than the youth themselves believe that students like to work in their free time according to 61% of the university experts. --- Youth TV Preferences Currently, television occupies a significant place among the means of spending leisure time for all categories of the population, including young people. Our research in 2020 showed that among the television preferences of modern youth, entertainment shows and television series are leading (40% and 36.4%, respectively). Half of the girls surveyed chose entertaining TV shows as their favourite type of TV viewing. Among young men, this figure is only slightly more than 30%. Also, more than half of the girls spend their leisure time watching television series, while sports TV shows are more popular among boys. Feature films are in third place in popularity -25.5% of respondents are ready to watch TV for them ( Every fourth of those who answered, as it turned out, does not watch TV at all (25%). Apparently this is due to the fact that your favourite movie or program can be watched on the Internet, moreover, viewing is not interrupted by advertising and you can choose a convenient time. Moreover, the number of those who do not watch TV, in principle, among boys is almost three times higher than among girls. Almost every fifth of the respondents watches the news, which indicates interest in the events taking place around them, although this indicator is not as high as we would like. However, it can be assumed that the share of those interested in what is happening in the country and in the world is much higher, just the majority of young people prefer to receive information from the Internet. Political programming and TV concerts turned out to be one of the most unpopular television genres (4% for each option). The author's programs were chosen for viewing by only one person from the whole set. Considering television genres in the preferences of various social groups, one should pay attention to the fact that the most popular television genre according to the survey results -entertainment television shows -is the least popular among students. This group of young people mainly prefers to watch television series. And such a genre as feature films is most popular among working people. It is interesting to note that among the unemployed category of youth, there is a high demand for viewing the news. In other groups, people are less interested in what is happening. Most likely, due to the large amount of free time, this category of people manages to monitor the environment in the world. Comparing the results of our research with the studies of television viewing of young people in other cities of Russia, we saw that the television preferences of young people in different cities differ quite seriously. In part, this is due to the different wording and the suggested choice of answers. For example, the genre of television series was indicated in the list of answers in our study of the leisure time of young people in small towns in the Chelyabinsk region, in contrast to other research works. And this answer option was chosen by a fairly large number of young people -36% of the respondents. --- Digital Technologies in Organizing and Conducting Youth Leisure There are a lot to new forms of leisure activities especially among young people. In 2020, the answer to the question: "What do you prefer to do in your free time?" was: "I surf the Internet" was the leading one. This was indicated by 70% of respondents, "sitting on the internet" meant any activity not related to school or work: using social networks, chatting, watching videos, etc. The results of the survey in 2020 showed that secondary school students and students prefer to spend time on the Internet. In the 2000s, the popularity of the Internet in the structure of youth leisure was not yet very high: in 2003, only 20% of respondents spent time on the Internet, and by 2007 the share was 41%. In a 2009 study of young students of (Aryamova, 2012). The significantly lower proportion of young people who spent time on the Internet a decade and a half ago is primarily due to the lower availability of this form of leisure. In addition to spending free time on the Internet, young people use the digital space to get information about the possibilities of organizing their leisure time. Thus, it was the Internet that turned out to be the most popular source of information on cultural and leisure activities in small towns of the Southern Urals (Zlatoust, Kopeysk, Korkino, Yuzhnouralsk). More than 50% of respondents learn information from the Internet (websites, social networks). The second most popular is advice from acquaintances, friends, and relatives of informants (48%). Comparing the results obtained with the results of studies of students in the Chelyabinsk region, which were carried out by the authors in the first decade of the 21st century, showed that the popularity of the Internet in the structure of youth leisure was not so high, but even then it had an undoubted tendency to increase. So, in 2003, only every fifth respondent spent time on the Internet, and by 2007 the share of those had increased to 41%. And according to a study conducted by the authors within the framework of monitoring in 2009 among students of the Chelyabinsk region, only 18% of schoolchildren and 27% of students in secondary specialized institutions spent their time on the Internet. In 2007-2008. The teenagers' visit to Internet cafes did not arouse much interest either. In the general ranking of favourite activities in free time, visiting Internet cafes is in 13th place. An insignificant share of the respondents (4.8% of students, 4.1% of schoolchildren, 4.0% of college students and 3.5% of college students) indicated that they spent their free time there. It is noteworthy that young men are more interested in Internet cafes than girls, respondents aged 15-16. According to our survey, information about current cultural events taking place in their city is generally available to everyone, but more than 40% of respondents noted that such information does not always reach them. When analysing the cases, we found out that young people would like to receive information about events in their city more quickly, in addition, the channels of information dissemination caused dissatisfaction: "often information about city events is made by sticking up [advertising posters on advertisement stands], I don't look...". Based on the answers received, we can conclude that information channels in the city are not working effectively enough. Nowadays, it is typical for young people to regularly visit social networks. Thus, according to a nationwide sociological survey conducted by the Institute of Sociology of the Russian Academy of Sciences in 2018, the overwhelming majority (91.8%) of young Russians communicate on social networks, visiting them on average 269 days a year, that is, at least five days a week (Gorshkov, 2019). In addition to communicating on social networks, young people reflect their leisure in them: they take photos during their holidays and post them on Instagram, VKontakte, noting friends and the geolocation of the institution. We monitored 100 accounts of young South Urals young residents on Instagram, since it is easier to make a selection of publications on this social network than on others. Particular attention was paid to geolocation tags, which marked the places where the photo was taken. As a result of monitoring, a tendency was revealed -to spend their free time in another city. For example, cities such as Kopeysk and Korkino are satellite cities of Chelyabinsk, except for trips to work or study, young people from these cities travel to Chelyabinsk for their leisure time: go to a concert, walk around the city centre, visit a cafe, etc., making posts about this in social networks, as a rule, posts about leisure in Chelyabinsk appear from Friday to Sunday. Young residents of Yuzhnouralsk and Zlatoust do not go to Chelyabinsk for the purpose of rest so often (taking into account the distance), they try to combine the trip with other things. For example, when analysing the cases, one of the informants said that he travels to Chelyabinsk to a doctor, at the same time he definitely devotes time to rest "I go to a cafe to sit, there is more choice or to a shopping centre". The second tendency revealed in the course of the analysis was the spread of a relatively new form of leisure: the rest of schoolchildren in shopping and entertainment centres (trade centres, Malls) Moreover, such leisure is not always associated with shopping. Schoolchildren most often noted themselves in geolocation at food courts in the shopping mall. During the in-depth interviews, we found out that a large number of teenagers gather in the malls, and many of them spend their time aimlessly. Informants reported that they do not always have money to buy food in food courts, but, nevertheless, young people gather here in companies, despite the emerging conflicts with food court security services. Among the student youth in the geolocations of Instagram, institutions were also in the lead, where you can not only sit, drink alcohol, but also dance. --- How do Young People Rest More? The second most popular way to spend leisure time was walking with friends. Among secondary school students 79.2%, students 65.7%, working students 73.1%. A frame analysis of the cases showed that walking with friends for secondary school students is not only a form of entertainment, but also "a way to relax from parents and school". According to the informant, (16 years old): "at home I am controlled by my parents, at school by teachers, and only on the street I can be myself, forget about studying". Despite the busy schedule of study, work and other types of employment, young people have time to devote it to their friends. It can be assumed that it is extremely important for young people to communicate with their friends, to realize their need for social recognition. Looking at the choice of people with whom respondents prefer to spend their free time, we see that most of them choose to spend time with friends. Almost 25% of respondents like spending time with their family and relatives. 10% prefer to spend time alone. The most undesirable choice was spending time with work colleagues or classmates -5%. Females prefer to relax with relatives much more often than males, more than 40% of the surveyed females chose this option, and males much more often choose to spend their free time with friends. Working youth prefer to spend time with family or relatives (67%). A significant part of the respondents (primarily young students) devote their free time to sports. Among secondary school students, sports was chosen by 45.8%. Non-working young people are the least active in sports (only 15.1% of respondents). We believe this is due to the fact that sports activities require financial expenses (club membership, the purchase of equipment etc.). However, in comparison the results of 2020 with the results of surveys of students in 2003-2009, the share of young people engaged in physical culture and sports has significantly decreased. Among the students surveyed in those years, the vast majority did this regularly or occasionally. The study showed that reading was also popular. In the questionnaire, we clarified that we mean reading as a form of recreation, which implies an independent choice of books. 37.5% of secondary school students and 29.6% of working young people read "for fun". Students read the least (15.7%). Students explained that there is little time for reading due to study. Sports as a leisure activity are more common among male respondents, and reading is more common among females. We believe that in general, the popularity of sport is a positive trend. Our assumption that working young people most often have the opportunity to relax in cafes, bars and restaurants was confirmed -more than 25% of working young people chose this format of leisure. Schoolchildren (4.2%) and students (8.6%) rarely choose this type of recreation. We believe that this is due to the small amount of money that young people have. Schoolchildren said that going to a cafe with friends is leisure, and with parents "this is different, it's for eating, nothing special", showing how the perception and interpretation of the same action (visiting a cafe) differs depending on the company (parents or friends). Approximately the same situation has developed with visiting night clubs -this type of recreation was chosen mainly by working students (23.1%), but working young people are less likely to visit such institutions (9%). According to one of the informants (female 25 years old), "when I was studying in college, I constantly went dancing, now I have a family, a child, what clubs?". Even those respondents who do not have not family responsibilities explain the desire for a quiet leisure activities by the fact that they are tired from work and spend a lot of time on household chores. The lack of interest in night clubs among secondary school students and students (0%) can be explained by age restrictions on visiting such entertainment venues. 20% of secondary school students chose to walk alone as a leisure activity. One of the informants said (female, 17 years old) "I like to go out with my friends, but sometimes I like to go out alone, think, especially if there are problems at school or I have a fight with someone". Almost 35% of working students and 30% of non-working youth chose cinema as a leisure activity. Since the research was conducted in small cities, the types of leisure options are limited. Film premieres almost coincide with the dates in larger cities, and tickets are affordable, so even unemployed young people can afford this type of leisure. Creative work is most often done by working young men and young women (13%). "After work, I go to rehearsals, my friends and I have our own music group, we play all sorts of things, just like that, not for money or fame, just like it" (male, 27 years old). The least popular options for them were gambling and board games, participation in social activities and visiting theatres and museums. When analysing the factors that influence the choice of the form and place of leisure for young people, it turned out that the proximity of the place of their recreation to their place of residence is not a significant factor, perhaps this is due to the fact that the survey was conducted in small cities, in which recreation places are mainly concentrated in the central part of the city, and it is quite easy to reach by public transport. A curious trend was that for residents living in the central part of the city, the factor of the distance of the place of leisure from home is more significant than for residents of the suburbs and towns. Since the main array of entertainment facilities is located in the central part of the city, this criterion should be more significant for people living far from the centre than for the population of the central part of the city. Another important criterion for choosing a leisure activity is the price. Its importance among the unemployed and working people is due to the fact that the first category has limited resources and can't afford large expenses, and the second category devotes most of its time to making money and is aware of the cost. Among secondary school students, the fewest respondents agreed with the importance of prices for leisure activities. This is easily explained by the fact that they do not earn money themselves, they get money for entertainment from their parents, and Almost 20% of the respondents believe that advertising has a significant impact on the choice of a place for their leisure activities. About 40% of respondents recognized the importance of having promotions / bonuses at leisure venues. The share of people who recognize the importance of this factor is much higher among the female respondents (58%). Only half as many males share this view. --- Conclusion The results revealed that young people now, no less than at the beginning of the 2000s, feel the need for recreation, entertainment, and communication that is not related to work or study. Most young people have enough for them free time during the day. The function of leisure as an opportunity for self-development is losing its popularity among young people, and the desire for entertainment, getting momentary pleasure becomes the basis for leisure activity. The modern young generation often prefers passive forms of leisure. Television is no longer one of the primary forms of leisure, activities in the media space, including social networks, the internet in general and computer games are replacing it. In our opinion, this is a natural process, since the digitalization of society, which changes the worldview and values of people, especially affects the lifestyle of the younger generation, which is more susceptible to external influences than other age categories. In order to assess the positive and negative consequences of this process, it is necessary to conduct a separate study. Our research also revealed a number of problematic areas related to the organization of leisure activities for young people in small cities of the Southern Urals. It is important to improve the quality and accessibility of such forms of cultural leisure as museums and theatres, and this should become one of the leading directions of regional policy and local governments. Thanks to the constant sociological monitoring of youth leisure, satisfaction of leisure activities, it is possible to solve problems arising in the field of organizing cultural recreation in a timely manner.

This article presents the results of research on the impact of digitalization on the leisure of modern youth. The factors in the transformation of leisure activities, the role of leisure practices in the lives of modern Russian youth are revealed. Our conclusions are based on data from the studies independently conducted by the authors from 2003 to 2020. About 2,000 respondents were survived annually. Quantitative survey data were supplemented by the results of interviews with experts, the youth of the small and medium-sized cities of Chelyabinsk Region, frame analysis of case studies and the content analysis of the social networking accounts of young people, secondary analysis and document analysis. The analysis established common practices of youth leisure, identified the impact of digitalization on the changing shape of the leisure activities and identified the destructive effects of the Internet on the leisure of today's youth of the South Urals.

Monadology and ethnography: Towards a Tardian Monadic Ethnography The restoration of sociologist, criminologist, social psychologist and judge Gabriel Tarde to the pantheon of social theory seems to be steadily under way. Almost forgotten after succumbing to Émile Durkheim as the leading figure of French sociology, his work has recently enjoyed a remarkable comeback. Very little attention had been paid to Tarde's work until the French re-edition, from 1999 onwards, of most of his work under the direction of Éric Alliez. Since then, Tarde has been praised by Bruno Latour (2002;2005) as the precursor to Actor-Network-Theory; taken up by Maurizio Lazzarato (2004) as the foundation for a new politics of multiplicity; his great controversy with Durkheim reconstructed (Viana Vargas et al., 2008) and even re-enacted; 1 and his work been the subject of at least one conference 2, special issues in various academic journals 3, and an edited book (Candea, 2010a). The recent translation into English of Tarde's 1895 masterpiece Monadologie et Sociologie (Tarde, 2012), 'the most metaphysical of the works of the most philosophical of sociologists' (Alliez, 1999;cited in Lorenc, 2012), adds an important piece to this revival. Tarde's monadology -which radically reconfigures the Leibnizian ontology it is based upon -is arguably one of his most remarkable theoretical contributions. The aim of this article is to explore its far-reaching implications for the conceptualisation and study of the social, and to relate those to the practice of ethnographic research. The article is organised as follows: in the next section, Gabriel Tarde's rediscovery and some of his key ideas will be briefly discussed. Then, his monadology will be presented in some detail. This is followed by an ethnographic interlude, which will be used in turn to inform the'monadic ethnography' proposal developed in the following section. The article concludes with a wider consideration of Tarde's interest for ethnography in relation to recent debates around a so-called 'ontological turn' in the social sciences. --- Rediscovering Gabriel Tarde A good part of the recent literature on Tarde revisits his intellectual dispute with Émile Durkheim over the nature and scope of sociology (e.g. Alliez, 2004;Candea, 2010a;Latour, 2002;2005;Mucchielli, 2000;Viana Vargas, 1998). It was, we are told, a foundational moment for the discipline. Durkheim's victory would not only explain sociology's becoming over the 20 th century, but also the strangeness that reading Tarde produces today and his near disappearance from the history of social theory. For those who sympathise with his arguments, Tarde represents the path not taken, and going back to his work the possibility of plotting a line of flight from sociology (and indeed the social sciences) as we know it today (Candea, 2010b). This new wave of interest in Tarde has proved fruitful in reactivating several of his intellectual endeavours, including a statistics of the fluctuations of belief and desire (Barry, 2010;Didier, 2010); an 'economic psychology' or'science of passionate interests' (Latour and Lépinay, 2009); the theorisation of 'contagion'and 'virality' in relation to media (Sampson, 2012); or a monadic'sociology of associations' (Latour, 2005;Latour et al. 2012). In this article, I will avoid the controversy with Durkheim, already well discussed in the literature cited above, and focus instead on Tarde's monadology, as seen from an ethnographer's point of view. In doing this, I wish to continue a line of work advanced by Matei Candea (2010b) and Eduardo Viana Vargas (2010), who already identified the specific potential of Tarde's monadological hypothesis for ethnographic research. More broadly, the'monadic ethnography' proposal outlined in this article is closely related to assemblage theory and actor-network theory (ANT hereafter), which is the reason why I want to engage in a partial introduction to the work of Tarde via Deleuze and Latour. 4 In 1968 Gilles Deleuze published Différence et Répétition -a book whose title already indicates an unequivocal Tardian filiation. In a long footnote, he praises Tarde's work and concludes: 'All of Tarde's philosophy may be presented... as a dialectic of difference and repetition which founds the possibility of a microsociology upon a whole cosmology' (Deleuze, 1994: 314). There are two focal points of interest here. The first relates to Tarde's thesis that repetition, adaptation and opposition are -in that order -the fundamental dynamics that govern all phenomena (Tarde, 2000). In the case of human societies, he argued that this triad took the specific form of imitation-innovation-opposition, the idea being that it is through and by repetition (imitation) that difference is produced and distributed (as either innovations or oppositions):...it is because certain social phenomena, such as a dogma, phrase, scientific principle, moral maxim, prayer, industrial process, or the like, tend to spread in a geometrical ratio by imitative repetition, that they interfere with one another in a felicitous or infelicitous manner. That is, the discordant sides of their nature come together in certain minds, giving rise to logical or teleological duels, which constitute first germs of social oppositions (wars, competitions, and polemics); while the harmonious sides of their nature come together in the mind of the genius, or sometimes even in the ordinary mind, producing true logical syntheses, inventions, and fruitful originations, which are the source of all social adaptation. (Tarde, 2000: 63-64) The quote above -apart from a fitting introduction to Tarde's peculiar writing style -hints at Tarde's 'infinitesimal' or'molecular' approach, the second focal point of interest for Deleuze, further developed with Félix Guattari in A Thousand Plateaus. Tarde's analyses insist in tracing those imitations, oppositions and innovations back to their origin in the infinitely small, where the key to the big and regular lies. As he put it: '...instead of thus explaining lesser facts by greater, and the part by the whole, I explain collective resemblances of the whole by the massing together of minute elementary acts -the greater by the lesser and the whole by the part'(2000: 35). Deleuze and Guattari crucially argued that this form of'microsociology' was not, as Durkheim claimed, devoted to the individual, but to 'flows' or 'waves' instead: Imitation is the propagation of a flow; opposition is binarization, the making binary of flows; invention is a conjugation or connection of different flows. What, according to Tarde, is a flow? It is belief or desire (the two aspects of every assemblage)... [In] the "molecular" realm of beliefs and desires... the distinction between the social and the individual loses all meaning since flows are neither attributable to individuals nor overcodable by collective signifiers. (Deleuze and Guattari, 1989: 241, emphasis in original). I will come back to the issue of 'belief' and 'desire' and the individual/collective dualism below, in relation to Tarde's monadology. But before doing so, I will end this section with a brief introduction to Bruno Latour's engagement with Tarde. In a series of key texts (2002; 2005; 2010; 2012), Latour has reclaimed Tarde as a precursor to ANT, on the grounds that he had anticipated -by almost 100 years -the idea of a'sociology of association': [Tarde] vigorously maintained that the social was not a special domain of reality but a principle of connections; that there was no reason to separate "the social" from other associations like biological organisms or even atoms... Above all, he considered the social as a circulating fluid that should be followed by new methods and not a specific type of organism. (Latour, 2005: 13) Because of this general principle of society as association, Tarde's sociology ignored the nature/society divide as well as the micro/macro distinction -two of ANT's most important contributions to social theory. Regarding the former, Tarde argued that society is a universal form of association, shared between human and non-human entities, and hence highlighted the inadequacy of starting from a divide between natural and social facts à la Durkheim. With regards to the micro/macro distinction, Tarde's relational and infinitesimal approach invalidates it to a large extent, in the sense that the'small' reveals itself to be richer and more complex than the 'big', which in fact is a mere 'extension' of some of its components (Latour, 2002). I will now turn to Tarde's monadology, where these counterintuitive positions are explained in detail. --- Gabriel Tarde's monadology 'The monads, children of Leibniz, have come a long way since their birth' (Tarde, 2012: 5). Monadology and Sociology's (hereafter MS) first sentence already introduces an important caveat: these monads are rather different from Leibniz's. The term monad, from the Greek monas, a 'unit' or a 'one', has a long tradition in philosophical writings, dating back to the Pythagoreans, but it was Leibniz who popularised it in the early 1700s. In the modern sense it is used to describe: '(1) a simple, irreducible, and sometimes indestructible entity; and (2) the minimal unity into which the cosmos and all composite things in it can be resolved; yet (3) containing within itself, in contrast to material atoms, powers and relations of which it is itself the source' (Loemker, 2006: 324). Traditionally, a monadology is a'metaphysical system that interprets the world as a harmonious unity encompassing a plurality of such selfdetermining simple entities' (Loemker, 2006: 324). In other words, classic monadologies entail a theory of cosmic harmony, based on the coordination of a multiplicity of monads, themselves considered active substances. In Leibniz, monads are differentiated and differentiating, but also closed ('monads have no windows', as he famously put it [1898: 219]) and subject to a higher principle of harmony ensured by the existence of the supreme Monad, or God (see Viana Vargas, 1998). Tarde's monadology departs from this philosophical tradition in key ways. It is, first of all, secular; there is no God and no pre-determined harmony. It is, also, an open and infinitesimal system, where monads affect each other and are themselves composite entities. MS's central hypothesis (p. 28) is that 'everything is a society, that every phenomenon is a social fact': from cells to stars, both living things and inorganic beings, they are all societieswhere'social' and'society' refer, as Latour indicated, to a form of association rather than any particular entity. It is worth examining how he arrives at this remarkable inversion of both Spencer's'society as organism' thesis (1906) and Durkheim's first rule of the sociological method, 'to consider social facts as things'(1982). First, Tarde argues that reality is made out of monads, which, contrary to Leibniz, are not the simple elements that make up aggregates, but composite entities themselves. Drawing from findings in physics, cellular theory and chemistry, he argues that science has repeatedly shown how so-called individual bodies (e.g. planets, cells, parasites) were in fact 'a multiplicity of distinct elements linked to each other in the same way as they are linked to the elements of other aggregates' (MS,6). It is only because we look at things from a distance and they are unknown to us that we believe them to be 'indistinct, undifferentiated, and homogenous... But everywhere where a scientist digs beneath the indistinction which is apparent to us, he discovers an unexpected treasury of distinctions' (MS: 24). The progress of science, he says, continues to break more and more of these indivisible bodies into 'highly complex constructions, furnished with a specific architecture and animated by highly varied internal movements' (MS: 9). In fact, it is in the infinitely small where increasing complexity is found and where actions originate, 'emanating from a multitude of agents who are so many invisible and innumerable little gods' (MS: 25) (Tarde also calls them 'hidden workers'). We know this, Tarde argues, because we can extrapolate what we know from being able to observe human society from within: If we look at the social world, the only one known to us from the inside, we see agents, men, much more differentiated and more sharply characterized as individuals, and richer in continual variations, than are the mechanisms of government or the systems of laws or of beliefs, or even dictionaries or grammars, and this differentiation is maintained by their competition. A historical fact is simpler and clearer than the states of mind of any of its actors. (MS: 37) The example above points at two essential characteristic of'monads' or'societies': first, that difference (or heterogeneity), and not identity (or homogeneity), is at their core (Viana Vargas, 2010); second, that because these entities are the result of the relations between their components or, more exactly, a provisional reflection of the hegemony of some of them over others, they are unstable. Tarde talks about the 'internal revolts' that break apart 'all great regular mechanisms', provoked by the fact that their constitutive elements, the soldiers of these diverse regiments, the temporary incarnation of their laws, always belong only by one aspect of their being to the world they constitute, and by other aspects escape it... The attributes which each element possesses in virtue of its incorporation into its regiment do not form the whole of its nature; it has other tendencies and other instincts which come to it from its other regimentations; and, moreover... still others which come to it from its basic nature, from itself, from its own fundamental substance which is the basis of its struggle against the collective power of which it forms a part. (MS: 47) This sense of internal agitation and instability is linked to the basic drive of monads: they are powerless in isolation, so they tend to associate; they are endowed with belief and desire, so they aim to conquer and possess other monads: Each monad draws the world to itself, and thus has a better grasp of itself. Of course, they are parts of each other, but they can belong to each other to a greater or lesser extent, and each aspires to the highest degree of possession; whence their gradual concentration; and besides, they can belong to each other in a thousand different ways, and each aspires to learn new ways to appropriate its peers. Hence their transformations. They transform in order to conquer; but, since none will ever submit to another except out of self-interest, none can fully accomplish its ambitious dream, and the sovereign monad is exploited by its vassal monads, even as it makes use of them. (MS: 57-8). Like unsettled particles in search of a companion, monads are 'bundles of possessive agencies' eager to possess each other, objects of'reciprocal desires and beliefs' (Debaise, 2008). And this mutual possession, the very definition of society for Tarde, does not tend towards a pre-established harmony à la Leibniz; it remains a continuous movement of shifting hegemonies and possessions. Relations of 'possession' or 'having' are indeed more fundamental and offer a much more promising footing for social theory than the old philosophy of 'being', says Tarde. Being is a crude measure of existence: it cannot be quantified, it is folded upon itself. Possession, on the other hand, reinforces the priority of relations in the constitution of reality and, as the fundamental drive of monads, is universally applicable. As he puts it in an oftenquoted passage: All philosophy hitherto has been based on the verb Be, the definition of which was the philosopher's stone, which all sought to discover. We may affirm that, if it had been based on the verb Have, many sterile debates and fruitless intellectual exertions would have been avoided. From this principle, I am, all the subtlety in the world has not made it possible to deduce any existence other than my own: hence the negation of external reality. If, however, the postulate I have is posited as the fundamental fact, both that which has and that which is had are given inseparably at once. (MS: 52). I will come back to the implications for ethnography of this conceptualisation of monads as unstable and heterogeneous assemblages of infinitesimal complexity defined by their relations of possession. But first, let me provide some empirical material to inform the discussion. --- The Black Cap This ethnographic interlude is taken from a study of the conflict surrounding the eviction and demolition of the Can Ricart factory in Barcelona, approved by the City Council as part of a rezoning plan affecting the city's industrial district (Poblenou). I conducted fieldwork between 2005 and 2007, a period in which, among other things, a wide alliance to save the factory was formed, evictions were resisted and alternative plans produced, negotiations with the Council were held, most workshops ended up being displaced, and part of the factory was listed as industrial heritage. The following situation took place in June 2005, when most of the businesses still had not left the factory and were indeed resisting eviction orders and trying to negotiate better compensation payments. 5 Daniel Iracheta founded his precision machining company in the 1950s. He started in Mariano Aguiló Street, in the Poblenou neighbourhood, and when business grew he moved into a bigger workshop in San Joan de Malta Street. He had to move again in 1974, because of the City Council's plans to extend the Diagonal Avenue, which however took twenty-five years to complete. He found a suitable space in Can Ricart, which had by then been divided into several workshops. Seven-hundred square meters, said Vicente, enough for his forty machines and workers. The dog, which had been barking all along, finally stopped as I paid him some attention. Vicente's office sat above the shop floor, and through the window one could see and hear the workers and the machines downstairs. Daniel Iracheta died in 1995, aged seventy-one, and the company, now managed by his wife Emilia, changed its name to Iracheta Ltd. This had huge implications for the current situation, Vicente continued, because the landlord didn't accept the continuity of the previous indefinite lease and made them sign a new five-year contract, which had by now expired and was renewed on a monthly basis. Compensation payments were based on the length of the contract, so they were not only facing eviction, but the possibility of having to close. They needed about 50 million pesetas to move to a new location, and had only been offered eight million. I'm not an actual employee, Vicente clarified, but I've been helping out with the accounting since my father-in-law died and my mother took charge. I used to work in a bank, so I know about these things. I retired a few years ago. My daughter does work for the company, he said, as Belén came in and briefly joined the conversation. The relationship with The Car Company had profoundly shaped Iracheta since they had started working for them in the seventies. From the layout of the workshop to the accounting practices, everything had been adapted to the manufacturer's requirements and protocols. In fact, in 1990 they were awarded the Y9 Quality Status certification, which recognized their excellence as suppliers and would soon be a requirement to work for The Car Company. Vicente pointed at the wall, where the certificate hung. It's a Total Quality Management system, said Belén, it covers everything since you pick up the phone until the part is delivered. We've got a Quality Manual which specifies all the processes: making a budget, ordering materials, storing parts, packing things for delivery... It allows you to prove that the part has been produced as you are supposed to. There is a record of everything we do: the measurements that were taken from the sample for quality control, the data generated by the statistical software where you introduce them, the certificates of treatment and painting, and so on. And not only with respect to manufacturing, said Vicente, there are requirements about minimum stock (for example to prevent disruptions in case of a strike), and a storage and labelling system that allowed inspectors to come and check the production without even asking. Down in the shop floor Vicente and Belén showed me the different machine-tools, the storage area and some of the parts they made. Among them was a small black cap, the last part they made for The Car Company. It covers the screw of the rear window handle in the Model Qiu, he said. It's a hinged window, the glass doesn't go up and down, but instead has a lever mechanism to push it out. Already the old Model Alegria used that part. We have been making it for twenty-five years now, exclusively for the whole world. It may not look like it, but it is a very complicated process, Vicente told me with a smile. I invited him to continue. The raw material, a 0.50 mm thick metal sheet, comes from The Northern Foundry. Following the technical specifications of The Car Company, it has to be custom-made using a special alloy. Engineers, he said, don't think how much money they would save if they followed standards. The Northern Foundry is a big company, the minimum order they take is 20,000 kilograms. The sheet comes in rolls of 2,000 meters, which are loaded into the machine. First, the progressive stamping press cuts a circular shape, leaving two small tips; it then squashes it, constipates the bottom, and finally cuts the two tips that held the piece. It all happens continuously, do you know what I mean? Pum-pum-pum. We used to have one of those presses, but it is like anything, you have to be an expert to get it right. My father-in-law was a mechanic, and the progressive presses are for professionals. But he wouldn't have got the job if he didn't have the press, so he, a gutsy person, just bought one. He later realised he couldn't master the process, so he asked other nearby workshops to do it for him. It had to be done very precisely, it took him some time to find someone who could actually make it properly. He finally gave the job to some guy called Fernández, who had two fine presses and was also based in the neighbourhood. But I have to tell you something, he continued, the part bears little resemblance with the original drawing by The Car Company. I mean, the cap is still a cap, but the angles, the radii and all these things are quite different. The stuff they drew could not be manufactured. Vicente laughed. Sure, they may be good drawings, but they're impossible to make, especially at the price they want to. The cap was supposed to be more rounded, but with those measures there was no way to attach it to the screw. It just couldn't be hammered onto it. Or maybe it could, but then the press would need sixteen operations instead of four, and this would mean making the part a lot more expensive. So we had to modify the angle, make it more open, because a sharp edge generates resistance. The engineers at The Car Company never wanted to admit it, that's just how they are, but hey, it's the only way the cap can be assembled. There was this time when their manager was on sick leave and we were told by his substitute to make the caps properly. When the old manager later came back and saw the parts, he went What the hell have you done? They'll be returned! And indeed they were. So we had to make them badly again. He even showed us all the paperwork and said that nobody had ever ok'd the changes, but everyone turned a blind eye because they knew that was the only way to get them done at that price. Once machined, Vicente continued, we take the parts to thermal treatment to increase their resistance. They have to be tempered, they need to be springy so that you can hammer them onto the screw later on. Once tempered, the parts have to be stripped off the husk of the treatment, otherwise the paint won't stick. The painting by the way is done in Alicante by cataphoresis. First, they give the parts a degreasing bath, to remove the oil from the previous processes and make sure they are perfectly clean. Then they are placed in racks that are immersed in the electrocoating bath, which works electrically and takes the exact amount of paint for each piece, microns. This is done by girls, said Vicente, because they have skinny fingers and are faster putting the parts in the rack. Once painted, the parts are brought here and we take them to a polyester treatment. This is expensive, because you have to put the parts one by one in another rack and spray-paint them. It also has to be the exact amount of paint. The cataphoresis and the polyester should add between 70 and 100 microns. It can't be more than that or the piece won't fit later on. There was once a purchasing manager at The Car Company who said from now on electrocoating only, to save costs. You can't see the polyester treatment, and all the interior black metal parts look good with the cataphoresis paint only, so she thought it was worth trying. But these caps are external parts, and the sun eats them and they rust. So we went back to the polyester treatment. In fact, in the beginning, at the time of the first Model Alegria, they were chromium-plated. But that was too expensive, it cost more than 40 pesetas per piece. That's when they decided for cataphoresis plus polyester. Now they cost about 25 pesetas. But our share is minimal, he said, 1.20 pesetas, the same as the raw material; electrocoating plus transport to Alicante cost about 6 or 7 pesetas, polyester treatment 12... Vincent laughed. Yes, yes, I know, it's a very laborious process, totally irrational. Other manufacturers, like Peugeot, have a much simpler system, a single piece of plastic with the cap already included in the hinge, and an inside screw. But you know what happens? That in these large companies is very difficult to change things. No one dares. We used to supply the caps directly to The Car Company, because we had the Y9 certification. But it all changed in the eighties. If a car had say 50,000 parts, they probably had 20,000 providers. And managing that was crazy, Purchasing Managers and Departments everywhere. So at some point they decided they weren't going to buy parts anymore, but entire assembled components instead. They reduced the number of providers by a factor of 20 and also asked them to demand their subcontractors the Y9 or ISO9000 certification. I think this was pioneered by Super López, the guy who left General Motors for Volkswagen, remember him? So, anyway, we stopped supplying the cap directly to The Car Company and started selling it instead to Peninsula Glassware, which were assembling the windows. The Car Company basically forced them to buy the part from us. Thing is, Peninsula Glassware was later bought by a French multinational, Saint-Germain, and I remember the new people at quality control were reluctant to accept the caps at first, because they didn't correspond to the original drawings. We had to explain to them that that was the way they had to be and that we were not going to change it. In fact, they never wanted the cap in the first place, because it is hammered onto a metal screw which has the exact diameter of the window hole, and sometimes the glass breaks when assembling it. They wanted the whole part to be made of plastic, like the Peugeot one, because if you force a plastic screw against the window hole, you scrape the plastic instead of breaking the glass. It's even more airtight. But The Car Company has never wanted to change it, maybe because the Model Qiu is supposed to be discontinued soon. In fact, when we still sold the cap directly to them, my father-in-law had to make special nylon hammers for them, so they didn't crush the caps when hammering them onto the screw. He even put a magnet inside the hammers, so the workers could grab the caps more easily. Vicente nodded his head. It was lunchtime and the workers were leaving the workshop. Most of them would go home during the two-hour break. Belén joined us again. There are fewer and fewer businesses that want to work for the car industry these days, she said. It's very competitive, too competitive. Prices go up, expenses go up, but they still force you to be 5% cheaper each year. Do what you need to do, they say, you already know the part you're making. So each year we have to somehow sort it out... We used to make fifteen parts for them, and now we only do this one. It's because of that; as time goes by you lose your profit margin. You know what happens?, interrupted Vicente, you have to make a lot of parts to compensate for the yearly discount. You start with a good price, but after three or four years you are just breaking even on it. And after another three years you're losing money on it. But if you make more parts, and invoice for say 500 million pesetas and manage to make a 10% or 5% profit, then that's fine. You lose money on some, and make money on others, but overall you are still making a profit. Of course, this is a cheap and convenient system for them: they know they are paying a fair price on 25 to 30% of the parts; and they also know you are barely breaking even on 30% of the parts and losing money on the rest. But they have managed to lower the total price by 20%. If you only make a few parts you just can't make it. This is what happened to us, and that's why my father-in-law eventually stopped manufacturing the other parts for them. They didn't like that at the headquarters in Germany, and they wanted to stop all trade with us. But of course, the cap is basically our design. I mean, we're the ones who have the original drawings, not of the part itself, which at the end of the day it's their design, but the ones for the stamping press. They could draw them again, but it would cost them a lot. And since the Model Qiu is, so to speak, an endangered species, it's not worth it. I looked again at the black cap, only a couple of centimetres of size and a few grams of weight, before heading to the canteen. --- Towards a monadic ethnography I: object In this section the proposal for a'monadic ethnography' will be outlined, drawing both from the theoretical discussion and the example presented above. My aim is to explore what it would mean for ethnographic research to take as its starting point Tarde's monadology, that is, how his 'pansocial' (Lorenc, 2012) ontological hypothesis -'everything is a society'may be translated into methodological strategies for doing ethnography. As it will become apparent,'monadic ethnography' adopts and combines Deleuzian and Latournian appropriations of Tarde's work. Researching monads means, first of all, that our objects of study are better understood as relational assemblages or entanglements. That is, as 'constellations of singularities' (Deleuze and Guattari, 1987), multiplicities made up of heterogeneous parts in relation. This in turn means that, if interrogated with sufficient sensitivity, elements are also likely to reveal themselves as composite entities ad infinitum. In other words, every monad is, potentially, a monad of monads. This explains Tarde's 'infinitesimal' approach to sociology, understood as a descent into the'small', which is more complex -and indeed explains -the 'big'. As Latour et al. have put it commenting on this issue, Tarde's hypothesis is that 'there is more complexity in the elements than in the aggregates, or stated a bit more provocatively that "The whole is always smaller than its parts"' (Latour et al., 2012: 2; see also Kwa, 2002;Law, 2004). This is an image I find useful to the extent that it conveys a sense of the monads' counterintuitive internal architecture. But as explained above, Tarde's monadology is a system of open relations where monads relate to other monads and where, at least in principle, everything could be connected to everything. This means that in addition to consisting of a multiplicity of parts, monads are also likely to be part of larger multiplicities 6. Therefore, what we identify as a monad at each stage of our analyses is to a large extent an effect of our position and perspective with regards to the relational web we are tapping into. The story of the black cap provides a good example of this. First, the cap can only be understood through the association of workers, machines, production protocols, clients, raw materials, transport networks, etc. If inspected closely, it would not be hard to see in each of those entities monads-within-the-monad: the technical drawings, the production instructions, the contracts and the suppliers could all become entanglements in themselves, subject to the same dynamic of heterogeneity, coordination and instability as the car part itself. Moreover, in exploring the connections that shaped the black cap, such disparate processes as Barcelona's renewal plans, new industrial groupings in France and changes in the supply chain for a certain car company based in Germany became entangled. Importantly, the relations and connections between and within monads are modulated by their basic drive, their 'avidity'. This

