Sex in the shadow of HIV: factors associated with sexual risk among adolescents in a community-traced sample in South Africa

Background: Sub-Saharan Africa is home to 85% of the worldâs HIV-positive adolescents: an estimated 1.3-2.2 million 10-19 year olds. Adolescents living with HIV face multiple sexual and reproductive health risks: unwanted pregnancies and the risk of mother-to-child-transmission, risk of infecting partners, co-infection with other STIs, and the rising but undocumented risk of re-infection by potentially resistant HI-virus strains. Using contraception, especially condoms, is particularly challenging for all adolescents. It is even more difficult for HIV-positive adolescents due to HIV-related factors such as learning their HIV-positive status, withholding or disclosing their HIV-status to sexual partners, and accessing services in the home, clinics, and schools. This thesis aims to understand which factors shape sexual risk-taking among HIV-positive adolescents to inform the development of interventions that promote safe sexual practices in this population. Methodology: This thesis applies a socio-ecological model to investigate factors associated with sexual risk-taking among HIV-positive adolescents. It consists of three stand-alone papers: a systematic review and two quantitative papers based on a cross-sectional epidemiological and aetiological study of unprotected sex among HIV-positive adolescents and community controls in South Africa. Paper 1 is a systematic review of rates, correlates, and interventions to reduce sexual risk-taking among HIV-positive adolescents and youth in sub-Saharan Africa. Paper 2 looks at associations between HIV-status knowledge and disclosure and protective sexual practices in the cross-sectional study sample. Paper 3 explores the relationship between various social protection provisions and unprotected sex among HIV-positive adolescents. The candidate co-developed and conducted a community-traced study of adolescents in the Eastern Cape, in South Africa: 1,060 HIV-positive adolescents and 467 community controls. HIV-positive 10-19 year old adolescents were recruited from 53 government facilities in a health sub-district with antenatal HIV prevalence of over 30%. 90.1% of the eligible sample was traced, with only 4.1% refusing to take part. Community controls were neighbouring or co-habiting 10-19 year old adolescents, 92% of whom agreed to take part. Voluntary informed consent was obtained from adolescents and caregivers in the language of their choice: English or Xhosa. Questionnaires were administered by trained research assistants using mobile devices (tablets) with adolescent-friendly graphic content to ensure participant interest and reduce participant burden through skip-patterns. The systematic review (Paper 1) included studies located through electronic databases and grey literature. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines were followed. Quantitative studies reporting on HIV-positive participants (10-24 year old) included data on at least one of eight outcomes (early sexual debut, inconsistent condom use, older sexual partner, transactional sex, multiple sexual partners, sex while intoxicated, sexually transmitted infections, and pregnancy). Only studies conducted in sub-Saharan Africa were included. The candidate and a second author independently piloted all processes, screened studies, extracted data independently, and resolved any discrepancies. Due to variance in reported rates and correlates, no meta-analyses was conducted. The systematic review informed the analyses conducted for the two quantitative papers. Analyses for Papers 2 and 3 used condom use at last sexual encounter (dichotomised either as safe sex/ abstinence or unprotected sex) as the outcome, controlling for a series of covariates. Analyses used SPSS 22 and STATA 11. For each paper, the hypothesised factors were entered as independent variables in multivariate logistic regressions controlling for potential confounders. Based on the findings of the systematic review, gender moderation analyses was run entering a 2-way interaction term of gender*correlate in multivariate logistic regressions, controlling for covariates. Marginal effect models explored the effect of combinations of risk/ protective factors. Predicted probabilities for safe sex/ unprotected sex were computed for different two- and three-way combinations of the independent variables, controlling for covariates significantly associated with the outcome. Paper 2 tested the effect of three types of disclosure on protective sexual practices: (i) knowledge of oneâs own HIV-positive status, (ii) disclosing oneâs HIV-status to a partner, and (iii) knowing a partnerâs HIV-status. It compared HIV-positive status aware adolescents (n=794) with the rest of the sample (n=733). Paper 3 investigated associations between nine types of social protection provisions and unprotected sex. In line with UNICEFâs definition, social protection was defined as any provision aimed at preventing, reducing and eliminating economic and social vulnerabilities to poverty and deprivation among HIV-positive adolescents. The nine social protection provisions tested by the analyses included âcashâ and âcareâ factors accessed in the home, school, and community. Results: Paper 1 â âSexual Risk-Taking among HIV-Positive Adolescents and Youth in Sub-Saharan Africa: A systematic review of prevalence rates, risk factors, and interventions.