Introduction

Gender-based violence (GBV) against women and girls is a persistent public health threat that puts women’s health and wellbeing at risk and violates their human rights. GBV can take different forms including physical, sexual and emotional abuse [1]. The World Health Organization (WHO) estimated that globally, in 2018, 1 in 3 women ever experienced physical or sexual violence, or both, by an intimate partner or non-intimate male partner, with little change in this proportion over the last decade [1–3]. Evidence points to violence starting early, with 1 in 4 ever-partnered adolescent girls and young women (AGYW) aged 15–24 years globally having already experienced physical or sexual violence by a partner [1,3]. The African region is among the WHO regions with the highest lifetime prevalence of violence, with 36% of women aged 15–49 years ever experiencing physical and/or sexual violence from a partner or non-partner [1,3]. Studies from multiple settings in sub-Saharan Africa (SSA), including Kenya and South Africa, also indicate high levels of exposure to GBV among AGYW [4–12].

Violence against women has far-reaching negative physical, mental, sexual and reproductive health effects. These include physical trauma or injuries; depression, post-traumatic stress and anxiety disorders, and suicide; unintended pregnancies, abortions, miscarriages and still births; and increased risk of acquiring sexually transmitted infections (STIs) [13–17]. Furthermore, a recent Global Burden of Disease study ranks intimate partner violence (IPV) (any behaviour within an intimate relationship that causes physical, psychological or sexual harm to those in the relationship) as the second most common factor contributing to disability-adjusted life years globally in women aged 20–24 years [18,19].

Exposure to GBV has a causal association with incidence of HIV infection [20]. Multiple studies have documented pathways between exposure to violence and women’s vulnerability to HIV [9,16,21,22]. Such pathways include direct infection from forced sexual intercourse and associated genital trauma. Proximate determinants include limited control over the circumstances of sexual intercourse including negotiating condom use, reduced access to HIV prevention services, and characteristics of violent male partners such as having multiple and/or concurrent sexual partners and a higher likelihood of HIV-positivity [7,8,10,23–25]. More distal, structural drivers including poverty and economic stresses, gender inequalities, prevailing social norms, and misogyny, create the conditions that enable violence and HIV acquisition to persist.

Increasing attention has focused on addressing the structural drivers of exposure to violence among women. Recent evidence has shown the potential of interventions combining economic strengthening and gender-transformative programming in reducing IPV and HIV risk behaviours among women [26]. Intervention with Microfinance for AIDS and Gender Equity (IMAGE) combined gender training and community mobilization with microfinance and reduced IPV by 55% among women aged between 15 and 49 in rural South Africa [27]. Building from this design, the MAISHA trial conducted in Tanzania subsequently demonstrated a reduction in physical and/or sexual IPV among women aged 33–46 years [28]. However, the evidence is less clear among young women [29,30]. A government-implemented cash plus intervention in Tanzania that combined life-skills training, mentoring and microfinance was effective at reducing female participants’ experiences of sexual violence and male perpetration of physical violence [31]. While some interventions have shown positive results on reducing IPV and GBV outcomes, others aimed at addressing the structural drivers of exposure have not been effective. For instance, the adolescent girls empowerment program (AGEP) in Zambia and the ‘Girl Empower’ program in Liberia had no effect on the acceptability of IPV and protection from sexual violence, respectively [32,33]. A few interventions have aimed to reduce violence among AGYW, but results have been mixed [29,30]. Furthermore, several interventions reporting reductions in violence, such as IMAGE [34], included AGYW but did not publish age-stratified results for those aged under 25 [30].

