Improved Geocoding of Cancer Registry Addresses in Urban and Rural Oklahoma

Between 1997 and 2013 (the included study years), approximately 23% of addresses in the Oklahoma Central Cancer Registry (OCCR) were not geocoded to the address level. Addresses in rural counties were geocoded with poorer quality, preventing the instructive geographic research that informs policymaking.

To improve the accuracy of the geocodes, we first utilized the United States Postal Service's LACSLink database to correct addresses; specifically, to convert old rural route-based addresses to modernized Enhanced 911 (E911) addresses. We created custom geocoders using regional E911 reference data sets and used existing national scope geocoders of NAVTEQ and the North American Association of Central Cancer Registries. We attempted to geocode 5,102 addresses, which are either regular street addresses or rural route addresses. In the process, we evaluated and tabulated performances of the address correction. Accordingly, we first tabulated how well each geocoder could geocode original and LACSLink corrected addresses. We then documented the overall performances of geocoders based on pairwise comparisons.

We were able to geocode 1,945 addresses out of this data set using 5 distinct geocoders. We observed that the LACSLink correction and E911 data were useful in the specific purpose of geocoding rural addresses, as found in the literature.

We conclude that both LACSLink correction and E911 data were useful for improving geocoding of cancer records, many of which were in rural areas. Future directions include further validation of the geocoding and plans to conduct spatial exploratory data analysis to generate hypotheses related to the distribution of cancer in Oklahoma.