Methods for modeling and mapping spatial variation in disease
risk continue to motivate much research. In particular, spatial
analyses provide a useful tool for exploring geographical heterogeneity in health outcomes, and consequently can yield clues as to disease aetiology,
direct public health management and generate research hypotheses. This article presents a Bayesian partitioning approach for the analysis of individual level geo-referenced health data. The model makes few assumptions about the underlying form of the risk surface, is data adaptive and allows for the inclusion of known determinants of disease. The methodology is used to model spatial variation in neonatal mortality in Porto Alegre, Brazil.