The DREAMS (Determined, Resilient, Empowered, AIDS-free, Mentored and Safe girls) Partnership established by the United States President’s Emergency Plan for AIDS Relief (PEPFAR) was developed primarily to reduce HIV incidence among AGYW in 10 SSA countries by tackling the multiple, underlying causes of AGYW vulnerability to HIV. This includes addressing GBV, incomplete schooling and socioeconomic vulnerabilities, and empowering AGYW and their communities through a holistic package of evidence-based biomedical, behavioural and social protection interventions [35]. The DREAMS core package integrates violence prevention education into school-based and community-based interventions, including Start, Awareness, Support, Action (SASA!), Stepping Stones and Vhutshilo, all curriculum based interventions that aim to raise awareness, support and action around violence in the community [34,36–38]. DREAMS also provides post-violence care services, social asset building interventions and other interventions that aim to empower AGYW, strengthen the families of AGYW economically, and mobilize communities [35,39]. These interventions are delivered by DREAMS implementing partners (IPs) who are organizations experienced in providing sexual and reproductive health and/or HIV-related services in the recipient countries. A recent publication reported the uptake of DREAMS interventions in the three study sites [40]. The most accessed interventions in all study sites were HIV testing, social protection, and school-based interventions. There was higher uptake of HIV testing and contraception among older AGYW while younger AGYW more frequently accessed social asset building and school-based interventions. There were differences in uptake by site for interventions such as contraception, parenting and social protection programmes, and the provision of PrEP. Among the DREAMS implementation countries are Kenya and South Africa, which are considered HIV endemic countries where the risk among AGYW is high. Recent evidence from DREAMS program in South Africa shows a relationship between IPV and the HIV care and treatment cascade including awareness of HIV status, ART uptake and viral suppression [41].

Despite increasing evidence of efficacy from trials in controlled conditions, the effectiveness of multi-component interventions addressing violence against AGYW is limited in real-world circumstances [26]. This study assessed the impact of DREAMS on experiences of physical, emotional, and sexual violence among AGYW in three settings in Kenya and South Africa over three years of DREAMS implementation.

Methods

Results

Cohort retention

In Nairobi, 852 AGYW aged 15–22 years were followed-up in 2019 out of 1081 enrolled into the cohort in 2017 (79% retention rate). In uMkhanyakude, 1712 AGYW were followed-up in 2019 out of 2184 enrolled in 2017 (78%). In Gem, 1018 AGYW were followed-up out of 1171 enrolled in 2018 (87%). Cohort flow diagrams are presented in Fig 1. The main reasons for loss to follow-up in all settings were out-migration from the enumeration area and participants not being reached.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Cohort flow diagrams of AGYW aged 13–22 years followed up between 2017 and 2019 in Nairobi, uMkhanyakude and Gem.

https://doi.org/10.1371/journal.pgph.0001818.g001

S1–S3 Tables show loss to follow up (LTFU) status of AGYW by individual characteristics at enrolment in Nairobi, Gem, and uMkhanyakude. AGYW who were invited to DREAMS in 2017, and in school (Nairobi and uMkhanyakude) or with secondary education (Gem), were more likely to be followed-up in 2019. Those who were older at enrolment and had ever had sex or a pregnancy (Kenyan settings), were less likely to be followed-up. In Nairobi, AGYW who reported any recent violence in 2017 were less likely to be followed-up in 2019 than those who had not experienced violence, while there were no differences in LTFU by violence in Gem or uMkhanyakude.

Characteristics of AGYW

Overall characteristics at enrolment of AGYW who were followed-up in 2019 are shown in Table 1. Higher proportions of younger than older AGYW were enrolled in Nairobi (55% vs 46%), Gem (61% vs 39%) and UMkhanyakude (57% vs 43%). The majority of AGYW in Nairobi (63%) and UMKhanyakude (79%) were in school. About one third of AGYW in all sites reported being food insecure (34% in Nairobi, 23% in Gem, 31% in UMkhanyakude). While socioeconomic status (SES) was evenly distributed in Nairobi, there were more AGYW in the low SES group in Gem (42%) and UMkhanyakude (36%). Characteristics of AGYW stratified by whether they were a DREAMS beneficiary by 2018 are presented in S4 Table. The majority were classified as DREAMS beneficiaries (74% Nairobi; 57% Gem; 53% uMkhanyakude). High proportions were in school, more so among DREAMS beneficiaries compared to non-beneficiaries (68% vs 51%, Nairobi; 88% vs 69%, uMkhanyakude). DREAMS beneficiaries were also younger compared to non-beneficiaries (e.g., 59% vs 42% were aged 15–17 in Nairobi), more frequently reported food insecurity (e.g., 26% vs 17% in Gem) were of lower socio-economic status (e.g., 39% vs 32% low SES, uMkhanyakude), and had never had sex (e.g., 69% vs 56% Nairobi).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of DREAMS beneficiaries and non-beneficiaries at time of cohort enrolment in Nairobi, Gem and uMkhanyakude.

