Issues in the mapping of two diseases.

Recently, there has been increased interest in the geographical modelling of two or more diseases. In this article, we consider a number of issues relating to such an endeavour including the standardization process and the comparison of univariate and bivariate disease mapping models. A principle motivation for the examination of two or more diseases is to discover similarities or dissimilarities in the geographical distribution of risk. In this article, we propose a proportional mortality approach to give clues to areas of similarity and dissimilarity. A secondary aim of bivariate modelling is to 'borrow strength' between diseases in order to provide better estimates of risk in each area. We will illustrate various modelling strategies using incidence data from 1996 to 2000 on lung and bladder cancer in Washington state.