Data collection.

We conducted a cross-sectional secondary data analysis using logistic regression and moderation testing with data from two data sources collected during the final 2018–2019 waves of the respective studies. The first dataset included young women from the HPTN 068 study, a randomized controlled trial assessing cash transfers conditional on school attendance for HIV prevention among AGYW. The HPTN 068 study methods have been published elsewhere [26]. Briefly, females who met the inclusion criteria—including being aged 13–20 years, residing in the AHDSS villages, not pregnant or married, and enrolled in secondary school—were recruited between March 5, 2011 and December 12, 2012 [26]. Written consent or assent for young women and parent/guardian written consent (required if participant <18) was completed in the language of their choice (xi-Tsonga or English). Between 2011 and 2015, annual questionnaires were administered, available in the local language, xi-Tsonga, or English. Detailed information on socio-demographics, relationship history, sexual reproductive health and HIV prevention behaviors, and mental health was collected. Biological samples to determine HIV-serostatus were collected at each visit. Two post-intervention visits were conducted between 2015–2016 and 2018–2019, for which participants re-consented. For this analysis, inclusion was restricted to women aged 18–25 years who participated in the final 2018–2019 post-intervention visit and reported living in an AHDSS village at the time of this visit. A separate attrition analysis of the HPTN 068 cohort found no significant differences between participants who provided data for this last time point and those who did not in terms of baseline risk and protection factors or treatment arm [37].

Additionally, data was used from a 2018 cross-sectional population representative post-intervention survey of the Tsima CM Study, a cluster randomized-controlled trial that evaluated a community mobilization intervention designed to change gender norms and reduce HIV risk [27]. Between August –December 2018, 1,182 men and women AHDSS residents aged 18–49 years were sampled and written informed consent was obtained from all participants before administering the survey. Further details of the study design and results have been published elsewhere [38]. For this analysis, we used three scales from the CM population-based surveys that aligned with the theoretical conceptualization and framing of community-level resilience promoting factors and social resources.

Measures.

Transactional sex outcome: We constructed a binary ever/never variable, with participants classified as exposed if they reported engagement in any type of transactional sex. To create this lifetime measure, we aggregated data from each study and post-intervention visit where participants were asked yes/no questions about their three most recent sexual partners (e.g., Did you feel like you had to have sex with [partner] because they gave you money? …because they gave you things?). All marital relationships were considered non-transactional. Thus, participants were coded 1 if they reported ever engaging in any transactional sex with a non-marital partner at any study visit.

ACE exposure predictor: We constructed a dichotomous variable using validated measures from the WHO’s ACEs international questionnaire (ACE-IQ) that included 10 items spanning seven distinct ACE categories [39]. Participants were asked about five of the categories, including emotional, physical, and sexual childhood abuse and witnessing physical or emotional abuse in the household, that may have occurred during their first 18 years of life during the 2015–2017 post-intervention visit. Additionally, we created two separate measures for maternal and paternal death before age 18 by assessing responses during annual study visits about the death of a parent and the participant’s age at time of their parent’s death. If a participant’s parent died before age 18, they were classified as having experienced an adverse experience of maternal and/or paternal death during childhood (Full questions in S1 Table). Participants who self-reported ≥1 ACEs were reported as ACE exposed. To assess cumulative exposure and dose response, we calculated a summed ACE score (0–7) with a higher score indicating exposure to more ACE categories. Scores were categorized into three groups—0 ACEs, 1–2 ACEs and ≥3 ACEs—to align with previous research and enable comparison of findings [40,41].

Resilience moderators: Five scales were used to measure resilience-promoting factors at the individual (n = 1), interpersonal (n = 1), and community/village (n = 3) level. Each community-level variable represents a village’s average score across all survey respondents (regardless of demographics) using sample weights to match respondents to the target population of village residents [42]. All resilience measures were collected during the final 2018–2019 data collection waves.

Internal Resilience: We asked participants to self-rate their psychological resilience using the 25-item Connor-Davidson Resilience Scale (CD-RISC) [43]. The CD-RISC assessed individuals’ tenacity and self-efficacy, tolerance of negative affect, adaptability and secure relationships, and sense of purpose, control and meaning. Items score on a 5-point Likert scale with sum totals ranging from 0 to 100. Higher total scores reflect greater ability to cope with adversity and accept change positively [44]. The CD-RISC has demonstrated validity extensively, including among SA adolescents [45,46]. In this study, the Cronbach α was 0.94, indicating excellent internal consistency.

