Controlling for migration effects in ecological disease mapping of prostate cancer

Analytical disease mapping aims at linking control variables, the exposure to potential risk factors and to protective elements with the geographical distribution of disease events. Its goals are (a) to identify areas of atypical low and high disease risk and (b) to enhance our understanding of the underlying disease generating process. However, along with many other obstacles in analytical disease mapping, interregional migration masks for degenerative diseases with a long latency period any direct regional association between the observed disease rates and risks factors. Prostate gland cancer mortality rates of the 508 State Economic Areas (SEA) for the reporting period 1970–1994 are linked in an ecological regression model to proxy variables of putatively influential factors. A semi-parametric geographical filtering approach that uses the 1965–1970 interregional census migration flows among the SEA as underlying spatial structure is applied to control for confounding migration effects. This approach allows, despite some simplifying assumptions, to assess direct effects of potential risk factors on the spatial variability of prostate cancer and to visualize underlying spatial dimensions of the disease generating process. Results indicate a shift in relevance of the incorporated exogenous variables.