Associations between WASH-related violence and depressive symptoms in adolescent girls and young women in South Africa (HPTN 068): a cross-sectional analysis

N/A

Objective There is a lack of research on experiences of WASH-related violence. This study aims to quantify the association between experience or worry of violence when using the toilet or collecting water and depressive symptoms among a cohort of young women in South Africa. Methods Data are from visit 3 of the HPTN 068 cohort of adolescent girls in rural Mpumalanga Province, South Africa. Participants (n=1798) included in this analysis were aged 13-21 at baseline. Lifetime experience of violence or fear of violence when using the toilet and collecting water was collected by self-report; depressive symptoms in the past week were measured using the Center for Epidemiological Studies Depression Scale (CES-D). We used G-computation to calculate the prevalence difference (PD) and prevalence ratio of depression (CES-D score >15) associated with each domain of violence, controlling for baseline covariates. Findings A total of 15.1% of respondents reported experiencing violence when using the toilet; 17.1% reported experiencing violence when collecting water and 26.7% reported depression. In adjusted models, those who reported experiencing violence when using the toilet had an 18.1% higher prevalence of depression (95% CI: 11.6% to 24.4%) than those who did not experience violence when using the toilet. Adjusted prevalence of depression was also higher among those who reported violence when collecting water (PD 11.9%, 95% CI: 6.7% to 17.2%), and who worried about violence when using the toilet (PD 12.8%, 95% CI: 7.9% to 19.8%), as compared with those who did not report these experiences. Worrying about violence when collecting water was not associated with depression after adjusting for covariates. Conclusion Experience of WASH-related violence is common among young women in rural South Africa, and experience or worry of experiencing violence is associated with higher prevalence of depressive symptoms. Trial registration number NCT01233531; Post-results.

Data for this study are from the HIV Prevention Trials Network (HPTN) 068 cohort of adolescent girls and young women in rural Mpumalanga Province, South Africa, a longitudinal cohort established in 2012 to estimate the effect of cash transfers, conditional on staying in school, on HIV incidence. Participants were eligible if they were between the ages of 13 and 20 years, enrolled in grades 8–11 at a participating public school, unmarried, not pregnant at the time of enrollment, able to read, had parents or guardians able to open a bank account, and resided in the Medical Research Council/Wits University Agincourt Health and Socio-Demographic Surveillance System (AHDSS) study site. The AHDSS study site is in a rural area of Mpumalanga Province, South Africa that is characterised by high HIV prevalence, high poverty, and migration for work.20 Most households lack access to piped water in their dwellings, and sanitation is rudimentary.20 All households with eligible adolescent girls and young women in the study area were recruited. A total of 2533 participants enrolled and were followed annually for up to 4 years. At each study visit, participants completed interviewer-administered surveys on a wide variety of domains that included economic activities, health behaviours, health knowledge, and attitudes towards social norms. Sensitive items, such as sexual behaviour and mental health, were completed by the participants themselves via Audio Computer Assisted Self Interview (ACASI), where participants listen to questions and response categories through headphones and select their responses. Prior research has found higher reporting of sensitive issues via ACASI as compared with interviewer-administered surveys.21 Participants’ heads of households completed surveys about household composition and wealth at each visit. Full details on the study recruitment and procedures, including a full description of the sample22 and primary trial outcomes,23 have been published elsewhere. The primary outcome of interest for this study is depression, assessed using the Center for Epidemiological Studies Depression Scale (CES-D).24 The CES-D is a 20-item measure that assesses symptoms of depression over the past week, with frequency of experiencing each symptom as rarely/none of the time, some of/a little of the time, occasional or a moderate amount of time, or all of the time. Scores can range from 0 to 60; in keeping with the literature, we used a cut-off of 16 or greater as an indicator of depression.25 Our exposures of interest are WASH-related violence, assessed across four domains: experience of violence when collecting water, experience of violence when using the toilet, fear of violence when collecting water and fear of violence when using the toilet. Direct experience of violence was assessed by the following question: ‘How often have you experienced violence when collecting water?’ and fear of experiencing violence was assessed by the following question: ‘Do you ever feel concerned or worried about experiencing violence when using the toilet?’ Participants were categorised as being exposed to experience of violence or fear of violence if they responded ‘Just a few times’, ‘Regularly/about once a week’ or ‘Every day’, as opposed to ‘Never’. Though both direct experiences with violence and fear of experiencing violence were assessed at the same time point, based on the wording of the question, we assume that direct experience of violence precedes fear of experiencing violence. We used a directed-acyclic graph to identify a minimally sufficient set of literature-based confounders available in our study. We identified the following sociodemographic covariates of interest: age at time of survey,11 12 26 maternal and paternal education,26 orphan status,26 household food insecurity in the past 30 days,11 18 27 decile of household capita consumption12 26 and any negative events experienced by the household in the past 12 months18 27 (assessed via the household survey). Negative events reported in the household survey included experiences such as death or serious illness of a household member, loss of livestock or crop failure, job loss or loss of government grants or loss or destruction of property. We also controlled for toilet type9 11 and household water source.18 Data for this analysis are drawn from visit 3. We limited our analysis to this time point as only 35.5% of the enrolled sample participated in visit 4 given planned study exit due to graduation from high school. While this data is cross-sectional, experiences or fear of violence was assessed as a lifetime measure, and we assume that those experiences precede the depression measure, which evaluates depression symptoms experienced in the past week. We also assume that experience of violence precedes fear of violence. Records with missing data on parental education were treated as a separate category in analysis; missing data on household food insecurity (n=7), decile of total household per capita expenditure (n=8) and orphan status (n=17) were directly extracted from prior household survey visits. We used G-computation to calculate the predicted marginal prevalence difference of depression and the predicted marginal prevalence ratio of depression associated with each individual domain of WASH-related violence. All models adjusted for age at time of visit, maternal education, paternal education, orphan status, household food insecurity, decile of household capita consumption, negative household experiences, and trial arm, and accounted for clustering by village by using the nonparametric cluster bootstrap to calculate 95% Wald-type percentile-based CIs from 500 resamples. Models assessing fear of violence when collecting water and fear of violence when using the toilet additionally adjusted for prior experience of violence when collecting water and when using the toilet, respectively. Models assessing experience or fear of violence when collecting water additionally adjusted for household water source; models assessing experience or fear of violence when using the toilet additionally adjusted for household toilet type. All analyses were performed using Stata V.15.28 All studies, including the HPTN 068 trial, conducted in the AHDSS study site receive permission to undertake research activities from a forum comprised of community and village leaders. Findings from the main trial were communicated to the community via community meetings and factsheets. Additional details on community involvement are available via the study site’s Public Engagement Office.29

• jayaweera, ruvani t.
• goin, dana e.
• twine, rhian
• neilands, torsten b.
• wagner, ryan g.
• lippman, sheri a.
• kahn, kathleen
• pettifor, audrey
• ahern, jennifer