Social Support: We used the 10-item Social Provisions Scale to assess the extent to which participants reported familial and proximal resilience-promoting social environment resources were available to them [47,48]. Items covered: emotional attachment; advice or informational guidance; tangible help; opportunity for nurturance; and sense of belonging to a group [49]. Response options included a 3-point Likert scale. Summed scores ranging from 10 to 30 were calculated, with higher scores reflecting higher levels of social support. Cronbach α = 0.87 in this study.

Social Cohesion: This 6-item scale assessed people’s perceptions about their village’s community-level connectedness conceptualized as promoting individual resilience, including items about the extent members of their community trust each other, share values, and look out for one another [50,51]. Summary scores were calculated for each respondent, which were then aggregated to create a mean village score for each community.

Community Consciousness: This 17-item scale was used to assess how well village residents work together to address village-level issues and solve problems. Individual respondent summary scores were calculated, followed by an aggregated mean score for each village.

Organizations and Networks: This measure asked residents 11 yes/no questions about which types of community organizations (e.g., youth, church, cultural, etc.) support and work to improve people’s lives in their village.

Control variables: Based on prior findings from the parent study and preliminary analyses, we controlled for: age of participant; highest level of educational attainment grouped into three categories [(1) some secondary/high school education, (2) completion of Grade 12 matric/high school, (3) completion of tertiary education)]; and means-tested household social grant recipient as an indicator for socio-economic need [52]. Duplicative confounders (e.g., social grant receipt and food insecurity) were excluded to avoid multicollinearity and maintain parsimony.

Quantitative data analysis.

All statistical analyses were conducted in SAS v.9.4 [53]. Descriptive statistics was used to check for potential errors and outliers, to summarize the socio-demographic characteristics, and to estimate the prevalence and patterns of ACEs. We used chi-square tests to assess relationships between ACE group categories and demographic variables, and t-tests estimated associations between ACE exposure and mean resilience scores. Missingness was less than 1% on each of the variables used in the models; all models were estimated using complete cases. We used logistic regression to estimate unadjusted and adjusted odds ratios of the links between ACE exposure and transactional sex. A sub-group analysis model to estimate women who reported 1–2 ACEs vs. ≥ 3 ACEs was also fitted. All adjusted models controlled for age, education, and social grant receipt. To conduct the moderation analyses, we built our models via the following steps: 1) Linked each participant to their corresponding village-level measures based on their residence; 2) Mean-centered all continuous variables; 3) Fit separate adjusted multi-variable logistic regression models with each resilience measure added as a second predictor to ACE-exposure; and 4) Added the interaction term to assess moderation effects between ACE exposure and each individual resilience measure. Following Hayes’ [54] moderation testing approach, probing was planned for interactions that were statistically significant.

Study design and data collection.

In consultation with two local youth partners and a broader Stakeholder Working Group (SWG), we conducted a qualitative study using photovoice, a participatory group-based and visual methods approach, to examine how young women understand and experience resilience in the context of ACEs and how these formative experiences influence their sexual relationships and HIV prevention behaviors. Following a study design workshop and photovoice training with the SWG, we collaboratively developed a recruitment strategy that addressed the sensitive nature of recruiting women with ACEs histories. Women were recruited between June 23 – July 13, 2022. Youth partners and SWG members used flyers and word of mouth to conduct outreach through their professional and personal networks, referring interested individuals to field workers. To protect women’s privacy, field workers contacted potential participants via WhatsApp using discreet language (avoiding mention of research or HIV) to arrange an informational call. During these calls, field workers described the study and conducted screening using a standardized checklist. Participant inclusion included sexually active young women ages 18–25 years living in an AHDSS village who self-reported ≥1 ACE (same questions as quantitative analysis; S1 Table) and were going to remain in the area for the study duration. Those who met the inclusion criteria and expressed willingness to participate were invited to attend a group informational session and informed consent process. Women who chose to participate provided written informed consent in xi-Tsonga or English.

Five weekly study sessions took place between July 13-August 10, 2022. Sessions were facilitated in xi-Tsonga by seasoned field workers, and at times, the Principal Investigator (PI), an American and SA permanent resident, through the use of simultaneous translation. The PI and field workers each have extensive experience working in HIV with young SA women, which facilitated our ability to create a trusting environment conducive to sharing authentic stories and critical reflection. During orientation, we introduced the theme of resilience using culturally-relevant concepts like Ubuntu—an SA term meaning “I am because we are” and signifies communal resilience and interconnected strength—to describe the concept of community-level resilience. We also asked participants to define resilience themselves, with some using idioms, such as “She then decided to ‘wear her big panties’ and started taking treatment” (P1) to describe the internal resilience of a friend after seroconverting [55]. All participants were loaned digital cameras and trained both in the mechanics of using the cameras and how to take photos safely, ethically, and with consent/assent, including obtaining signed media release forms before taking identifying images of any individuals (S2 and S3 File) [56]. To address potential social harm, a trusted social worker was identified by our youth partner for anyone wanting a referral.

