IntroductionThere are two million HIV‐positive adolescents in southern Africa, and this group has low retention in care and high mortality. There is almost no evidence to identify which healthcare factors can improve adolescent self‐reported retention. This study examines factors associated with retention amongst antiretroviral therapy (ART)‐initiated adolescents in South Africa.MethodsWe collected clinical records and detailed standardized interviews (n = 1059) with all 10‐ to 19 year‐olds ever initiated on ART in all 53 government clinics of a health subdistrict, and community traced to include lost‐to‐follow‐up (90.1% of eligible adolescents interviewed). Associations between full self‐reported retention in care (no past‐year missed appointments and 85% past‐week adherence) and health service factors were tested simultaneously in sequential multivariate regression and marginal effects modelling, controlling for covariates of age, gender, urban/rural location, formal/informal housing, maternal and paternal orphanhood, vertical/horizontal HIV infection, overall health, length of time on ART and type of healthcare facility.ResultsAbout 56% of adolescents had self‐reported retention in care, validated against lower detectable viral load (AOR: 0.63, CI: 0.45 to 0.87, p = 0.005). Independent of covariates, five factors (STACK) were associated with improved retention: clinics Stocked with medication (OR: 3.0, CI: 1.6 to 5.5); staff with Time for adolescents (OR: 2.7, CI: 1.8 to 4.1); adolescents Accompanied to the clinic (OR: 2.3, CI: 1.5 to 3.6); enough Cash to get to clinic safely (OR: 1.4, CI: 1.1 to 1.9); and staff who are Kind (OR: 2.6, CI: 1.8 to 3.6). With none of these factors, 3.3% of adolescents reported retention. With all five factors, 69.5% reported retention.ConclusionsThis study identifies key intervention points for adolescent retention in HIV care. A basic package of clinic and community services has the potential to STACK the odds for health and survival for HIV‐positive adolescents.

Background Little evidence exists to comprehensively estimate adolescent viral suppression after initiation on antiretroviral therapy in sub-Saharan Africa. This study examines adolescent progression along the HIV care cascade to viral suppression for adolescents initiated on antiretroviral therapy in South Africa. Methods All adolescents ever initiated on antiretroviral therapy (n=1080) by 2015 in a health district of the Eastern Cape, South Africa, were interviewed in 2014–2015. Clinical records were extracted from 52 healthcare facilities through January 2018 (including records in multiple facilities). Mortality and loss to follow-up rates were corrected for transfers. Predictors of progression through the HIV care cascade were tested using sequential multivariable logistic regressions. Predicted probabilities for the effects of significant predictors were estimated by sex and mode of infection. Results Corrected mortality and loss to follow-up rates were 3.3 and 16.9%, respectively. Among adolescents with clinical records, 92.3% had ≥1 viral load, but only 51.1% of viral loads were from the past 12 months. Adolescents on ART for ≥2 years (AOR 3.42 [95%CI 2.14–5.47], p< 0.001) and who experienced decentralised care (AOR 1.39 [95%CI 1.06–1.83], p=0.018) were more likely to have a recent viral load. The average effect of decentralised care on recent viral load was greater for female (AOR 2.39 [95%CI 1.29–4.43], p=0.006) and sexually infected adolescents (AOR 3.48 [95%CI 1.04–11.65], p=0.043). Of the total cohort, 47.5% were recorded as fully virally suppressed at most recent test. Only 23.2% were recorded as fully virally suppressed within the past 12 months. Younger adolescents (AOR 1.39 [95%CI 1.06–1.82], p=0.017) and those on ART for ≥2 years (AOR 1.70 [95%CI 1.12–2.58], p=0.013) were more likely to be fully viral suppressed. Conclusions Viral load recording and viral suppression rates remain low for ART-initiated adolescents in South Africa. Improved outcomes for this population require stronger engagement in care and viral load monitoring.