Estimation of HIV burden through Bayesian evidence synthesis

Planning, implementation and evaluation of public health policies to control the human immunodeficiency virus (HIV) epidemic require regular monitoring of disease burden. This includes the proportion living with HIV, whether diagnosed or not, and the rate of new infections in the general population and in specific risk groups and regions. Estimation of these quantities is not straightforward: data informing them directly are not typically available, but a wealth of indirect information from surveillance systems and ad hoc studies can inform functions of these quantities. In this paper we show how the estimation problem can be successfully solved through a Bayesian evidence synthesis approach, relaxing the focus on ‘best available’ data to which classical methods are typically restricted. This more comprehensive and flexible use of evidence has led to the adoption of our proposed approach as the official method to estimate HIV prevalence in the United Kingdom since 2005. AMS 2000 subject classifications: Primary Bayesian inference.