Bayesian Modelling and Mapping of HIV Infection Rate in Thailand

Abstract

Disease mapping and statistical modeling of incidence/prevalence play important roles in epidemiology to display the spatial risks on a map and to explain the causal pattern between the disease outcomes and the potential risk factors. HIV infection is still a major public health problem in Thailand. Bayesian hierarchical method was proposed to fit with the HIV mapping data and to cope with the HIV modeling incidence among risk factors. A useful source of data information is retrieved from the NAP (National AIDS Program), collected by the National Health Security Office (NHSO) in Thailand 2013. The parameters are estimated by Bayesian mixed effect model via mean-variance adaptive Gauss-Hermite quadrature as a type of Bayesian hierarchical model and using the AIC, BIC, and DIC criteria to select the best fitted model. The best fitted model is in a form of interaction effect model in which combination of gender and age, sex worker (SW) and men who have sex with men (MSM), also people who inject drugs (PWID) and age, can jointly explain the HIV infection rate. HIV infection rate is higher at male and aged 15–24 years than other age groups, and higher at unsafe sex in MSM and SW group than others, and higher among PWID aged 15–24 years than other age groups. The top four provinces with the highest risk (HIV infection rate> 8.9%) were Nong Bua Lam Phu, Chumphon, Udon Thani, and Samut Prakan, respectively.