Modelling of geographic cancer risk factor disparities in US counties

Abstract

The goal of this research is to create a theoretical framework for the identification of cancer risk factor disparities and address the recognition of geographic patterns in these factors. 34 secondary variables covering the entire US at the county level in 2010 were analyzed, both individually and grouped (theoretically and statistically), in relation to the mortality to incidence ratio (MIR) for all cancer sites. An a priori assessment and a principal components analysis (PCA) were used to group variables to test societal constructs. OLS and geographically weighted regressions (GWRs) were used to assess influence of both individual and grouped variables against the MIR. The theoretical grouping of variables showed little change in predictive capability of OLS models. In GWR model, there was marked improvement over the OLS. Maps produced using local R2 showed clear regional patterns of influence between the indicators and the MIR. Both the theoretical model and the justification for a spatial approach to cancer risk factor disparities were shown to be effective in this paper. The link between this suite of indicators and the health outcomes is clear, and supports the idea that a full representation of the SES landscape should be used to both predict health outcomes and to assess policy options for improving these outcomes. With the presence of definitive regional patterns and clear connections between the MIR and societal groupings, the findings from this research suggest a need to shift to a more comprehensive and spatial approach to cancer disparities research.