MSM with HIV: Improving prevalence and risk estimates by a Bayesian small area estimation modelling approach for public health service areas in the Netherlands

Abstract

Despite close monitoring of HIV infections amongst MSM (MSMHIV), the true prevalence can be masked for areas with small population density or lack of data. This study investigated the feasibility of small area estimation with a Bayesian approach to improve HIV surveillance. Data from EMIS-2017 (Dutch subsample, n = 3,459) and the Dutch survey SMS-2018 (n = 5,653) were utilized. We applied a frequentist calculation to compare the observed relative risk of MSMHIV per Public Health Services (GGD) region in the Netherlands and a Bayesian spatial analysis and ecological regression to quantify how spatial heterogeneity in HIV amongst MSM is related to determinants while accounting for spatial dependence to obtain more robust estimates. Both estimations converged and confirmed that the prevalence is heterogenous across the Netherlands with some GGD regions having a higher-than-average risk. Our Bayesian spatial analysis to assess the risk of MSMHIV was able to close data gaps and provide more robust prevalence and risk estimations.