Together with participants, we generated three photo assignments that aligned with the SEM approach used in the quantitative analysis. Weekly topics explored: 1) Impact of ACEs on psychological resilience and HIV prevention behaviors; 2) Interpersonal resilience (e.g., social support) and sexual relationships; and 3) How government facilitates young women’s resilience (through provision of social resources). Each week, participants took and shared a photo, followed by a group discussion using the SHOWED method – a structured series of sequential questions that utilizes participant-generated photos to facilitate critical reflection from a surface level understanding of the image to a societal root-cause analysis [57]. See S4 File for an example facilitation guide with tailored SHOWED questions. A final synthesis session in August and separate coding workshop with the participants in October 2022 helped guide the iterative analysis phase. Sessions were recorded with participants’ permission, and simultaneously translated and transcribed into English weekly to facilitate member checking at each subsequent session. Additionally, a locally trained third-party researcher fluent in xi-Tsonga and English checked the transcripts against the original audio recordings for accuracy and quality.

Qualitative data analysis.

We used thematic analysis and composite sequence analysis approaches to facilitate an understanding of how ACEs influence women’s life course development [58]. Steps in the analysis process included: 1) Preliminary read of all transcripts, conducted quality and member checks; 2) Developed codebook guided by photo assignments and thematic groupings from participants’ coding workshop; 3) Analyzed photos and transcripts, including the PI and a local researcher double-coding and developing summary and analytic memos in Dedoose, a web-based research software; 4) Discussed emerging thematic patterns and discrepancies, refined the codebook and finalized coding; 5) Developed code report summaries and matrices to further identify patterns and interpret findings. Beyond member checking, triangulation was achieved during a community forum and dissemination event in November 2023 that included participants, members of the site’s Community Advisory Board and the project’s SWG, Agincourt leadership, and local partners. With support from the PI and facilitators, participants presented the findings—sharing displayed photos accompanied by illustrative quotes. As such, we confirmed the results through a robust discussion between the study participants, community stakeholders and research staff attendees.

ACEs and their effects instilled resilience at an early age.

Participants described the myriad material, psychological, and health effects of ACEs on their life trajectories. One salient theme highlighted the ripple effects of maternal death, which often left participants food insecure, lacking stable housing, and vulnerable to physical and emotional abuse and neglect by extended family and alleged caregivers. These struggles led girls to take on the primary caregiver role for their younger siblings, as they sought to protect them from the same harmful exposures. The photo in Fig 2 was taken to capture the weight of the daily realities due to neglect and orphanhood:

Download:

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

Fig 2. “My struggle”.

https://doi.org/10.1371/journal.pone.0341216.g002

“Life was hard for me because the situation I was in forced me to mature faster than my actual age. I had to take care of my siblings at the age of eight, I was a child myself but a mother to my two siblings. My father also did not care about us, it’s like he almost forgot that we existed…I’m showing that young child who is carrying wood on her head to demonstrate my experience, the wood represents the burden that I was carrying. I was forced to carry huge responsibilities that were beyond my abilities at a young age.” (P6)

For the women who lacked parental support and protection, resilience was frequently associated with self-reliance:

Most of the time I would be alone taking care of myself while my mother was away [working in another city]…I was forced to learn to cook for myself at a young age…I had challenges both at home and at school, I had to face these tough challenges at an early age but because I knew what I wanted, I became resilient. I learned to take care of myself and I did not give up on going to school regardless of the challenges I was facing.” (P1)

Beyond stories of overcoming obstacles to meet their basic survival needs, participants also spoke of the psychological toll of ACEs. Women described the emotional fallout, including anger, depression, and suicidal thoughts, caused by their trauma. For P6, who faced a series of spiraling abuses following the death of her mother, she described it as:

“this anger is like an animal that lives inside you, an animal that needs to be constantly fed and when it’s hungry, you go crazy.” (P6)

Others described how ACEs compelled them to become withdrawn and resist forming new attachments:

“I grew up a bubbly child, I loved meeting new people and socializing but now I’m no longer open and welcoming to people…I probably did not deal with my trauma because I no longer have friends and I have become more of a loner.” (P1)

Of note, some participants expressed how emotional abuse and witnessing abuse inflicted longer lasting pain and deeper wounds to heal than physical abuse:

