A1114
Title: A Bayesian evidence synthesis model for estimating Hepatitis C prevalence among people who inject drugs in England
Authors: Pantelis Samartsidis - University of Cambridge (United Kingdom) [presenting]
Daniela De Angelis - University of Cambridge (United Kingdom)
Matthew Hickman - University of Bristol (United Kingdom)
Abstract: Hepatitis C (HCV) is a blood-borne virus affecting the liver and can lead to acute liver damage, cirrhosis, and death. The introduction, since 2015, of effective drugs has led many countries to commit to elimination. Estimating prevalence among people who inject drugs, the most affected by HCV, is crucial to monitor progress towards elimination, but poses challenges. First, information on prevalence exists in many data sources, including surveillance and bio-behavioral surveys. The datasets are heterogeneous in terms of the populations that they target and the sample size. Second, prevalence trends post 2015 differ by geographical region due to varying levels of treatment and by risk group due to differences in healthcare engagement. Third, in many data sources, behavioral characteristics are not recorded, hindering risk group classification. Finally, there is little information about the relative proportions of risk groups in the population, which is required to evaluate overall prevalence. To address these issues, a Bayesian evidence synthesis model is proposed that estimates prevalence by year, region, and risk group. A full characterization of uncertainty is obtained by accounting for potential misclassification of individuals in risk groups, and by modelling the relative proportions of risk groups using survey data. To tackle computational concerns, an integrated nested Laplace approximation is developed within Gibbs algorithm. The method is applied to data from England.
