Condom use correlates among youth living with HIV in South Africa: lessons for promoting safer sex

ABSTRACT Identifying factors associated with condom use can help inform the design of sexual reproductive health interventions for young people living with HIV. Data were collected from 294 sexually active youth in South Africa aged 14–24 years, living with HIV and aware of their status. Logistic regression found condom use was associated with partner status, disclosure, treatment adherence, knowledge, alcohol use, enacted stigma, and age at known status. Seventy-seven percent reported condom use at last sex, with higher odds among those who: disclosed their status to their last partner (OR = 2.13, 95% CI = 1.09–4.15), had not missed any ART doses in the previous week (OR = 2.67, 95% CI = 1.35–5.27), knew to use condoms when both partners have HIV (OR = 2.65, 95% CI = 1.34–5.25) and learned their status by age 12 (OR = 2.89, 95% CI = 1.30–6.43). Lower odds of condom use were among participants with an HIV-positive partner (OR = 0.36, 95% CI = 0.18–0.73), those who experienced recent stigma (OR = 0.55, 95% CI = 0.31–1.00), who drank at least 1 day per month (OR = 0.43, 95% CI = 0.22–0.84) and females (OR = 0.41, 95% CI = 0.20–0.84). Condom education and services promoting stigma management and safe disclosure could increase condom use among young people living with HIV. Lower condom use among adolescents reporting non-adherence to treatment underscores the critical need for effective interventions.


Introduction
An estimated 7.9 million people in South Africa are living with HIV − 14% of the country's population and more than any other country in the world (Simbayi et al., 2019).Approximately 800,000 of these cases are among 15 to 24-year-olds (Zanoni et al., 2016).South Africa reflects broader trends in sub-Saharan Africa (SSA), where incidence is high among adolescents and youth, especially girls and young women (Joint United Nations Programme on HIV/AIDS, 2020; Simbayi et al., 2019).Expanded access to early infant diagnosis services and antiretroviral therapy have also resulted in an increase in the number of children with perinatally acquired HIV who survive to adolescence and become sexually active (Ferrand et al., 2009).Only 59% of all sexually active young people in South Africa report using a condom at last sex and 45% report consistent condom use (Simbayi et al., 2019).Recent systematic reviews highlight higher rates of unprotected sex among youth living with HIV (YLHIV) in SSA (Toska et al., 2017;Zgambo et al., 2018).YLHIV who engage in unprotected sex are vulnerable to unplanned pregnancy and sexually transmitted infections.In the era of Undetectable = Untransmittable (Cohen et al., 2011;Eisinger et al., 2019;Rodger et al., 2016), unprotected sex in the absence of viral suppression also increases risk for HIV transmission between sexual partners, including HIV superinfection (Redd et al., 2013).Sexual risk reduction among YLHIV is therefore a critical component of epidemic response strategies.
Recent systematic reviews highlight the paucity of evidence on factors associated with unprotected sex among YLHIV (Ssewanyana et al., 2018;Toska et al., 2017;Zgambo et al., 2018).Fifteen to twenty-four-year-olds have lower rates of treatment retention and viral suppression than adults (Simbayi et al., 2019;Zanoni et al., 2016), yet only a few studies in SSA have investigated the potential link between HIV-related factors, such as antiretroviral therapy, and unprotected sex among YLHIV.Disclosure to sexual partners is generally low among YLHIV, and mixed findings have been reported regarding its association with safer sex, particularly in relation to condom use (Thoth et al., 2014).Two recent studies among adolescents living with HIV in South Africa found disclosing one's HIV status to a partner was not associated with condom use (Kidman & Violari, 2020;Toska et al., 2015).Disclosure is complicated by HIV-related stigma (Smith et al., 2008); stigma compounds health risk behaviors among YLHIV generally (Zgambo et al., 2018) and is understudied (Ssewanyana et al., 2018).Furthermore, studies conducted with YLHIV often do not account for mode of transmission or whether participants know their HIV status, which limits understanding about sexual risk behaviors and potential strategies for intervention (Ssewanyana et al., 2018).This analysis responds to that gap by identifying factors associated with condom use among sexually active YLHIV to inform the development of effective risk reduction interventions in SSA.It uses baseline data from a larger cohort study designed to evaluate a structured support group intervention and other services for young people living with HIV in South Africa.The findings may be used to optimize sexual risk reduction strategies for this and similar populations.