This article outlines the project of a 'monadic ethnography' based on Gabriel Tarde's monadology. Tarde's key contention is that 'everything is a society', i.e. that the world is made up of composite and relational entities of infinitesimal complexity called monads. These assemblages of heterogeneous elements engaged in relations of mutual possession constitute the object of study of 'monadic ethnography'. Their analysis, in turn, has a series of methodological and formal implications, including a transformation of concepts of scale, spatiality and temporality and the need to find representational strategies suitable for conveying the monads' dynamic qualities. A fieldwork example which discusses the making of a car part in a small workshop based in the Can Ricart factory in Barcelona is provided. Throughout the article, the idea of 'monadic ethnography' is discussed in relation to the recent rediscovery of Tarde's work, the work of Bruno Latour and Gilles Deleuze, and the socalled 'ontological turn' in the social sciences.

also likely to be part of larger multiplicities 6. Therefore, what we identify as a monad at each stage of our analyses is to a large extent an effect of our position and perspective with regards to the relational web we are tapping into. The story of the black cap provides a good example of this. First, the cap can only be understood through the association of workers, machines, production protocols, clients, raw materials, transport networks, etc. If inspected closely, it would not be hard to see in each of those entities monads-within-the-monad: the technical drawings, the production instructions, the contracts and the suppliers could all become entanglements in themselves, subject to the same dynamic of heterogeneity, coordination and instability as the car part itself. Moreover, in exploring the connections that shaped the black cap, such disparate processes as Barcelona's renewal plans, new industrial groupings in France and changes in the supply chain for a certain car company based in Germany became entangled. Importantly, the relations and connections between and within monads are modulated by their basic drive, their 'avidity'. This 'eagerness' means that monads engage in a constant dynamic of mutual possession: they will try to conquer or assimilate other monads, and will in turn be approached and taken by others. Their desire for hegemony will sooner or later be met by another monad's competing avidity, resulting in a never-ending process of coupling and decoupling and a permanent battle for hegemony. Using more contemporary terms, there is no escape from relations of power in the study of monads, understood as a circulation rather than a location 7. Monads are unstable arrangements, not so much in the sense of being permanently at risk of falling apart, but in the sense that their crystallisation at any given time is the result of a provisional equilibrium between forces. Vicente and Belén's narrative provided clear evidence of how power modulated the relations between the parts: the German headquarters, for instance, exerted an increasing hegemony by reshuffling the supply chain at one point, demanding the ISO9000 standard, and in the form of a 5% annual discount rate. Rather than a metaphysics of cosmic harmony in the vein of classic monadologies, Tarde's neo-monadology is an ontology of universal struggle. The above remarks suggest that doing monadic ethnography consists, to a large extent, in delving into this potentially infinite unfolding of relations and new monads, and charting a path through their avidity, mutual possession, and struggle for hegemony. Monads are bottomless, chasmic; monadic ethnography an abyssal endeavour. 8 The monads' characteristics also suggest the need to deploy a diachronic form of analysis to fully grasp them. In relation to this, Georgina Born has convincingly argued about the need to incorporate Tarde's theory of imitation-opposition-innovation in the discussion of his monadology, noting that his method is concerned with 'analysing not only the elementary structures of process..., but their cumulative outcome in historical trajectories of variation or transformation, stability or stasis'(2010: 235). The black cap, for example, revealed itself to be a continuous variation whose stability can only be described as a synchronic fiction. The part was a constant succession of transformations in the association between materials, providers, procedures, etc. --- Collaboration as well as competition between monads (or 'adaptation' and 'opposition' in Tarde's terms) animated these transformations: the client demanded a lower price, the glass factory lobbied for a change in materials, the manufacturer adapted the design to the budget... And yet, the part did not become something else altogether (see Law 2002). Accounting for its endurance is just as important as accounting for its continuous change. In fact, the story of the black cap is a remarkable tale of resilience (in the form among other things of several're-alignments' of the parties involved) on the face of the radical transformations brought about by the vertical integration of European car manufacturing. The Tardian monad is a highly appropriate concept to describe this combination of variation and obduracy. 9 Born further argues that the study of monads indeed requires 'an expanded analytics of temporalities' that goes beyond the'monotemporality of becoming, which can flatten out and pre-empt investigation of the multiplicity of time' (Born, 2010: 243). The idea here is that monads are made of heterogenous temporal relations, trajectories, and rates of change. In other words, monads are palimpsestic, an amalgam of temporal sediments; doing monadic ethnography calls for an investigation of their traces, the equivalent of a geological crosssection. 10 The black cap, for example, was traversed by the multiplicity of temporalities enacted by its different parts: the long duration of the materials, the shorter duration of quality standards and machine tools; the high rate of change within transport networks and the (relative) stability of the workforce... These different temporal trajectories and histories of change left some traces, which ethnographic fieldwork revealed. This immanent conceptualisation of time has a parallel in the way that monadology approaches space. Rather than situating monads in a pre-existing and fixed Euclidian matrix, it is perhaps more appropriate to study how monads enact their own conditions of spatial im/possibility (this is an argument well-articulated in ANT: see Law, 2002: 92;Latour, 2005). Instead of a stable geography of scales and its concomitant vocabulary of micro and macro, local and global, monadology relies on a topology of connections and intensity, in which spatial arrangements are performed and achieved. The black cap tale was not about local processes affected by global dynamics, but about the enactment of a set of relations that produced a given spatial arrangement. In other words, the 'post-Fordist' restructuring of the car industry was not above and afar, but rather a concrete and localised wave of changes within the relational organisation of the production. --- Towards a monadic ethnography II: form This brief account of how neo-monadology transforms our objects of study already suggests the basic shape of an ethnographic approach towards them. In effect, the research of Tardian monads implies a certain monadisation of their research. Questions of form are, then, central to the debate. Monadic ethnography relies on formal strategies that can adequately convey the transformation of its object of study outlined above. 11 One of the defining characteristics of ethnography, at least in its classic sense, is that it operates from within the society, group or setting it studies and through interaction with its actors. Monadic ethnography also relies on being conducted from the inside -it is only through relations that we can study relations, as Strathern and others have argued. Moreover, it is only from within that we can overcome the illusion of perfectly stable and finished entities and start to discern their immense internal complexity and agitation. Studying monads from within is therefore an active rejection of the 'view from above' (or 'nowhere'), that is, an external, superior, detached perspective (see Haraway, 1988). Monadic ethnography is about looking at the world from the monad's point of view -or, as Viveiros de Castro (2003) would put it, seeing the world the monad sees. Latour et al. have argued along the same lines that monads are 'a point of view on all the other entities taken severally and not as a totality... The "whole world", as Leibniz said, would be "grasped" or "reflected" through this idiosyncratic point of view' (Latour et al., 2012: 7). This is an important issue I will come back to in the conclusion. Tarde's monadology implies a multiplicity of possible points of view; consequently, discovering and adopting such perspectives is part of the monadic ethnographer's job. Again, the story of the black cap can help us advance in the discussion. I constructed the text as a non-verbatim conversation for three reasons. First, it was a way of letting Vicente and Belén's understanding and description of the production of the car part drive the narrative -instead of mine, or any external point of view. The 'dialogic' (Bakhtin, 1981) form adopted, inspired by Peter Weiss' The Aesthetics of Resistance 12 (2005), was an experiment in removing all (academic) commentary and taking seriously the way the informants told the story and performed the theoretical work themselves and in their own vernacular. The absence of a second-level, 'authorised' voice that clarifies, ratifies, explains, etc. was my way of recognising that the extraordinary story of the black cap, with all its monadic quality, belonged to them; it was their 'theory'. It took me a while to recognise that there was a remarkable isomorphism at play: Tarde's monadology had triggered an interest in exploring objects as relational assemblages that eventually led me to the black cap; Vicente and Belén, in turn, deployed a rather monadological strategy in their account of its fabrication, which I set out to reproduce in my writing. I will return to the important question of whose ontology we talk about when we talk about monadology in the conclusion. Secondly, removing quotation marks and dialogue lines highlighted the constructed nature of the text. The fragment above is not the transcription of a taped fieldwork interview; it is a re-construction of notes and recollections of several conversations, which I assembled and gave to Vicente and Belén to read and comment. Although the'style' was solely my choice, the production of the text involved discussions in which the anonymity of the sensible elements was decided, my inaccuracies corrected, further detail provided, and the overall tone and direction of the narrative agreed. However limited our collaboration was, I think it points at an important methodological strategy and a logical consequence of the commitment to adopt the monad's point of view as opposed to an external one. Vicente and Belén acted as the spokesmen for the black cap, as it were; it was only coherent with the principles of monadic ethnography outline above to enter a dialogue with them. Thirdly and very briefly, deploying a dialogic format was also a matter of interpellation: I wanted to put the reader in the position of the attentive listener that I had occupied during fieldwork. The significance of these three gestures can probably only be gasped by briefly considering what not trying to engage with the monad from within would look like. The black cap story could have easily been framed as an example of the global post-Fordist transformation of the car industry which started in the late seventies. I could have provided a general description of such transformation as discussed in the specialist literature and then use fragments of my fieldwork to illustrate how it was experienced locally in the context of my study. This would not only reduce the actors' own theory of the reality they inhabit to mere side evidence, but also betray the fundamental monadological insight that the key to understanding the post-Fordist transformation lies in the small, in the myriad actions of its tiny components. I would like to conclude this section by making clear that I am not arguing that dialogic writing is monadic ethnography's only or even preferred formal strategy. Far from it. My intention was simply to make explicit the choices I made in this particular instance of monadic description, as a result of trying to accommodate the object and the situation I was studying. I would not want to prevent the development of other forms of monadic ethnography by setting out formal guidelines. --- Conclusion: ontology, monadology, ethnography This article has explored the shape ethnography may take if one was to accept Gabriel Tarde's bold monadological hypothesis, namely that 'everything is a society'. For Tarde (and later on for Latour),'society' and'social' describe a universal form of association between heterogeneous elements, rather than a specific human aggregate. The resulting relational assemblages, or'monads', are characterised by their infinitesimal complexity, irreducible heterogeneity, their openness and avidity toward other monads, and their chronically unstable condition. Tarde's monadology is based on the primacy of difference over identity, the interest in relations of possession or "having" rather than a concern over "being", and the centrality of hegemony (both within and between monads) in the conceptualisation of relationality. This particular combination of claims we have come to associate with Deleuzian-and ANT-inspired perspectives is, in my opinion, one of the reasons why the experiment of (re)thinking ethnography through Tarde's monadology is worth considering: as I have attempted to show in the previous sections, compared to some versions of ANT, monadic ethnography places a greater emphasis on relations of power and temporal dynamics; compared to some appropriations of Deleuzian theory, it is less speculative and more committed to fieldwork, to making sense of these assemblages from within, to 'following the actors'. Monadic ethnography may be in a position to balance out Deleuzian and Latourian readings of Tarde and negotiate a tricky yet productive in-betweenness. Inasmuch as Tarde's monadology constitutes a strong ontological affirmation with regards to the make-up of the world, it may be productive to relate the project of a monadic ethnography to recent debates around the so-called 'ontological turn' in the social sciences -particularly within anthropology and Science and Technology Studies (STS). After all, the usefulness, appropriateness and politics of this growing interest in metaphysical questions in the context of social research cannot be taken for granted. Although a full account of this ongoing debate is outside the scope of this conclusion, it may be useful to briefly revisit some key positions within it. 13 In anthropology, the question of ontology arises out of the perennial challenge of appropriately attending to systems of thought and theories which are other to ours, perhaps even incompatible. For Eduardo Viveiros de Castro, one of the early proponents of a becoming-ontology for the discipline, the language of ontology is important for one specific and, let's say, tactical reason. It acts as a countermeasure to a derealizing trick frequently played against the native's thinking, which turns this thought into a kind of sustained phantasy, by reducing it to the dimensions of a form of knowledge or representation, that is, to an "epistemology" or a "worldview" (2003: 18). The need to 'take seriously' the ontologies of the peoples one studies, that is, not as 'worldviews' but as 'worlds that are viewed', not as 'opinions' but as 'objectively experienced worlds', is at the basis of Viveiros de Castro's (2010; 2011) groundbreaking work on Amerindian perspectivism -an ontology he defines as 'perpendicular' and incommensurable to ours. Whereas Western ontology is based on the assumption that only one world/nature exists, which can however be interpreted or represented in many different ways (mono-naturalism, multiculturalism), perspectivism inverts this and establishes that there is one culture (shared among humans, spirits, animals and some plants) and many natures (mono-culturalism, multinaturalism). All entities see ('represent') the world in the same way, through the same categories ('food', 'home', etc.) but the world they see is different (blood for us, beer for jaguars; mud for us, ceremonial palace for tapirs). The importance of encountering this radically other ontology for Viveiros de Castro lies in its capacity to unsettle Western ontology, in particular its self-perceived superiority. By acknowledging the existence of alternative ontologies and engaging with them as proper metaphysical systems, anthropology -argues Viveiros de Castro -may finally be in a position to close its 'karmic cycle' and take on its new mission as the 'theory-practice of the permanent decolonisation of thought'(2010: 14) or, in other words, to become 'the science of the ontological self-determination of the world's peoples'(2003:18). In sharp contrast, Phillipe Descola's work (2013) has engaged with the multiplicity of ontologies found in ethnographic studies as a problem of classification, and consequently developed a taxonomy based on how similarities and differences between humans and nonhumans are established across two distinct dimensions, 'interiority' and 'physicality'. Western ontology, which he calls 'naturalism', distinguishes humans from non-humans based on their internal difference, namely that only humans have a soul, consciousness, subjectivity and language. This ontology, dominant in the West since the 17 th Century, co-exists with three others: 'animism', 'totemism' and 'analogism'. For Descola, perspectivism is a particular 'type' of animism, where the distinction between humans and non-humans is based on their physicality, i.e. their physiological, perceptual and sensory-motor features. In short, in the work of Descola, the language ontology is an instrument for developing a'relativist universalism' in anthropology -rather than, as Viveiros de Castro argued, a 'bomb' in the discipline's headquarters (see Latour, 2009 for a review of a debate between the two). Yet another distinguishable form of engagement with ontology has been articulated by STS scholars. In a recent review of the 'ontological turn' in their field, Woolgar and Leza<unk>n (2013) argue that while the 'turn' may indeed be seen as a move away from epistemology ('the object of inquiry is the very existence or being of entities, not merely the modes of knowing pre-existing entities'), this distinction has rarely been stable within STS, a perspective whose main concern is the analysis of how things come to be what they are, how objects and subjects are enacted and performed -in other words, 'how in practice, and in detail, particular ontologies are achieved'. Rather than committing to a particular ontology or 'theory of what there is', our field's current curiosity about ontologies and their enactment is best understood as a way of extending its idiosyncratic critical sensibility -an appreciation of fluidity in seemingly stable entities, a recognition of difference beyond claims to singularity (and vice versa), a reluctance to take the world at face value -to the realm of the ready-made, to the world of those entities whose being might seem most unproblematic and ordinary. (2013: 336) Where, then, stands monadic ethnography in all this? In my opinion, it represents a fourth mode of engagement with ontology, different from the three above and yet with selective affinities to all of them. This is because the monadology it is founded upon is both an ontology in itself (a theory of the make-up of reality based on monads and their relations) and a method for studying the monads' own ontologies. Tarde's monadology is therefore both an alternative to hegemonic naturalism (another ontology), and 'a way of making sense of a world woven from multiple ontologies' (Lorenc, p.c.) (a privileged route onto other ontologies). At the same time, Tarde's definition of monads as more or less fragile achievements implies a sustained interested in how they come to be, how they attain and maintain their ontological status. In his afterword to MS, Theo Lorenc develops the idea that Tarde's monadology implies a continuity between the ontological activity of the researcher and that of her objects: the scientific or metaphysical theories we develop to explain reality would not be foreign to the elements, but on the contrary, would be profoundly continuous with the cosmic plans which form their own most intimate reality. Each element has, and in some sense is, an ontological theory of its own. Thus, Tardean metaphysics could be described as an ontology of ontologies: the universe is woven from the theorizing activity of its innumerable elements. (Lorenc, 2012: 92) With this fractal scheme of self-similarity, Tarde elegantly addresses the ontological duplicity inherent to his argument. When it comes to doing ethnography, I would argue this duplicity is not so much a problem as a basic fieldwork condition: the people we work with also have their theories about the make-up of the world, and the relationship between theirs and ours is an empirical matter. In this article, an instance of remarkable isomorphism was presentedbut it is not hard to imagine scenarios of conflict, incommensurability or plain misunderstanding 14. Faced with a similar predicament, Henare et al. (2007) argued that 'hosting' alternative cosmologies does not necessarily mean replacing the old 'episteme of all epistemes' with a totalising 'ontology of all ontologies', but rather'recuperat[ing] a facility [for concept production] informants may already have' and developing a methodology that might generate a'multiplicity of theories'(2007: 16). This approach in turn echoes Latour's argument that ANT analysts should only possess an infra-language 'whose role is simply to help them become attentive to the actors' own fully developed meta-language, a reflexive account of what they are saying' (2005: 49) -a move that as he acknowledges can be considered a revitalisation of ethnomethodology's programme: the study of the procedures by which society is produced by its members, and the theories they use to make sense of it (Garfinkel, 1967). With regards to monadic ethnography, I would not be disappointed if the engagement with ontology translates into a minimal and rather humble aspiration: to become a form of research that, starting from a monadological understanding of reality, is open to engaging with the monads' performative power at all levels, including their theoretical and ontological work. Monadic ethnography may then become a battleground for the competition between theories, or a version (or a variation) of the monads' own theorising work. On occasion, it may become a vehicle for the transformation -perhaps even the possession -of our conceptualisations by those of the monads. In any case, an ethnographic approach based on the respect towards and attentiveness to the actors' own theories seems to me a valuable first step for this newly born creature; it carries with it an invitation to devise forms of theoretical collaboration and co-production with the people we work with and a call for an ethnographic practice which is seen, at its core, as an inventive activity, a machine for distributing new ideas, concepts, theories, beings. --- Humaines, no. 3, 2000; Multitudes, no. 7, 2001; and Economy and Society, vol. 36, no. 4, 2007. 32 4 See Candea (2010b) for an excellent overview of Tarde's work written with an anthropological sensibility. The recuperation of Tarde within social psychology is also remarkable and offers an alternative path into his ideas (see López and Sánchez-Criado, 2006;Lubek, 1981;Van Ginneken, 1992). 5 All the names of individuals and businesses have been anonymised, with the exception of Iracheta's family. When negotiating the publication of the text, and on the face of their disappearance as a family business, they choose to appear under their real names as an act of memory. 6 I thank Theo Lorenc for this formulation. 7 The parallels with Foucault's'microphysics of power' are evident, and were already pointed out by Deleuze (2006). 8 This tentative vocabulary, developed in dialogue with Theo Lorenc, is an antidote to metaphors of 'flatness' routinely used to refer to relational ontologies, and which in our opinion betray the vertiginous verticality of Tarde's ontology. 9 See Harman (2009) on the difficulty the concept of 'actor-network' has on this issue. 10 It falls outside the scope of this text to explore the relationship between Tarde's monads and the work of his successor as Chair of Philosophy at the College de France, Henri Bergson. Bergson's (1962) notion of the virtual coexistence of several times in a single plane is here particularly relevant (see also Deleuze 1988). 11 Granted, the question of finding an appropriate formal strategy for our ethnographic tales is not specific to monadic ethnography, and ever since the publication of Writing Culture (Clifford and Marcus, 1986) discussions around the literary grounds of ethnographic strategies of representation are commonplace. I have been inevitably influenced by these debates, even though I read anthropology in context of active rejection of the discipline's 'post-modern drift'.