https://doi.org/10.1371/journal.pgph.0001818.t001

Levels of violence

Fig 2 shows the levels of different forms of violence reported by AGYW over the evaluation period by age group. The proportion of AGYW who self-reported experiencing any form of violence was higher in Nairobi than in Gem and uMkhanyakude (e.g., 29% vs. 18% vs. 19%, respectively, in 2019) (Fig 2). In Nairobi and uMkhanyakude, a decline was observed in the proportions of AGYW reporting violence between 2017 and 2019 (Fig 2). The proportion who self-reported experiencing any form of violence fell from 41% to 29% in Nairobi, and from 35% to 19% in uMkhanyakude. Levels of violence reported remained stable in Gem at 18% in 2018 and 2019. Emotional and physical violence were the most commonly reported form of violence (in 2019, 22% and 15% in Nairobi; 12% and 11% in uMkhanyakude; 10% and 11% in Gem, respectively), while levels of sexual violence were under 10% in 2019. However, the proportion of AGYW self-reporting sexual violence was higher in Nairobi in 2017 (15%) and 2019 (9%) than in Gem and uMkhanyakude (3–6%). In Nairobi, a slightly higher proportion of AGYW aged 18–22 years reported experiences of violence compared to those aged 15–17 years, particularly for sexual violence (11% vs 7%). In Gem and uMkhanyakude, levels of violence reported by AGYW aged 13–17 and 18–22 years were similar.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Violence trends among AGYW in Nairobi, Gem and uMkhanyakude between 2017 and 2019.

https://doi.org/10.1371/journal.pgph.0001818.g002

Impact of DREAMS on experience of violence

Descriptive results and multivariable logistic regression.

Table 2 presents the proportions reporting violence in 2019 by DREAMS invitation status, and adjusted odds ratios (aORs) for the associations between invitation to DREAMS and experience of violence. Proportions reporting recent violence were very similar among those invited to DREAMS and those not invited, with small absolute differences of 1–2 percentage points and the largest increase observed was 6% (e.g., 17% of those invited to DREAMS and 13% of those not invited reported physical violence in Nairobi).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Association between DREAMS and experiences of violence among AGYW followed up in 2019 by site using conventional logistic regression.

https://doi.org/10.1371/journal.pgph.0001818.t002

Across all settings, there was very little evidence of an association between DREAMS and overall levels of violence: adjusted odds-ratio (aOR) of 0.9 (CI 0.6–1.3) in Nairobi; aOR 0.9 (CI 0.6–1.2) in Gem; and aOR 1.0 (CI 0.8–1.3) in uMkhanyakude. In Nairobi, there was weak evidence of greater odds of experiencing physical violence among DREAMS beneficiaries compared to non-beneficiaries (aOR 1.4, (CI 0.9–2.3)). S5 Table presents an expanded results of the logistic regression by invitation to DREAMS and age-group for the three study sites. For older AGYW in Gem, there was weak evidence of lower odds of experiencing any form of violence among DREAMS beneficiaries compared to non-beneficiaries (aOR 0.6, (CI 0.4–1.1), p = 0.09). Site was largely unattributed to any notable difference in findings (see online S5 Table).

In terms of other factors associated with experiences of any form of violence in the adjusted multivariable logistic regression models, violence among AGYW in Nairobi was associated with food insecurity (aOR 1.4, CI (1.0–2.0), p = 0.03), ever having sex (aOR 1.9, CI (1.1–3.3), p = 0.02), and ethnicity (more likely among the Luo and Luhya ethnic communities compared to other groups, p = 0.004) (S6 Table). In Gem, AGYW with lower socio-economic status were less likely to experience any violence than those with higher status (aOR 0.6, CI (0.3–1.1), p = 0.05). In uMkhanyakude, none of the characteristics examined were associated with violence.