“Watching your mom getting abused by the man that claims to love her can affect you in a negative way. From my experience, emotional pain takes [more] time to heal compared to physical pain.” (P2)

In contrast to stories of household abuse and neglect, a few participants credited their parents’ positive parenting for teaching them how to cope and adapt in the face of life adversities. Fig 3 represents one participant’s appreciation for her own mother’s perseverance by depicting her mother gardening in what seems to be unfertile dirt. Despite a life of hardship, her mother remains determined to cultivate the land, symbolizing her hope for her daughter’s future. The photographer noted: “this picture demonstrates that my mom sacrificed a lot for me to have the life that I have today.” (P10) Only one woman spoke of being raised in a two-parent household and its positive impact:

Download:

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

Fig 3. “Let’s Grow South Africa Together”.

https://doi.org/10.1371/journal.pone.0341216.g003

“Growing up was difficult but I’m grateful for my parents because they always taught us how to deal with life challenges…they were very open to talk to us about everything…My parents instilled confidence and independence in me. As a result, I can face any difficult situation that I come across with every day…I am resilient because my past experiences made me the stronger person that I am today.” (P8)

As participants reflected on the impact of ACEs on their lives, they reckoned with the complexity of the consequences: “my childhood affected my ability to be resilient in both good and bad ways.” (P1) While their resultant self-reliance and perseverance were highly admired life skills they benefited from, they often came at the cost of their mental well-being, which in turn, had lasting effects on their sexual relationships as they aged into young women.

Resilience helped women avoid and also sustain transactional sex relationships.

Hardship and deprivation were the backdrop for many participants’ narratives about ACEs, resilience, and transactional sex relationships. Longing to escape poverty, some participants described transactional sex as a worthwhile path to what women described as a “soft life” (P1, P2, P5) – a lifestyle beyond mere economic survival or security, but also inclusive of material and economic ease. Yet, this lifestyle often had consequences that included abuse, with one participant declaring that being with “a sugar daddy means accepting to be abused.” (P6) Several participants’ photos captured this idea of enduring intimate partner violence (IPV) within transactional sex relationships, such as Fig 4 which poignantly depicts a woman being beaten by her partner as she does household chores. The participant shared:

Download:

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

Fig 4. Man beating woman while she cleans.

https://doi.org/10.1371/journal.pone.0341216.g004

“I took this picture because it shows some resilience… Even though she is being abused she is not thinking of leaving the house because she thinks that [he] loves her. She doesn’t even know where to go so she rather stays.” (P4)

This example highlights how some women perceived resilience as perseverance to tolerate abuse for the sake of economic survival and optimally a “soft life” of material comfort. Participants used this term commonly; many believing that women in their community would “rather suffer in a mansion with all the nice things [they] want than having peace at the shack” (P6) and “as long as you wake up to warm bacon [all laugh].” (P10)

In contrast, other participants were adamant that resilience entailed taking actions that prioritize bodily (physical and mental) self-preservation and health over material benefits or social status. One participant summed up this view during member checking, noting that accepting abuse reflects a lack of resilience:

What I was saying is that I’m not going to endure the pain of being abused, I think there is no resilience [among women who accept abuse] but, on her [P4 as cited above] statement, resilience is there, as she rather stays than going back home to poverty. She accepts being beaten up at the end of the day because she wants to have access to a good lifestyle… like having access to a cold drink. [However,] I would rather leave than to be beaten up and abused but she accepts to be abused for the sake of living ‘soft life’. (P2, emphasis added)

For P2 and others, they disagreed with the notion that enduring abuse for economic benefit signifies resilience. Instead, they attributed their resilience cultivated during childhood as providing the inner strength and determination to navigate out of unhealthy relationships and break their cycle of abuse, even if it meant sacrificing economic security or desirable lifestyles:

“I don’t want to endure pain for the sake of a relationship. I will end a relationship that threatens my physical or emotional health any day, any time. I have been resilient my whole life.” (P6)

The desire to minimize further trauma was additionally characterized through examples of emotional detachment with sexual partners. Women were keenly aware these relationships served a transactional purpose and accepted them as such. The same participant continued, describing how she cycles through boyfriends, using relationships for gifts and to be able to afford socializing. Even if the men become emotionally attached, she does not:

“I cannot show affection towards my sexual partner. I feel like I am emotionally unavailable for the relationship. I am not mentally invested, and it feels like I’m in that relationship for fun and nothing serious… He honestly tries to show me love but I can’t love him back. It was similar situations with my ex-boyfriends... I lost my mother, so it doesn’t hurt when men leave me.” (P6)

Even among the non-violent relationships, women discussed links between childhood trauma and subsequent mistrust: “I think [trauma] affects us negatively because of the lack of trust…I am also in a healthy relationship, but I do not trust my boyfriend” (P10). Another participant described her willingness to remain in a loveless transactional relationship as enacting resilience, given the financial support it affords her children. She depicted her poor living conditions in Fig 5, an un-staged image of her empty fridge that she entitled “month end” highlighting the eager wait for her partner’s first of the month paycheck to buy groceries for her children:

Download:

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

Fig 5. “Month End”.

https://doi.org/10.1371/journal.pone.0341216.g005

P9:“I also stayed in a relationship with the father of my baby because I needed financial support to take care of the baby. The relationship was not working but I couldn’t break up with him.”