This article outlines the project of a 'monadic ethnography' based on Gabriel Tarde's monadology. Tarde's key contention is that 'everything is a society', i.e. that the world is made up of composite and relational entities of infinitesimal complexity called monads. These assemblages of heterogeneous elements engaged in relations of mutual possession constitute the object of study of 'monadic ethnography'. Their analysis, in turn, has a series of methodological and formal implications, including a transformation of concepts of scale, spatiality and temporality and the need to find representational strategies suitable for conveying the monads' dynamic qualities. A fieldwork example which discusses the making of a car part in a small workshop based in the Can Ricart factory in Barcelona is provided. Throughout the article, the idea of 'monadic ethnography' is discussed in relation to the recent rediscovery of Tarde's work, the work of Bruno Latour and Gilles Deleuze, and the socalled 'ontological turn' in the social sciences.

Introduction Tobacco use is the single most preventable cause of death in the United States (1). Although the 2014 Surgeon General's report, The Health Consequences of Smoking -50 Years of Progress, indicates that the prevalence of current cigarette smoking is on the decline, the report emphasizes the need to further monitor patterns of use for all tobacco products, particularly as disparities in use persist and alternate forms of tobacco use are increasing in popularity among youths (1). Hookah tobacco smoking, for example, has increased tremendously (2,3). Hookah, also known as water pipe or shisha, is a device used for smoking tobacco and other substances. Hookah smoking involves passing tobacco smoke through water before inhalation (4). In a typical 1-hour hookah smoking session, hookah users inhale approximately 90,000 mL volume of smoke, which is substantially more smoke than the smoke from 1 cigarette (500-600 mL) (5,6). The charcoal used to heat the tobacco can raise health risks by producing high levels of carbon monoxide, metals, and cancer-causing chemicals (6). One session of hookah use contains approximately 200 puffs of smoke, which exposes users to 3-to 6-fold higher levels of carbon monoxide and 46-fold higher levels of tar than from a single cigarette (6,7). Globally more than 100 million people use hookah regularly (8). However, given the recent proliferation of hookah cafes worldwide, this estimate is likely to increase (6). In the past decade, 2,000 to 3,000 new hookah cafes opened in the United States alone (9). Until recently, few studies focused on hookah smoking, and this practice was not considered a serious health problem (7,10). Although studies have begun to examine hookah use among college-aged students, the related socio-behavioral risk factors are largely unknown (2,11). Reports do, however, suggest hookah smoking is increasingly popular among youths in the United States (11,12). Thus, the objective of this cross-sectional study was to determine the prevalence of hookah use at a large urban university in south Florida and to describe the knowledge and practices associated with hookah use among university students. --- Methods This study used a cross-sectional survey design. Subjects were recruited from a convenience sample of students who were attending a large urban university in Florida during the 2011-12 academic year. To be eligible for participation, participants had to be enrolled as a graduate or undergraduate student at the University of South Florida (USF) during the spring 2012 semester. To assess the prevalence of hookah use, a cluster sample of 478 students was recruited. Students were asked to take a self-administered survey that included questions on demographics; current hookah smoking (defined as hookah use in the past 30 days); hookah use during their lifetime; associated risk factors; and knowledge, attitudes, beliefs, and practices regarding hookah use. Questions were designed based on a literature and expert review. Before data collection, all study instruments were field-tested with a sample of 20 students and revised on the basis of the test results. To determine the required size for the overall sample and to calculate cluster sizes (one-stage cluster sampling), we used the Centers for Disease Control and Prevention's EpiInfo v.6 Statistical Calculator's "population study option" (parameters: total population size, expected frequency, and worst acceptable). More than 30 natural clusters were sampled, including all the university's colleges, 2 libraries, the student center, on-campus dormitories, and the fitness center. The study was reviewed and approved by the university's institutional review board. All data were analyzed in SAS 9.3 (SAS Institute Inc). First, demographic characteristics of hookah users and nonusers were compared using <unk> 2 and Fisher exact tests. Current and lifetime use prevalence was evaluated for the entire sample and subgroups. Finally, unadjusted and adjusted analyses using logistic regression modeling were conducted to identify associated factors for hookah use. Analysis results were adjusted for age, sex, education, cigarette smoking, alcohol use, owning a personal hookah, having a friend who smoked hookah, and proximity to hookah lounges. --- Results --- Sample description In total, 478 participants were interviewed, of which 261 were women (54.6%) and 217 were men (45.4%) (Table 1), a ratio consistent with the university's enrollment in 2011-12 (women, 56%; men, 44%; <unk> 2 P =.540). Most (78.8%) were undergraduate students, a finding also consistent with the USF enrollment (undergraduate, 74%; graduate/postgraduate, 22%; <unk> 2 P =.781). Participants were of varied ethnicities including white (33.4%), Asian (21.3%), Hispanic (17.4%), and black (12.8%). The sample ethnicity was statistically different (P =.001) from the university's enrollment in 2011-12, which was reported as white (60%), Hispanic (17%), Asian (6%), and black (11%). --- Current and lifetime prevalence of hookah smoking Current prevalence of hookah use was 16.3% (95% CI, 13.0-19.7) (Table 2). More men than women reported current hookah use (22.2% men vs 11.5% women). Students of Middle Eastern descent reported the highest prevalence of current hookah use (24.4%), followed by white (17.3%), Asian (16.3%), Hispanic (12.5%), and African American (3.6%) students. There was no difference in the prevalence of hookah smoking between undergraduate (15.6%) and graduate (15.8%) students. The prevalence of lifetime (ever use) of hookah in our sample was 54.4% (95% CI, 50.0-58.9) with 49.4% of women and 60.6% of men reporting ever using a hookah. Lifetime use by race/ethnicity was 68.2% for white students, 62.0% for Hispanic students, 53.5% for Middle Eastern students, 42.3% for Asian students, 28.8% for African American students, and 20.0% for Native American students. Lifetime use among undergraduates was 53.8% and among graduates was 54.1%. Additionally, prevalence of lifetime use was 2-fold higher among participants who owned a private hookah (96.2%) than among those who did not (49.5%). --- Knowledge, attitudes, and practices of hookah smoking When shown a picture of a standard hookah, most (95.8%) participants recognized the image. When asked about the harmfulness of hookah smoking, 74.6% indicated that hookah smoking is harmful for health. However, 12.6% reported hookah smoking was not harmful, and 12.8% reported they were unsure of the harmfulness. Most (50.6%) participants also indicated that cigarette PREVENTING CHRONIC DISEASE VOLUME 11, E214 PUBLIC HEALTH RESEARCH, PRACTICE, AND POLICY smoking is more dangerous than hookah smoking. When asked to identify the sources from which they received information on hookah harmfulness, participants cited no formal means of acquiring data about hookah safety or harm; they instead reported using their own judgment (70.3%) or acquiring information from friends (25.9%) or the Internet (22.4%). When asked about the presence of hookah bars or lounges, 92.2% of participants reported having a hookah bar or lounge within a 10-mile radius of their residential area. Those who had not smoked hookah were asked if they would ever consider hookah smoking in the future. Almost 30% reported they would. Of those who reported they would consider hookah smoking in the future, reasons cited for this were the time together with friends, the fun associated with the activity, the pleasant atmosphere, the social acceptability, and the perception that hookah smoking was a healthier alterative to cigarette use. Most (85.3%) respondents reported having a friend who smokes hookah. More than 30% of the sample reported having a friend who owns a hookah, while 11% of respondents reported owning one themselves. Current hookah usage was 6-fold higher among participants who owned a private hookah (60.8%) when compared with those who did not (11.0%). Also, prevalence of current hookah use was 4-fold higher among cigarette smokers (45.6%) than among those who did not smoke cigarettes (11.3%). Current hookah users were asked how often they smoke hookah in a given month. Most (55.3%) reported smoking hookah once in a month, with fewer smoking hookah twice a week (19.7%), once a week (19.7%), or every day (5.3%). When asked about length of exposure, almost half of current hookah smokers reported smoking hookah for longer than 1 hour per session, with 32.4% smoking for 61 to 120 minutes and 13.5% smoking greater than 120 minutes per session. Another 40.5% of participants reported spending between 30 and 60 minutes a session and 13.5% reported smoking for less than 30 minutes. --- Factors associated with hookah smoking Unadjusted and adjusted logistic regression modeling was conducted to determine the independent association between socio-demographic factors and hookah use as well as knowledge, attitudes, and practices regarding hookah smoking (Table 3). In unadjusted analysis, sex and race were significantly associated with increased odds of hookah smoking. Men were 2.2 times more likely than women to use hookah (odds ratio [OR], 2.2; 95% confidence interval [CI], 1.33-3.64). When compared with whites, African Americans (OR, 0.18; 95% CI, 0.04-0.77) had significantly lower risk of hookah smoking; the lower risk among Hispanics (OR, 0.68; 95% CI, 0.31-1.50) was insignificant. The increased risk among Middle Easterners (OR, 1.05; 95% CI, 0.56-1.96) was not significant. Cigarette smoking, alcohol use, hookah ownership, and having a friend who uses hookah were other significant risk factors in the unadjusted model. However, in the adjusted model, only cigarette smoking and hookah ownership remained significantly associated with hookah use (controlling for sex, alcohol use, having friends who use hookah, and having a hookah bar or lounge within 10 miles of one's residence). Specifically, cigarette smokers were 4.52 times more likely than nonsmokers to use hookah. Moreover, hookah ownership increased 10.67-fold the risk of hookah use. --- Discussion Findings from our study suggest the current prevalence of hookah use among college students at USF is 16.3%. This rate is consistent with prevalence studies conducted with college-aged youths (11,13,14). Findings are also consistent with those of other studies, including higher rates of smoking among men than among women, higher prevalence among non-Hispanic whites than among other races, and higher prevalence among smokers than among nonsmokers (13,14). The prevalence of ever using hookah (54.4%) found in our study was higher than estimates reported by Fielder and colleagues (15) in New York (about 45%) and by Smith and colleagues (14) in California (about 25%). The value of our study lies in the descriptive nature of the knowledge, attitudes, and practices of the current student population, which shed light for potential preventive actions in university settings. In particular, our study indicates 2 important factors that must be addressed to curtail hookah smoking: the social nature of hookah use and misperceptions regarding risk and harmfulness of hookah smoking reported by nearly a quarter of our sample. Other studies pointed to the importance of these factors as predictors of increased prevalence of hookah smoking during college (11,14). However, specific interventions are needed to dispel each different belief of the population at risk. For instance, most respondents acknowledged that hookah smoking has harmful effects, findings consistent with other studies (10). Yet more than half of the sample believed hookah smoking to be a safer alternative to cigarette smoking and, regardless of their perception of harmfulness, almost a third of the sample reported they would consider hookah smoking in the future. This finding suggests that knowledge regarding the harmful effects of hookah use may not be the most substantial barrier to preventing hookah smoking and that correct- ing the misperception of hookah as a safer alternative may be a more appropriate target for health education efforts. Further investigations into risk factors for hookah smoking initiation and continuation are needed (14). --- PREVENTING CHRONIC DISEASE The reasons why individuals believe hookah smoking is a safer alternative to cigarette smoking are unclear. However, there is no evidence that the effects of the tobacco are less serious if a water pipe is used (16). In our study, more than half of the sample perceived cigarette smoking as more dangerous than hookah smoking. Yet data suggest that hookah smoking, because of higher levels of carbon monoxide and tar, poses a considerably greater health hazard than cigarette smoking (6,9). Because of the filtration mechanism, smoke that emits from a hookah is softer and lighter and has a more pleasant smell than smoke emitted from a cigarette. This may lead smokers to a false belief that hookah smoking is safe or harmless (17). Traditional antitobacco campaigns focus on cigarettes, cigars, and smokeless tobacco with little to no attention to hookah or other nicotine delivery systems, potentially further spreading this false belief. One notable reason that students consider hookah smoking is the social nature of the activity, which may be a difficult barrier for health education programs to overcome. Hookah smoking brings with it numerous challenges not combated by previous antismoking campaigns, including the social aspect of relaxing with friends and the fun associated with this activity. The high rate of hookah device ownership in our study suggests that hookah smoking may be moving from an occasional activity to a regular habit. However, we could not find studies conducted in the same setting with which to compare our study findings. In Florida, efforts have been made to track tobacco use (including hookah) among high school students through population-based surveys (18), but data for college students are not gathered. Efforts are beginning at the national level to track tobacco use (including hookah) among college students in the United States, but institutionally based studies are scarce. The absence of local data is a limitation for direct local action. This situation should prompt universities to monitor alternative forms of tobacco use, including hookah, e-cigarettes, and emerging alternatives, in local student health assessments. Our findings suggest the need for population-based studies to examine unintended effects of current regulations on alternative forms of tobacco use. For instance, although Florida's Clean Indoor Air Act regulates "any lighted tobacco product," (19) many hookah lounges are licensed as retail tobacco stores, exempting them from the act. This loophole in the law creates the appearance that hookah smoking may be permitted in places where cigarette and cigar smoking is prohibited, which may be perceived as ac-knowledgment that hookah smoking is safe for the public. Under current regulations, hookah can be purchased from a hookah lounge or an online shop without age restriction and hookah products are widely available in convenience stores at lower prices than cigarettes, making hookah an available and viable alternative to cigarettes. Additionally, the lack of regulations on the proximity of hookah smoking cafés to university settings is also concerning. Some have recommended regulations to prevent hookah establishments from operating near high schools or colleges, enforcement of strict identity checks, and the taxing of hookah products as potential ways to combat the rising trends in hookah use (12,20). More studies, both qualitative and quantitative, are needed to examine the role of popular media in promoting new hookah establishments near universities to evaluate whether hookah is being portrayed as a "healthy" alternative to smoking. As with any research initiative, this study is not without limitation. Although numerous instruments for measuring hookah use have been reported (21), we could not find a standardized tool available to estimate the prevalence of hookah use. Thus, our instrument was designed based on available data in the literature. Although this study makes an important step in developing such a tool (ie, our instrument was pre-piloted, piloted, and field tested, and revised multiple times before data collection), assessment of factorial validity and cross-validation in large samples was outside the scope of our study. Future studies should attempt to assess the psychometric properties of available research instruments on larger samples. Another limitation is the cross-sectional nature of our data. Specifically, our findings suggest cigarette smoking is associated with hookah use. However, because of the cross-sectional data, such association should be taken with caution and must be further explored in longitudinal studies that can characterize risk more appropriately. More studies are needed to understand the link between cigarette usage and hookah usage. Future studies should compare the frequency and intensity of usage between hookah and cigarette smoking. For instance, we were not able to assess whether students were substituting cigarette use with hookah use or whether they were using both smoking methods in tandem. Cigarette smokers may indeed constitute a high-risk group for hookah smoking, in which case additional efforts should be aimed at cigarette smokers to increase their awareness of the harmfulness of all alternative smoking mechanisms, including hookah. Conversely, hookah smoking may be a gateway for smoking. Our data do not permit such distinction, and we recommend that panel studies be conducted in university settings to uncover why prevalence changes are occurring. Future studies should attempt to disentangle the relationship between concurrent cigarette smoking and hookah use. Because of the cross-sectional nature of the data and the lack of comparison data from previous years, we could not assess trends. Because current trends in water pipe use in other universities indicate that water pipe use is increasing among college students (11,14,22), future studies must attempt to examine hookah usage longitudinally and by year in university. --- PREVENTING CHRONIC DISEASE Our findings indicate a need to monitor knowledge, attitudes, and practices related to hookah use among college students. Misperceptions associated with hookah use are a starting point for the development of health behavior change interventions. On the basis of our findings, we recommend that public health messaging consider misperceptions regarding hookah use as a safe alternative to cigarettes and target youths and college students as well as the general public. Educational campaigns must be designed to address misunderstandings regarding risks associated with hookah smoking and should be inclusive of other recreational tobacco use and nicotine delivery devices, including e-cigarettes, e-hookah, and other emerging devices, as well as the regulation of sales and marking of these devices. Future studies should attempt to gain a deeper understanding of the social and behavioral determinants of hookah use and determine the incidence of hookah use in representative samples of college and high school students. Tobacco control activities should include prevention of water pipe tobacco use in university settings. --- Tables --- Author Information Corresponding Author: Jaime Corvin, PhD, MSPH, CPH, 13201 Bruce B. Downs Blvd, MDC 56, Tampa, FL 33612. Telephone: 813-974-6690. E-mail: [email protected]. Author Affiliations: Shams Rahman, Lissette Chang, Selamawit Hadgu, Abraham A. Salinas-Miranda, University of South Florida, Tampa, Florida. --- PREVENTING CHRONIC DISEASE

Although hookah smoking is becoming a source of tobacco use among college students in the United States, little is known of the students' knowledge, attitudes, and practices regarding hookah use. This cross-sectional study was aimed at determining the prevalence of hookah use and describing social and behavioral factors associated with hookah smoking among university students in a large urban university in Florida.A convenience sample of 478 undergraduate and graduate students was recruited. Lifetime use and current use was evaluated. Logistic regression modeling was used to assess the independent association between study covariates and hookah use.Prevalence among students of having ever used hookah during their lifetime was 54.4%. Hookah use within the past 30 days was 16.3%. Hookah use was significantly associated with cigarette smoking (odds ratio [OR], 4.52; 95% confidence interval [CI], 2.13-9.60) and hookah ownership (OR, 10.67; 95% CI, 4.83-23.66) but not with alcohol use (OR, 1.73; 95% CI, 0.74-4.04). Findings also suggest hookah is perceived as a safer alternative to cigarette smoking. Almost 30% of those who never smoked hookah reported they would consider smoking hookah in the future.Hookah smoking is popular among college students. Misperceptions associated with hookah use indicate a starting point for developing health behavior change interventions. Future studies should investigate social and behavioral determinants of hookah use and determine the incidence of hookah use among college and high school students. Tobacco control activities should include prevention of hookah tobacco use in university settings.

