P16.13 Epidemiology of hiv prevalence among urban women in tanzania

Abstract

Background Estimating the distribution of new HIV infections according to identifiable characteristics is a priority for programmatic planning in HIV prevention. We propose a mathematical modelling approach that uses robust data sources to estimate the distribution of new infections acquired in the generalised epidemics of sub-Saharan Africa and validate it against cohort data. Methods We developed a predictive model that represents the population according to factors powerfully associated with risk: gender, marital status, geographic location, key risk behaviours (sex-work, injecting drug-use, male-to-male sex), sero-discordancy within couples, circumcision and ART status. Incidence inference methods are applied to estimate the short-term distribution of new infections by group. The model is applied within a Bayesian framework whereby regional demographic and epidemiological prior information is updated, where possible, with local data. We validated and trained the model against cohort data from Manicaland (Zimbabwe), Kisesa (Tanzania) and Rakai (Uganda). Building on the results from the acquisition model we infer likely sources of transmission. The model was applied to six countries in the region to investigate potential differences in incidence patterns. Results Without training using the site-specific data, the model was able to predict the pattern of new infections with reasonable accuracy: 95% credible intervals were substantially overlapping and the rank ordering of groups with new infections was consistent. With training using group-specific data on new infections, the accuracy of predictions for subsequent rounds of data improved further and credible intervals narrowed. When applied to the six countries in the region the model showed variation in the distribution of infections between and within countries consistent with the data on prevalence. Conclusions It is possible to accurately predict, the distribution of new HIV infections acquired using data routinely available in many countries in the Sub-Saharan African region. This validated tool can complement additional analyses on resource allocation and data collection priorities.