Comparison of Bayesian Spatio-temporal Models of Tuberculosis in Makassar, Indonesia

Abstract

Abstract South Sulawesi province ranks sixth-highest in tuberculosis (TB) in Indonesia. Makassar ranks the highest in South Sulawesi. Spatio-temporal modelling can identify the areas with high risk as well as the temporal relative risk of disease. We analysed the tuberculosis cases data from Makassar City Health Office for 15 districts over seven years from 2012 to 2018. Seven models of Bayesian Spatio-temporal (BST) Conditional Autoregressive (CAR) were applied by using the measures of goodness of fit (GOF) namely, DIC and WAIC. The results showed that BST CAR localised model with G = 3 has the lowest DIC and BST CAR adaptive has the lowest WAIC. Based on the preferred model (Bayesian ST CAR localised with G=3), Panakukang district had the highest relative risk of TB in 2012, 2013, and 2014, while Makassar district had the highest relative risk of TB in 2015, 2016, and 2017. Mamajang had the highest relative risk of TB in 2018.