Authors: Jennifer Alford-Teaster*, Geisel School of Medicine - Dartmouth College, Julie Weiss, Geisel School of Medicine - Dartmouth College, Fahui Wang, Louisiana State University
Topics: Medical and Health Geography, Applied Geography, Field Methods
Keywords: Cancer Services Areas (CSA), Hospital Service Area (HSA), GIS, network community detection, localization index (LI).
Session Type: Poster
Start / End Time: 3:05 PM / 4:45 PM
Room: Lincoln 2, Marriott, Exhibition Level
Presentation File: Download
We present preliminary results of an automated network-based community detection method to construct spatial units of Cancer Service Areas (CSA) in the Northeast Region of the United States (U.S.). Cancer services were pulled from Medicare claims and included cancer-directed surgeries, chemotherapy, and radiation therapy. A network was constructed for the cancer service volumes between 5,969 zip code areas in the Northeast U.S. The network optimization method defined CSAs by maximizing service volumes while minimizing volumes between CS. The algorithm, automated in Geographic Information Systems (GIS), yielded a maximum Localization Index (LI) for a given number of CSAs. LI reflected the proportion of measured utilization that occurs within the service area and was a critical index of assessing “goodness” of the units. The method accounted for user-defined constraints such as minimum CSA size (e.g., population). Our preliminary study defined 13 CSAs in the study area. The regions yielded the global maximum modularity value for the network, a quality measure for network community detection. In comparison to existing health service area units, such as the comparable Hospital Referral Regions (HRRs) of the Dartmouth Atlas, the derived CSAs have several advantages: (1) a significantly higher average LI value, (2) more compact in geometric shape, and (3) more balanced in-service area size. The automated, data-driven, and scale-flexible method is adequate for health professionals to define CSAs as the spatial unit of analysis of cancer-specific care delivery. The results will allow for assessment of geographic footprints, population characteristics, and per capita oncologist supply across CSA.