Abstract
BACKGROUND: Adolescent girls and young women (AGYW) aged 15-24 years are more likely to acquire HIV than their male counterparts, and well-targeted prevention interventions are needed. We developed a method to quantify the risk of HIV acquisition based on individual risk factors and population viral load (PVL) to improve targeting of prevention interventions. SETTING: This study is based on household health survey data collected in 13 sub-Saharan African countries, 2015-2019. METHODS: We developed a Bayesian spatial model which jointly estimates district-level PVL and the probability of infection among individual AGYW, aged 15-24 years, based on individual behavioral/demographic risk factors and area-level PVL. The districts (second subnational level) typically comprise the areas of estimation. The model borrows strength across countries by incorporating random effects, which quantify country-level differences in HIV prevalence among AGYW. RESULTS: The combined survey data provided 52,171 questionnaire responses and blood tests from AGYW, and 280,323 blood samples from all respondents from which PVL was estimated. PVL was-by far-the most important predictor of test positivity [adjusted odds ratio (aOR) = 70.6; 0.95-probability credible interval 20.7-240.5]. Having a partner with HIV increased the odds of testing positive among AGYW who were never (aOR = 12.1; 7.5-19.6) and ever pregnant (aOR = 32.1; 23.7-43.4). The area under the cross-validated receiver-operating characteristic curve for classification of test positivity was 82%. CONCLUSION: The fitted model provides a statistically principled basis for priority enrollment in HIV prevention interventions of those AGYW most at risk of HIV infection and geographic placement of prevention services.