Procedures and participants
Study participants were enrollees in a curriculum-based socio-educational support group for YLHIV called Vhutshilo 3, which was implemented by community-based organizations serving orphans and vulnerable children and youth (OVCY) in urban communities of Gauteng and KwaZulu Natal provinces and funded by the U.S. President's Emergency Plan for AIDS Relief (PEPFAR).Vhutshilo 3 participants were identified and recruited by care workers from local community-based organizations from clinics serving HIVpositive patients as well as YLHIV enrolled in their case management services.
Participation in Vhutshilo 3 groups was voluntary with no incentives provided.Group facilitators at selected sites informed participants aged 14-24 years about the opportunity to take part in the cohort study and referred them to survey staff for the informed consent process.Data collection for this analysis began in June 2019 and ended in March 2020, when group sessions were paused due to COVID-19 precautions.The Social Behavioral Institutional Review Board at Tulane University (study reference no: 2018-1736) and Pharma Ethics Independent Research Ethics Committee in South Africa (study reference no: 181021579) approved this study and authorized eligible children under age 18 to consent to their own participation.
Study participants completed the survey following the second or third support group session held within 1 month of the first session, using paper questionnaires written in English (the language used throughout the support group).Trained staff from TK Research Consulting, a South Africa-based research organization unaffiliated with the program implementers, administered the survey.Each question was read aloud by survey staff, while participants marked their responses on individual copies of the questionnaire.Survey staff clarified the meaning of questions upon request using a reference copy of the questionnaire locally translated into isiZulu and Setswana.To ensure confidentiality and minimize social desirability bias, coded identifiers were used in place of names on the questionnaire, and each completed questionnaire was sealed in an individual envelope by the participant before returning it to survey staff.Participants received refreshments and a small monetary incentive (ZAR 20, approximately US$1.20) upon completing the survey.

Measures
Participants' age at the time of the survey was calculated using the date of birth recorded during registration for the OVCY program; gender was also obtained from program registration records.Age was dichotomized as 0-17 and 18+ to simplify the interpretation of results and reflect a threshold at which many young people in South Africa leave school and begin to exercise greater independence.
Respondents were asked if they had sexual intercourse in the previous 6 months and whether they used a condom the last time they had intercourse.They were also asked whether they knew their most recent partner's HIV status, what that status was, and whether their most recent partner knew the respondent's status at the time.Endorsement of sero-concordant condom use was assessed by asking, 'If both partners are HIVpositive, do they need to use a condom during sexual intercourse?'.
Adherence was assessed by asking, 'In the last 7 days, how many days did you miss taking one of more of your ARV pills?' with space to write in the number or select 'I'm not taking ARVs'.Local program staff suggested using the term ARV (antiretrovirals), indicating it was commonly used by adolescents to describe antiretroviral therapy.The time frame was chosen based on research conducted with pediatric populations which found that a seven-day recall demonstrated superior reliability (Dolezal et al., 2003) and was significantly associated with viral load test results (Usitalo et al., 2014).Cluver et al. (2015) also reported on the adaptation in South Africa of a similar measure for studying adherence among adolescents.Participants were classified as adherent if they did not miss any dose in the past 7 days.Those who indicated they had never taken ARVs or had stopped taking it more than 30 days prior were likewise classified as non-adherent in accordance with South African (National Department of Health, 2019) and WHO recommendations (World Health Organization [WHO], 2017) for universal initiation within 1 week of HIV diagnosis.
Other measures used in the analysis included alcohol use, enacted stigma, and the age at which YLHIV learned their HIV-positive status.The age at which participants learned their status was categorized dichotomously into 0-12 and 13+ years, in accordance with previous studies and guidelines for disclosure to perinatally infected adolescents (Cluver et al., 2015;WHO, 2011).Respondents were considered to use alcohol if they reported drinking 'about once or twice a month' or more often in the previous 6 months.Participants who responded affirmatively to, 'In the last 6 months, has anyone treated you badly because of your HIV?' were considered to have experienced recent enacted stigma.