need to provide accurate and accessible scientific information to British Columbians, the responsibility for issuing public health guidelines, and the ability to persuade citizens to follow them. In this paper, we focus specifically on the role of the first-person plural pronoun "we" in constructing a multi-faceted relational ethos [Condit, 2019] for Henry that negotiates the hybrid identities of scientific expert, authoritative regulator, and motivational community leader. Our rhetorical analysis shows how this occurs through the shifting and ambiguous ways in which "we" communicates a sense of communal togetherness when used inclusively to refer to all British Columbians but also rhetorically reinforces a hierarchical separation between BC's public health experts and citizens when used exclusively to refer to public health authorities. Examining the specific meanings and functions of "we" within Henry's COVID-19 communication can help science and health communication researchers as well as practitioners better understand both the powerful and potentially problematic roles that this small but significant word can play in fostering relations of trust and identification with public audiences. Our theoretical approach integrates research on the importance of public trust in experts and public officials during crises with the rhetorical concept of ethos and socio-linguistic research on the ambiguous meanings and functions of first-person plural pronouns. We begin by reviewing this approach before introducing our case study in more detail, explaining our method, and presenting our rhetorical analysis of the primary meanings and functions of "we" in Henry's COVID-19 communication. We close by reflecting on the diverse and ambivalent ways in which this small but significant linguistic device contributes to Henry's relational ethos and on the relevance of our study for the fields of public health and science communication. --- Theoretical framework --- Trust in science and risk communication research The importance of communicating in ways that foster relationships of trust between citizens and experts, especially during times of crisis, is well-established within the fields of science and risk communication [Siegrist, 2019;Siegrist & Zingg, 2014]. Trust is a key factor in determining public perception of risk and compliance with public health measures, especially in contexts of high uncertainty [Hunt, Wald, Dahlstrom & Qu, 2018;Siegrist & Cvetkovich, 2000;Siegrist & Zingg, 2014]. In general, science and risk communication frameworks outline elements such as empathy, expertise, competence, integrity, goodwill, and dedication as crucial elements for demonstrating the communicator's trustworthiness [Besley, Lee & Pressgrove, 2020;Reynolds & Quinn, 2008;Mihelj, Kondor & <unk>t<unk>tka, 2022]. Survey research conducted during the early stages of the pandemic confirmed that public trust in science and politicians was an important predictor for accepting COVID-19 public health measures in Germany [Dohle, Wingen & Schreiber, 2020]. Mihelj et al. [2022]'s study showed the importance of political independence in shaping public perceptions of experts' trustworthiness within newer Western democracies. In Canada, a May 2020 survey showed that citizens considered public health officials the most trustworthy source for accurate, timely information about the pandemic, with British Columbians indicating the highest level of trust among provinces [Waddell, 2020]. Despite some erosion, Canadian trust in public health officials remained strong throughout the first year of the pandemic [Leger, 2021]. Drawing on Siegrest and Zingg's [2014] research, Khosravi [2020] found that audiences will better hear and act on COVID-19 public health messaging if they perceive the communicator as aligned with their own values and intentions, and hence trustworthy. Tworek, Beacock and Ojo [2020] claim that effective "democratic" public health communication during the first six months of the pandemic did "more than communicate facts"; it sustained and built community by fostering social trust among citizens, and between governments and publics Bucchi [2021, p. 31]. similarly argues that the pandemic has emphasized the importance of science and crisis communication that builds mutually trusting relationships among citizens, experts, and institutions. Two recent COVID-19-related experimental studies have explored the role of first-person plural pronouns in fostering relations of trust and community between public health communicators and their intended audiences. Tian, Kim and Solomon [2021] maintain that "the use of first-person plural pronouns within messages may both reflect and bolster a sense of community and communal orientation to coping " Tu, Chen and Mesler [2021, p. 590]. likewise claim that using the pronoun "we" in COVID-19 messaging can build relationships, foster collective identity, and encourage cooperative behaviour [p. 575]. --- Ethos As rhetorical critics, our approach complements research that investigates public perceptions of experts' trustworthiness by focusing on the communication techniques that experts use to engender trust in their implied audiences. By contrast with audience reception studies, rhetorical criticism focuses on how situated communicative actions address and, in so doing, constitute audience identities and values [Charland, 1987] as well as the communicator's identity and values. In this paper, we focus on the role of "we" in the construction of a trustworthy and credible character for Henry in relation to the kind of audience that BC's public updates simultaneously encouraged and presupposed. In rhetorical terms, we thus are concerned with the role of "we" in constituting Henry's ethos as well as developing a better understanding of the ethos of public health and science experts in general. The concept of ethos affords a rich and generative way to think about how public health officials strive to develop relationships of trust with public audiences through their communication practices. Within the Aristotelian tradition, ethos refers to the speaker's character as constructed by their discourse [Condit, 2019, p. 179]. As Aristotle explains, "Persuasion is achieved by the speaker's personal character when the speech is so spoken as to make us think him [sic] credible" [Aristotle, 2004, p. 7]. Notably, the proof of ethos was considered especially persuasive in situations that lack certainty or consensus [Ofori, 2019, p. 55], as has clearly been the case throughout the COVID-19 pandemic. The concept of ethos includes three main components: phronêsis, aretê, and eunoia. Phronêsis refers to how the speaker communicates "good sense, practical wisdom, sagacity, expertise, and intelligence" [Smith, 2004, p. 10]. Aretê refers to the speaker's virtuous nature or "excellence of character" as demonstrated through the speech and as defined by the socio-cultural context [Ofori, 2019, p. 56]. Eunoia refers to the speaker's expression of goodwill -of friendship and empathytoward the audience; it is the aspect of ethos that attends most explicitly to cultivating a relationship with the audience [Miller, 2003]. While all three components potentially contribute to generating audience trust and respect for the speaker, Condit's [2019] study of Anthony Fauci and Thomas Frieden's public health communication during the 2014-15 Ebola crisis foregrounds the centrality for public health officials' ethos of attempting to cultivate an empathetic, affiliative relationship with audiences during times of health crisis. Drawing on classical and contemporary rhetorical theory, Condit [2019] proposes a relational concept of ethos "as the activation, rebuilding, or maintenance of relationships among different social positions: publics and institutions" [p. 177]. They argue that in the context of health crises like pandemics, ethos forms an essential "relational pivot" or "bridge" between public health leaders and publics [p. 186]. For Condit, as for us, analysing how specific public health experts construct ethos in this relational sense is important, not because of their rhetorical successes or failures as individuals, but because their communication necessarily enacts an institutional ethos "bound to the positionality of public health officials as scientists-who-direct-public-health-policy." Whether consciously or not, through their rhetorical enactment of this positionality, public health officials construct socio-politically situated relationships between citizens and health authorities [Condit, 2019, p. 185]. As Beason's [Beason, 1991] research on "signalled ethos" in business speeches shows, the pronoun "we" may play a subtle but significant role in cultivating a strong relationship with audiences. Beason claims that this pronoun fosters identification or "similitude" with audiences, creating a sense of affiliation and building a sense of community [p. 131]. In their discussion of effective communication practices for science researchers, Varpio [2018] likewise underlines how the pronoun "we" contributes to the writer's ethos by communicating a sense of mutual identification, cohesion, and community [p. 208]. Similarly, in their study of TED talks, Scotto di Carlo [2014] argues that the use of "we" helps breach the barrier between scientific experts and lay audiences [p. 604] while Luzón [2013] maintains that "we" in science blogs promotes dialogic involvement and enacts a dual author identity of specialist and civic scientist. --- Socio-linguistic research on "We" The research on "we" reviewed above emphasizes the pronoun's role in constructing the speaker's relational ethos by evoking a sense of identification and community. However, socio-linguistic scholarship provides a fuller understanding of the complex and sometimes ambiguous meanings and functions of "we." Unlike some languages, English does not differentiate formally between inclusive and exclusive senses of "we" [Harwood, 2005;Scheibman, 2014]. Inclusive "we" refers to both the speaker and the addressee, while exclusive "we" does not include the addressee. For example, in the context of COVID-19 when a public health officer states "we must all follow public health measures," "we" refers inclusively to both speaker and audience whereas the statement "we are increasing testing capacity" refers exclusively to "we" in public health. Exclusive "we" sometimes designates an individual speaker as in cases of pluralis maiestatis (the "royal we"e.g., the statement "We are not amused" attributed to Queen Victoria) or pluralis modestiae (author's plural -e.g., in a single-authored research paper, using "we" to refer to the individual author) and sometimes, in a true-plural sense, designates a group that includes the speaker and others, but not the addressee (as in the above example of "we" in public health) [De Cock, 2016]. Thus, the meanings and functions of "we" in English are inherently heterogeneous, ambiguous, and context-dependent [Du Bois, 2012;Makmillen & Riedlinger, 2021;Scheibman, 2014]. The fuzziness of the exclusive/inclusive divide can be situationally useful and strategically deployed in academic, political, and healthcare contexts. In scholarly writing, authors can use "we" inclusively to index mutual knowledge and construct audience involvement but "we" also may refer exclusively to the author and their potential co-authors or, as noted above, to a single author [Harwood, 2005;De Cock, 2016]. In their study of research articles by Mäori scholars, Makmillen and Riedlinger [2021] have shown how the blurred categories of "we" can extend beyond a basic inclusive/exclusive distinction to include a range of shifting identities and social relations as researchers and community members, thus inviting "multiple identification through ambiguity" [p. 11]. Likewise in political discourse, first-person plural pronouns can foster identification, alignment, and disalignment [Bucholtz & Hall, 2005] and contribute to the construction of subject positions, group relations, and group identity [Downing & Perucha, 2013, p. 380]. Uses of "we" in political discourse may alternate between or ambiguously combine exclusive reference to the speaker's specific group (e.g. a government institution or political party) and broader inclusive references to a national "we" (i.e. the country and its people). The pronoun can function strategically "as a distancing effect to avoid [or diffuse] responsibility on certain issues" [Downing & Perucha, 2013, p. 388] and it also can suggest "a shared moral stance" between political leaders and public audiences [Bucholtz & Hall, 2005, p. 604]. Citing Billig's [1995] concept of "banal nationalism," Proctor and Su [2011] maintain that, precisely because of its "everyday" quality, "we" functions as an unremarkable but important "flag" within a nationalistic register [p. 3252]. Socio-linguistic research on healthcare uses of "we" has focused primarily on doctor-patient communication. In this context, "we" may refer exclusively to the healthcare provider or a healthcare team, communicating distance between provider and patient and indicating that the speaker belongs to a group that the audience does not [Rees & Monrouxe, 2009]. "We" also may refer inclusively to provider and patient, signaling their connection, but inclusive "we" can also obscure power asymmetry and it may permit healthcare providers to feel they are inclusive, when in fact they are not [De Cock, 2016;Rees & Monrouxe, 2009]. De Cock [2016] also discusses a hearer-dominant "we" which refers not to the speaker but, exclusively, to the addressee, as in "How are we feeling today?" This "we" can demonstrate empathy and establish solidarity, but it may also be perceived as condescending or impolite, and as maintaining, rather than diffusing, the relationship's power asymmetry [De Cock, 2016;Rees & Monrouxe, 2009]. Combining socio-linguistic insights on the pronoun "we" with science and risk communication research on trust and rhetorical theories of ethos provides a generative framework for understanding the complex and ambiguous ways this pronoun may function in fostering relationships between public health officials (or other expert science-risk communicators) and their implied citizen audiences. In the case study analysis that follows, we demonstrate this complexity by exploring the multiple roles that "we" plays in the rhetorical construction of Henry's ethos during BC's COVID-19 updates. --- Case study context Appointed Provincial Health Officer in 2018, Henry is British Columbia's senior public health official. Working closely with the BC Centre for Disease Control (BCCDC) and drawing on her prior experience in infectious disease management, Henry led the province's COVID-19 response. At a national level, Canada's pandemic response and communication were led by the country's Chief Public Health Officer, Dr. Theresa Tam. However, because healthcare is a provincial jurisdiction in Canada, each province developed its own pandemic plans and communications. In some provinces, political leaders and/or other scientific experts played central roles in official COVID-19 communication. For example, Ontario's official updates were delivered mainly by the Premier Doug Ford and Health Minister Christine Elliott while the leader of the province's COVID-19 Science Advisory Table, Dr. Peter Jüni, provided frequent (but less official) media interviews and public health information. In BC, however, Henry acted as the province's principal communicator throughout the pandemic while Premier John Horgan and Health Minister Adrian Dix played less central communication roles. Under BC's Public Health Act, Henry also had the authority to declare a public health emergency and issue public health orders [H. Henry & Henry, 2021, p. 108]. Similarly to other provinces, BC's public-facing communication occurred primarily through regular public updates streamed live by the CTV News channel. For the period of our research (mid-March to end of December 2020), Henry delivered almost daily in-person updates until the end of May, and approximately two to three updates weekly thereafter. Lasting approximately 15 minutes each, these updates typically began with an introduction by Health Minister Dix, followed by Henry's report on new cases, community outbreaks, and case status by health region.1 After this report, Henry normally talked for an equal, if not greater, amount of time about the civic roles and responsibilities of British Columbians during the pandemic. A media question period occurred after the prepared portion of each update. Dubbed the "calming voice in a sea of coronavirus madness" [Picard, 2020] and "BC's Great Communicator" [Smith, 2020], Henry was widely applauded both within and beyond British Columbia for her style of communication [Little, 2020;Marsh, 2020;Picard, 2020;Porter, 2020;Smith, 2020;Tworek et al., 2020]. In their report Democratic health communications during COVID-19, Tworek et al. [2020] describe Henry as "an exceptionally good communicator," commending her for "being gentle and compassionate, building trust, expecting good faith, and offering clear scientific information in a consistent, understandable, reassuring way" [pp. 60, 17]. For health journalist Picard [2020], Henry's public briefings comprised "a master class in crisis communication." During the first months of the pandemic, she became a cultural icon, celebrated on social media, as well as in public art murals, musical tributes, and fashion [Picard, 2020;Hinks, 2020;Wainman, 2020]. Instead of threatening disciplinary enforcement, Henry's main communication strategy was to encourage citizens to follow public health protocols by ensuring they properly understood "what they need to do and why" and trusting that they would willingly fulfill their "mutual responsibilities" [Porter, 2020]. According to Tworek et al. [2020], BC's response aimed "to cultivate trust (among citizens as well as between government and public)" to strengthen public solidarity, collaboration, and resilience [p. 59]. As we have argued in our previous study of BC's COVID-19 communication, the province's strategy enacted a (neo)communitarian ideology focused on citizens' civic duty to voluntarily take responsibility for the health of their community [Spoel, Lacelle & Millar, 2021]. In this paper, we investigate more fully the role of the first-person plural within this pro-social rhetoric.2 As Henry herself explained at the 2020 Canadian Science Policy Conference, her communication team consciously decided to use inclusive expressions such as "what we are going through" and "here's what we can do" to help foster feelings of togetherness and to "drive the altruistic response" [B. Henry, 2020]. --- Method Our study used the method of rhetorical analysis which, in general, applies rhetorical theory and concepts to specific instances of language use to generate textured insights about the situated forms, meanings, and functions of this language use. The specific instances of language use in our case study consisted of the 131 live-streamed public updates wherein Henry was the primary speaker, delivered between March 16, 2020 and December 31, 2020. This time frame represents (1) the first wave of the pandemic in BC from mid-March to roughly the end of May; (2) the easing of restrictions between June and August; (3) the second wave which began in September, peaked in late November, and decreased modestly during December. The first-personal plural pronoun usage analysed in this paper is drawn from data collected as part of our larger project examining how BC's COVID-19 communication during the first parts of the pandemic rhetorically constituted "good" covid citizens. This data consists of manually transcribed excerpts from BC's public updates which contain content related to pro-social behavior and civic values for how BC citizens should respond to the pandemic as well as to the role of public health officials in guiding this response. After identifying all uses of "we" through a word search, 3 lead author Spoel categorized them into three main types of referent based on their meanings within the context of specific statements: 1. occurrences of inclusive "we" that refer both to Henry (or to Henry and her Public Health team) and to her implied audience; 2. occurrences of exclusive "we" that refer either to Henry and her Public Health team as a group or to Henry singularly, but do not include the audience; and 3. occurrences of ambiguous and/or shifting inclusive and exclusive referents for "we" (see Table 1). Following review by authors Millar and Lacelle the research team collaboratively refined these categorizations. Referents were identified as ambiguous and/or shifting whenever more than one member of the research team considered them such. Through ongoing team discussion and grouping of "we" statements into separate thematic files in our secure google drive, we iteratively coded each of these main categories into eight sub-themes (see Table 2). Statements that exemplify more than one theme were included in all relevant thematic groupings. Table 2. Sub-themes. inclusive "we" exclusive "we" ambiguous/shifting "we" "we" as a province and community together "we" as research experts "we" as epistemically ambiguous and shifting "we" as those who act together and who have acted together "we" as health professionals "we" as rule-maker and rule-follower "we" as subject of modal verbs "will," "can," "must," and "need to" "we" as regulatory authority In the following analysis, we provide examples and analysis of each of these main categories and sub-themes of "we" usage and draw on our theoretical framework to discuss their meanings and functions in the rhetorical enactment of Henry's relational ethos. --- Analysis Inclusive "We" The most obvious and prominent use of "we" in Henry's discourse is an inclusive type, referring to both speaker and audience. In this primary usage, "we" seems intended to generate a sense of togetherness and solidarity that supports an ethos of communal strength and coping. It evokes Henry's affiliation with her audience and supports her character as a leader who can bring people together in an empathetic and motivating way. These meanings are reinforced by how the term "we" often occurs in association with the terms "community" or "communities" and "together," as well as the terms "province" and "BC." The excerpts included in Table 3 illustrate how "we" typically refers to all British Columbians. Henry also frequently articulates variations of the ubiquitous COVID-19 slogan "we're all in this together," especially in briefings delivered during the first wave (March to June 2020). Using "we" in this way suggests 1 we all can stand proud knowing we're doing our bit and that we are holding the line for our families and our communities here in BC (Apr 4) 2 we need to continue together to do our part to protect our province (Jun 23) 3 we can and must continue to protect our families, our loved ones, and our communities across the province and to do this we need to continue to do this together (Jun 15) 4 we are in this together / we're all in this together (recurring) 5 we are standing united (Apr 18) / we are united (Aug 31, Sep 3) 6 we are all in the same storm (Nov 25) solidarity between Henry and her implied audience, as well as -importantlyamong audience members themselves. Consonant with Billig's [1995] concept of "banal nationalism," this usage fosters a sense of provincial pride that motivates citizens to "hold the line" against COVID-19. Rather than simply stating a static condition of what "we" are, Henry more often links this communal "we" with action, with what "we" are doing or need to do. These expressions emphasize both the ability and the obligation of "we" to work together and do "our part" for "our province" (see Table 4 below). Linking "we" with verbs and phrases denoting action reinforces a dynamic ethos of communal strength and coping. "We" as those who act together 1 we are working together (Apr 2) / we're all working on this together (Apr 11) 2 we're doing our bit... we are holding the line for our families and our communities (Apr 4) 3 we are making a difference together (Apr 18) / together we are making a difference (Dec 17) 4 we're getting through this together (Apr 21) 5 the work that we are all doing (June 9) / the work we're doing together (Oct 13) 6 we are making a difference through our actions (Sep 24) 7 together we are using our layers of protection (Dec 17) "We" as those who have acted together 1 the groundwork that we have done together (Apr 27) 2 we have shown that we can work together (Apr 29) 3 we've worked hard to flatten our COVID19 curve in BC (Jul 6) 4 the measures we have taken collectively (Aug 6) 5 the work we have all done together (Sep 3) 6 we have done our part (Oct 22) Following the initial phase of the province's pandemic response, Henry also often refers, using the past tense, to what "we" have done (see Table 4). By commending how "we" have acted together in the past, these kinds of statements construct Henry's ethos as a leader who recognizes and appreciates the efforts of her constituents. References to what "we" have already done also establish a shared history of the community's ability to do whatever is needed to control the pandemic, with the motivational implication that if "we" have been able to do this in the past, "we" can do it again. The frequent articulation of "we" with the modal auxiliary verbs "will," "can," "must," and "need to" when describing the actions of British Columbians adds more layers to how inclusive "we" constructs an ethos of communal action and coping (see Table 5 --- below). Table 5. Sample Excerpts: "We" as Subject of Modal Verbs "Will," "Can," "Must," and "Need to". Modal verb "will" (72 occurrences) 1 we will get through this together (recurring) 2 we will work together to support each other (Apr 8) / we will work together and we will get through (Nov 27) 3 we will make it through this together (May 16) 4 we will do this together (May 23) Modal verb "can" (119 occurrences) 1 we can all make a difference (Mar 24) 2 we can do it together (Mar 27) / we can do this, we can get through this together (Sep 10) 3 we can work together (Apr 29) 4 together, we can make it through this (May 12) Modal verbs "must" and "need to" (63 and 345 occurrences) 1 we need to work together to continue to make a difference (Apr 29) 2 we need to continue to do all we can to protect our families and our communities (May 21) 3 we must all do our part to protect ourselves and our communities (June 11) 4 we need to continue to do our part together (Oct 22) 5 this is the road we must walk and we must walk it together (Nov 7) 6 we need to push back on COVID-19 united and together (Nov 30) With statements that include the expression "we will" (see Table 5), Henry both presumes and enacts a strong leadership ethos by authoritatively declaring what "we" will do -assuming "we" continue to "work together" and "do our part." The modal "will" communicates Henry's judgment of what will likely be the case, rather than the absolute certainty of what is going to occur [Palmer, 2001]. "We will" thus functions as a confidence-building rallying call for the audience to continue complying with public health measures. In the recurring expression "we can" (see Table 5), the modal "can" functions epistemically to establish what is possible in terms of communal action, as well as dynamically to index the ability of "we" to successfully perform the called-for actions [Palmer, 2001]. Here again, we see how expressions with inclusive "we" function in a highly suasory way, by commending "our" capabilities and asserting Henry's confidence that collectively "we" are able to do what is needed to successfully navigate the pandemic. The modals "must" and "need to" occur even more frequently in Henry's descriptions of the actions "we" do "together," especially during the second halves of her briefings (see Table 5). Rather than emphasizing what is likely and possible or what "we all" have the ability to do, these modals merge a sense of empirical necessity with moral obligation [Palmer, 2001], thus framing required action as, also, conscientious action [Palmer, 2001;Giltrow, 2005]. These statements address an inclusive "we" whose health and safety requires the community as a whole to "push back on COVID-19 united and together." They also constitute "we" as a community with an inherent sense of ethical duty to act "together" to "protect our families and communities." These various uses of inclusive "we" foreground arête and eunoia by signalling Henry's empathetic affiliation with her audience, predicated on their shared identity of proud and committed British Columbians who have an inherent pro-social moral disposition motivating them to "work together" on behalf of "our community." Conversely, the relational ethos that inclusive "we" communicates can also be interpreted as socially distancing. Henry's authoritative assertions of what "we" are, will, can, and must do index a hierarchical relationship between her position as public health-community leader and the audience's position as those her discourse seeks to motivate. In a patronizing, hearer-dominant sense [De Cock, 2016], inclusive "we" thus can be seen as encouraging mainly identification among audience members, rather than between Henry and the audience. --- Exclusive "We" Along with these types of inclusive "we," an exclusive type of "we" also occurs consistently in Henry's COVID-19 updates. These instances of "we" are exclusive because they appear to refer to Public Health experts and herself but not to the audience of British Columbians to whom the updates are addressed. Instead of joining speaker and audience, this form of "we" distinguishes Public Health knowledge and actions from those of citizens. Exclusive "we" occurs most frequently in the "facts and figures" portion of the regular public briefings, as well as in longer modeling updates delivered approximately once a month. Our analysis of exclusive "we" shows several distinct though interconnected meanings. Exclusive "we" may refer to: i) Henry and her Public Health team as scientific experts and researchers; ii) BC's Public Health professionals as healthcare providers and managers; or iii) Henry singularly as the province's legally-authorized public health regulator. --- "We" as research experts The most obvious uses of exclusive "we" occur in recurring phrases referring to the pandemic-related data and scientific knowledge which Henry presents (see Table 6 below). This usage is especially prominent during the longer data-intensive modeling updates. Generally, the exclusive "we" of public health expertise occurs in conjunction with verbs pertaining to scientific knowledge and processes, such as "know," "understand," "learn", "monitor," "watch," and "collect" (see Table 6). These statements characterize "we" as having specialized knowledge (e.g., concerning genomic sequencing) and as conducting careful and diligent processes of COVID-19 data collection, analysis, and modeling. As well, when used with the verb "learn," exclusive "we" often highlights the evolving nature of scientific knowledge about the virus, as seen in excerpts 1 and 3 (Table 6). With statements like these, Henry explicitly acknowledges the reality of scientific uncertainty, while also emphasizing the extent of what "we have learned." These kinds of explanations contribute to her ethos by demonstrating honesty and humility to her audience, as well as a commitment to continuous learning. As a whole, uses of exclusive "we" like those illustrated in "We" as regulator A third type of exclusive "we" foregrounds a legal-political ethos of authoritative governance. This "we" refers to Henry's identity as the province's public health regulator with the legal authority to declare public health emergencies and issue orders. Of all uses of exclusive "we," this "we" constructs the greatest distance and power asymmetry between Henry and her implied public audience. Table 8. Sample Excerpts: "We" as Regulatory Authority. 1 we have issued an order [for travelers to self-isolate] I'm in a class of people which is just the legal way of saying that this is voluntary unless you don't do what we say and then we have the ability to enforce an order so it is something that we take very seriously and we are willing to enforce if we need to (Mar 17) 2 we would not be easing these restrictions if we did not feel we could do so safely (May 19) 3 we have established clear rules for safe social interactions here in British Columbia (Jun 29) 4 issuing orders is not something that we do lightly, it is our last resort (Sep 8) 5 orders are... what we need to put in place right now to address those areas where we were seeing transmission (Nov 9) Variations of "we put in" to describe issuing orders we are putting these measures in place (May 9) the measures that we've put in place (Aug 13) the provincial orders we have put in place (Oct 26) we put in the restrictions (Dec 23) we put in our most recent public health orders (Dec 23) For example, in excerpt 1 of Table 8 (above), an exclusive "we" issues orders that citizens ("you") are required to follow. Although the word "voluntary" evokes a non-coercive ideology of social trust and somewhat mitigates the authoritarian character of "we," the phrase "unless you don't do what we say" clarifies that what "we say" is, in reality, a non-negotiable regulation. Later in September, Henry's communicative style similarly justifies and mitigates the imposition of renewed public health regulations when she states that "issuing orders is not something that we do lightly, it is our last resort." By early November, as the second wave took hold, she reluctantly confirmed that "orders are... what we need to put in place right now" (Nov 9). Here, "we" is characterized as obliged to take action in a way that is not arbitrary but situationally necessary and responsive. Unlike uses of "we" to refer to the province's whole Public Health team, when Henry uses "we" with reference to "the measures we have put in place" or to an order "we have issued," this is best understood as a singular "royal we." As BC's Provincial Health Officer, Henry was the individual official who had the authority to issue, as well as rescind, public health measures. Although she occasionally uses "I" in these authoritative statements (see excerpt 1, Table

This paper investigates the multiple meanings and functions of the pronoun "we" in COVID-19 public updates by British Columbia's acclaimed Provincial Health Officer Dr. Bonnie Henry in 2020. Our rhetorical case study shows how "we" contributes to Henry's relational ethos by attempting to foster a communal identity with her implied audience while also distinguishing public health expertise, actions, and authority from citizens' knowledge and actions. Ambiguous uses of "we" blur the line between the knowledge and responsibilities of "we" in public health and "we" as citizens. Overall, our rhetorical analysis demonstrates the significant but ambivalent role this pronoun can play in building relations of social trust among citizens, experts, and institutions within public health and science communication contexts and it suggests the importance of judicious pronoun usage when communicators strive to foster these relations.

