GEOSPATIAL APPROACHES FOR UNDERSTANDING THE ROLE OF RESIDENTIAL MOBILITY AND AREA-LEVEL FACTORS IN COLON CANCER SURVIVAL DISPARITIES.

Abstract

A primary reason geospatial approaches are important in cancer research is that health and disease are shaped not only by factors such as age, race/ethnicity, genes, and clinical care but also by the environment where individuals work and act. While the use of geospatial approaches in cancer research is growing, several limitations remain. For example, for most population-based studies, cancer patients' neighborhood environments are based on only a single location derived from the residence at the time of diagnosis.This dissertation aimed to address this limitation by using a unique dataset of colon cancer patients diagnosed in New Jersey that include residential histories obtained through a data linkage with LexisNexis, a commercial data collection company. By incorporating residential histories, I moved beyond a cross-sectional approach to examine how residential histories and socio-spatial mobility can change a patient’s geographic context over time and influence survival. To demonstrate the application of these data in this dissertation, I completed three case studies. In the first case study, I compared whether including residential histories changed the risk of death estimates by neighborhood poverty compared to the traditional approach when including only the location at the time of diagnosis. Results suggested that the risk of death estimates from neighborhood poverty were generally similar in strength and direction regardless of residential histories inclusion. This finding was likely a result of minimal socio-spatial mobility of colon cancer patients (i.e., patients generally moving to census tracts with similar poverty levels). The second study aimed to compare the geographic risk of death estimates when using single location and residential histories in spatial models. Results overall showed that the geographic patterns of the risk of death estimates were generally similar between the models. However, not accounting for residential mobility resulted in underestimated geographic risk of death in several areas. This finding was related to the fact that approximately 35% of the colon cancer patients changed the residency, and 12% of the initial study population left New Jersey after the diagnosis. In the third case study, I examined whether landscape characteristics (e.g., built environment) were associated with the risk of death from colon cancer independent of individual-level factors, residential mobility, and neighborhood poverty. The results indicated that an increasing proportion of high-intensity developed-lands substantially increased the risk of death, while an increase in the aggregation and connectivity of vegetation-dominated low-intensity developed-lands reduced the risk of death. These findings suggested that places lacking greenspaces could have worse access to recreational sites that promote physical activity. Overall, this dissertation expands our knowledge about the geographic disparities in colon cancer in New Jersey. It also provides specific examples of integrating residential histories and remote sensing-based products into cancer disparities research. Including residential histories opens up new avenues of inquiry to better understand the complex relationships between people and places, and the effect of residential mobility on cancer outcomes. Combining multiple socio-demographic and environmental domains to estimate the neighborhood effects on cancer outcomes will increase our potential to understand the underlying pathways.