F2:“How does that relate to resilience?”

P9:“I was resilient because I didn’t love him, but I had to stay with him because I needed his money…I had no other choice.”

Conversely, some women adamantly avoided transactional relationships despite the temptations of the economic relief it would afford. For them, resilience manifested through their steadfast resistance to transactional sex and the feared consequences it may incur. For instance, one woman who had suffered multiple ACEs (n = 5) and had to leave school at an early age to care for her siblings while her father worked in another province, relished that:

“What makes me happy is just because I never involved on the situations [i.e.,transactional sex] that make myself to live a soft life while it will end in tears or the things that I should regret at the end.” (P5)

Another participant recounted living next door to her neglectful father who re-married:

“Sometimes I would see my father eating nice food while I did not have food to eat. The sad thing is that my father was not taking care of me…This was a painful experience for me, but I had to be resilient because I did not want to end up doing bad things like sleeping with men so that I can live good life like my father’s children. I can say that I am where I am today because of resilience.” (P4)

As a whole, these examples illustrate the complex, and at times, conflicting perspectives women had on resilience and its role in facilitating outcomes; some taking resilient action to minimize exposure to transactional sex relationships and subsequent HIV risk and related health consequences while others entered and sustained transactional relationships for the more immediate economic benefits.

Lacking social resources following ACEs left women vulnerable to transactional sex.

Another strong theme was the absence of parental and social support growing up—often a direct result of being orphaned, neglected, abused—which combined with insufficient sexual education in schools left women vulnerable to unsafe transactional sex. Several participants commented that “we lacked people who could teach us about the consequences of having sex, like getting HIV or falling pregnant at an early age.” (P2) One participant further described how parents use physical abuse in lieu of informational support and guidance that could help young women better navigate sexual relationships as they transition to young adulthood:

“Our parents failed to tell us the truth. Instead of them telling us the truth, for instance let’s say I slept at my boyfriend’s place and come back home in the morning, my mom would opt to beat me up instead of teaching me about the consequences of my wrong doings. I think that is what has affected us during childhood…Most of us ended up becoming mothers at an early age because of that.” (P3)

Peers often filled these informational voids and encouraged transactional sex by focusing selectively on the benefits and omitting the risks: “our peers will only tell you that it is fine to do it [transactional sex]. They will talk about the exciting part only, but they will not tell you about the consequences of it.” (P6) Blurring the lines between peer pressure and peer support, participants described the practice of “tagging along” where women play match maker to help their friends secure transactional sex partners:

“Let’s say I am in a relationship with a rich guy who does everything for me. My friend on the other side has a boyfriend who does not afford material things…Because we are friends when they ask things from me, I will tell them to ‘tag along’ whenever I go out with my boyfriend so that they can meet his friends who are also rich. In this way, my friends will be pressured to have relationships with my boyfriend’s friend because of the pressure of wanting to be like me. Most girls are in transactional relationships because of pressure from their friends.” (P1)

Importantly, a few participants acknowledged the resilience-promoting role of poverty alleviation and social protection programs for meeting their economic needs. Citing social grants and home-based care specifically, these participants credited these programs with reducing women’s reliance on transactional sex relationships with “blessers”, a colloquial SA term used to refer to older, often married wealthy men who “bless” younger women with gifts, money, and other gains in exchange for sexual favors:

Many households lack basic needs such as food, but the government has tried to address some of these challenges through home-based care. [They] support people with food and some also provide school uniforms for those in need, especially young women so that they do not go around looking for support from blessers.“ (P9)

Yet, despite these social provision measures, most participants felt they were on their own when it came to navigating life challenges, having become accustomed to not having social support during childhood. When asked about how non-governmental organization and village-level community resources help women prevent HIV, one participant replied simply, “I don’t relate with the community [so] it’s difficult to think about this question.” (P2) Thus, while community resources were often perceived as non-existent or inaccessible, transactional sex relationships were readily within reach and available to women.