Socio-Spatial Factors Driving to Cancer Disparities
Advisor: Dr. Susan L. Cutter
While cancer rates have shown promising trends over the last few decades, not all populations have experienced the same levels of decrease in cancer incidence in mortality rates. Identifying populations suffering from the impacts of the disparities has become a major goal in cancer research.
Most research has focused on the influence of single variables on cancer disparities or on small-scale case studies. Using the information from these analyses, the research conducted in this dissertation tests the relationship of selected variables to an outcome measure, the mortality to incidence ratio (MIR) in search of spatial relationships between the indicators and the MIR. The goal is to identify influential variables in addition to determining whether variables consistently express the same influence over the MIR.
In order to achieve the goal, three separate analyses are conducted. A regression model is run first, using thirty-four potential variables as independent variables and the MIR as the dependent variable. A principal components analysis is run second to test the contribution of indicators to the variance in the set and measure correlations amongst the indicators. A second regression tests the predictive ability of the PCA-grouping and the contribution of each group to the MIR. A path analysis is conducted as well to determine how factors influence each other and interact to yield cancer outcomes. The third and final step involves a sptial analysis to establish regions where disparities exist as well as identifying differences in the contribution of variables to the disparities.
The findings of the research reveal a complex interaction of variables and a level of dependence between the aggregated groups. Health indicators, when considered separately, are the best predictors of cancer fatality, but they are highly dependent on the presence of certain community, social, and economic factors. In addition, the most significant finding was that aggregated groups of indicators did not have a consistent influence over cancer outcomes. There were differing levels, and in some cases a complete reversal, of spatial correlation between the groups and the MIR dependent on place. The most likely association on the nature o finfluence is related to the urban/rural divide. This finding suggests a need to shift to sub-regional analyses.