Review and empirical comparison of joint mapping of multiple diseases
Samuel Manda, Richard G. Feltbower, Mark S. Gilthorpe · Southern African Journal of Epidemiology and Infection · 2012
There has been a substantial amount of recent modelling work in multivariate disease-mapping models in epidemiology. These models provide information on similarities, as well as differences, on the effect of risk factors among diseases. Additionally, they can be used to identify disease-specifc risk factors, which would otherwise have been masked by established factors. The purpose of this article is to provide a review of the biostatistics literature, by comparing four joint disease-mapping models. In particular, multivariate intrinsic conditional autoregressive (ICAR) and multivariate multiple membership multiple classifcation (MMMC) models, as well as, shared-component and proportional mortality models are compared, with regard to similarities and differences between the assumptions and inferences. As an illustration, the four different models are ftted to population-based oesophagus and stomach cancer data. These two cancers share common risk factors associated with smoking, and diet or alcohol consumption. The four methods produce similar substantive fndings. However, the shared component model adds more versatility in answering more substantive epidemiological questions, than the other three models. The review article provides a useful reference for epidemiologists, and public health researchers and planners, who are interested in disease mapping. In order to promote the uptake of these models, and aid their implementation by non-experts, WinBUGS® codes used to ft the models are provided as supplementary material.