't do what we say" clarifies that what "we say" is, in reality, a non-negotiable regulation. Later in September, Henry's communicative style similarly justifies and mitigates the imposition of renewed public health regulations when she states that "issuing orders is not something that we do lightly, it is our last resort." By early November, as the second wave took hold, she reluctantly confirmed that "orders are... what we need to put in place right now" (Nov 9). Here, "we" is characterized as obliged to take action in a way that is not arbitrary but situationally necessary and responsive. Unlike uses of "we" to refer to the province's whole Public Health team, when Henry uses "we" with reference to "the measures we have put in place" or to an order "we have issued," this is best understood as a singular "royal we." As BC's Provincial Health Officer, Henry was the individual official who had the authority to issue, as well as rescind, public health measures. Although she occasionally uses "I" in these authoritative statements (see excerpt 1, Table 8), typically she uses "we" when referring to the "putting in" of orders. This diminishes the impoliteness of imposing regulations on citizen behaviour, while also diffusing individual responsibility for the imposition. --- Ambiguous and shifting "We" Thus far, we have looked at instances where inclusive and exclusive meanings of "we" are fairly easy to distinguish. Additionally, throughout the updates, Henry makes statements which blur the distinction. This occurs both through the inherent ambiguity of specific instances of "we" as well as shifts within single statements of what appears to be, on one hand, an exclusive meaning and, on the other, an inclusive meaning. In this section, we focus on how these shifts and ambiguities occur in relation to the kinds of knowledge and action attributed to Public Health and those attributed to all British Columbians. --- Epistemic ambiguities and shifts Who "we" refers to when used as the subject of epistemic verbs such as "see," "look," "learn," "understand," and, especially, "know" is quite often ambiguous. While some occurrences of the frequent collocate "we know" seem to refer exclusively to public health expertise, at other times it is unclear whether Henry is referring to what "we" as experts know or whether she is referring to, and thus claiming, a kind of common knowledge among all British Columbians. For instance, in the excerpts about what "we know" in Table 9 (below), does "we" refer to knowledge possessed by Public Health experts or to what Henry presumes as being common knowledge? Table 9. Sample excerpts: Who is the "We" Who Knows? 1 we know... that COVID-19 is in all of our communities (May 26) 2 we know that children... are less likely to get COVID-19 (Jun 4) 3 we know that across the country we are continuing to see cases (Jun 4) 4 we know that COVID-19 is going to be in our communities for some time to come (Aug 4) 5 we know that different interactions require different precautions (Sep 3) 6 we know that when these COVID-19 safety plans are followed... that we don't see transmissions (Oct 26) A slightly different blurring takes place within single statements where "we" in one instance quite clearly refers to all British Columbians but in another seems to be more or less exclusive to health experts. In these utterances, inclusive "we" is often associated with either the status or the actions of the provincial community as a whole, whereas ambiguously exclusive/inclusive forms of "we" are associated with knowledge (see Table 10 below). Table 10. Sample Excerpts: "We" Who Know and "We" Who Act and Experience. 1 we are trying to keep you informed... and [ensure] that we all have the information that we need to take the actions to protect our families and our communities (Apr 13) 2 this is the updated model that we have used to help us understand what we are doing, how we are getting along with physical distancing (Jun 4) 3 we estimate that we are at somewhere about 65% of normal... so we still have it in our hands and in our actions to bend our curve back down (Sep 3) 4 we know the respiratory season is coming, and we need to all continue to make those choices that keep our communities. In excerpts 1-3 (Table 10), the shift between exclusive and inclusive referents is fairly easy to distinguish: "we are trying to keep you informed" is exclusive and "we all have the information that we need" is inclusive; "the updated model that we have used" is exclusive and "how we are getting along" is inclusive; "we estimate" is exclusive and "we still have it in our hands" is inclusive. In these excerpts, the knowledge-making activities of "we" as Public Health experts provide insight on how "we" as British Columbians are faring and reinforce the importance of all British Columbians continuing to take the "actions" recommended or required by Public Health. The asymmetry of this relationship between public health expertise and citizen action is obscured by the use of the same pronoun to refer to both groups. In other statements, although the "we" who is experiencing the pandemic and taking action seems inclusive, it is more ambiguous whether the "we" who knows is exclusive or inclusive. For instance, in excerpt 4 (Table 10), the "we" who needs to take action ("make those choices") is clearly inclusive but the "we" who knows that respiratory season is coming could be exclusive or inclusive -is Public Health communicating its expert knowledge to citizens, or is Public Health asserting that this is common knowledge? A similar ambiguity occurs in excerpt 5. Increasingly during the second wave, the "we" who knows seems to be the same as the inclusive "we" who acts. This occurs most evidently through the recurring phrase "we know what we need to do" (see excerpts 6-7, Table 10). That "we" who "know" are the same as "we" who "do" becomes fully evident with Henry's statement near the end of December: "Every person in our province now is a COVID-19 prevention expert. We know what we need to do to protect those around us." This characterization of citizens as "prevention expert[s]" potentially bolsters Henry's relational ethos because it praises her audience's knowledge and fosters identification between the exclusive "we" of Public Health experts and the inclusive "we" of all British Columbians. However, the kind of expertise Henry attributes here to citizens is limited to preventive behaviours. As well, linking the claim that "every person" is a "prevention expert" with the statement "we know what we need to do" positions citizens as having both epistemic and behavioural responsibility for managing the pandemic, while implicitly diffusing Public Health responsibility. "We" as rule-maker and rule-follower Our analysis also revealed shifting and ambiguous uses of "we" in statements which combine the exclusive "we" of Public Health in its regulatory (rather than knowledge-making) identity with an inclusive "we" designating those who are expected and required to follow regulations (see Table 11 below). Table 11. Sample Excerpts: "We" Who Make the Rules and "We" Who Follow the Rules. 1 we're asking people to take voluntary steps to help us in our community and while they are voluntary there is an expectation that we will do our civic duty to do our best around this to proactively protect our communities and our families (Mar 18) 2 we need to continue to take these actions that... we've directed you to take (April 2) 3 we are making restrictions... around the importance of making sure that if we are ill we stay home (May 26) 4 we are following the rules that we have established for those safe social interactions (July 6) 5 we need to all continue to do the right thing from the get-go so that we don't need to use the enforcement measures that we have for our use now (Aug 24) 6 we know that the majority of people in BC are following our public health orders that we have in place (Dec 29) Excerpt 1 (Table 11) moves between exclusive and inclusive meanings when discussing the "steps" that Public Health expects, though has not yet ordered, citizens to take. The initial "we" refers in an exclusive sense to what Henry and BC's Public Health authority are asking citizens to do, while the second "we" refers inclusively to what all British Columbians are expected, or being asked (pseudo-voluntarily) to do. The statement "we need to continue to take these actions that... we've directed you to take" (excerpt 2) similarly shifts between the communal "we" of all British Columbians who take actions and the exclusive, regulatory "we" who issues directives to the communal "we." The other excerpts in Table 11 likewise ambiguously combine an exclusive regulatory "we" with an inclusive regulated "we." Though logically disconcerting, using the same pronoun to refer to those who regulate and those who are regulated blurs the contrast and power asymmetry between these two groups, giving the impression that the "we" who makes rules and the "we" who follows rules are one and the same. It draws attention away from Henry's singular, authoritative ethos as Public Health regulator and reinforces her identification with the dominant communal "we" discussed in the first part of this analysis. It also implies that the rules themselves have been generated by, or at least belong to, this communal "we," rather than being imposed by an external authority. --- Concluding discussion Although at first glance the pronoun "we" may seem like a small and therefore inconsequential feature of Henry's public health communication, our analysis shows that it plays a significant role in constituting her ethos and constructing strong but shifting and ambiguous relationships with her implied public audience. While the particularities of our case study cannot be simply generalized to other contexts, our findings nonetheless both confirm previous research on the persuasive power of "we" to foster a sense of communal identity between expert and lay audiences (eg Carlo, Luzon, Tian, Tu, Beason, Varpio) and also show that, conversely, it can function to delineate hierarchically the exclusive identity of experts or health officials as distinct from the non-expert audiences they aim to address. The referential fluidity of "we" allows it to contribute in multi-faceted, complex ways to the rhetorical construction of ethos and to activating both symmetrical and asymmetrical "relationships among different social positions" [Condit, 2019, p. 177]. In our view, this rhetorical complexity indexes the hybrid set of identities and exigencies that inform the communication practices of public health officials during times of crisis. Most notably, we see how use of inclusive "we" can cultivate a pro-social, communal sense of solidarity as a motivational basis for acting "together" to protect the health of all community members. We think that inclusive "we" contributes to Henry's eunoia by signalling empathetic affiliation between herself and the audience as well arête by communicating a strongly communitarian ethic that she both presumes and asks the audience to share. Notably, the call to collective action that pervades Henry's updates characterizes inclusive "we" not only as capable of performing the necessary actions, but also as situationally and morally obligated to do so, thus conferring both behavioural and ethical responsibility on British Columbian citizens. Paradoxically, while this framework draws on the identity of an altruistic, communal "we" as the motivation for citizen action, the required behaviours are individual. "Doing our part" does not really mean "we" together doing one thing; it means, more precisely, each one of "us" acting, at an individual level, according to public health guidelines. In this sense, the communal ethos that inclusive "we" helps to generate obscures the essentially individualistic nature of the required behaviours. Additionally, Henry's repeated exhortations that "we are all in this together" and "we need to work together" also implicitly constitute the citizens who make up this communal "we" as all equally responsible for and capable of acting according to public health guidelines. As critics have noted, this framing of the situation deflects attention from the significant socio-structural differences in citizens' pandemic experiences and inequities in their capacities to effectively implement public health guidelines [Bowleg, 2020;Sobande, 2020;Speed, Carter & Green, 2022]. By eliding these important differences, inclusive "we" in Henry's discourse, and likely the discourse of other public health officials, risks inscribing an exclusionary form of pandemic privilege. The less dominant but nonetheless substantive uses of exclusive "we" throughout the public updates likewise indicate this pronoun's diverse inflections in rhetorically shaping Henry's relationship with British Columbian citizens. By contrast with the sense of proximity between speaker and audience signalled by inclusive "we," exclusive "we" functions to distinguish Henry and her Public Health team from the audience in terms of expertise, healthcare provision, andmost strongly -regulatory authority. Although these uses foreground an asymmetrical relationship between speaker and audience, they nonetheless can be seen as contributing to phronesis, by demonstrating the expertise that "we" possess and share with the audience, and to arête, by emphasizing the diligent, careful, and responsive work that "we" as healthcare providers perform for citizens. The hierarchical distance between speaker and audience manifests most evidently in the use of "we" to designate Henry in her capacity as the province's Public Health regulator. Here, the singular "we" contributes to Henry's ethos by signalling this high status, while also mitigating the imposition of the orders she issues and diffusing responsibility for those orders. While inclusive and exclusive uses of "we" are typically fairly easy to determine in Henry's discourse, the boundaries between the two are also quite frequently blurred. Whether intentional or unintentional [Harwood, 2005], this blurring ambiguates who exactly "we" are, and therefore who knows what and who is responsible for what. The conflation of exclusive and inclusive "we" strengthens Henry's relational ethos by implying shared knowledge, values, and responsibility in a way that dissolves boundaries between experts and non-experts, between public officials and citizens, and between governing bodies and those they govern. However, we would argue that these ambiguously exclusive/inclusive uses foster the impression of a symmetrical, dialogic relationship between Public Health and citizens that obscures its inherent power asymmetry. These uses also disperse responsibility for knowledge, action, and regulations into the broader communal "we." As a whole, our analysis shows that the pronoun "we" can contribute in vital and varying ways to the rhetorical construction of relational ethos as a "bridge" between public health officials and their implied citizen audiences [Condit, 2019]. At the same time, our study suggests that the role of "we" in building relations of social trust among citizens, experts, and institutions [Bucchi, 2021] may also be ambivalent and problematic. An important limitation of our analysis is that it only illuminates the kinds of audience identities and relationships that are evoked by Henry's own discourse; an audience reception study of how British Columbians responded to this discourse (e.g., on social media) would therefore complement our study in valuable ways. Further research in other public health and science communication contexts internationally could likewise enrich and complicate the field's understanding of the role pronouns play in constructing ethos and expert-citizen relationships. Overall, we think that our study indicates the value for the fields of science and health communication of investigating how trust and relational ethos may be fostered through the rhetorical practice of pronoun usage by experts and public authorities. Importantly, our study nuances the dominant view that "we" functions primarily (if not entirely) to encourage affiliative identification between speaker and audience, by illustrating how it can also inscribe asymmetrical forms of rhetorical distance as well as how apparently inclusive forms of "we" may problematically and possibly unethically obscure this distance in some situations. For science and health communicators, our study suggests that "we" should be used judiciously, with careful attention to the distinction between inclusive and exclusive meanings and the different kinds of relationships between experts and citizens that these meanings imply. --- Authors Philippa

This paper investigates the multiple meanings and functions of the pronoun "we" in COVID-19 public updates by British Columbia's acclaimed Provincial Health Officer Dr. Bonnie Henry in 2020. Our rhetorical case study shows how "we" contributes to Henry's relational ethos by attempting to foster a communal identity with her implied audience while also distinguishing public health expertise, actions, and authority from citizens' knowledge and actions. Ambiguous uses of "we" blur the line between the knowledge and responsibilities of "we" in public health and "we" as citizens. Overall, our rhetorical analysis demonstrates the significant but ambivalent role this pronoun can play in building relations of social trust among citizens, experts, and institutions within public health and science communication contexts and it suggests the importance of judicious pronoun usage when communicators strive to foster these relations.

Introduction Type 2 diabetes (T2D) is one of the leading causes of death and disability in the USA [1]. Low-income racial and ethnic minorities are disproportionately affected by T2D, experiencing higher incidence, prevalence, comorbidity (e.g., hypertension, dyslipidemia), and complications (e.g., blindness, end-stage renal disease, limb amputation) compared to non-Hispanic whites [1][2][3]. According to the 2014 National Diabetes Statistics Report, racial and ethnic minorities are almost twice as likely to be affected by diabetes, with even greater disparities among racial and ethnic minorities living in poverty [1, 4•]. These disparities have been largely attributed to patient-level lifestyle factors (e.g., poor diet, physical inactivity [5,6]) and health system factors (e.g., disparities in healthcare access and medical care [7]). Subsequently, efforts to prevent and control T2D within these populations have targeted access to clinical services and/or behavioral risk factors in clinical settings. However, clinical interventions have often demonstrated mixed findings and improvements have been difficult to sustain over time [8]. Others have cited environmental conditions as a primary driver of health disparities and have targeted features of the built environment [9,10]. The built environment refers to the humanmade or modified structures used in daily life, as compared to naturally occurring aspects of the environment [11••]. Numerous studies have demonstrated relationships between the built environment and health [9,10]. For example, a "walkable" neighborhood is positively associated with increased physical activity and reduced prevalence of obesity [12,13]. Access to healthy food within the built environment has been associated with reduced diabetes distress and improved diabetes self-care [14]. Signs of physical disorder within the built environment (e.g., dilapidated housing, few recreational spaces, broken glass on the ground) have been associated with unhealthy behaviors (e.g., physical inactivity, poor dietary habits) and poor health outcomes (e.g., poor diabetes control, hypertension) [11 --- The Role of Social Disorder Although physical disorder and features of the built environment are critical to understanding environmental effects on T2D, social disorder may play an equally important role among low-income racial and ethnic minorities [11••, 15•, 16•]. Social disorder, the disruptive social conditions that influence health, may take the form of crime, noise, loitering, public drinking, drug use, conflicts, and panhandling-any of which may create a sense of unrest or danger [17•]. There are two likely mechanisms through which social disorder may impact diabetes-related outcomes. First, perceptions of social disorder may discourage residents from engaging in healthy behaviors (e.g., safety concerns may limit outdoor physical activity, perceived discrimination may influence assessments of which grocers are welcoming), thus increasing the risk of diabetes onset and poor diabetes control. Second, perceived fear and chronic stress can lead to allostatic load-the overactivity or underactivity of allostatic systems (e.g., hypothalamic-pituitary adrenal axis, immune system), with resultant inflammation and stress hormone patterns that can increase the risk of obesity and diabetes [17•, 18•]. We note, however, that impressions of social disorder may be subjective and variable. For instance, seeing a group of people conversing outside might motivate one person to take a walk (e.g., engage in neighborly socialization, thus mitigating stress response mechanisms) and a different person to stay inside (e.g., avoid perceived dangerous loitering, thus triggering stress response mechanisms). Social disorder is usually conceptualized within the social environment, defined as the immediate social setting in which people live, including interpersonal behaviors, cultures, and attitudes [19]. However, we believe that the current understanding of social disorderlimited to the immediate social setting-does not fully capture the many dimensions of disorder experienced among low-income racial and ethnic minority communities. In this paper, we propose a multilevel ecosocial model (Fig. 1) to describe potential mechanisms through which social disorder may contribute to diabetes disparities in lowincome racial and ethnic minority communities. We also discuss the results of a narrative literature review, conducted to identify evidence supporting or refuting the proposed associations between social disorder (broadly defined) and diabetes-related health outcomes within vulnerable communities. We conclude by providing recommendations for future directions in research and interventions. This paper provides a theory-driven summary of some of the most important literature in the field and describes how major themes inform our multilevel conceptual model on social disorder and diabetes-related health outcomes. --- Methods We searched PubMed, PsychInfo, Scopus, and SocIndex databases using a combination of MeSH terms and keywords: type 2 diabetes mellitus, obesity, cardiovascular disease, disparities, community, social environment, built environment, neighborhood disorder, social disorder, physical disorder, prejudice, discrimination, racism, policing, crime, violence, residential segregation, social cohesion, social capital, and collective efficacy. We primarily included papers published from 2010 to 2015 to highlight the most current research. We did not include studies evaluating social network theory or network analysis, as these studies were outside the scope of this review. It is worthwhile to note that, given the focus on social disorder, we excluded studies describing positive social aspects (e.g., social support) and health. We also make a distinction between social disorder and general definitions of social determinants. Social determinants refer to the place-related conditions that influence health, both positive and negative [20, 21••]. Social disorder, as defined in this paper, refers to the disruptive social conditions on all levels, including place, which may influence health. --- Conceptual Model Figure 1 summarizes several plausible mechanisms through which social disorder may contribute to diabetes disparities. First, structural and neighborhood-level influences, such as economic systems, macrosocial inequalities, and residential segregation, mutually reinforce each other. Economic deprivation can result in spatial inequalities in the built and social environment (Fig. 1, left), which can, in turn, reinforce residential segregation and concentrated poverty. Second, neighborhood-level influences can impact individual behaviors and processes. The built environment (e.g., a lack of healthy food options) can impede healthy dietary intake (Fig. 1, top), or the social environment (e.g., perceived safety concerns) can impede physical activity (e.g., by decreasing outdoor exercise within the community) (Fig. 1, bottom): both paths strengthen the pathways to obesity and comorbid disease. Alternatively, noise in the environment and stressors related to perceived threats of violence or perceived interpersonal discrimination (e.g., in interactions with police) could disturb sleep, thereby activating allostatic systems (Fig. 1, right) and facilitating obesity risk. Finally, personal and social health beliefs/behaviors operating at the individual and interpersonal level are related to health (Fig. 1, right). For example, interpersonal racial discrimination and social comparison can produce stress, which can subsequently result in the adoption of unhealthy eating behaviors and physical inactivity; both stress and maladaptive behaviors can lead directly to poor health. --- Structural Features of Social Disorder Fitzpatrick and LaGory suggest that obesity and related outcomes are a consequence of larger social structures, as evidenced by the highest rates of obesity within US states that have the lowest income, least education, and highest percentages of racial/ethnic minorities [21••]. Indeed, racial and ethnic minorities often live in communities affected by interrelated macrosocial and economic inequalities (Fig. 1, far left) with downstream consequences on health. --- Macrosocial Inequalities Low-income minority communities are more likely than white communities to have been shaped by large-scale biased policies and practices of governments (i.e., federal, state, municipal), institutions (e.g., banks), and associations (e.g., in real estate). For example, the 1968 National Advisory Commission on Civil Rights was created to understand the underlying causes of the 1967 race riots in major US cities. The report identified white racism and failed policies in housing, education, and social services at the state and federal level as the main causes of urban violence [22]. The tools that furthered racial segregation included racially restrictive covenants on housing, mortgage redlining, and urban renewala policy practice from 1949 to 1974 that displaced hundreds of thousands of households, the majority of which were nonwhite and poor [23]. During the 1990s and 2000s, in an environment of urban gentrification, some areas where mortgage capital had been withheld in earlier periods actually became flooded by high-risk variable rate mortgages and predatory lending schemes [24••, 25]. This influx of risky contracts facilitated a wave of foreclosures [24••], thus furthering the existence of social and physical disorder within lowincome minority communities. Currently, the presence of race-based retail redlining (i.e., stores and/or businesses choosing not to serve certain communities based on racial/ethnic composition) may contribute to limited healthy food options in a given neighborhood [26]. --- Economic Inequalities Beginning before and persisting beyond the Civil Rights era, large-scale economic and social forces have contributed to social disorder in urban neighborhoods [27]. For instance, over the past half-century, the USA lost hundreds of thousands of manufacturing jobs, leaving a landscape of economic and social devastation in many urban communities [28,29]. The homes of unemployed workers sat deteriorating or abandoned. The bars, religious congregations, and meeting halls that had served as social networking sites eventually collapsed, socially and sometimes physically. For many residents, it felt like work "disappeared" [30]. For others, jobs that paid a living wage in manufacturing were replaced by service-sector jobs with wages that could not sustain a household. Such economic transformation was formative in the current status of many communities with high proportions of unskilled workers who would have previously been employed in the manufacturing economy-quite often communities of racial/ethnic minorities. --- Neighborhood Organizational Features of Social Disorder Consequently, these communities can be characterized by economic deprivation and residential segregation (Fig. 1, center left). Place of residence is strongly associated with social position, socioeconomic status, and race/ethnicity. As such, the organizational features of place and its characteristics may be an important mediator between race/ethnicity and health disparities, particularly in chronic diseases such as diabetes. --- Residential Segregation Residential segregation is a neighborhood characteristic that can perpetuate health disparities [30]. It can be a consequence of micro-level discrimination (e.g., real estate steering practices) or broader systematic practices and structural policies (e.g., mortgage redlining, public housing policy). Racial segregation, even after adjustment for individual and familyrelated factors, has been associated with myriad health conditions, including obesity and T2D [31•, 32•, 33, 34••]. One observational study evaluated a racially integrated, economically homogenous community in Baltimore and found that disparities in the prevalence of diabetes dissipated (10.4 % blacks, 10.1 % whites) compared to national estimates (10.5 % blacks, 6.6 % whites) [35•]. Other research has challenged the notion that racial segregation is related to health. For example, Piccolo et al. examined potential associations between neighborhood racial composition and diabetes prevalence in Boston and found a large variation in the prevalence of T2D between Boston neighborhoods that could not be reasonably explained by racial/ ethnic composition or specific neighborhood attributes, such as socioeconomics, safety, or neighborhood disorder [15•]. Mixed findings in the literature about the relationship between residential segregation and health outcomes may indicate that a racially homogenous neighborhood by itself, regardless of which racial group is being evaluated, is not inherently an adverse feature [36]. Rather, health risks may develop when patterns of residential segregation are associated with poverty and neighborhood deprivation. --- Neighborhood Deprivation Neighborhood deprivation refers to neighborhood organizational features such as economic disadvantage, unemployment, poor education options, and substandard housing conditions [37, 38•, 39]. Thus, neighborhood deprivation reflects both social disorder (the focus of this paper) as well as physical disorder (as part of the built environment). Neighborhood deprivation (e.g., concentrated poverty, residential instability) has been associated with negative health outcomes, such as obesity, diabetes, and cardiovascular disease, as well as intermediate measures such as medication treatment adherence [31•, 34••, 40, 41]. One observational study examined the role of neighborhood poverty on racial disparities in diabetes prevalence (blacks vs. whites), using data from the US Census and the National Health and Nutrition Examination Survey (NHANES). Findings indicated that although neighborhood-level racial composition itself did not increase the odds of having diabetes, concentrated neighborhood poverty was associated with increased odds of diabetes prevalence for blacks (both poor and non-poor) as well as poor whites [34••]. An experimental study, part of the Moving to Opportunity (MTO) housing intervention, examined obesity and diabetes outcomes when households were given the opportunity to move from a high-poverty neighborhood to a low-poverty neighborhood [32•]. Individuals with the opportunity to move to a low-poverty neighborhood were less likely to become obese or develop T2D over the 14-year cohort study. Neighborhood deprivation is often reflected in a built and social environment that offers few healthy resources (Fig. 1), thus increasing individual risk for unhealthy behaviors [38•, 42-45]. Such neighborhoods, for example, may be less likely to have grocery stores and recreational facilities (components of the built environment) and might also be less likely to have community cooking classes or walking groups (products of social environmental factors such as social cohesion). --- Neighborhood Social Environment Features of Social Disorder The literature suggests that the neighborhood social environment (Fig. 1, center right) may affect health and contribute to disparities within low-income and minority communities [40]. Disorder in the neighborhood social environment includes factors such as poor social cohesion and collective efficacy, neighborhood crime and safety concerns, and policing and incarceration practices, which can promote unhealthy behaviors and induce physiological distress mechanisms that negatively impact health (Fig. 1, far right). --- Lack of Social Cohesion and Collective Efficacy Social cohesion is the "degree of connectedness and solidarity that exists among people living in a defined geographic area" [24••, 46•]. Collective efficacy is "social cohesion combined with shared expectations for social control, " a willingness of community members to trust, look out for each other, and intervene for the common good [24••]. The interpretation of collective efficacy has varied over time and between cultural groups. For example, what were perceived as "slums" full of social disorder (e.g., people sitting out on the porch) in the 1940s have been reinterpreted by some as brimming with collective efficacy; what some analysts interpreted as loitering, Jane Jacobs termed "eyes on the street" that foster community safety [47]. Recent literature suggests that low collective efficacy is associated with a variety of health outcomes that cluster at the neighborhood level, including T2D and cardiovascular disease Collective efficacy may reinforce positive attributes of a community, as well as counteract some of the negative factors. It may, for instance, serve as a buffer between large-scale detriments (e.g., racism, economic flux, poverty) and the community, and consequently reduce individual stress and protect health (Fig. 1, far right). A study of working class Italian Americans in Roseto, Pennsylvania, suggested that the common ground of cultural homogeneity fostered an "unconditional interpersonal support" that was associated with reduced stress and positive health outcomes [36,49]. Neighborhoods with high collective efficacy tend to have a higher degree of connectedness and interpersonal trust, which allows residents to advocate for themselves and reinforces health-promoting behaviors, such as walking and exercise [21••]. There is also evidence that collective efficacy might encourage healthy behaviors even in the presence of physical disorder (e.g., litter, loitering) or the existence of impediments (e.g., access to health foods). Using 10-year data from a population-based cohort study, Christine et al. found that higher levels of social cohesion in low-income neighborhoods was associated with a lower incidence of diabetes [50•]. Alternatively, in some neighborhoods with poor collective efficacy, the addition of businesses (e.g., grocers that offer healthy food options, fitness centers) or other amenities may not be understood as accessible for long term residents (e.g., due to price, brand selection) and can even lead to the displacement low-income groups by higher income groups (i.e., gentrification) [51], thereby perpetuating poor health outcomes among lowincome residents. --- Crime and Resident Safety Concerns Racial/ethnic minorities are more likely to live in neighborhoods associated with higher rates of crime and social instability [42]. Although we found no studies directly examining associations between neighborhood crime and diabetes, several studies indicate that perceptions about safety may influence or mediate obesity or obesity-related disease (Fig. 1, right). First, perceptions of crime may increase a resident's level of stress; chronic stress has been shown to increase allostatic load and stress hormones (e.g., cortisol) that increases the risk of developing T2D [17•, 52•]. Second, in high-crime neighborhoods, residents may be less likely to participate in physical activity (e.g., walking) [53], thus increasing an individual's risk for obesity and T2D. Third, perceived safety concerns may reduce a resident's willingness to leave their neighborhood to access amenities that promote health (e.g., parks, grocery stores, medical care) [54•]. For example, one study using California Health Interview Survey (CHIS) data found that people who reported living in an unsafe neighborhood were more likely to report delays in filling prescription medications [54•]. Fourth, the daily activities of residents who live in neighborhoods with high rates of crime can be tied to policing practices. For instance, controversial policing strategies, such as "stop and frisk, " may deter a sense of community identity and have downstream adverse consequences for collective efficacy and social order [55•, 56-59]. The idea, that the very practices thought to address social disorder (e.g., policing to counteract crime) may actually cause stress in particular contexts, underscores that various factors in our model may have counteracting effects (Fig. 1). --- Individual and Interpersonal Features of Social Disorder At the individual and interpersonal level, racial/ethnic minorities-especially in low-income communities-can experience social disorder through both externalized and internalized racism (Fig. 1, right). Externalized racial discrimination can take on overt forms, such as harassment, violence, and mistreatment-or it can take on more subtle forms, such as limited employment opportunities, poor quality medical care, or unconscious bias [60•, 61]. Internalized racism can take on the form of poor self-esteem, self-stereotyping, and social comparison (i.e., evaluating self-worth by comparing and contrasting personal attributes) [61,62]. Both reflect an important type of psychosocial stressor that can adversely affect health status [60•, 61]. --- Interpersonal Racial Discrimination Racial/ethnic minorities who report interpersonal racial discrimination are at increased risk for a variety of stress-related health outcomes and chronic diseases [61, 63•, 64]. Exposure to racial discrimination can lead to increased allostatic load, which has been linked to increased risk of obesity, hypertension, insulin resistance/diabetes, neuronal damage, immune disorders, and mental health disorders [18•, 52•, 61]. Moreover, individuals who report racial discrimination may be more likely to engage in negative coping strategies, such as unhealthy eating and cigarette use, thus increasing their risk for diabetes and diabetesrelated complications [52•, 64]. One study using Behavioral Risk Factor Surveillance System (BRFSS) data reported that, in unadjusted models, self-reported racial/ethnic discrimination in healthcare was associated with diabetes quality of care (e.g., HbA1c testing, foot examinations) and complications (e.g., diabetes-related foot disorders, retinopathy) [60•]. --- Social Comparison Recently, social comparison, a process of evaluating social worth based on comparing and contrasting personal attributes, has been described as a potential form of internalized racism [62]. Prior studies suggest that the activation of negative stereotypes among minority or stigmatized groups can create expectations, anxieties, and reactions that can adversely affect health [61]. One study of low-income Brazilians explored the relationship between social comparison and diabetes self-care, showing that social comparison among adult primary care patients often precipitated a negative social identity (e.g., "being ill, " "being different, " "being embarrassing"), which impacted both personal habits (e.g., medication adherence, nutritional intake) and experiences with diabetes (e.g., "being diabetic is leading a difficult life") [62]. Thus, among marginalized populations (e.g., low-income racial/ethnic minorities) who are at increased risk for negative social comparisons, the addition of chronic diseases such as diabetes may enhance perceptions of stigma and worsen social comparisons, with subsequent adverse health consequences. --- Discussion Racial and ethnic disparities in diabetes remain a serious public health issue. Although prior studies have investigated behavioral and environmental mechanisms that contribute to such disparities, we sought to collect and organize studies as part of a broader conceptual model describing the role of social disorder in diabetes-related health outcomes. We present evidence from the literature that social disorder can contribute to the disproportionate risk, prevalence, and outcomes of diabetes in low-income racial/ethnic minority communities. The notion that "place matters" -that where people live can determine their health statusis not new. Over the last decade, research has begun to show that place can be either protective or detrimental for health behavior and outcomes [11 However, place may be particularly meaningful for low-income racial/ethnic minority communities. For some, place represents an accumulation of social-structural factors, neighborhood features, and interpersonal interactions that produce social disorder. Racial and ethnic minorities disproportionately live in places with significant social disorder, which may meaningfully impede pathways to positive health outcomes. Thus, our expansion of the social disorder construct can serve as a guide to better inform points of intervention. Recently, there has been renewed interest in clinical interventions that address the social determinants of health. Health Leads is a non-profit organization that matches patients' unaddressed social needs to appropriate resources (e.g., housing, employment, support groups, food programs) [65]. CommunityRx employs local high school youth to conduct an annual inventory of businesses and organizations on Chicago's South Side [66, 67•, 68•]. The CommunityRx database integrates with electronic medical records to generate personalized, condition-specific referrals (HealtheRx) to resources near the patient's home [67•, 68•]. In 2015, CMS announced a 5-year, $157 million program to test a model called Accountable Health Communities (AHC) [68•]. This model combines screening for social needs with community service navigation and capacity-building to ensure that patients are not only connected to appropriate services, but community-based service partners are aligned and have adequate capacity to meet community needs [68•]. Health sectors are also participating in cross-sector collaboration to enhance health system-community linkages, and play a role in community development strategies to promote health. A recent report for the Robert Wood Johnson Foundation Commission to Build a Healthier America described emerging strategies to support cross-sector partnerships [69•], including the promotion of a "culture of health" as the new subjective norm. New policies enacted by the Affordable Care Act encourage health systems to consider not only individual-level patient outcomes, but also the population health of their communities. Health policy changes have bolstered preventive, non-clinical interventions that target the social and economic barriers to population-based health. Low-income racial and ethnic minorities will particularly benefit from interventions that target neighborhoods shaping individual risk [41,70,71]. One study of HOPE VI, a new generation of public housing, demonstrated an increase in neighborhood-based physical activity among housing residents [41]. However, the mass relocations and destruction of existing communities facilitated by HOPE VI may have countervailing effects on social cohesion and collective efficacy [72]. This illustrates the need for models, like the one we present here, to understand the complex interrelationships between social disorder and health, and the potential for unintended health consequences from health policies and interventions. Finally, the increased availability of individual-level technology (e.g., mobile phones, dashboard cameras) has increased documentation and public awareness of racially unjust policing and incarceration practices, especially in low-income minority neighborhoods [56][57][58][59]. Although few studies have directly examined the impact of neighborhood policing on health outcomes, resources are being developed to establish data sources in this field that may better inform law enforcement policies related to health. In 2015, President Obama launched the Task Force of 21st Century Policing to better understand issues surrounding neighborhood law enforcement, including its impact on mental and physical health [73•]. The Task Force also started a Police Data Initiative to increase data collection, improve transparency, and foster community trust [73•]. As we develop evidence-based strategies to address the health of at-risk communities, it will be increasingly important to address the impact of social disorder on activities that shape prevention and control of diabetes and other chronic diseases. It will also be important to systematically test interventions at multiple levels (i.e., structural, neighborhood, interpersonal, individual) where social disorder functions. Only then can we begin to align multilevel strategies-building on race, place, and poverty-to improve what matters most to minority patients in poor communities: daily life.

