Modeling Disease Risk By Spatial Statistical Approaches

Disease mapping is often carried out to investigate the geographical distribution of disease. The data of disease maps are typically in the form aggregate counts. In the past several decades, the modeling spatial variation in disease risk is studied carefully by the researchers of statistics, and many productions are arrived in both academic fields and applied fields.Disease relative risks tend to be similar in neighboring area, and common approach is to use random effects models that allow estimation of relative risk in an area to borrow strength from neighboring areas, thus producing more stable estimation. In this paper, we provide new models in order to investigate the extent of spatial variability. The structure of this paper is as follows.Firstly, we introduce spatial statistics and provide reviews of the development history of disease mapping in epidemiology. We present the basic concepts and some properties which are correlative with this paper.Secondly, we provide a new method in which the whole of risk in areas is considered as an underlying continuous risk surface. We model the log relative risks as a Gaussian random field so as to estimate not only individual area relative risks but also the whole of relative risk function. We approximate the distribution of area relative risks to provide an analytically tractable form.Thirdly, we provide non-spatial and spatial models using random effects approaches and regression models, also we describe the Possion-lognormal model and the joint model, consider choice of prior distributions.Finally, we illustrate our methods with simulated data using Matlab software and sample data using MCMC method.