A comparative analysis of frequentist and Bayesian approaches to estimate dengue disease transmission in Bandung-Indonesia

Abstract

Dengue disease is one of the major health concerns in Indonesia because of its high number of cases. There is no specific treatment to cure people infected by this mosquito-borne disease and prevention is the best strategy to avoid infection. A study of dengue disease mapping is essential to depict the high-low risk areas. This study aims to estimate the relative risk of dengue disease in Bandung, West Java, Indonesia, using a frequentist statistics approach i.e. the Standardized Morbidity Ratio (SMR) and the Bayesian statistics approach, i.e. Poisson-gamma and the Besag, York and Mollie (BYM) models. Dengue disease data for Bandung from 2014-2016 are investigated and the results are displayed in graphs and maps. Further, two Bayesian models are compared using the Deviance Information Criterion (DIC) to decide the model which best fits the data. Based on the smallest DIC value, the Poisson-gamma model was found superior to that of the BYM. Further, results show that the most critical areas affected by dengue disease are located in south-west Bandung; and the highest risk of dengue disease in Bandung for the period of 2014-2016 occurred during the rainy season (November-December).