The recent resurgence of social and civic disquiet in the USA has contributed to increasing recognition that social conditions are meaningfully connected to disease and death. As a "lifestyle disease," control of diabetes requires modifications to daily activities, including healthy dietary practices, regular physical activity, and adherence to treatment regimens. One's ability to develop the healthy practices necessary to prevent or control type 2 diabetes may be influenced by a context of social disorder, the disruptive social and economic conditions that influence daily activity and, consequently, health status. In this paper, we report on our narrative review of the literature that explores the associations between social disorder and diabetes-related health outcomes within vulnerable communities. We also propose a multilevel ecosocial model for conceptualizing social disorder, specifically focusing on its role in racial disparities and its pathways to mediating diabetes outcomes.

Introduction Researchers in international luxury study culturally situated understandings of luxury across countries (Christodoulides and Michaelidou, 2022;Veloutsou et al., 2022), including cultural differences between Western countries and East Asia (Chadha and Husband, 2010;Schmitt and Lee, 2015;Seo et al., 2015;Shukla et al., 2015;Wong and Ahuvia, 1998). The emphasis on East Asia reflects the region's growing importance to the luxury industry. Forecasts predict that it will become the world's largest market for personal luxury goods, raising US$ 330-370bn in 2025 and growing at an annual rate of 10% (D'Arpizio et al., 2021). Consultancies expect China to reach more than half the global market value of luxury goods by 2025 (IPSOS, 2020;McKinsey, 2019;Statista, 2021). Because of its increasing commercial importance, luxury research focusing on China is growing (Jiang and Shan, 2018;Jin et al., 2015;Sinha, 2014). Current research uses the country, China, as the unit of analysis (Jin et al., 2015;Sinha, 2014), comparing it with other countries (Shukla et al., 2015;Yang et al., 2018). This approach suggests that Chinese luxury consumption is homogeneous, even though research shows nuance and complexity (Godart and Zhao, 2014;Jiang et al., 2022;Rovai, 2014Rovai,, 2016;;Siu et al., 2016;Zhang and Zhao, 2019). Consequently, one gap in the literature is the lack of research on the internal dynamics of Chinese luxury consumption. The current amount of luxury research on China does not reflect its commercial importance. Veloutsou et al. (2022, p. 385) observe that "[a]lthough China is en route to becoming the largest luxury market globally, it is interesting that its research output on international luxury marketing is almost five times smaller than that of the United States, and three times smaller than the research originating from the U.K." Whereas the previous quote refers to the output of luxury research by Chinese authors, a broader case can be made about the need for more research about China's luxury consumption dynamics due to its economic relevance for the luxury industry (McKinsey, 2019). This study focuses on the segment of post-1990s young Chinese adults whose consumption behavior differs from that of their parents and international peers because of their economic prowess, always-online presence, international mobility and taste for luxury consumption (Chung et al., 2009;Godart and Zhao, 2014;Jiang et al., 2022;Ngai and Cho, 2012). As the luxury consumption practices of the post-1990s generation in China lack sufficient research, this study proposes the following research question: What are the new forms of luxury consumption among the post-1990s generation of Chinese young adults? The paper reports on an in-depth qualitative study of how young adult luxury consumers who self-identify as Chinese engage in new forms of luxury consumption; specifically, their preferences for luxury brand collaborations. Table 1 contains examples of collaborations between luxury brands and odd non-luxury partners. This study contributes to unconventional luxury (Thomsen et al., 2020), new forms of luxury consumption (Christodoulides et al., 2021) and international luxury research (Veloutsou et al., 2022). Unconventional luxury emphasizes experiential (Holmqvist et al., 2020a;von Wallpach et al., 2020) and digital (Creevey et al., 2022;Holmqvist et al., 2020b) aspects of luxury. This paper extends this field by proposing alternative understandings that constitute the markers of singularity and exclusivity in luxury. Unusual collaborations constitute convey rarity and exclusivity because the collections will not repeat. For international luxury (Veloutsou et al., 2022), this study breaks down a generational approach to Chinese consumption; hence, adding nuance to existing research that approaches the country as a unit of analysis. Building upon the generational divide in Chinese luxury consumption (Byun et al., 2020;Godart and Zhao, 2014;Jiang et al., 2022;Zhang and Zhao, 2019), this paper conceptualizes how young Chinese consumers use luxury to navigate transient social media, ephemeral digital possessions and fleeting experiences. --- Conceptual framework The conceptualization of luxury in the literature is evolving (Christodoulides and Wiedmann, 2022). Early researchers studied the characteristics of luxury goods because these products do not conform to the economic logic of supply and demand. Luxury goods offer signals of social status, prestige and exclusivity (Grossman and Shapiro, 1988). Economists studying luxury goods focused on pricing, as items such as private jets and yachts were prohibitively expensive for most people. More recently, luxury has been conceptualized as a business strategy (Hoffman and Coste-Maniere, 2012), in which luxury managers create a sense of exclusivity by embeddding meanings of singularity and uniqueness into commercial offerings. For example, by emphasizing heritage, provenience and aesthetics, luxury brands sell cosmetics and personal care products as luxury items even if these product categories are not intrinsically luxurious. As a result, luxury managers use tools, including a high price tag, to convey a sense of exclusivity, the characteristic without which luxury cannot exist (Kapferer and Bastien, 2009, p. 315). Table 2 proposes a glossary of concepts. --- Conventional luxury The conventional understanding of luxury in the literature conceptualizes luxury goods as high-quality products sold at high prices for wealthy buyers (Ko et al., 2019), conveying superior quality (Dubois and Duquesne, 1993). Luxury brands are crucial for the luxury strategy as they are symbols imbuing authenticity, prestige and exclusivity to build a sense of uniqueness worthy of commanding premium prices (Ko et al., 2019, p. 406). As a business strategy (Kapferer, 2015;Kapferer and Bastien, 2017), luxury conveys a sense of singularity through the following three dimensions: (1) timelessness, (2) inaccessibility and (3) tradition. Timelessness refers to the permanence of luxury over time, resisting the whims of fashion fads and the passage of time. "The perfect luxury product should in fact be timelessness, a concept that is antithetic to that of obsolescence called on by price reduction" (Hoffman and Coste-Maniere, 2012, p. 94). Whereas fashion is associated with changing trends, luxury has been considered timeless, because durable materials, fine artisanship and classic design signal wealth permanence. The symbols of luxury encapsulate lasting wealth and ageless The cosmetics brand launched a brand extension to celebrate "singles' day" Online gaming and digital culture Source(s): Own creation based on the press releases by the brands ("Balenciaga j Fall 21 j Press", 2021; Galvan, 2018;Lozano, 2019;Wu, 2023) Table 1. --- Examples of luxury design collaborations --- Unconventional luxury brand collaborations esthetics. For example, precious stones and rare metals signal that an object will retain financial value over time and classic designs signal resilience to changing tastes. Exclusivity is "one of the keywords of luxury" (Kapferer and Bastien, 2012, p. 85). There is no luxury without "the feeling of exclusivity" (Kapferer and Bastien, 2012, p. 51), which is "the feeling of having one's social standing reinforced through the brand, the feeling of being a special person" (Kapferer and Bastien, 2012, p. 20). Veblen's theory of the leisure class is foundational in luxury as it describes how members of the upper class use luxurious goods and leisure activities to show their wealth (Veblen, 2009). Today, luxury brands remain tokens of status, and thus, "the conventional understanding of luxury remains centered on how wealthy consumers display signals to convey social exclusivity" (Holmqvist et al., 2020a, p. 503). Inaccessibility operationalizes exclusivity by using taste regimes and insider codes to establish social distinctions and to create boundaries between insiders and outsiders. The notion of "taste regimes" means the concerted judgments of esthetic qualities, consecrated through cultural competence and social credit (Arsel and Bean, 2013). For example, coveted groups use taste regimes, like art appreciation, to signify social class. When taste regimes remain stable through time, they become tradition, which is an important dimension of luxury (Holmqvist et al., 2021). It codifies symbolic meanings, solidifying them through rituals that pass down through time (Langlois, 2001). Signifiers can become traditions because they solidify as they pass from the past to the present, reifying unwritten rules. Research shows the importance of tradition for luxury consumption (Schade et al., 2016) because the enactment of long-standing social rituals encapsulates the present as an outcome of the past (e.g. heritage and country of origin). --- Unconventional luxury Unconventional luxury is a recent stream (Thomsen et al., 2020) that studies alternative ways to convey exclusivity and uniqueness in digital and experiential contexts. For example, digital possessions can be perceived as luxurious even though they lack the conventional markings of permanence (Holmqvist et al., 2020b), and experiences can be luxurious by enabling hedonic escape from daily routine (Holmqvist et al., 2020a). At its core, unconventional luxury studies the paradox of rarity. Whereas most people should be aware of luxury brands, the offering must appear out of reach (Kapferer and Bastien, 2017;Kapferer and Valette-Florence, 2022). Luxury brands cast an aura of unattainable rareness by evoking prestige that is aspirational and singular. For Kapferer (2012), one way to create rarity is the cult of the designer, in the name of the creator and its singular esthetics convey a sense of uniqueness. New forms of luxury consumption are emerging in the sharing economy (Christodoulides et al., 2021;Holmqvist et al., 2020b). Some examples include co-ownership and second-hand luxury (Bardhi et al., 2020). A puzzling aspect of luxury in the sharing economy is its temporality (Christodoulides et al., 2021, p. 93), like leasing a supercar for an evening or posting fleeting experiences on Instagram. Luxury brand extensions at accessible prices are other emerging forms of luxury (Kapferer and Laurent, 2016). This democratization of luxury requires unconventional strategies to create and maintain a sense of rarity (Rosenbaum et al., 2021), including personalization, limited collections and partnerships. --- Luxury consumption in China: a generational divide The luxury industry is increasingly focusing on China. For the consultancy firm McKinsey, "Chinese consumers are now the engine of worldwide growth in luxury spending" (Kim et al., 2019, p. 1), driven in part by a cohort of consumers "born between 1980 and 2015 [...] are reshaping global luxury" (McKinsey, 2019, p. 1). This generational divide is a uniquely Chinese phenomenon (Jiang and Shan, 2018;Ngai and Cho, 2012) that stems from 1990s economic reforms and the one-child policy. China's economic reforms under Deng Xiaoping during the 1990s incentivized foreign investment, including luxury-related investment (Rovai, 2014(Rovai,, 2016)), in special economic zones in Shenzhen, Zhuhai and Xiamen. Attracted by rapid urbanization and the concentration of wealthy individuals in a handful of cities, global luxury brands such as Herm es, Cartier and Chanel opened stores in first-tier cities such as Shanghai, Beijing, Guangzhou and Shenzhen. The one-child policy also shaped China's social fabric. This policy is a government program that started in 1979 to restrict Chinese families to one child each; hence, shaping a generation for which one child is the center of attention of the extended family (Cameron et al., 2013;Fong, 2016). One child is the sole recipient of emotional and financial resources in the family. In 2016, the two-child policy replaced the one-child policy. In 2021, the policy allowed three children. During the 1990s, media reports called the segment "Little Emperors" and "Little Empresses" to emphasize how parents and grandparents offered undivided attention and financial support to a single child (Byun et al., 2020;Fastoso et al., 2018). Contrary to its ostensive meaning, the little emperors' segment is not a group of entitled children. The cohort encompasses young adults from across society who are not necessarily narcissists. What they do have in common is the opportunity to express themselves in ways that their parents could not-through consumption (Cappellini and Yen, 2013). From a market segmentation perspective (Diaz Ruiz and Kjellberg, 2020), the organic or feral categorizations that analysts use to classify Chinese consumers are relevant to understand how the label "little emperors" relates to local market conditions. Media, journalists and analysts coined the term to represent local realities without employing formal market research techniques. However, even though the segment is not backed by formal --- Unconventional luxury brand collaborations market research, it is insightful because it contains the cultural interpretations by China's longtime observers. Analysts often subdivide the cohort into two: an affluent post-1980s generation consolidating their careers and the post-1990s generation characterized by social media savviness (McKinsey, 2019). This study focuses on the post-1990s generation, as they navigate influencers, online media, video games and designers, following trends that shift continuously via social media apps such as WeChat, Douyin, Weibo and Xiaohongshu. Research into the consumption patterns of young Chinese adults (post-1990s) shows that the ownership and display of designer brands constitute a form of social capital needed to advance in society (Kim et al., 2019). One reason is that consumption in China still reflects a "Confucian worldview" in which community is essential (Rovai, 2016, p. 77). Social signifiers position people under the social hierarchy of a reference group (Mainolfi, 2020, p. 291). However, younger people struggle in a digitally connected world because the social signifiers shift per the relentless stream of online content and social media influencers. Luxury brands targeting young Chinese consumers are pivoting to online culture, which is not luxury's traditional area of expertise (Holmqvist et al., 2020b). However, brands are innovating digital strategies. For example, Balenciaga released digital skins and virtual accouterments for the videogame Fortnite (Parkes, 2021), and Louis Vuitton designed the look of a popular character in the videogame League of Legends (Favis, 2020). A pivot online to collaborations with videogame franchises and digital pop culture constitutes an ongoing reconfiguration of luxury that the literature is yet to conceptualize fully. Along with their digital savviness, international mobility is relevant. Industry reports show that Chinese young adults make luxury purchases abroad, "a majority of them, about 70%, will be doing their luxury spending overseas, as a result of an increasing affinity for outbound travel and the price differential resulting from China's import tax regime" (McKinsey, 2019, p. 4). --- Method This study is qualitative with an interpretive approach (Doz, 2011), as part of a learning activity in the context of a master's class on cultural consumption. Researchers in management education proposed that merging case study research and student learning in class would be productive (see course research in Christensen and Carlile, 2009). Theory development and teaching can benefit from course research. Enrolling students as course researchers enriches theory building and the student's learning because students shift from being recipients to co-creators of learning. Under the close supervision of the second author, course researchers conducted the fieldwork as part of a postgraduate course focusing on new forms of consumption in Asia. The investigation required students to conduct, analyze and interpret semi-structured interviews with young adult luxury consumers self-identifying as Chinese. Students discussed the interviews in the context of the literature and presented their findings in course assessments. Students used reflective notes to propose new consumption phenomena. In addition, they learned to develop semi-structured interview guides as part of an in-class workshop. In turn, the authors collaboratively designed and vetted the participant sampling criteria, approved the interview guide and aligned the data-collection procedures to uphold ethics and data transparency. The project sought and received ethics approval according to the Australian code for the responsible conduct of research prior to fieldwork. Recruitment followed a purposive sampling approach (Gentles et al., 2015), which is a common qualitative sampling technique that uses pre-established judgment criteria to choose participants. Course researchers each had to identify two prospective respondents meeting the criteria of the media descriptions: young adults born in or after the 1990s, self-identifying IMR 40,7 as Chinese and consumers of luxury brands. In addition, respondents had to be knowledgeable about unconventional luxury brand collaborations. As a result, the potential participants were screened using four main criteria: (1) 18-30 years old; (2) identifies as Chinese; (3) in the last two years, has purchased at least one luxury fashion product from a pre-defined list of popular brands (Gucci, Louis Vuitton, Chanel, Armani, Rolex, Yves Saint Laurent, Dior, Herm es, Burberry, Versace, Fendi, Balenciaga, Moschino, LeSportsac, Miu Miu, Valentino) and ( 4) is familiar with recent unconventional luxury collaborations from a pre-defined list (LV 3 LOL, Sims 3 Moschino, Gucci 3 Donald Duck, LV 3 The Adventures of Zoooom with Friends, LeSportsac 3 Hello Kitty, Miu Miu 3 Disney, luxury styling in Animal Crossing: New Horizons). Course researchers proposed potential respondents and selected respondents who met these criteria. After eliminating three respondents who did not identify as Chinese, the final sample resulted in eighteen interviews. Current guidelines for sample sizes for saturation in qualitative research shows that researchers generally achieve saturation with a range of 9-17 respondents (Hennink and Kaiser, 2022). Table 3 gives a description of the respondents. Most of the participants have pursued opportunities to study and live abroad in Canada, Australia and the United States of America. The fieldwork included semi-structured interviews with a predesigned guide that covered participants' life narratives, their experiences purchasing personal luxury goods and their views and experiences of unconventional luxury design collaborations. The interview guide further explored the personal and cultural significance of luxury. A semi-structured depth interview format enabled interviewers to cover a set range of topics and provided a basic scaffold for the interview interaction. At the same time, the semi-structured approach allowed interviewers to uncover new themes and issues that are important to participants and that extend beyond the researchers' pre-understanding. To facilitate an improvisational dimension of the interview, as part of their pre-fieldwork training, each interviewer prepared and practiced interviewing and probing during an inclass workshop. This approach aligns with the flexibility and sensitivity that is a hallmark of qualitative research and promotes a deeper understanding of emerging forms of luxury (Christodoulides and Michaelidou, 2022). To assure data transparency, all interviews were audio recorded and transcribed. Each course researcher submitted a data package that included the consent form, audio recording and verbatim transcription for each interview. Interview data totaled 739 audio minutes and 314 single-spaced pages of transcription. Table 4 contains an illustrative selection of quotes. Following Christensen and Carlile (2009), course researchers participated in weekly inclass discussions, wrote reflective essays and joined one-on-one consultations. The analysis proceeded through a back-and-forth movement between three levels of analysis (Doz, 2011). The first is a holistic understanding of each participant's narrative through general themes. In the second level, the codes were grouped and re-grouped to build commonalities, variations and tensions across the interviews. For example, while all participants had purchased luxury before and were familiar with unconventional luxury collaborations (Table 1), contestations emerged about whether these luxury collaborations were of "good taste." The third level brought our emergent understanding into conversation with the extant luxury literature, leading us to conceptualize unconventional luxury collaborations as a set of three oppositional meanings, which we discuss below. --- Findings --- An ongoing reconfiguration of meanings of luxury Whereas the conventional understanding of luxury in the literature emphasizes timelessness, we find that our young Chinese respondents value luxury offerings in which rarity emerges Unconventional luxury brand collaborations through somewhat odd collaborations. Even though the branding literature establishes that brands should find matching partners (Michel and Willing, 2020), our respondents enthusiastically discuss how luxury brands are exciting when they collaborate with unconventional partners. Collaborations become a source of rarity for a generation that is accustomed to luxury consumption. As Paula recalls, luxury consumption is so intertwined in daily life that "everyone" seems to own luxury brands. A high level of luxury use means that the luxury brand cannot be a source of distinction, but instead, it becomes the norm. Our respondents are familiar with luxury brands from an early age. For instance, Paula reflects upon her upbringing in the province of Zhejiang, China, where even middle-school pupils wear designer backpacks to school. This high penetration of luxury consumption does not mean that everyone is from the upper social class. Instead, it means that even parents who cannot entirely afford luxury brands must equip their children with these items. The broad access to luxury brands affects what luxury means to Chinese society. Whereas the luxury strategy in the West intends to emphasize social distinctions, the role of luxury changes when most people around the consumer own and display luxury brands. The social pressure transforms luxury consumption into a tool for fitting in with the group, rather than standing out. In turn, brands must find ways to create a sense of uniqueness and rarity, including via design collaborations, limited editions and collectibles that appear to increase the perceived rarity of an offering. I got luxury stuff from a really young age. I think, when I grew up, my mum loves to have a lot of bags and shoes and all that stuff. So, I think a lot of that is influenced by her. I wouldn't be able to pinpoint an exact age, probably primary school, but my mom gave me luxury backpacks, that kind of thing. (...) Right now, I am a really big fan of Balenciaga. I really like their style, it's very alternative and very different. But it's also quite simple in some of their pieces, so I do like that. They are everywhere online. They popped up on my feed and I think 'Hey! that looks quite interesting.' Joanne I am careful with the kind of image I want to portray. To me having a cartoon character is more like fun and young, it is for teenagers. But the kind of vibe I want to portray is like calmer and more professional, more mature. When I go to the office, I don't think I would bring a bag with a cartoon character on it, because I want to be taken seriously Monica Yeah, I think that luxury brands really want to seduce youngsters because their purchasing power is so strong. I think in this generation, they are used to getting everything from their parents. (...) and they are all now playing Fortnite, LoL (League of Legends), and Honor of Kings, so no surprise Caitlyn I did buy a collaboration from Louis Vuitton before as well but it's Supreme and Louis Vuitton. [...] When they collaborate, it's like a cool and high-end kind of thing Source(s): Own creation based on the authors' fieldwork For instance, whereas the conventional codes of luxury emphasize timelessness, our respondents discuss luxury through ephemerality. While conventional luxury constructs meanings of exclusivity through taste distinctions, our respondents find luxury brands relatable through their lens of Chinese culture and youth-oriented activities. Finally, while luxury brands tend to emphasize their tradition, our respondents emphasize playfulness. --- Ephemeral vs. timeless Unconventional collaborations are an alternative source of rarity because they are transient and ephemeral. Ephemeral consumption means that activities and offerings constitute, disperse and reconstitute in ways that are recognizable but never identical (Diaz Ruiz et al., 2020, p. 1019). Luxury brand collaborations are ephemeral when they shift from luxury's focus on permanence. Instead of a durable lasting object, collaborations are rare because they are transitory, existing only briefly. Limited editions will not be repeated. As Ian describes, "I like the idea because it's limited; everyone wants to buy the product, which is like special because you only got limited numbers, so I think it's like a special meaning." Therefore, brief collaborations add a layer of ephemeral uniqueness because they are available for a limited time. In the literature, conventional luxury accentuates fine materials, craftsmanship and design to confer the impression of permanence. For example, a Tiffany jewel is valuable because its timeless design and quality materials mean that the object will retain financial value in the future. However, designer collaborations can make other offerings seem irreplaceable. For example, luxury brand Balenciaga's collaboration with The Simpsons and Fortnite has been received positively (Parkes, 2021), even if the partnering brands may lack a clear fit. Collaborations between luxury brands and odd partners convey hype because of the implicit promise that they will exist once. One of our respondents, Emma, finds luxury brand collaborations exciting. She views her Gucci purchase as rare because it is the outcome of an ephemeral collaboration with a well-known art museum. She considers it important that others recognize all three brands: Gucci, the Metropolitan Museum of Art and New York. I bought one t-shirt, Gucci; they collaborated with MET Museum in New York. I think that was the first item that I bought from Gucci. The t-shirt price in Gucci, sometimes, I felt like it was a little higher than what I am thinking about, but because it is a collaboration with the MET Museum, I bought it. Because they printed words on the t-shirts, and they printed the whole building on the t-shirt. I felt like that was special. And yeah... especially the color because that t-shirt is pink! [...] I am the type of person I don't care what [the words printed in the t-shirt] say. Words from other people. [The] only [thing] I am thinking about is [if] I wear it, and it makes me happy, that's it. (Emma, interview) In the literature, luxury should resist fads (Kapferer and Bastien, 2017), and yet collaborations work through fads to build a sense of rarity. Luxury brands appeal to the younger generation through messages and icons that are familiar to the shifting tastes of online culture. I think adding, like... collaborations with cartoons or anime could be cute. [...] And also, I think many of my friends like the collaborations with cartoons that brands like Louis Vuitton did. I have many friends that are, they are very rich, and they buy them a lot, like very frequently. Interviewer: So, could you say that they buy it because it's cute or 'cause it's in the luxury category? I think for both reasons, and also, for me more of the latter, more like because of its luxury. [...] I think that cobrand add[s] a new element to the luxury brand. (Harry, interview) If design collaborations are too common, or last for too long, then the illusion of singularity dissipates. Design collaborations infuse the idea of uniqueness because the object is hard to find, which implies that not all consumers may purchase them. Only those in the know can hope to obtain the collaboration. --- Trendy vs. inaccessible Our data shows that the rise of luxury brand consumption in China does not necessarily reflect the taste regimes of the upper class. Whereas luxury brands in Europe and the United States draw from the codes of art and high culture associated with the upper social class of Western societies, we find that our Chinese respondents draw from the ever-shifting trends of online culture. Luxury references in China include cultural heritage, online culture and global entertainment. These markings signal popular accessibility rather than inaccessibility. I like it because, it is just like, for example, Hello Kitty. It's a people's favorite back to when they're in child, you know. So, I think it's a really good combination of, like, a child's favorite image with those luxury brands. (Ian, interview) By trendiness we mean that luxury works through a paradox in which increasing the public relatedness of the offering also increases its perceived rarity. Previously, the literature on luxury distanced luxury brands from popular culture (Veblen, 2009). Yet our respondents observed how luxury brands draw inspiration from and embed themselves with, popular online culture to make their offerings relatable. One strategy seems to be to appeal to childhood. One example is Gucci's collaboration with Disney, which received multiple mentions. The importance of trendiness of brand collaborations remains unclear. It is possible that the constant flow of unpredictable partners may be detrimental to a luxury brand's carefully curated image. For example, Maddie reflected upon how a collaboration with a video game with a "toxic" community may create a negative image for the brand. However, she refers only to the specific collaboration; otherwise, she speaks positively of Louis Vuitton. Sometimes, some collaborations have bad connotations. The reason I would not buy a Louis Vuitton League of Legends bag is... I support the business and play the game, but it is because the community is not very good, it is very toxic, like, you don't want to be holding something like that. (Maddie, interview) The lack of fit between odd partners may be a challenge, as popular franchises may reduce the aspirational aura of luxury brands. In her interview, Caitlyn explains that unconventional collaborations risk dispelling a luxury brand's aura of inaccessibility. For unconventional luxury collaborations, a playful tone flirts with a sort of rebelliousness that sometimes irks or outrages older audiences for whom cultured, refined taste signals tradition. The juxtaposition between playfulness and tradition may create something unique, but it also conjures the risk of alienating conservative consumers of luxury because the design does not maintain its value over time. Harry acknowledges that unconventional luxury collaborations may be too childish for audiences for whom luxury equals tradition and class. People probably see you as children instead of adults. [...] Most of my clothing [is bright, fresh, and colorful], except for the suits and the shirts. For those more formal ones, I would buy them from either Armani or Valentino or Gucci. And for the rest, just like normal style ones; [I] buy those ones with fresher designs and more like, I don't know how to describe, but like street style. [...] What I think is, like LV, Chanel, or Gucci, or maybe Givenchy. Yeah, of course, they must have their own classical styles, because that's where their things get started, and they have to keep that trend. Because that was classical. So, I mean it is like the CD, I mean Christian Dior, people nowadays still remember the New Look of that suit, so that's the milestone of the whole fashion design story. [...] They want to have something new to break through like old-fashioned limits. (Harry, interview) For some young consumers, new bolder designs mark milestones of design history that break apart from old conventions. When luxury brands choose design collaboration partners correctly, the result blends elements of timelessness with the uniqueness of a one-time outcome that should not return. In her interview, Maddie reflects upon this juxtaposition, explaining that her mum buys luxury she describes as "cute," but she also describes how a playful collaboration may be "tacky, childish, wrong, and even embarrassing." I find it a little bit tacky. [...] Honey, think about it. I'll give you an example. You buy that Doraemon Gucci sneaker; will you be wearing this at 28 [years old]? Maybe, maybe, it could be a small chance, but at 33? There is pretty much no chance unless you're a real hipster mum. --- Discussion This study conceptualizes the emergence of a new form of luxury consumption characterized by short-lived, ephemeral collaborations between luxury brands and odd or unusual partners. This section conceptualizes unconventional luxury brand collaborations, starting IMR 40,7 with a definition. Luxury brand collaborations are partnerships between luxury brands in which their images, legacies and values intertwine. Luxury brand collaborations are unconventional when luxury brands partner with non-luxury brands or designers seemingly on the opposite spectrum in terms of design, esthetics, positioning and values. Today's unprecedented collaborations include luxury fashion houses partnering with streetwear brands, high-end designers collaborating with celebrity pop stars, artisan analog brands with digital games and global brands working together with local designers. Indeed, it appears that in luxury branding, opposites attract. Previous research presented an overview of new forms of luxury consumption (Christodoulides et al., 2021). This paper extends this research by conceptualizing unconventional luxury brand collaborations through the following three dimensions: ephemeral, trendy and playful (Table 5). Ephemerality is a characteristic of being transitory, existing only briefly (Diaz Ruiz et al., 2020;Holmqvist et al., 2020a) uniqueness by promising that each collection is rare because it will not be repeated. This notion challenges Hoffman and Coste-Maniere (2012), who assert that luxury products should be timeless and instead supports Bardhi et al. (2020), who conceptualize luxury as "liquid," or in a continuous state of flow. Moreover, ephemeral luxury consumption has been conceptualized before, but from the consumer perspective as a hedonic escapist moment (Holmqvist et al., 2020a). Extending Holmqvist et al. (2020a), this paper conceptualizes