Statistical methods
Of 1337 individuals who completed the baseline survey and knew their HIV status, 315 reported having sexual intercourse in the past 6 months, but 21 of those were excluded from the analysis due to missing data on the outcome or independent variables.This resulted in an analytic sample of 294 sexually active YLHIV.Except for age, the independent variables in the model were limited to those with statistical significance to conserve degrees of freedom with this limited sample.The frequencies were generated using SAS version 9.4.Logistic regression models were estimated using Stata IC version 14 with the Logit command.Standard errors were corrected for clustering of respondents in 93 Vhutshilo 3 groups using the cluster option.
Condom use at last intercourse was regressed on nine variables representing age greater than 17 years, female gender, HIV status known before age 13, respondent disclosed their HIV status to their last partner, last sexual partner was HIV positive, aware that condoms should be used when both partners have HIV, ARV adherent in last week, recent experience of HIV-related stigma, and consumed alcohol at least monthly in the previous 6 months.

Results
Of the 294 respondents in the analytical sample, 58% were female and 50% were under age 18; respondents had a mean age of 17.8 years (see Table 1).Almost two-thirds of the respondents learned they had HIV after the age of 12 (65%).Slightly less than half were currently taking ARVs and had not missed a dose in the previous 7 days (48%).Fortyone percent consumed alcohol at least once per month in the preceding 6 months, and 29% experienced enacted stigma during this period.About three quarters (77%) of subjects used a condom the last time they had sexual intercourse and 59% disclosed their status to their most recent sexual partner.
Table 2 shows the results of the logistic regression model.While age was not associated with condom use at last sex, respondents who learned their HIV status before age 13 had almost triple the odds of using a condom (OR = 2.89, 95% CI = 1.30-6.43).Adhering to an ARV regimen was associated with greater than two-fold higher odds of using a condom (OR = 2.67, 95% CI = 1.35-5.27).Awareness that condoms are necessary when both partners are HIV-positive was associated with more than double the odds of using a condom (OR = 2.65, 95% CI = 1.34-5.25).Respondents who disclosed their HIV status to their partner had double the odds of using a condom (OR = 2.13, 95% CI = 1.09-4.15).
Females had 59% lower odds than males of using a condom at last intercourse (OR = 0.41, 95% CI = 0.20-0.84).Drinking alcohol at least 1 day per month was associated with 57% lower odds of condom use (OR = 0.43, 95% CI = 0.22-0.84).One-fifth of respondents had recently engaged in sex with an HIV-positive partner, and the odds of condom use were 64% lower for these encounters (OR = 0.36, 95% CI = 0.18-0.73).Recent enacted stigma related to one's HIV status was also associated with lower odds of condom use (OR = 0.55, 95% CI = 0.31-1.00).

