Conference is now cancelled
A Bayesian evidence synthesis to estimate trends in HIV prevalence
Anne Presanis (MRC Biostatistics Unit, University of Cambridge)
The number of people living with HIV (PLWH) who remain unaware of their infection is impossible to observe directly, requiring instead estimation. Since 2005, annual official estimates of the total number of PLWH in England, both diagnosed and undiagnosed, have therefore been based on a Bayesian synthesis of multiple data sources, linked to the parameters to be inferred through a network of model assumptions, including a hierarchy over space and time. For several demographic and risk groups, the number of people in each group, the prevalence of HIV, and the proportion of HIV infections that are diagnosed are all estimated. Following a Value of Information analysis for the 2012 version of the model, in which we estimated expected reductions in uncertainty, expressed as a loss, from learning specific parameters or collecting data of a given design, the HIV prevalence model has been further refined to make more comprehensive and effective use of the available data, resulting in estimates of recent trends in HIV prevalence in England. We describe the key concepts underlying this evidence synthesis and estimate that the number of undiagnosed infections halved from 13,500 (9,800-20,200) to 6,900 (4,900-10,700) over 2012-2017. This decrease corresponds to a steady increase in the proportion of people living with HIV aware of their infection, from 84% (95% credible interval 77-88%) to 92% (89-94%), demonstrating particularly that England reached the UNAIDS 90-90-90 targets in 2016.