A0903
Title: Bayesian mapping of mortality clusters
Authors: Andrea Sottosanti - University of Padova (Italy) [presenting]
Abstract: Disease mapping techniques study the distribution of various disease outcomes within a territory, aiming to identify areas with unexpected mortality rate changes, analyze connections across diseases, and divide territories into clusters based on disease incidence or mortality levels. The focus is on detecting spatial mortality clusters that occur when neighbouring areas within a territory exhibit similar mortality levels due to multiple diseases. Identifying both spatial boundaries and responsible diseases is crucial, yet existing methods fail to address this dual problem effectively. To overcome these limitations, Perla is proposed, a multivariate Bayesian model that clusters the areas in a territory according to the observed mortality rates of multiple causes of death, also exploiting the information of external covariates. Perla incorporates the spatial structure into clustering probabilities using the multinomial stick-breaking, and it exploits suitable global-local shrinkage priors to ensure that the detection of clusters is driven by concrete differences across mortality levels while excluding spurious differences. An MCMC algorithm is developed for posterior inference with closed-form Gibbs sampling for nearly all model parameters, requiring minimal tuning. The effectiveness of the methodology is demonstrated by analyzing mortality levels in two distinct territories: the Padua province in northeastern Italy and the counties on the east coast of the U.S.
