Population Projection in the Context of HIV/AIDS: Application of Bayesian Melding to a Cohort Component Projection Model with HIV

This presentation will describe preliminary work addressing two topics: 1) estimation and calibration of the traditional cohort component projection model using maximum likelihood (MLE) and Bayesian melding (BM) methods, and 2) adapting that same model to add a disease process and estimate disease-related parameters. The specific application relates to HIV/AIDS in Africa and builds on earlier work by Patrick Heuveline. An HIV-enabled cohort component projection model is fit to a variety of data derived from populations in East Africa using MLE and BM and the results are compared, and age-specific HIV epidemic indicators are estimated and presented. Finally, future uses of the model and methods including probabilistic projections of population size and structure and epidemic indicators such as age-specific incidence and prevalence are discussed.