Estimated causal impact of DREAMS

The estimated causal effect of DREAMS on experiences of violence among AGYW is presented in Table 3 and Fig 3. There was no evidence of a difference attributable to DREAMS in the overall levels of violence experienced in 2019. In Nairobi, it was estimated that proportions of AGYW who experienced any form of violence would be similar under the counterfactual scenarios that all AGYW were DREAMS beneficiaries and that none were DREAMS beneficiaries (28.7% compared to 27.8% respectively, difference +0.9% with CI (-6.5%, +7.7%)). For Gem, the equivalent percentages were 17.3% versus 19.7% (-2.4% (-7.6%, +2.8%)) and for uMkhanyakude 18.7% and 19.1% (-0.4% (-4.3%, +3.5%)). In Gem, among the older age group, there was a reduction of 4.7% estimated if all AGYW were DREAMS beneficiaries (16.8%) versus if none were DREAMS beneficiaries (21.4%), with weak statistical evidence to support this difference (CI (-12.8%, +3.3%)).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Estimated causal impact of DREAMS on different experiences of violence among AGYW in Nairobi, Gem, and uMkhanyakude.

https://doi.org/10.1371/journal.pgph.0001818.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Estimated causal effect of DREAMS on experiences of violence among AGYW followed up in 2019 by site.

https://doi.org/10.1371/journal.pgph.0001818.t003

There were some notable findings by violence form. In Nairobi, it was estimated that proportions of AGYW who experienced recent physical violence would increase from 10.7% to 16.0% if all AGYW were DREAMS beneficiaries (difference +5.3% with CI (0.2%,10.0%). This rise was evident in both the younger and older age groups, with weak statistical evidence in sub-analyses by age-group. In Gem, there was borderline evidence of a decrease in experiences of sexual violence, attributable to DREAMS, among older AGYW (difference -2.7% with CI (-6.9%, +1.1%). For all other forms of violence and age sub-groups, there was no evidence of a causal impact of DREAMS.

Compared to the primary analyses presented above, results were very similar in sensitivity analyses that used alternative approaches to control for confounding (S18 Table).

Discussion

This study assessed the impact of DREAMS’ multi-component intervention on experiences of violence among AGYW in ‘real-world’ non-trial conditions in rural and urban settings in Kenya and South Africa. Our results revealed that recent violence perpetrated by males was common among AGYW. Applying a causal inference analysis, we found little evidence of DREAMS’ impact on such experiences within three years of implementation.

Across the three diverse study settings in 2019, about 20–30% of AGYW reported experiencing at least one form of violence inflicted by a man in the previous 12 months. These estimates are broadly consistent with the WHO Africa regional estimates for 2018 of 36% prevalence of lifetime IPV and/or non-partner violence among all women aged 15–49, and country-specific prevalence estimates of recent IPV among ever- partnered women aged 15–49 years for Kenya (23%) and South Africa (13%) [3]. Recent studies conducted in Kenya, South Africa, and other sub-Saharan African countries indicate persistently high levels of GBV among women [4–7]. Levels of violence reported in our study were highest in informal settlements in Nairobi. An earlier cross-sectional study, conducted in the same setting, to examine the association between violence victimization and expectations for achieving aspirations, reported that one-third of adolescent girls aged 11–15 had experienced at least one form of violence [6]. A similarly high level of recent IPV (30%) was reported at baseline among young women aged 18–30 living in urban settlements in Durban, South Africa, before intervention implementation [60]. High vulnerability driven by livelihood insecurity and poor access to services in urban informal settlements has been linked to poor health outcomes, including exposure to violence and HIV, which may partly explain the high prevalence of violence we observed in Nairobi [6,60,61]. This interpretation is further supported by higher levels of violence reported among those who were food insecure or of lower socio-economic position in our Kenyan analyses.

We found little evidence of impact of the multi-component DREAMS intervention package on experiences of violence across our study settings. Challenges with intervention implementation may be one of the reasons explaining the limited impact. For instance, a study conducted in South Africa to assess the effect of community mobilisation on men’s IPV perpetration reported contextual challenges as a barrier to successful implementation [62]. This is consistent with process evaluation reports which have reported contextual factors such as loss to follow-up of AGYW who either migrated or were unable to keep up with the interventions; or the volunteers participating in mentoring sessions due to better economic opportunities. Furthermore, there were some interruptions in programming in the first year of implementation in Kenya due to political interruptions brought about by national elections. Among the DREAMS core package of interventions, those targeting GBV reduction include parenting/ caregiver programs, school-based HIV and violence prevention, and community education and norms change [35]. Available evidence indicates low uptake (<10%) of parenting/caregiver and community mobilisation interventions targeted at young men and community members, in contrast to the very high levels of participation of AGYW in girl-centered interventions [51]. Interventions targeted at the community and partners of AGYW have been shown to be effective at reducing GBV by changing attitudes and promoting equitable social norms against GBV, especially among men. For example, the ‘Raising Voices SASA!’ kit and Safe Homes and Respect of Everyone (SHARE) interventions implemented in Uganda were associated with lower IPV incidence and more supportive community responses to female victims [34,63]. However, these examples come from trial conditions with high fidelity, and more time and training may be needed to embed and adapt community-based norms-change and GBV interventions that were newly introduced into these settings with DREAMS investments [34,52,54,64].

