Disease mapping using mixture distribution.

BACKGROUND & OBJECTIVES Data on infectious diseases like tuberculosis (TB) have been analyzed in the past without giving adequate attention to spatial variations. Earlier studies also attempted to display disease status of sub regions, usually census tracts, by categorizing them into quartiles, that helps the authorities to identify high- or low-risk areas. This approach is based mainly on binomial and Poisson models for disease data, and the recent attempts focus on using mixture models of Poisson distribution. We carried out this study to find wards of Madurai Corporation having high risks for TB disease, to develop a model of mixture of Poisson distributions for the number of cases and to classify each ward to one of many risk groups for TB disease, and to represent spatial distribution of TB incidence in Madurai city. METHODS produced the observed counts of TB patients in 72 wards of Madurai Corporation. The number of risk groups and the Poisson parameters of each group were found by maximum likelihood approach using the computer package C.A.MAN (Computer Assisted Mixture ANalysis). Bayesian methods were used to associate each ward to a particular risk group. The results were geographically presented in maps by using ArcView mapping software. RESULTS Using binomial model, 26 wards were categorized as high risk wards, and with mixture model approach 15 wards showed standardized morbility ratio (SMR) >1. The wards along river Vaigai and densely populated wards had high risk. INTERPRETATION & CONCLUSION Our findings demonstrate the usefulness of the mixture models for disease data with geographical variations.