Probabilistic Projections of HIV Prevalence Using Bayesian Melding

Leontine Alkema, University of Washington
Adrian Raftery, University of Washington
Samuel J. Clark, University of Washington

The Joint United Nations Programme on HIV/AIDS (UNAIDS) has developed the Estimation and Projection Package (EPP) for making national estimates and short-term projections of HIV prevalence based on observed prevalence trends in antenatal clinics. Understanding uncertainty in its projections and related quantities is important for more informed policy decision making. We propose using Bayesian melding to assess the uncertainty around the EPP predictions. Prevalence data as well as information on the input parameters of the EPP model are used to derive probabilistic HIV prevalence projections - a probability distribution on a set of future prevalence trajectories. We relate antenatal clinic prevalence to population prevalence and account for variability among clinics using a random effects model. We discuss the results of the Bayesian melding procedure for Uganda, where prevalence peaked at around 28% in 1990; the 95% prediction interval for 2010 ranges from 1% to 7%.

  See paper

Presented in Session 83: Statistical Modeling Issues in Population Research