Purpose -This study conceptualizes a form of luxury consumption in which luxury brands collaborate with unconventional non-luxury partners. These unconventional luxury brand collaborations are growing in popularity among Chinese luxury consumers of the post-1990s generation. Luxury brands are exploring new branding strategies due to the growing commercial importance of Chinese luxury consumers. Design/methodology/approach -An in-depth qualitative study informs this paper. Interviews with young adult luxury consumers self-identifying as Chinese reveal a growing interest for luxury brands that collaborate with odd partners in social media and online culture. Findings -Unconventional collaborations between luxury brands and non-luxury partners catalyze shifting meanings of luxury through the following juxtapositions: ephemeral instead of timeless, trendy rather than inaccessible, and playful in contrast with traditional. First, young Chinese consumers construct luxury meanings through ephemerality, like digital possessions, social media fame and fleeting experiences. Second, luxury meanings emerge in trendiness among social media influencers and online culture rather than in the seemingly inaccessible taste regimes of the upper class. Third, younger consumers appreciate fun, rebellious and over-thetop aesthetics in luxury brands. Originality/value -The study contributes to the nascent field of unconventional luxury by conceptualizing how unusual, odd and unexpected collaborations constitute new forms of luxury consumption. The shifting meanings of luxury consumption that this study conceptualizes raise new opportunities and challenges for luxury brands. One of such is the release of limited collections with non-luxury partners seemingly at the opposite spectrum of design, image and values. Moreover, the study adds nuance to the understanding of luxury consumption among young Chinese consumers.

that in luxury branding, opposites attract. Previous research presented an overview of new forms of luxury consumption (Christodoulides et al., 2021). This paper extends this research by conceptualizing unconventional luxury brand collaborations through the following three dimensions: ephemeral, trendy and playful (Table 5). Ephemerality is a characteristic of being transitory, existing only briefly (Diaz Ruiz et al., 2020;Holmqvist et al., 2020a) uniqueness by promising that each collection is rare because it will not be repeated. This notion challenges Hoffman and Coste-Maniere (2012), who assert that luxury products should be timeless and instead supports Bardhi et al. (2020), who conceptualize luxury as "liquid," or in a continuous state of flow. Moreover, ephemeral luxury consumption has been conceptualized before, but from the consumer perspective as a hedonic escapist moment (Holmqvist et al., 2020a). Extending Holmqvist et al. (2020a), this paper conceptualizes ephemerality as a luxury branding strategy. In contrast with previous research asserting that planned obsolescence is the antithesis of luxury (Hoffman and Coste-Maniere, 2012;Kapferer and Bastien, 2012), unconventional luxury brand collaborations demonstrate that short-lived initiatives are possible. Trendiness is the ability of luxury brands to navigate the constantly shifting social dynamics and preferences of social media to capture novelty and remain popular in the prevailing styles and fashion, especially online. Unconventional luxury brand collaborations allow luxury houses to complement their images of tradition and wealth with the allure of passing fads without compromising their own brand image. Extending Holmqvist et al. (2021), this paper conceptualizes how unconventional partnerships may open alternative paths to navigate between luxury innovation and tradition in digital environments. Moreover, this paper adds new dimensionality to unconventional luxury (Rosenbaum et al., 2021;Thomsen et al., 2020) by demonstrating that luxury brands must choose partners wisely to ensure that the collaboration is trendy and reinforces luxury's aura of exclusivity. Suitable partners for luxury brands have high street credibility and capture the defining spirit or mood of current times. One example is the collaboration between Fendi and the American shapewear brand Skims, focusing on body positivity and inclusive sizes (Fendi 3 Skims). The collaboration responds to growing frustration with body sizes in luxury brands (Scaraboto and Fischer, 2013). Fendi can tap into the body-positive market with the collaboration while maintaining an aura of exclusivity in their regular collection. Playfulness opens an opportunity for luxury brands to explore new boundaries (Diaz Ruiz and Makkar, 2021). Unconventional luxury brand collaborations enable luxury brands to use design choices to connect with youthful audiences who have various interests. For instance, through eSports, video games, anime and pop culture, luxury brands delve into a new playground. The notion of playfulness extends previous research on unconventional luxury (Thomsen et al., 2020) because it provides alternatives to appeal high-net-worth individuals through their hobbies. Unconventional luxury brand collaborations do not replace conventional luxury; instead, they complexify luxury. Figure 1 shows the interplay of conventional and unconventional elements in luxury brand collaborations. Ephemerality, trendiness and playfulness are tools that luxury managers can use to enhance the conventional repertoire of luxury strategies. The unconventional elements in this paper complement conventional elements of luxury, like tradition, exclusivity and timelessness. Both sets of tools complement each other to raise the expectation of the rarity and unexpectedness of limited collections. Luxury brand managers must design collaborations to convey a sense of novelty, uniqueness and exclusivity. --- Implications for luxury branding in China The literature on luxury in China is developing into a distinctive research stream because of China's commercial importance for the luxury industry (McKinsey, 2019). This study finds evidence of the re-stratification of Chinese society, as the post-1990s generation navigates a heavily digitalized online culture. Young adult consumers are using luxury in ways that their parents did not, like using luxury to stand out from their elders while simultaneously fitting in with an always-online group characterized by catching up through an endless stream of newness and fleeting fame on social media. By playing with unconventional luxury offerings, Chinese luxury consumers assemble contrasts that create even more nuances in the luxury status game, becoming tastemakers that luxury brands cannot afford to ignore. To connect with younger consumers, luxury brands cannot take themselves too seriously and become a little bit playful in their offerings. Luxury brand collaborations offer an opportunity to create a dialog with consumers that value fun, without alienating more traditional consumers. The resulting offering creates uniqueness by balancing the cultural accessibility of icons of popular culture with the aura of exclusivity of luxury brands. Moreover, luxury strategists can use brand collaborations to create legitimacy in the Chinese market by choosing partners wisely and restricting these collaborations to one-off appearances. For instance, collaborations with local designers and influencers can help to build cultural legitimacy for foreign brands. Brand strategists interested in the Chinese market may be aware of cultural differences, such as face-saving, gift-giving and collectivism (Wong and Ahuvia, 1998). For the younger generation, luxury consumption involves navigating the ebbs and flows of social media, internet fame and passing fads. To connect with the Chinese post-1990s generation, luxury brands must balance ephemerality and playfulness to create exciting offerings by emphasizing the uniqueness and irreplaceability of limited collaborations. --- Future research agenda Unconventional luxury brand collaborations should open exciting research opportunities for the luxury, branding and international marketing fields. This study contributes primarily to unconventional luxury (Thomsen et al., 2020) and new forms of luxury consumption (Christodoulides et al., 2021). However, the qualitative nature of this study and our emphasis on China prevents us from generalizing to the whole luxury industry. Further research needs to build on and validate its effects on the global luxury market. For example, unconventional collaborations may be strategies that non-luxury brands can use. Brand researchers may wish to investigate successful brand collaborations with odd or unusual partners, which seemingly lack fit. In fact, unconventional brand collaborations may challenge the orthodoxy of brand fit (van der Lans et al., 2014), which tends to focus on brand similarities, not complementarities. Table 6 introduces opportunities for further research. Whereas China's importance to the luxury industry is evident, most reports refer to its purchasing power and appetite for shopping luxury goods (D'Arpizio et al., 2021;IPSOS, 2020). China's importance today is as a shopper, not as a producer or tastemaker. However, China's role can change. By imposing unique preferences and taste, Chinese consumers can influence the development of trends and offerings. Future research can study whether Chinese consumers establish new consumer trends, potentially influencing taste globally. --- Unconventional luxury brand collaborations As China's economic prowess grows, researchers are addressing the particularities of luxury consumption in the Chinese way (Rovai, 2016). In turn, we call for future researchers to approach Chinese consumption as nuanced rather than monolithic. More research is needed to understand consumer segmentation that emerges organically (Diaz Ruiz and Kjellberg, 2020). In other words, the study of consumer culture in China requires nuance. A sociohistoric approach to Chinese luxury consumption can open avenues to understanding it better. For example, we know little about China-specific subcultures, brand communities and consumer tribes. --- Conclusion As the Chinese market for luxury consumer goods grows, generational dynamics affect luxury meanings. The younger generation of digitally connected Chinese consumers is shaping new forms of luxury consumption. Their demand for unconventional luxury brand collaborations recasts luxury meanings from timeless, inaccessible and traditional into ephemeral, trendy and playful. For example, luxury brands join seemingly odd partners at the opposite of the spectrum to build exciting but limited initiatives that build a sense of novelty and rarity because they will not be repeated. Unconventional brand partnerships can potentially rejuvenate luxury's consumer base.

Purpose -This study conceptualizes a form of luxury consumption in which luxury brands collaborate with unconventional non-luxury partners. These unconventional luxury brand collaborations are growing in popularity among Chinese luxury consumers of the post-1990s generation. Luxury brands are exploring new branding strategies due to the growing commercial importance of Chinese luxury consumers. Design/methodology/approach -An in-depth qualitative study informs this paper. Interviews with young adult luxury consumers self-identifying as Chinese reveal a growing interest for luxury brands that collaborate with odd partners in social media and online culture. Findings -Unconventional collaborations between luxury brands and non-luxury partners catalyze shifting meanings of luxury through the following juxtapositions: ephemeral instead of timeless, trendy rather than inaccessible, and playful in contrast with traditional. First, young Chinese consumers construct luxury meanings through ephemerality, like digital possessions, social media fame and fleeting experiences. Second, luxury meanings emerge in trendiness among social media influencers and online culture rather than in the seemingly inaccessible taste regimes of the upper class. Third, younger consumers appreciate fun, rebellious and over-thetop aesthetics in luxury brands. Originality/value -The study contributes to the nascent field of unconventional luxury by conceptualizing how unusual, odd and unexpected collaborations constitute new forms of luxury consumption. The shifting meanings of luxury consumption that this study conceptualizes raise new opportunities and challenges for luxury brands. One of such is the release of limited collections with non-luxury partners seemingly at the opposite spectrum of design, image and values. Moreover, the study adds nuance to the understanding of luxury consumption among young Chinese consumers.

Introduction This paper analyses the role of the third sector (Boccaccin 2013) in the integration of persons with psychic disabilities and their ability to network with the other actors who, for various reasons, are involved in the job placement provided for by the targeted employment. For this purpose, the main national and international evidence will be presented, with a focus on the Italian local realities of Veneto and Lazio. The chapter, more or less extensively, deals with all the methods and actors involved in labour inclusion, together with the system of targeted employment and related projects, making specific reference to the role of social cooperation and in particular the one in Rome, using a research included in the text, entitled "Psychiatric disability and work: a possible duo? Socio-medical policies, territorial governance and labour inclusion" (Giangreco 2009). This research also gives specific space to the role and importance of parental associations, which prove to be a positive actor since they are configured as a pressure group for the implementation of active and inclusive labour policies addressed to the intellectually and mentally disabled. However, the chapter concludes by providing the reader the possible scenarios of social cooperation in Italy. Day Care Centres and Social Cooperatives Compared Differences and Critical Issues: The Italian Case Some preliminary clarifications are necessary for a correct reading of the contribution. The inclusion of mentally ill in the world of work and, more generally, in socialization contexts, is often the prerogative of day centres and social cooperation bodies. Day care centres are little geared towards the entrepreneurial dimension in unprotected and sheltered contexts, unlike social cooperation, which makes it the focus of its activities. It is possible to state that, except for type D day care centres that are more oriented towards the labour dimension, the other types (A, B, C) tend to reproduce dynamics that can be traced back to the institutionalization paradigm (Di Santo 2013), in the belief that the labour dimension is the exclusive preserve of social cooperatives. Day care centres generally offer a possibility of work that is strongly linked to the institution, often with therapeutic aims, generating a vicious circle with the subject they are trying to cure. At this point, a first element of differentiation with social cooperatives is evident, which, although specialized and linked to well-defined economic balances, tend to let their users experience situations within the ordinary labour market, albeit with strong limitations. Here are the results of some research that report this trend. As emerges from a survey conducted in the Veneto region (Lascioli et al. 2006), about 20% of the subjects considered have a has a stay in the cooperative of more than five years, while 50% have an average time of about two years. Furthermore, it is --- 2 shown that the guarantee function of cooperatives is greater for those individuals whose productivity level is excessively low compared to able-bodied employees. According to research carried out by the Centro Studi CGM (Maiello 1997), people who manage to find work outside the social enterprise are only 14.6% of the sample. To this meagre percentage, it should be added that those individuals have severe disabilities. Another trend that is spreading within B-type cooperatives is a form of temporary insertion in non-protected contexts, in order to favour the person's insertion in the ordinary market. In general, the world of social co-operatives tends to make both the function of guarantee and that of transitionality coexist; where this is not possible, they specialize, some protecting one orientation, others another. Co-operatives that tend to defend their guarantee function run the risk of not favouring the person who turns to them, since the action remains bound to the context of the same, just like day centres. On the other hand, those who adopt the transitional regime risk rendering their intervention futile if they do not prove capable of networking with the other actors involved in job placement. A further negative aspect, linked to the pre-employment training that can take place in cooperatives, is the fact that the trainee is often not called upon to perform a future task, but only employed as an ordinary worker, without actually being one. To this picture it should be added that in many cases, even within cooperatives, it is preferred to employ staff whose disadvantage, although present, has little influence on the work to be performed. Within social enterprises there are two factors that optimize the success of work placements: the first is the degree of experience gained by the cooperative and the second is the presence of a person responsible for work placements for these categories. Some researchers have shown a negative correlation between poor work skills and co-operatives specialized in a certain target of discomfort. In fact, the presence of people with different types of disadvantage improves performance as it stimulates coping skills. This thesis is also supported by the study carried out by McKinsey & Company (Koller et al., 2010), which asserts that the coexistence of people with different types of disadvantage hinders the formation of specific needs. Moreover, the same study emphasizes that starting number of placements too high, not only leads to leads not only to a decrease in performance, but also to greater difficulties in post-employment. The social cooperative only fulfils its purpose if it is able to enhance the skills of the disabled person, in accordance with their inclinations and training. In most cases, however, this is not possible, as the cooperative does not perform highly specialized work for itself or for others, thus offering limited choice and career possibilities. The legislator, in the Legislative decree 276/03, intended to recognize, through article 14, the value of the work carried out by social enterprises for the integration of disadvantaged people, offering them the chance to be introduced into the world of work for at least once, despite their objective difficulty in entering the ordinary production cycle. The article is intended to make up for the shortcomings in the application of Article 12 of Law 68/99 concerning provisions on targeted employment, to be understood as the insertion of persons with disabilities into the world of work. On the basis of this analysis it can be hypothesized that type A and B social cooperation may be faced with three possible scenarios in the future (Giangreco 2009) <unk> Transitionality: progressive cancellation of social enterprises in favour of other more efficient subjects; <unk> Consolidation of the niche: recognition of certain areas in which only social cooperation can act, under a monopoly regime; <unk> New Community enterprise: legitimization of its organizational and management form by means of transversal policy initiatives, aimed at creating connections between the social enterprise and other subjects, not only economic, operating on the territory. Currently, the employment models pursued by the social enterprise that are most widespread in Europe concern transitional employment, i.e. temporary experience on the job, self-financed and permanent employment, with the aim of creating stable jobs, professional insertion with permanent subsidies, sheltered employment aimed at the severely disabled, and finally socialization through productive activities, which does not aim to insert subjects into the ordinary labour market, but to produce socialization through it. Operating 3 The adoption of one of the four models of work inclusion outlined above significantly characterizes the work of the co-operative itself and the way in which it conceives work inclusion for people with disabilities in a broad sense. Moreover, their inclusive capacity is influenced by three orders of variables, such as: <unk> Structural variables: internal characteristics of the company concerning its ability to accommodate workers with disabilities, specifically the degree of organizational maturity of the same, which in turn requires a clear definition of management roles and responsibilities, desiring specialization in a single type of disadvantage and consequently requiring a high percentage of disabled people within its workforce. A facilitating aspect of this process is the small size of the enterprises, which accentuates the propensity to develop networking with other related bodies and organizations, all variables influenced, in the final instance, by the level of governance (Bertin 2013) adopted by the cooperatives considered. <unk> Strategic variables: these concern the institutional mission carried out by the cooperative. They regard the decision made by almost all of these social enterprises to occupy a marginal position in the sphere of active labour policies. In fact, social cooperatives have historically always preferred to interweave stable relations with social and health services, neglecting relations with specific institutions for the activation of employment policies and with companies belonging to the ordinary market. The absence of a common project on a territorial basis has a strong impact on the difficulties of these companies in employing disabled people. <unk> System variables: are related to the policies that regulate the relationships between the local public administration, the labour market and the co-operatives themselves. It is possible to hypothesize that a weak use of legislative instruments in favour of social enterprises has consequently weakened their ability to guarantee stable employment to their users. Taking into account the variables outlined above, research on social cooperation in the Rome area has set itself the objective of trying to identify the strategies they adopt in the professional fulfillment paths they propose to offer (Giangreco 2009). This led to the identification of three distinct types of cooperatives with structural characteristics, identities and objectives that refer to different ways of conceiving and addressing the problem of labour inclusion of individuals with psychic distress. From the totality of type B cooperatives operating in the Rome area, those most active on the issue of inclusion were isolated and considered. Twenty-three were contacted, of which only twenty actually participated in the survey. Then, index variables were created in order to bring out the peculiar characteristics of the companies considered, with regard to: organizational development, level of specialization in the field of psychiatric disability, level of receptiveness, and propensity to relationality. From the intersection of the variables considered, three different types of social cooperation therefore emerged: <unk> Refractory: although they show a high degree of organizational development, they do not seem to express a willingness to hire people with mental disorders internally, limiting themselves exclusively to guaranteeing a residual quota of places for apprenticeship experiences. In this regard, from some excerpts of the interviews carried out in the survey, co-operative managers state that they only hire disadvantaged people with adequate skills and for particular contexts, such as single mothers, exconvicts, ex-drug addicts, but not psychiatric patients. In fact, in line with what has just been reported, it emerges that these enterprises, in order to increase their productivity, prefer to hire disadvantaged, but not disabled people, as they are engaged in activities linked to public bodies, while reserving a small space for the private sector as well. Within the sample considered there are also large and small cooperatives. The larger ones compete in size with private companies, use subsidized financing, and are almost never in relation with employment centres, while the relationship with the Department and Mental Health Centres is strong (Innocenzi 2007). <unk> Ineffective: the cooperatives belonging to this typology were conceived by operators of the Mental Health Department and are based at Day Care Centres. They are highly specialized in the category of disadvantage that can be traced back to psychic discomfort and in Day Care Centres they generally initiate training activities, running training workshops. They are considered fundamentally ineffective because, although they have a high degree of receptiveness, attributable to a high number of psychic subjects employed, they are not in a position to carry out new recruitments, for reasons due to organizational and strategic factors. In the first case, the discontinuity in the work offered is due to only temporary orders offered by public bodies. In the second one, difficulties are linked to purely welfare interventions and therefore by their non-training nature. These cooperatives are ultimately considered ineffective because they have not developed an 'entrepreneurial project' and are therefore forced to passively submit to events. They tend to polarize resources towards the Day Care Centres to which they belong. This, while on the one hand is an advantage, on the other hand constitutes a strong constraint to access for those disabled people who demonstrate high degrees of productivity. For these reasons, all Day Care Centres belonging to this typology do not fully respond to the original mission of labour inclusion. They are also: young enterprises, lacking an organizational chart, with a mainly informal organization and a low number of working members, both able-bodied and disabled. Moreover, due to the characteristics considered, they are placed, among cooperatives, at the lowest step of organizational development, showing a lower propensity to relationality, being detached from all the contexts aimed at guaranteeing inclusion. <unk> Inclusive: this type of social enterprise does not show prejudice towards the mentally ill, unlike refractory enterprises (Innocenzi 2007), however, being saturated with personnel, it is unable to recruit new staff, unless it gains a new market space. Inclusive co-operatives are well-established realities, which mainly deal with psychic distress and are also inclusive from an organizational point of view. Unlike ineffective enterprises, there are no structural deficits. Most of their economic income comes from public orders, as well as from small private handicrafts that shelter them from the risk of discontinuity in the employment of workers. More than 50% of those working in cooperatives are disabled, confirming the inclusive dimension of this enterprise. They also have a strong propensity to operate in networks, building positive synergies with other social enterprises, companies and consortiums in a strategic way to promote the labour inclusion of their users. A further ally of this type of cooperatives is the Department of Mental Health, with which they share objectives and projects. These enterprises show a willingness to positively involve all political, institutional and contextual actors in the disabled person's life in order to make him/her part of a process that can lead to strong active policies in this sense. In contrast to ineffective cooperatives, the link they establish with all mental health actors triggers a virtuous circle between the resources employed and the actors involved in reinvesting them. However, this type of enterprise mainly complains of economic difficulties, which are only partially alleviated by a progressive development of private activity. The total number of users considered in the survey is 192, distributed in the different types of social cooperation as follows: 138 in inclusive enterprises, 37 in ineffective ones, 17 in refractory ones (Giangreco 2009). In line with what has been said so far about inclusive enterprises, they prefer the open-ended contract form (N = 100), guaranteeing however, for all the other subjects considered, part-time contracts (N = 23) or occasional work (N = 15). On the other hand, social enterprises of the refractory and ineffective type prefer the form of part-time work, equally distributed between fixed-term and open-ended contracts. The choice of part-time work is justified by two orders of reasons: the first, functional, concerns the presumed inability of the disabled person to work more time; the second, instrumental, is related to purely economic reasons, since, by working more hours, the disabled person involved would lose the possibility of receiving disability benefits. Alongside proper work inclusion, the types of social enterprise considered so far also offer the possibility to carry out traineeships, with a double purpose: the evaluation of the disabled person's abilities and the learning, for them, of new expendable skills. Inclusive co-operatives use this situational training instrument in a positive way, by formulating an assessment in order to establish how, for how long and where to employ the users. In refractory social enterprises (Innocenzi 2007), this tool, on the other hand, is only used to improve the degree of skill possessed by the subject with a view to future employment. However, in both types of enterprise, the apprenticeship is not used in a biased manner, as the duration is approximately one year. As mentioned above, inclusive enterprises paradoxically have fewer possibilities of actually entering the ordinary labour market, despite this they manage to provide a greater number of work placements for their users by trying to overcome this limitation and maintaining a fair proportionality between the mentally ill, the other categories of disadvantage and the able-bodied. A balance between the different factors guarantees greater performance for the company and the worker who is part of it, it follows that a good choice in terms of planning and management of resources guarantees better reliability for the cooperative itself, compared to others that employ a high number of disabled human resources internally without a clear definition of the objectives to be achieved. The data in our possession suggest that the readiness for work inclusion of psychic individuals is strongly influenced by an identity factor. Inclusive companies place greater emphasis on their own identity, reference values and user needs. This aspect is not found in all the types of social enterprise considered and this has a negative influence on the aims pursued by the co-operative, if the latter is not able to work in a network with all the actors aimed at favouring inclusion, making choices dictated by private interests, the social dimension of the co-operative will not be real, but only a meaningless label. Negative aspects of the concept just described can be found in the following trends: the high turnover of workers with disabilities, the use of trainees as substitutes for the workforce, statements and reflections characterized by a lack of civic sense, the gratuitously offensive and discriminatory attitude towards particular categories of disadvantage, etc. The problematic aspect common to all the companies surveyed lies in identifying resources and objectives that correspond to their culture. From this point of view, it is necessary to have a system that constantly plans and verifies the objectives to be achieved, finding the appropriate means for the pursuit of its goals. All the cooperatives in the sample have a marginal position in the field of active labour policies, due to the following aspects: the structural backwardness of employment centres; the lack of dynamism of social enterprises in the Lazio region; the lack of local strategies on the part of the consortia system, which mainly provide technical and professional assistance services (accounting and tax operations and the carrying out of administrative paperwork) and services aimed at business development (consultancy for the finding of economic and financial resources and project development). The gradual reallocation of social enterprises to the centre of active employment policies could guarantee greater opportunities for growth with a consequent increase in the number of jobs reserved for the disabled. On the other hand, preserving the current status could only lead the Roman cooperation to survival, in a conservative perspective based on the maintenance of its positions within marginal niches. The real risk could therefore be, in a phase of shrinking public spending, that social cooperation would consider itself as one of the only subjects deputed to solving problems pertaining to its own sphere of competence, such as the employment of disadvantaged people, with the claim of creating development even in the absence of real links with the local production system. Waiting for the desired change, law 381/91 is the only one to protect the instrument of social cooperation. The partial application of this law does not allow the cooperatives examined to fully realize the inclusion of those concerned. This testifies to the fact that local administrations often prefer to achieve short-term savings through the instrument of contracting rather than directly entrusting their services to social enterprises. In fact, the increase of workers with psychiatric disabilities within Roman companies is due, in the first instance, to the ability of policy-makers and institutions to prepare the conditions for them to really take advantage of all the options available under the law. In this sense, the planning of assignments could certainly represent an effective way of matching, especially if the cooperative were organized in a consortium system, capable of sustaining over time the qualitative and quantitative scope of the productive outsourcing of ordinary companies. Indeed, only the right balance between internal and external factors could significantly contribute to enterprise development and the implementation of new jobs, guaranteeing a high level of social growth and related values. A final note: the evidence presented in this article predates the reform process of the third sector in Italy brought about by Law 106/2016; nevertheless the results remain at the centre of a debate that still retains its validity in the scientific field. --- Conflicts of Interest: The authors declare no conflict of interest

