Building better HIV models: a framework for incorporating evidence on structural determinants and interventions to estimate their impacts on HIV epidemics
Authors
James Stannah, Jorge Luis Flores Anato, Michael Pickles, Joseph Larmarange, Kate M. Mitchell, Adelina Artenie, Kostyantyn Dumchev, Serge Niangoran, Lucy Platt, Fern Terris-Prestholt Terris-Prestholt, Amteshwar Singh, Jack Stone, Peter Vickerman, Andrew Phillips, Leigh F Johnson, Mathieu Maheu-Giroux, Marie-Claude Boily
Abstract
BACKGROUND : Structural determinants (SDs) are social, economic, political, cultural, organizational, and environmental factors that shape HIV inequalities across individuals and populations. UNAIDS’ 10-10-10 targets aim to reduce exposure to SDs including punitive regulations, violence, stigma, and discrimination. Evidence-based dynamic HIV transmission models can help quantify the population-level impacts of SDs and structural interventions to inform decision-making. We appraised previous representations of SDs in HIV models to develop a new framework that supports the modelling and analysis of SDs.
METHODS : We performed a scoping review of HIV transmission dynamic modelling studies that modelled SDs, published until August 28, 2023, using Ovid Embase and Medline databases. To develop our framework, we considered how models represented exposures to SDs (statically vs dynamically) and reproduced causal pathways to estimate impacts on HIV, and the data they used.
RESULTS : We found 17 HIV modelling studies of SDs and/or structural interventions including incarceration of people who inject drugs (n=5) or Black men (n=2), violence against women (n=3), HIV-stigma (n=1), and homelessness (n=1), among other less well-defined exposures (e.g., “positive and negative attitudes”). Eight studies modelled SDs dynamically using granular exposure histories (e.g., current, recent, non-recent) that captured variation in duration and intensity of exposure. SDs mostly influenced HIV indirectly through simple causal pathways with single intermediate variables (mediators) – largely sexual/injecting partner numbers (n=9), mixing patterns (n=8), or condom use (n=6). Effects of SDs were mostly informed by cross-sectional data. Only seven studies fitted to observed trends in SDs or their effects on mediators/HIV. Using this, we developed our framework (Figure).
CONCLUSIONS : The representation of SDs in models could be refined to improve projections of their impacts and interventions using our framework. Fundamentally, this requires better inputs : more longitudinal studies investigating heterogeneity in SDs and their causal effects. Methods, findings, and limitations should be transparently communicated.