Hiv Prevalence Estimates From The Demographic And Health Surveys PDF Download

The DHS Program has supported the conduct of numerous large-scale HIV seroprevalence surveys. Some of these surveys used a testing strategy based on enzyme-immunoassays (EIA) and recent concerns were raised that this algorithm could have led to overestimation of HIV prevalence. The present report investigated the impact of potential misclassification of samples on HIV prevalence estimates for 23 surveys conducted from 2010-2014. Along with visual inspection of laboratory results, we examined how accounting for potential misclassification of HIV status through Bayesian latent class models affected prevalence estimates. Two types of Bayesian models were specified: one that only uses the individual dichotomous test results and a continuous model that makes use of the quantitative information of the EIA (i.e., their signal-to-cutoff values). Overall, we found that adjusted prevalence estimates roughly matched the surveys' original results, with overlapping uncertainty intervals, suggesting that misclassification of HIV status should not affect prevalence estimates in most surveys. Our analyses did, however, suggest that two surveys may be problematic; the Uganda AIDS Indicator Survey 2011 and the Zambia Demographic and Health Survey 2013-14, where prevalence could have been overestimated - the magnitude of which remains difficult to ascertain. Interpreting results from the Uganda survey is made difficult by the lack of internal quality control and potential violation of the multivariate normality assumption of the continuous Bayesian latent class model. In conclusion, despite limitations of our latent class models, our analyses suggest that prevalence estimates from most reviewed surveys are not overwhelmingly affected by sample misclassification.