Back

Evaluating Spatially Targeted HIV Interventions and Harm Reduction Services Among People Who Inject Drugs in a High-Burden Setting

Wang, J.; Clipman, S. J.; Mehta, S. H.; Srikrishnan, A. K.; Mohapatra, S.; Kumar, M. S.; Lucas, G. M.; Latkin, C. A.; Solomon, S. S.; Wesolowski, A.

2026-02-09 epidemiology
10.64898/2026.02.07.26345824 medRxiv
Show abstract

People who inject drugs (PWID) in India continue to experience high HIV incidence while coverage of HIV and harm reduction services within this population remains suboptimal in many settings, highlighting the need to identify novel service delivery points. To evaluate the effectiveness of spatially focused upscaling of interventions at observed venues where PWID injected drugs together, we developed an individual-based dynamic transmission model of HIV informed by detailed injection network, service engagement, and injection venue attendance data collected in a sociometric study of PWID (n = 2512) in New Delhi, India. HIV incidence was simulated for different spatial targeting strategies and with increasing service coverage at injection venues according to UNAIDS/UNODC goals. We identified significant decreases in predicted HIV incidence when deploying interventions at frequently visited injection venues (from 6.8 cases/100 person-years to 2.7/100PY for full service coverage at the most-visited venue, and further down to 1.3/100PY for 12 most-visited venues). Prioritizing the most visited venues stratified by spatial clusters provided services to a larger number of individuals versus prioritizing the overall most visited venues, suggesting that service expansion at venues that are spatially distinct with minimal population overlap has a slightly larger impact on reducing HIV incidence.

Published in Nature Communications (predicted rank #8) · training set

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.