Discussion
This study contributes to the limited evidence base for improved risk reduction programming among YLHIV in South Africa by identifying factors associated with condom use among participants enrolled in a structured socio-educational intervention.Less than one quarter of respondents reported being sexually active in the 6-months prior to the survey, with the majority reporting condom use at last sex (77%).The results support prior findings with respect to lower condom use among female adolescents and those who use alcohol (Gavin et al., 2006;Velloza et al., 2020).This study also presents new evidence specific to YLHIV: those who had experienced recent enacted stigma related to their HIV status and those in seroconcordant partnerships were less likely to use condoms, but endorsement of condom use for seroconcordant partnerships was positively associated with condom use.This study also found that condom use was higher among YLHIV who adhered to treatment, those who disclosed their status to a sexual partner, and those who learned they had HIV at younger ages.The results offer insights into how behavioral support interventions for YLHIV may be adapted to improve individual health outcomes and epidemic control.
The prevalence of condom use among sexually active YLHIV in this study (77%) is higher than in previous studies of similar populations from sub-Saharan Africa, which ranged from 18% to 71% (Toska et al., 2017).Earlier studies, however, generally included some participants who were unaware of their HIV positive status, a characteristic that has been associated with riskier sex practices (Obare & Birungi, 2010;Toska et al., 2015).The current study was also conducted at baseline among participants who self-selected to enroll in an HIV support group, and such engagement has been previously linked with reduced sexual risk taking (Toska et al., 2017) and may suggest better health seeking behaviors among this sample.
While age was not associated with condom use at last sex, YLHIV who learned their HIV status before age 13 were much more likely to have used a condom.Learning one's HIV status prior to age 13 is likely to be closely associated with perinatal acquisition of HIV.Although respondents were asked if they were infected perinatally, almost a quarter (23%) answered that they did not know.The variable also reflects knowledge of HIV status prior to sexual maturity, a factor which may influence future behaviors.Sherr et al. (2018) found that behaviorally infected adolescents living with HIV demonstrated greater mental health problems and substance abuse than perinatally-infected adolescents.Thus, lower condom use among those who acquire HIV behaviorally may reflect a continuation of risk behaviors that predate HIV infection.Differences in the behavioral profiles among young people living with HIV, based on the mode of infection, have important implications for intervention design and effectiveness (Sherr et al., 2018).
Self-reported antiretroviral therapy adherence among the sample of sexually active young people in this study was poor, with slightly more than half of participants missing at least one dose in the previous week.Self-reports of missing at least one dose in the last 7 days has been associated with viral loads more than 400 copies/ml (Usitalo et al., 2014), increasing the likelihood for onward transmission.Using the same measure, higher adherence has been reported among adolescents elsewhere in South Africa (Cluver et al., 2015); although past week non-adherence rates among YLHIV have been found to range from 6% to 96% (Ammon et al., 2018).This study found that ART adherence was positively associated with condom use, but other studies have found inconsistent associations (Bajunirwe et al., 2013;Bunnell et al., 2006;Diamond et al., 2005;Durham et al., 2018;Olley, 2008) which may be due to undetectable viral loads and/or a desire to conceive.Given the significance of viral load for both clinical outcomes and transmission risk (Fatti et al., 2014;Zanoni et al., 2016), lower condom use among those with poor adherence has serious implications for both individual health and epidemic control.
This study also offers new evidence about how seroconcordant partnerships may impact sexual behavior, an understudied topic among YLHIV (Gabbidon et al., 2020).
In these results, participants whose partners were also living with HIV were much less likely to use condoms.However, simply affirming the need to use a condom in seroconcordant relationships was associated with much higher odds of condom use, underscoring the potential value of expanding and evaluating sexual reproductive health interventions tailored to YLHIV (Toska et al., 2017).Intervention development research for PLHIV in South Africa identified similar barriers to condom use, including misperceptions about condom effectiveness and the need for seroconcordant couples to use them (Cornman et al., 2011).While it is acknowledged that efforts to promote condom use in SSA must move beyond a focus on women (Leddy et al., 2016), recent multicountry research illustrated that comprehensive knowledge on HIV & AIDS among female adolescents resulted in safer sex negotiation, suggesting that knowledge can reduce risk-taking (Frimpong et al., 2021).
