Spatio-temporal modelling in disease mapping

Disease mapping is an area of statistical research that is rapidly evolving. At the present time, it has become usual that Vital Statistics Agencies produce Atlases to display the mortality risk distribution in a region. Most of these published documents already make use of spatial models that correct the well-known deficiencies of the Standardized Mortality Ratio (SMR). Recently, interest has been on extending these models to incorporate jointly time trends and spatio-temporal interactions. Comparable maps in space and time can give valuable information not only about the present geographically localized disease problems but also on the evolution of these problems. Several proposals are being constantly given in the literature. Here, we adopt two of the most used: the spatiotemporal mixture model given in Böhning (2003), and that based on Bayesian analysis proposed by Bernardinelli et al. (1995). These models are compared in terms of how their specifications affect the pattern obtained. We illustrated our findings using a real data set of mortality due to colon and rectum cancer in males from Navarra, Spain.