The limited effect of DREAMS’ influence on gender attitudes and norms among men is evident in related analyses, which found little change over time in attitudes among large, representative samples of men in Gem and Nairobi [65]. Meanwhile, the proportion of AGYW with gender-equitable attitudes (toward sexual relationships and gender-based violence) increased annually during the DREAMS implementation period. Evidence from a general population-based survey done in the same community in Nairobi during 2019 also showed that around a fifth of young men in partnerships admitted to perpetration of violence, and that this was associated with less equitable attitudes towards gender norms [66]. Shifting ingrained inequitable social norms is recognised as challenging, for example due to resistance from patriarchal and traditional structures, and may take longer than the 2–3 year period of our evaluation. Recommendations for practitioners designing social norms interventions have recently been outlined, which may aid the design and adaptation of programmes in future [67]. The drawbacks of interventions that treat norms in isolation have been emphasized, and the importance of integrating norms change approaches, and those aimed at tackling GBV, with other individual, social, and structural approaches, as complex interventions such as DREAMS have aimed to do [67].

While some gender-transformative interventions aimed at youth in other contexts have been shown to be effective at addressing inequitable social norms and reducing GBV, other studies have shown limited impact [30,60,68]. For example, the Stepping Stones and Creating Futures intervention in South Africa combining educational gender-transformative curricula and life-skills training to strengthen livelihoods did not impact on experiences of IPV among young women after two years of implementation, although the intervention was effective at reducing men’s self-reported perpetration of IPV [29]. Although the Stepping Stones and Creating Futures intervention included participatory sessions with young out-of-school men, it did not include a broad community-based element as other more successful interventions such as SASA! have done [34]. Yet, there are also examples of interventions with a narrower focus, such as PREPARE in South Africa which incorporated school-based HIV and violence prevention education for girls and boys, health services, and safety programme components, that have been successful in reducing IPV victimization among adolescent girls [69]. While the evidence is, therefore, unclear on the necessity of including community-norms change components to reduce GBV among AGYW, this will be important in reaching the Sustainable Development Goals (SDGs) for health and gender equality. Indeed, a recent systematic review of evidence-based gender-transformative health interventions for children and young people globally concluded that tackling the societal and structural elements of restrictive gender norms, to elicit systemic change through social participation and multi-level, multi-sectoral action, was key to achieving the SDGs [70].

We observed an increase in reported physical violence due to DREAMS in Nairobi, and a trend in this direction for other violence forms in the same setting. One possible explanation is an increase in awareness, acknowledgement and disclosure of personal experiences of violence in this setting, and therefore increased reporting of violence, as a result of exposure to violence prevention education and empowerment curricula from the DREAMS intervention package. Another Nairobi-based study reported an increase in disclosure of experiences of sexual assault to others following an empowerment and self-defence skills intervention [71]. However, this would not explain the absence of an increase in other forms of violence in Nairobi (sexual and emotional), or increased reporting of violence among DREAMS beneficiaries in the other two settings. Alternatively, the finding may reflect a true increase in the violence experienced by DREAMS beneficiaries in Nairobi, as a result of participating in DREAMS. This may be explained using the backlash theory where there is resistance to violence-prevention interventions against women and girls, which may include increased risk of more violence [72]. For instance, in a related process evaluation of DREAMS in Nairobi, mentors reported conflict with some male partners who resisted the empowerment of their female partner in DREAMS safe spaces [52]. There are other examples of interventions seeking to reduce violence against young women in which impacts have been mixed across different violence outcomes, and an unintended increase in violence has been reported [73]. This phenomenon has been described in the social sciences literature, with the rationale that a man may inflict violence to re-impose authority over his wife/partner [73–75]. It is an unintended consequence of empowerment programmes which must be carefully considered in programme planning, to minimise harm and maximise safety of participants. It is more likely to occur in programmes that do not effectively reach the perpetrators of violence. Finally, in-depth analysis of qualitative data should be conducted to understand this exceptional finding, and whether it may be spurious.

