Fast Bayesian Classification for Disease Mapping and the Detection of Disease Clusters

Abstract

We propose a framework fast method for detecting clusters of disease based on generalized spatial scan statistics set in the context of Bayesian Hierarchical Models. The approach models clusters of disease as dummy variables as part of a Generalized Linear Model using advanced estimation procedures based on Laplace approximations. We discuss both the Binomial and Poisson cases, as well as mixed-effects models that can cope with overdispersed data. Models for dealing with zero-inflation and space-time clusters are also considered. Given the vast space of possible clusterings that needs to be explored, we use Integrated Nested Laplace approximations (INLA) to estimate the posterior marginal distributions for the parameters in the model, avoiding the use of computationally intensive methods such as Markov Chain Monte Carlo (MCMC). Cluster selection is performed with the Deviance Information Criterion (DIC), so that models with both fixed and random effects can be compared. We show the advantages of this approach using several case studies. In particular, we consider detection of clusters of disease in space and detection of clusters for zero-inflated datasets. 2 Virgilio Gómez-Rubio et al.