This paper focuses on the role of the third sector in supporting the placement of people with psychic disabilities and enhancing their ability to network with other subjects who are involved in a placement provided by the targeted employment system. Data analysis and key concepts were informed by the results of a master's thesis entitled "Psychic Disability: Comparing Job Placement Services". Data from national and international research that has dealt with the ability of these individuals to keep their jobs (Maiello 1997) in relation to coping skills (Koller et al. 2010) will be presented. Subsequently, the research conducted in the specific local contexts of Veneto and Lazio will be considered (Lascioli et al. 2006). Following this, there will be a focus on type A and B cooperatives in the Roman territory in which the variables underlying the enterprise are analyzed (Giangreco 2009) which implies a further subdivision into cooperative typologies: refractory, ineffective and inclusive. This research occurred just before the reform process of the third sector occurred by the law 106/2016

Introduction The COVID-19 pandemic had a devastating effect on numerous countries across the globe. It put a severe strain on health facilities and economies [1] and highlighted racial and socioeconomic health disparities. Marked by uncertainty, attempts at addressing the outbreak initially led to international lockdowns and large-scale quarantines without clear end dates. Increased cases of acute trauma or post-traumatic stress disorder during the pandemic demonstrate the severe impact of the mortality that ensued, in combination with economic instability and inevitable long-term lifestyle changes [2]. There is a large body of compelling evidence supporting the social determinants of health data as a strong predictor of health outcomes [3,4], and COVID-19 is not an exception. The review of the existing literature clearly shows the presence of COVID-19-related racial disparities in the United States, specifically in Black and impoverished populations [5,6]. Black individuals account for 12.4% of the American population and 14.3% of deaths by COVID-19 in the U.S. [7,8]. In comparison, White Americans comprise 61.6% of the population and 59.7% of deaths by COVID-19. This disparity is multifactorial and includes determinants such as poverty, access to healthcare, nutrition, and pre-existing co-morbid conditions [6]. Social factors that exist in Black and other minority communities, such as historically based mistrust of medical professionals, are also crucial to consider [9]. In order to fully assess the impact of such factors, it is important to understand both their development over time and their roles in individuals' present daily lives. Historic experimentation through unethical and non-consensual means on Black individuals in medical research in the U.S. include well-known examples, such as the infamous Tuskegee Syphilis Study [10]. The non-consensual use of tissue samples belonging to Henrietta Lacks and the subsequent lack of consent demonstrated by the publication of her genome and widespread distribution of her cell lines is another historic example [11]. For Indigenous communities, research studies are often based on settler colonial agendas that reinforce harmful stereotypes and serve as a source of medical mistrust in Native populations [12]. The implications of medical mistrust remain deeply rooted in the present day. For example, it has been shown that Black men with more frequent exposure to everyday racism and perceived racism in health care are more likely to delay preventative health screenings [13]. Vaccine hesitancy and lower COVID-19 vaccination rates for the Black population in the U.S. are also significant contributing factors to disparities in COVID-19 impact [14,15]. Access to vaccinations is only one aspect of this issue, as medical mistrust and misinformation regarding the vaccine may also contribute to a lower vaccine administration. With a clear role in presentday attitudes and health behaviors, it is crucial to understand historically based medical mistrust in Black and minority communities to effectively contextualize active disparities. Previous studies support the idea of a Matthew effect by which pandemic instability widens inequality by disproportionately affecting groups that are already disadvantaged [16]. In addition, cultural norms in predominantly Black communities that promote collective engagement through larger and more frequent group gatherings are associated with increased rates of COVID-19 diagnosis [17]. Environmental and geographic differences, such as urban environments, are also important to consider. Studies utilizing Geographic Information System (GIS) have found factors such as population density and asthma, suggestive of urban areas, to be related to higher rates of death by COVID-19 [18]. Spatial factors in urban environments that influence mobility and air quality have also been implicated with aggregates of COVID-19 cases [19]. Given the multi-factorial nature of health disparities, it is imperative to assess them at all levels ranging from local communities to a national scale. Documenting and evaluating the racial and economic disparities related to COVID-19 in particular is crucial for improving healthcare policy, assistive programs, and resource allocation at the local, state, and federal levels of affected communities. The focus of this study is to analyze the effects of COVID-19 specifically in the city of Gary, Indiana. Gary serves as a microcosm for impoverished, majority non-White, and urban communities across America. There is a large Black population, a high number of residents living below the poverty line, and a high unemployment rate [20]. The city is one of many in the U.S. that were severely impacted by deindustrialization. The closure of U.S. steel's largest manufacturing plant, Gary Works, in the 1960s left the city to economic collapse. Compounded with discriminatory housing policies from the remnants of segregation and Jim Crow laws, the population split as White Americans fled the city to find new jobs [21]. According to the 2020 Census, the total population in Gary is 69,093 people and 78% of the population is Black, compared to 13% nationally. The mean household income of Gary residents is USD 34,085, which is well below the national average of USD 69,717 [20]. Both the unemployment rate (44.4%) and the poverty rate (32%) in Gary are well above their respective national averages [20]. With the multitude of social determinants of health indices affecting Gary's population, we sought to elucidate the association of these factors with the number of COVID-19 infections and the severity of the disease, including hospitalizations and deaths. We specifically assessed racial and economic disparities by utilizing an age-adjusted race correlation and a population-adjusted zip code correlation to compare race categories and mean household income to COVID-19 infections and severity outcomes. We hypothesized that the Black population in Gary and zip codes with a lower mean household income would have a stronger correlation with the number of COVID-19 infections, hospitalizations, and deaths. O/E ratio effectively adjusts for population size, age distribution, and data with missing race. The O/E ratio is a commonly used method of data comparisons in the clinical setting [23]. --- Methods --- Statistical Analysis In the analysis of racial disparities, Pearson's chi-squared test was used to determine statistical significance, defined as a two-sided p-value of 0.05 or less. In this analysis, the expected values were calculated using the same formula for the expected count used in Section 2. In the analysis of the zip code distribution data, linear regression was used to model differences in COVID-19 severity and zip code-related variables. Linear regression analysis was used to determine the significance of differences in the population. Since these zip codes are adjacent to each other, the Pearson correlation coefficient was used instead of R 2 to describe covariance in order to avoid issues with the model assumptions. Exponential linear regression analysis was used to determine the significance of income distribution by the zip codes on COVID-19 outcomes. O/E ratios were converted to a log scale, and then linear regression was performed on logged O/E ratios. Since these zip codes are adjacent to each other, the Pearson correlation coefficient was used instead of R 2 to describe covariance in order to avoid issues with the model assumptions. --- Results --- Evaluation of the Number of Infections, Hospitalizations, and Deaths in the City of Gary The analysis of the number of positive cases showed that there was a higher number of infections in the non-Black population compared to the Black population (1.19 vs. 0.93, p <unk> 0.0001; Figure 1A). Exponential linear regression analysis was used to determine the significance of income distribution by the zip codes on COVID-19 outcomes. O/E ratios were converted to a log scale, and then linear regression was performed on logged O/E ratios. Since these zip codes are adjacent to each other, the Pearson correlation coefficient was used instead of R 2 to describe covariance in order to avoid issues with the model assumptions. --- Results --- Evaluation of the Number of Infections, Hospitalizations, and Deaths in the City of Gary The analysis of the number of positive cases showed that there was a higher number of infections in the non-Black population compared to the Black population (1.19 vs. 0.93, p <unk> 0.0001; Figure 1A). Despite lower infections, there was a statistically significantly higher number of COVID-19-related hospitalizations for the Black population compared to the non-Black population. After O/E normalization, there was a 3-fold higher number of COVID-19 hospitalizations in the Black versus the non-Black population (1.26 to 0.38, p <unk> 0.0001; Figure 1B). There also was a significantly higher number of age-adjusted deaths in the Black versus the non-Black population. After O/E normalization, there was a 2-fold higher number of COVID-19 hospitalizations in the Black versus the non-Black population (1.19 to 0.58, p <unk> 0.02; Figure 1C). --- Evaluation of the Effect of Zip Code Income on Infections, Hospitalizations, and Deaths in the City of Gary Age-group-adjusted COVID-19 hospitalizations versus median household income of zip codes in Gary, Indiana, are presented in Figure 2. As the median household income of a zip code decreases, the number of hospitalizations increases. Despite lower infections, there was a statistically significantly higher number of COVID-19-related hospitalizations for the Black population compared to the non-Black population. After O/E normalization, there was a 3-fold higher number of COVID-19 hospitalizations in the Black versus the non-Black population (1.26 to 0.38, p <unk> 0.0001; Figure 1B). There also was a significantly higher number of age-adjusted deaths in the Black versus the non-Black population. After O/E normalization, there was a 2-fold higher number of COVID-19 hospitalizations in the Black versus the non-Black population (1.19 to 0.58, p <unk> 0.02; Figure 1C). --- Evaluation of the Effect of Zip Code Income on Infections, Hospitalizations, and Deaths in the City of Gary Age-group-adjusted COVID-19 hospitalizations versus median household income of zip codes in Gary, Indiana, are presented in Figure 2. As the median household income of a zip code decreases, the number of hospitalizations increases. (46402, 46403,46404, 46406, 46407, 46408, and 46409) in the city of Gary were plotted against median household income by the zip code (red square, USD/year). Trendline represents the linear regression (R 2 = 0.63, p = 0.03). Significance was determined using a linear regression analysis. There were no correlations between zip code median household income and the number of positive cases or deaths (Figure 3A,B). (46402, 46403,46404, 46406, 46407, 46408, and 46409) in the city of Gary were plotted against median household income by the zip code (red square, USD/year). Trendline represents the linear regression (R 2 = 0.63, p = 0.03). Significance was determined using a linear regression analysis. There were no correlations between zip code median household income and the number of positive cases or deaths (Figure 3A,B). --- Infection Rates Correlate with the Zip Code Population Size As a zip code population size increased, the number of detected COVID-19 infections in that zip code also increased (Figure 4). This correlation was expected assuming that people in each zip code area had similar infection rates and COVID-19 testing availability. If any zip code area differed in either the availability of COVID-19 tests or infection rates, this would be reflected by a lack of correlation between the zip code population size and the number of COVID-19 infections. Trendline represents the linear regression (A: R 2 = 0.25, p = 0.25; B: R 2 = 0.06, p = 0.57). Significance was determined using a linear regression analysis. --- Infection Rates Correlate with the Zip Code Population Size As a zip code population size increased, the number of detected COVID-19 infections in that zip code also increased (Figure 4). This correlation was expected assuming that people in each zip code area had similar infection rates and COVID-19 testing availability. If any zip code area differed in either the availability of COVID-19 tests or infection rates, this would be reflected by a lack of correlation between the zip code population size and the number of COVID-19 infections. --- Discussion It has been documented in previous studies that COVID-19 disproportionality affects individuals of specific racial and economic demographics. The existing data demonstrate racial disparities in the impact of COVID-19, specifically in Black and impoverished populations in the United States [6,24]. This disparity is multifactorial and includes determinants such as poverty, access to healthcare, nutrition, and pre-existing co-morbid conditions. Trendline represents the linear regression (A: R 2 = 0.25, p = 0.25; B: R 2 = 0.06, p = 0.57). Significance was determined using a linear regression analysis. --- Infection Rates Correlate with the Zip Code Population Size As a zip code population size increased, the number of detected COVID-19 infections in that zip code also increased (Figure 4). This correlation was expected assuming that people in each zip code area had similar infection rates and COVID-19 testing availability. If any zip code area differed in either the availability of COVID-19 tests or infection rates, this would be reflected by a lack of correlation between the zip code population size and the number of COVID-19 infections. --- Discussion It has been documented in previous studies that COVID-19 disproportionality affects individuals of specific racial and economic demographics. The existing data demonstrate racial disparities in the impact of COVID-19, specifically in Black and impoverished populations in the United States [6,24]. This disparity is multifactorial and includes determinants such as poverty, access to healthcare, nutrition, and pre-existing co-morbid conditions. Significance was determined using a linear regression analysis. --- Discussion It has been documented in previous studies that COVID-19 disproportionality affects individuals of specific racial and economic demographics. The existing data demonstrate racial disparities in the impact of COVID-19, specifically in Black and impoverished populations in the United States [6,24]. This disparity is multifactorial and includes determinants such as poverty, access to healthcare, nutrition, and pre-existing co-morbid conditions. The current study evaluated whether Black and non-Black populations in Gary, Indiana, were impacted differently by COVID-19. The effects of income differences on COVID-19 infections and outcomes between zip codes in Gary were also evaluated. Three demographic variables from the U.S. Census data were significantly correlated to COVID-19 prevalence and severity. First, population size impacted the number of positive tests. Second, racial disparities exist in the number of infections, hospitalizations, and deaths of the Black population versus the non-Black population in Gary. Third, COVID-19 hospitalizations are significantly and negatively associated with zip code income. In this study, population size positively correlated with the number of COVID-19 cases, such that zip codes with greater populations had a higher number of infections. COVID-19 is transmitted from person to person by the direct inhalation of contaminated droplets or by close contact with an infected person [25]. In zip codes with larger populations, it is more difficult to practice social distancing and avoid close contact, increasing the likelihood of spreading COVID-19. Therefore, it is not surprising to find a higher number of positive cases in the more densely populated zip codes of Gary. The current study demonstrated a racial disparity in the number of positive COVID-19 cases in Gary, Indiana. A statistically significant difference was identified in the number of COVID-19 infections between the Black and non-Black populations, with the non-Black population having a higher number of infections. This finding is unexpected and differs from what has been reported in some previous studies. For example, Black individuals in Indiana's Marion County were four times more likely to test positive for COVID-19 than non-Black individuals. Counties with a higher Black population also had an increased number of positive COVID-19 cases [26]. One explanation for the dissimilarities between the current study and data reported by Hanson and colleagues [26] is that many testing centers in Gary became vaccination centers between the end of 2020 and the beginning of 2021. With fewer people seeking tests at the facilities, many were not tested until hospitalization when they became critically ill. This may explain the finding of the current study that the Black population has a lower infection rate, but a much higher hospitalization rate compared with the non-Black population. A second plausible explanation as to why the non-Black population in Gary had a higher infection rate than the Black population is differences in resource availability. This difference in resources can be quantified by comparing the median household incomes for the two groups. As of 2020, the median household income for the Black population in the United States was USD 45,870, while the median household income for Hispanics and white non-Hispanics was USD 55,321 and USD 74,912, respectively [7]. Income inequality has been further exacerbated by the COVID-19 pandemic. Black adults were found to be over three times as likely as White adults to report food insecurity, being laid off, or being unemployed [16]. A lower median household income and the greater likelihood of losing employment are additional barriers to obtaining COVID-19 testing for the Black population compared to the non-Black population. When comparing the unemployment rates, the Gary zip code 46402 with one of the highest hospitalization values had the unemployment rate of 21.7%. The zip code 46408 with the lowest hospitalization values had the unemployment rate of 14.7%. The assumption that the non-Black population was obtaining tests at a much higher rate than the Black population in Gary would explain the higher infection rate that differs from the previous studies. This is also supported by a study conducted in Illinois, which reported the total number of confirmed cases among Black and White individuals as almost even, despite 220,968 White individuals being tested compared to only 78,650 Black individuals [6]. Based on these, it is possible that many positive cases in the Black population were undocumented. The current study found racial disparities related to COVID-19 outcomes in Gary. A statistically significant difference was identified in the frequency of both hospitalizations and deaths between the Black and non-Black communities. There was a three-fold increase in the number of COVID-19 hospitalizations in the Black versus the non-Black population. There was also a two-fold increase in the number of COVID-19 deaths in the Black versus the non-Black population. These findings are consistent with evidence from previous studies. One study conducted at Montefiore Health System in the Bronx investigated hospitalizations and mortality rates of their Black and White patients prior to COVID-19 and during the pandemic. The study found that, during the COVID-19 period, the adjusted odds of death for Black patients were 1.6 times higher than that of White patients [27]. Another study researching racial disparities in COVID-19-related deaths across the entire United States found that only 20% of U.S. counties are disproportionately Black, despite accounting for 52% of COVID-19 diagnoses and 58% of COVID-19 deaths [28]. Pre-existing comorbidities also play an immense role in determining the severity of COVID-19 disease. Studies have found COVID-19 patients with cardiovascular disease, diabetes, congestive heart failure, hypertension, chronic kidney disease, and cancer to have a significantly higher risk of mortality compared to COVID-19 patients without these pre-existing comorbidities [29]. A higher prevalence of obesity, diabetes, hypertension, and chronic kidney disease has been shown in Black patients compared with White patients [30]. The Black population also experiences a drastically higher rate of food insecurity in the U.S. One study reported the rate of food insecurity in Indiana's Black population after the COVID-19 pandemic (55%) to be 134% higher than that of the White population and 308% higher than the rate in Indiana prior to the pandemic (14%) [16]. It is clear that the disparity in food insecurity already experienced by the Black population became significantly worse during the pandemic [16]. The increase in food insecurity contributes to difficulties in lifestyle changes that may lead to the further exacerbation of pre-existing comorbidities. It can be reasonably suggested that the higher number of COVID-19 deaths in the Black population in Gary may be related to the greater prevalence of pre-existing comorbidities and diminished socioeconomic opportunities to adhere to a healthy lifestyle. Some studies also point to vaccine hesitancy in the Black population contributing to a lower vaccination rate compared to the non-Black population in the U.S. The prevalence of COVID-19 vaccination hesitancy for adults in the U.S. was 26.3%, while the prevalence rate for the Black population was 41.6% [14]. There are multiple factors that could explain this significant difference in vaccine hesitancy [31,32]. Experience with racial discrimination and mistreatment in both healthcare and non-healthcare settings contributes to an increased vaccine hesitancy by patients developing medical mistrust [33]. Misinformation through mass media and social media is also a major contributor to vaccine hesitancy in minority communities [15]. Current data on vaccination support vaccination as a means to prevent contracting COVID-19 infection and significantly reduce the severity of the disease [34]. As the Black population receives the vaccine at a significantly lower rate, they are more likely to experience greater severity of symptoms and higher rates of mortality once they contract COVID-19. Vaccination hesitancy can be improved by providing culturally informed, consistent education and increasing access by distributing vaccines at local churches and community centers by trusted community members [15]. Furthermore, messaging to Black individuals from physicians of the same race has been shown to increase knowledge of COVID-19 symptoms and prevention methods [35]. The results from this study also found hospitalizations for COVID-19 to be considerably affected by income. A statistically significant correlation was found between the median household income of a zip code and the number of hospitalizations due to COVID-19 infection. Individuals in zip codes with a lower median household income in Gary are more likely to be hospitalized due to COVID-19 infection. Similar to the previous trends, there are multiple factors contributing to the finding of lower-income zip codes having more hospitalizations. A higher income offers a greater access to testing facilities and treatments, allowing patients to receive positive test results earlier and start treatment sooner [6]. Individuals with a lower income and less access to testing may not know of their infection until the disease significantly progresses and at which point hospitalization becomes inevitable. A higher income also offers a greater access to health insurance [36], preventative healthcare, and newer treatment options. In addition to beginning treatment relatively early, these patients may have a greater ability to take time off from work and impede the progression of the disease before hospitalization is required. Previous studies have found pre-pandemic food insecurity, housing insecurity, and unemployment status to be associated with most indicators of pandemic-driven economic instability [16]. These factors are likely to collectively contribute to a more severe disease progression and a greater likelihood of hospitalizations for individuals with a lower income. Historic restrictions in access to education and digital technologies in minority communities might significantly contribute to the income inequalities and mistrust in healthcare professionals as well [37]. Both of these factors greatly influence the outcome of COVID-19 infection by impeding access to reliable information and technology to make informed healthcare decisions. A recent retrospective observational study utilizing the COVID-19 Research Database to analyze factors contributing to telehealth inequities showed that patients with high school education or less were less likely to use telehealth than those with a bachelor's degree or higher [38]. This is important to consider in the context of the COVID-19 pandemic as telehealth visits became an increasingly popular way to access the healthcare system. It is evident that numerous factors contribute to the existing racial health inequalities in Gary that are further highlighted by the COVID-19 pandemic. Overcoming these inequalities requires coordinated effort from multiple local, state, and national organizations. As it was described in previously published studies, clinicians, public health professionals and researchers can all play a role in the advancement of health equity for Black Americans and other racial and ethnic minority communities [24]. Further directions for the current study include investigating which socioeconomic disparities specifically have the greatest impact on COVID-related health outcomes. We discussed several factors that may have contributed to the disparities that were found in this study, including limited access to testing, lack of health insurance and decreased access to healthcare overall, pre-existing comorbid conditions, vaccine hesitancy, and food insecurity. While these factors collectively contribute to disparities in COVID-19 impact, elucidating the associations of each factor would help to form targeted strategies to mitigate their impact. In addition to providing valuable insights to guide higher-level policy and resource distribution, this information could help clinicians to risk-stratify patients in their clinical practice. Overall, the current findings and future directions have the potential to characterize which socioeconomic factors should be specifically targeted at multiple levels to augment measures aimed at addressing the existing healthcare disparities. --- Conclusions The current study showed that: (1) population size affected the number of positive cases; (2) racial disparities are present in the number of infections, hospitalizations, and deaths in the Black population versus the non-Black population in Gary; and (3) COVID-19 hospitalizations have a significant negative association with zip code income. The critical assessment and evaluation of these results can assist the Gary Health Department in creating a more targeted COVID-19 response to better address the needs of the Gary community. In addition to addressing COVID-19 impact, informed measures would also help to prevent the exacerbation of these disparities by future outbreaks. It is important to note the stark disparities in COVID-19 impact between the neighboring zip codes found in this study. Negative differences in health outcomes for neighboring residents are concerning for factors that can be addressed by local interventions, and zip code data may serve as a strong starting point to develop these interventions [39]. As a microcosm for impoverished, majority non-White urban communities across the U.S., these findings should prompt further investigation by health departments into local trends to inform policy and resource allocation. Medical students and medical school faculty play a unique role in community health, and in collaboration with the local health departments, can promote health education in high-risk areas with strategically accessible vaccination sites, regular health screenings, and educational initiatives to promote overall healthy lifestyles. Campaigns to foster trust between medical professionals and community members would also aid in addressing hesitancy with vaccinations and other preventative or active treatment options. Connecting with grassroots organizations and community leaders is another way to build partnerships and promote sustainable change in minority communities. As research findings of studies on health disparities can be influential in the redistribution of preventative resources, novel treatments and educational initiatives to overcome health misinformation and mistrust and identifying and evaluating the causes of racial and economic disparities in healthcare are critical steps in addressing disparities in COVID-19 impact. --- Data Availability Statement: The data presented in this study are available on request from the corresponding author. Data was obtained from the Gary Health Department and are available after approval of the Gary Health Department. --- Institutional Review Board Statement: IUSM IRB review of the study was submitted for evaluation. It was determined that, since all data used in this study were de-identified, no IRB approval was needed. Informed Consent Statement: Patient consent was waived. Indiana University School of Medicine IRB review of this study was submitted for the evaluation. It was determined that since all data used in this study were de-identified, no IRB approval was needed. --- Conflicts of Interest: The authors declare no competing or financial interest. No financial disclosures was reported by the authors of this paper.

Background: The COVID-19 pandemic further exposed the prevalence of existing health disparities in Black communities in the U.S. The current study evaluates COVID-19 data collected in Gary, Indiana, from June 2020 to June 2021. We hypothesized that the number of COVID-19 cases, hospitalizations, and deaths were influenced by race and income. Methods: In collaboration with the Gary Health Department (GHD), we analyzed demographic data on COVID-19-positive cases. Results: Compared to Gary's non-Black population, age-and population-adjusted rates of hospitalizations and deaths in the Black population were 3-fold (p < 0.0001) and 2-fold (p < 0.05) higher, respectively. This is despite a higher infection rate (p < 0.0001) in the non-Black population. The median household income of a zip code was negatively correlated with COVID-19 hospitalizations (R 2 = 0.6345, p = 0.03), but did not correlate with infections and deaths. Conclusions: The current study demonstrates clear health disparities of income and race in the context of COVID-19-related infections and outcomes in the city of Gary. Indiana University School of Medicine Northwest and GHD officials can collaborate to utilize these data for the reallocation of resources and health education efforts in Gary's highly populated, low-income, and predominantly Black neighborhoods. It should also prompt further investigation into national health resource allocation.