Inferring population HIV incidence trends from surveillance data of recent HIV infection among HIV testing clients

Background: Measuring recent HIV infections from routine surveillance systems could allow timely and granular monitoring of HIV incidence patterns. We evaluated the relationship of two recent infection indicators with alternative denominators to true incidence patterns. Methods: We used a mathematical model of HIV testing behaviours, calibrated to population-based surveys and HIV testing services programme data, to estimate the number of recent infections diagnosed annually from 2010 to 2019 in Cote d'Ivoire, Malawi, and Mozambique. We compared two different denominators to interpret recency data: those at risk of HIV acquisition (HIV-negative tests and recent infections) and all people testing HIV positive. Sex and age-specific longitudinal trends in both interpretations were then compared with modelled trends in HIV incidence, testing efforts and HIV positivity among HIV testing services clients. Results: Over 2010-2019, the annual proportion of the eligible population tested increased in all countries, while positivity decreased. The proportion of recent infections among those at risk of HIV acquisition decreased, similar to declines in HIV incidence among adults (>= 15 years old). Conversely, the proportion of recent infections among HIV-positive tests increased. The female-to-male ratio of the proportion testing recent among those at risk was closer to 1 than the true incidence sex ratio. Conclusion: The proportion of recent infections among those at risk of HIV acquisition is more indicative of HIV incidence than the proportion among HIV-positive tests. However, interpreting the observed patterns as surrogate measures for incidence patterns may still be confounded by different HIV testing rates between population groups or over time.