The article by Aneja, Gross, Soulos, and Yu outlines applications of geographical information systems (GIS) software to oncology research.
The article by Aneja, Gross, Soulos, and Yu outlines applications of geographical information systems (GIS) software to oncology research. GIS software can enlighten the profession's understanding of the geographic distribution of resources, practice patterns, patients, incidences, and outcomes. The purely descriptive use of geo-coded data as it relates to cancer care and patient populations will definitely prove illuminating to both providers and researchers. The more difficult task of analysis attempts to identify causal relationships suggested by the data.
The authors provide a thoughtful summary of the history of GIS software and describe how data elements (eg, cancer incidence) can be mapped to counties and made available for descriptive or statistical analysis. The authors also note the specific statistical tools available to researchers using geographic-coded data, including spatial autocorrelation, geographically weighted regressions, and spatial interpolation. Important to each application are the limitations mentioned at the end of the article, primarily the temptation to infer causality between explanatory and dependent variables.
Specific applications of GIS in cancer research are categorized by the authors. The first application maps cancer incidence and mortality. The second looks at the supply side of the health care market, investigating the distribution of resources. The third application also involves the supply side of the market, focusing on practice patterns among providers. The fourth type of study the authors discuss uses GIS to examine potential disparities in resources or outcomes-related patient characteristics.
Taken together, the four types of studies explore the geographic distribution of potential patients and their characteristics, as well as of providers and practice styles. The integration of all four types of studies is critical to a comprehensive understanding of cancer care in the United States. However, inferences from cross-sectional studies that examine the geography of patient and supplier characteristics at a single point in time may well miss the story that led to the observed distribution. The authors note this in their discussion of cancer incidence and mortality, suggesting that GIS software should be used to depict areas where cancer "mortality is most rapidly changing."
The authors' observations about changes that take place over time suggest other research avenues. Exploiting both time series and cross-sectional variation in the geographically defined variables of interest allows the researcher to identify both within-region and between-region relationships. For example, is the geographic distribution of cancer-related death rates persistent through time within age and gender groups? Further, how have the reductions in cancer-related death rates at younger ages and the increases at older ages over the last 40 years been distributed geographically? Similarly, how has the geographic distribution of cancer care providers evolved over time? Is there evidence that potential patients move to areas with greater ease of access or more intensive interventions? Conversely, do we see cancer providers moving to areas with higher concentrations of potential patients? When it comes to offering policy prescriptions, the answers to these questions are vital.
The characteristics of cancer treatment regimens distinguish them from other types of health care interventions in terms of duration, follow-up, and specific resource or expertise intensity. These characteristics may lead to the observation that geographic distribution of resources or treatment protocols are less than preferable on some dimensions. However, when resources are expensive and specific expertise is scarce, we should not be surprised that the quantity and quality of cancer care varies along geographic dimensions.
The hospital referral regions and hospital service areas defined by the Dartmouth Institute can serve as examples of regional nodes of expertise in the diagnosis and treatment of cancer that have been defined with the possibility of identifying Cancer Care Referral Regions (CCRR). Developing such regions requires linkage of patients to providers, with the understanding that the providers may vary by diagnosis and/or treatment. One dimension of health care access that the authors did not mention is the geographic distribution of the sources of payment for health care in general and cancer care in particular.
The authors chose a perfect title for their survey of applying GIS software in the context of oncology research. They introduce the readers to the wide range of applications of geospatial data in oncology research and they remind readers of the limitations of real world data that are not based on a controlled experiment. Utilizing geo-coded time-series data on cancer incidence, cancer-related deaths, patient characteristics, and provider concentrations will hone analysts' understanding of the importance of geography in observed distributions and will refine research questions that may ultimately uncover causal relationships.
Financial Disclosure:The author has no significant financial interest or other relationship with the manufacturers of any products or providers of any service mentioned in this article.