Seminar - Value of Information in a Bayesian evidence synthesis to estimate HIV prevalence

Location: Doherty Institute Auditorium, 792 Elizabeth St, Melbourne

Presenter: Dr Anne Presanis, MRC Biostatistics Unit, University of Cambridge, UK

Abstract: Annual national estimates of the number of people living with HIV (PLWH) in England, particularly those who are unaware of their infection, have, for several years, been based on a Bayesian model that combines evidence from multiple sources of surveillance and other survey data. For several demographic and risk groups, the group size, HIV prevalence, and the proportion of PLWH aware of their infection are all estimated. This evidence synthesis overcomes the challenge of quantities such as undiagnosed prevalence being inherently unobservable, by allowing them to be estimated indirectly through the network of available data and assumptions on how the data relate to the unobserved epidemic characteristics. Assessing such a model in a “Value of Information” analysis, to understand which parameters are estimated with most uncertainty and which data sources have most influence on the final estimates, is important to prioritise what future data should be collected to reduce the uncertainty. Such an analysis for the 2012 version of the HIV prevalence model led to further development to make more comprehensive and effective use of sexual health clinic data. This updated model has been retrospectively applied to data from 2012 to 2017, to estimate recent trends in undiagnosed HIV prevalence in England. The proportion of PLWH aware of their infection steadily increased from 84% (95% credible interval 77-88%) to 92% (89-94%) over 2012-2017, corresponding to a halving in the number of undiagnosed infections from 13,500 (9,800-20,200) to 6,900 (4,900-10,700), and reaching the UNAIDS 90-90-90 targets in 2016.

Bio: Anne Presanis is a Senior Investigator Statistician at the MRC Biostatistics Unit, University of Cambridge, working in the “Statistical methods Using data Resources to improve Population Health” theme. Her research in Bayesian evidence synthesis lies at the interface of statistical inference, mathematical biology and epidemiology, motivated by substantive problems in infectious disease modelling, where her work can have substantial impact on public health policy. Anne’s interests in evidence synthesis methods include efficient model building and model criticism for Bayesian evidence synthesis models, including methods for detecting, measuring and accounting for conflicting evidence and bias in observational studies, and value of information methods. Her research has been motivated by problems such as estimating HIV prevalence, incidence and transmission; burden of hepatitis C virus; and severity of influenza.