More than half of the respondents in this analysis had disclosed their HIV status to their last partner, which accords with the 31%-74% disclosure rates reported in a recent systematic review (Toska et al., 2017).Disclosure of one's status to the most recent sexual partner was associated with increased odds of condom use at last sex in this study.While this finding is consistent with one study conducted among adults on antiretroviral therapy in Free State, South Africa (Booysen et al., 2017), it conflicts with null findings from two other studies among YLHIV in other areas of South Africa (Kidman & Violari, 2020;Toska et al., 2015), suggesting the need for more evidence about the potential role of disclosure in promoting safer sex.Geographic location has been found to influence disclosure patterns, as have other situational factors not accounted for in this analysis including risk of violence, social support and mental health (Mugo et al., 2021;Zgambo et al., 2018).Importantly, while protective for condom use in this study, HIV disclosure has also been associated with experiences of enacted stigma among people living with HIV in SSA (Hargreaves et al., 2018).Qualitative research from South Africa reported that people living with HIV in South Africa failed to use condoms due to the fear of experiencing stigma by disclosing their HIV status (Cloete et al., 2010).Correspondingly, the present study found that enacted stigma related to one's HIV status was associated with lower condom use.Thus, disclosure to sexual partners may not always mitigate sexual risk taking and could introduce other harms.Providers should be cautious about advocating for disclosure and consider an individual's unique circumstances and available support (Kidman & Violari, 2020;Toska et al., 2015).
This study has important limitations.A convenience sample of study participants recruited in an intervention setting may be more in need or more motivated to adopt healthy behaviors than non-participating YLHIV.These results should, therefore, be generalizable to other young people in South Africa engaged in similar interventions and aware of their HIV status, but not to all YLHIV.Also, these data were collected in an urban population which may differ in some important ways from rural populations (Thior et al., 2020).The cross-sectional study design precludes causal inference, but nonetheless identifies knowledge and skills to be incorporated in YLHIV support programming to improve condom use and other risk reduction strategies.While condom use at last sex is one of the most common measures applied in HIV research in SSA, additional markers reflecting consistent condom use would provide a more nuanced understanding of HIV risk (Fonner et al., 2014).The measures used here were self-reported; however, efforts were made to minimize social desirability bias (Kelly et al., 2013;Krumpal, 2013;Langhaug et al., 2010).Future research among YLHIV should incorporate additional indicators of sexual risk behavior not included in this relatively brief survey, such as the number of recent partnerships, age disparity, concurrency, as well as fertility intentions -which may affect decision-making about condom use.
Improved understanding of condom use behaviors among YLHIV is critical to the design of interventions supporting individuals' wellbeing and epidemic control.Given that 85% of all YLHIV reside in SSA (United Nations Children's Fund, 2023), attention to the culturally specific aspects of effective risk reduction is paramount.Findings from this study are directly applicable to the HIV response in South Africa.In addition to adherence support, organizations serving YLHIV should seek to provide comprehensive sexual and reproductive health interventions that include knowledge and skill-building components pertaining to condom use and psychosocial and disclosure support.Peer-led support groups for YLHIV in Uganda focused on providing health education and linkages to HIV and sexual and reproductive health services showed promise in reducing risk behaviors, including condomless sex (Vu et al., 2017).Greater expansion of these approaches targeting YLHIV is needed.
Tory M. Taylor, MPH, is the Technical Director for the USAID-supported Data for Impact (D4I) and Global Health Program Evaluation, Analysis, Research, and Learning (GH PEARL) projects.She is proud to have been part of Tulane University's Highly Vulnerable Children Research Center for many years previously.Her areas of technical expertise include survey research, program evaluation, and organizational capacity strengthening.Brian Luckett, MPH, PhD.With a MSPH in Biostatistics and a PhD in International Health and Development, Dr. Luckett's work has primarily focused on social and behavioral determinants of health.He has a background in research methods, epidemiology, economics and statistics and over 30 years of experience doing population-based research in south Louisiana and sub-Sharan Africa.Alexandra Spyrelis, MA, is a Senior Monitoring and Evaluation Specialist at DNA Economics.Her expertise lies in research, programme design, and monitoring and evaluation within the public health and social development sectors in South Africa.She is currently pursuing her PhD at Stellenbosch University.Johanna Nice, MPH, PhD, serves as an Assistant Professor in the Department of International Health and Development at Tulane University School of Public Health and Tropical Medicine.Her background is in designing monitoring and reporting systems for community-based initiatives and optimizing health programs and policies to benefit underserved communities.Her current research focuses on improving treatment adherence among adolescents and young adults living with HIV.

Table 1 .
Sample characteristics for all participants and by condom use at last intercourse

Table 2 .
Odds ratios for factors associated with condom use at last intercourse