â The systematic review (Chapter 4) found that, despite their heightened vulnerabilities and high rates of sexual risk-taking, there is a dearth of evidence on interventions which may help HIV-positive adolescents engage in safe sexual practices. The review included 35 studies, four of which were interventions aiming to reduce sexual risk-taking. The quality of the included studies was low with most studies (k=31) reporting findings from cross-sectional data. HIV-positive adolescents and youth reported high rates of sexual risk-taking, however findings were inconsistent about potential factors associated with sexual risk-taking. Factors consistently associated with sexual risk-taking in multivariate analyses included: food insecurity, living alone, living with a partner, and gender-based violence. No significant associations were reported for: rural residence, informal housing, anxiety, religious guidance, STI prevention knowledge, poor birth outcomes, orphanhood, parental monitoring, having a supportive family, social support, maternal education level, poverty, disclosing oneâs HIV-status to a partner, time on ART, ART adherence, receiving care at a hospital, opportunistic infections. However, most of the above associations were reported by only one study, therefore further analyses is needed to build the evidence base on potential determinants of sexual risk-taking among HIV-positive adolescents and youth. The included interventions consist of three individual- and group-based psychosocial interventions evaluated in three small-scale trials (n Quantitative data analyses of cross-sectional study data: Overall, adolescents in the full sample (n=1,527) reported high rates of sexual activity (34.9%) and high rates of unprotected sex (22%), with adolescent girls reporting higher rates of unprotected sex than boys (33% vs. 7%). In univariate ANOVA, rates of unprotected sex among those sexually-active were lower amongst ART-eligible HIV-positive adolescents (p±.001) compared to community controls, adolescent boys compared to girls (p±.001), and HIV-positive adolescents who knew their status compared to the rest of the sample (p±.001). Among sexually-active HIV-positive adolescents who knew their status, horizontally-infected adolescents reported more unprotected sex (64% vs. 44%, p±.001). Paper 2 âIs Knowledge Power? HIV-status disclosure and protective sexual practices among adolescent girls and boys in South Africa.â In Paper 2 (Chapter 6), analyses of the full sample of adolescents (n=1,527) found that knowing oneâs own HIV-positive status was associated with more safe sex (OR1.88, CI1.35-2.63, p±0.001), knowledge of a sexual/ romantic partnerâs HIV-status (positive or negative) was associated with less safe sex (OR0.45 CI0.29-0.69, p±0.001), while disclosing oneâs HIV-status to a partner was not associated with safe sex. Gender did not moderate the relationship between HIV-status knowledge and disclosure, and protective sexual practices. Among female adolescents (N=843), the patterns of association between disclosure (to and from) and safe sex were the same: HIV-positive status-aware adolescent girls were more likely to report safe sex than the rest of adolescent girls (OR1.86 CI1.26-2.74, p=0.002). When combinations of disclosure were considered, knowing oneâs own HIV-positive status was associated with higher rates of protective sexual practices, but it was not sufficiently protective when HIV-positive adolescents also disclosed to their partners or knew their partnerâs HIV-status. Paper 3 âSchool, Supervision, and Adolescent-Sensitive Clinic Care: Combination social protection and reduced unprotected sex among HIV-positive adolescents in South Africa.â Analyses in Paper 3 (Chapter 7) tested associations between social protection provisions accessed by HIV-positive adolescents in the home, school and clinics (n=1,060). Participants reported lower rates of unprotected sex if they received strong parental supervision (OR0.54, CI0.33-0.90, p=0.019), had free access to school (OR0.52, CI0.33-0.82, p=0.005), and received adolescent-sensitive care when accessing sexual health services (OR0.43, CI0.25-0.73, p=0.002). Gender acted as a moderator for the positive effect of adolescent-sensitive clinic care: the association between reduced reports of unprotected sex and better clinic care was stronger for girls than boys. Gender-disaggregated regression models confirmed these findings. Analyses then checked for multiplicative or additive effects: whether receiving more than one social protection provision had a greater effect on rates of unprotected sex. No multiplicative effects were found. In marginal effect models, receiving more than one social protection intervention had an additive effect on predicted probabilities of unprotected sex rates, controlling for covariates. This effect was even stronger among adolescent girls: without any social protection, 49% were likely to report unprotected sex; with 1-2 of the interventions 13-38%; and with all interventions only 9%. Conclusion: This thesis provides new evidence on what may support HIV-positive adolescents engage in safe sexual practices. The findings suggest that some factors may strongly support safe sex: knowledge of own HIV-positive status, good caregiver supervision, access to schools, and adolescent-sensitive sexual healthcare at clinics. These factors are located in various levels of the socio-ecological model, supporting the application of this theoretical model to understand the sexual health of HIV-positive adolescents in South Africa. These findings also highlight the greater vulnerability of adolescent girls, even when they are already HIV-positive, and propose potential interventions that may have positive effects in reducing these vulnerabilities: supported disclosure of their HIV-positive status, access to schools, good parental supervision, and adolescent-sensitive clinic care. By identifying key HIV-specific and HIV-inclusive programmes that may support safe sex, this thesis contributes to the problem theory for the sexual and reproductive health of HIV-positive adolescents.