Our results were generally consistent across age-groups of AGYW, except for Gem where there was a pattern of reduced violence among older but not younger AGYW. This is consistent with our related analysis of equitable attitudes towards violence-related gender norms, which showed more equitable norms in the older compared to the younger age group [65].

Descriptive analyses of trends over time suggested a large decline in the levels of violence between 2017 and 2019 in Nairobi and uMkhanyakude. The findings can be supported by the aging out theory that indicates that most GBV exposure or perpetration tends to reduce over time due to various reasons. A possible explanation for this observation in Nairobi is the differential cohort attrition by violence at enrolment, such that those who had reported violence at enrolment were more likely to be LTFU by 2019 than those who had not experienced violence at enrolment, which may have artificially deflated the proportions experiencing violence by 2019. In uMkhanyakude, the decline does not appear to be explained by cohort attrition. Here, it is possible that other (non-DREAMS) outreach campaigns against GBV in the community, organised by the municipality, could have influenced ‘background’ levels of violence experienced and reported by young women in the community. Additionally, the DREAMS violence-prevention interventions implemented in uMkhanyakude, Vhutshilo and Stepping Stones [36,38], targeted the wider communities and DREAMS may have heralded a wider awareness of GBV and social norms change.

Key strengths of this study were the representative selection of AGYW from three established DSS sites, harmonised data collection, and high cohort retention across the study settings. While cohort retention was high, it was differential by AGYW characteristics, possibly leading to a small degree of selection bias. A further limitation is that violence outcomes were self-reported and AGYW may have under-reported their experiences, particularly of sexual violence, due to feelings of shame or fear of blame or repercussion. However, overall experiences of recent violence were commonly reported and in-line with other reports in the literature. There is also a chance that the trained field interviewers may have been perceived to be DREAMS program staff rather than independent evaluation staff by respondents, potentially causing some social desirability bias in their responses. A degree of bias may also have been introduced if DREAMS invitees were more, or less, likely to report experiences of violence, as a result of their experience in the programme. Misclassification of exposure status could also have occurred if some AGYW did not report themselves as DREAMS invitees, potentially underestimating the proportion of DREAMS beneficiaries. This is unlikely in Kenyan settings where a single DREAMS IP invited AGYW to participate in DREAMS, although it could plausibly have occurred in uMkhanyakude. Furthermore, the tool did not collect information on other avenues with which violence against AGYW may have been perpetrated such as through technological abuse, stalking, financial abuse etc. which may also be a potential source of misclassification of exposure. Our results may not be generalizable to all DREAMS districts, but the diverse implementation contexts can offer insights for other settings implementing DREAMS.

Conclusions

Our study contributes important evidence to understanding whether complex interventions have an impact on reducing experiences of violence among AGYW in real-world non-trial contexts. We have shown that recent experiences of GBV were common among AGYW, especially in urban Kenya, and DREAMS had limited impact on reducing violence after 2–3 years of implementation. While there are a handful of efficacious multi-component interventions that reduce GBV towards AGYW from trial settings, there is need for more investment in implementation research to bridge the findings from efficacy to population-level effectiveness. Furthermore, tailoring programs to work in close partnership with communities could help address the differing perspectives and needs of men and women, and different age groups to improve program outcomes.

Supporting information

Acknowledgments

We acknowledge the contributions of Beatrice Maina, Jane Osindo, Silvia Njoki, Jaco Dreyer, Thembhlehle Zuma and Sammy Khagayi to the study. APHRC acknowledges the support and cooperation of residents of Viwandani and Korogocho for their participation in the data collection activities during the study. We are grateful to the field interviewers, research assistants and research administrators who made data collection activities a success. We also acknowledge the role of Daniel Carter during 2018–2019, who worked as a study statistician and as a member of the study team’s Data Analysis Working Group to co-develop methods of analysis and their programming in Stata. We also extend our appreciation to our research community including the community advisory boards in NUHDSS, Gem and uMkhanyakude.