Could there have been substantial declines in sexual risk behavior across sub-Saharan Africa in the mid-1990s?

• We examined the hypothesis of behavioral change driving HIV declines in Africa.
• We demonstrated the self-consistency and plausibility of this hypothesis.
• The average decline in the scale of sexual risk behavior across countries was 70%.
• The declines in risk behavior occurred over several years in mid to late 1990s.

BackgroundHIV prevalence is decreasing in much of sub-Saharan Africa (SSA), but the drivers of the decline are subject to much dispute. Using mathematical modeling as a tool for hypothesis generation, we demonstrate how the hypothesis that the drop in prevalence reflects declines in sexual risk behavior is self-consistent. We characterize these potential declines in terms of their scale, duration, and timing, and theorize on how small changes in sexual behavior at the individual-level could have driven large declines in HIV prevalence.Materials and methodsA population-level deterministic compartmental model was constructed to describe the HIV epidemics in 24 countries in SSA with sufficient trend data. The model was parameterized by national HIV prevalence and HIV natural history and transmission data. The temporal evolution of sexual risk behavior was characterized using established tools and uncertainty and sensitivity analyses on the results were conducted.ResultsDeclines in the scale of sexual risk behavior between 31.8% (Botswana) and 89.3% (Liberia) can explain the declining HIV prevalence across countries. The average decline across countries was 68.9%. The transition in sexual risk behavior lasted between 2.7 (Botswana) and 16.6 (Gabon) years with an average of 8.2 years. The turning point year of the transition occurred between 1993 (Burundi) and 2001 (Namibia), but clustered around 1995 for most countries. The uncertainty and sensitivity analyses affirmed our model predictions.ConclusionThe hypothesis that HIV prevalence declines in SSA have been driven by declines in sexual risk behavior is self-consistent and provides a convincing narrative for an evolving HIV epidemiology in this region. The hypothesized declines must have been remarkable in their intensity, rapidity, and synchronicity to explain the temporal trends in HIV prevalence. These findings provide contextual support for the hypothesis that changes in sexual behavior that materialized in the 1990s are a dominant driver of the recent decreases in HIV prevalence.