Abstract
The burden of cancer is a significant physical, psychological, and financial toll on individuals and populations. Cancer is more common in older populations, so advances in other areas of medicine will inevitably increase the numbers of cancer patients. The development of cancer is prolonged over years which makes it difficult to identify etiologic causes. And the treatment of the disease can also be quite protracted. In the case of advanced or aggressive disease, medical treatment is sometimes reduced to a series of clinical interventions that do little to change the timeline to mortality. Currently, our most effective tool in cancer control is prevention. Cancer risk, health behaviors, and medical care standards and access vary by geography. Epidemiologists can apply a spatial approach to epidemiology to see geographic patterns and test associations in the data based on geography in order to postulate about a community’s health, focus public health action, and choose suitable prevention interventions. The application of geographic information systems (GIS) and spatial analysis is a valuable tool for epidemiologists to address geographic discrepancies, which are often driven by race or social disparities in health. But relevant conclusions hinge on understanding the limitations of the data and the methods as well as the suitability of a spatial approach to epidemiologic research. The first part of this dissertation focuses on a specific application for cluster detection. This work indicates that clustering of disease with public health significance may be inadvertently missed if the analysis is conducted at only one scale. However, conducting spatial scan analysis at multiple scales often produces multiple, significant results that do not always overlap geographically. Until there is a practical tool to empirically assist in final result selection, manual visual assessment and local knowledge must be applied. The second part assesses the association between risk factors and the likelihood of a CRC cases being diagnosed in a cluster at high risk of late-stage at diagnosis. There was an unexpected association between area-based poverty and late-stage clustering with increasing poverty being less associated with late-stage at diagnosis cluster. This may be a “screening effect” driven by current cancer control efforts within the state which loosely correlates with state trends in CRC screening. The third part of this dissertation focuses on the data quality of cancer data and the impact on results. This work indicates that missing stage at diagnosis was spatially correlated, and that a common method of handling missing stage, removing those cases from analysis, may be the most biased method. And although misclassification of the sex code in registry data was not spatially correlated and, therefore, had limited influence on the results of spatial analysis, the impact of misclassified sex is dramatic on specific sites like male breast cancer. The theme of this dissertation is issues with application of geospatial techniques in cancer surveillance to address issues in cancer control. Relevant conclusions are that geospatial epidemiology demands a team science approach in order to adequately address software, methodological, and clinical issues that often are adapted from outside the traditional training of an epidemiologist. A geographic approach is a natural companion to population based research, but attention to data quality and relevance of results is required to make an important impact with these techniques.