Objective To estimate the potential HIV-impact of Avahan, the India AIDS Initiative, among targeted high-risk groups (including female sex workers (FSWs), their clients and men who have sex with men (MSM)) and the general population in different districts of Karnataka, Andhra Pradesh, Tamil Nadu and Maharashtra.
Design Impact evaluation involving mathematical modelling using detailed serial cross-sectional surveys on sexual behaviour and STI/HIV prevalence (IBBA) among targeted high-risk groups and the general population.
Methods A bespoke detailed deterministic model, parameterised with district specific IBBA data, was used to simulate HIV/HSV-2/syphilis transmission in high-risk groups and the general population in different districts. Latin hypercube sampling within a Bayesian framework was used to identify multiple parameter sets that reproduced multiple rounds of HIV prevalence data among FSWs, MSM and clients in all districts and the general population for some districts. The framework was used to test which of two hypotheses (H1 and H2) for time trends in consistent condom use (CCU) among FSWs derived from independent data sources, was more consistent with observed HIV trends, and if these trends could have occurred without post-Avahan increases in CCU (two null hypotheses were assumed—one being more (H0b) and less conservative(H0a)). The most likely CCU hypothesis was used to predict the intervention impact on HIV prevalence/incidence and HIV infections prevented.
Results Using the most likely CCU hypothesis for each district (H1), results so far suggest that the increase in condom use post-Avahan may have resulted in between 21 and 45% of new HIV infections being averted among FSWs in Mysore, Belgaum and Bellary respectively from 2004 to 2007. Similar results were obtained for clients but the absolute number averted was 2–8 fold more. Model projections (Abstract S7.3 figure 1) suggest that this has resulted the large decrease in HIV prevalence observed in these districts, and that this would not have occurred in the absence of Avahan. The syphilis treatment component alone prevented <9 and 13% of new HIV infections over 1 and 10 years. Impact projections for the general population and additional districts will be presented.
Conclusions These Bayesian modelling results, combined with observed HIV prevalence trends and evidence of successful implementation and scale-up of Avahan, provides plausible evidence that Avahan has reduced HIV transmission to a large extent among high-risk groups.