Use of Medicare claims data to measure county-level variations in the incidence of colorectal carcinoma.

Abstract

Population-based cancer registries that can be used to compare cancer incidence and mortality across regions of the U.S. are currently lacking. The authors conducted this study to validate Medicare claims as a measure of county-level colorectal carcinoma incidence among older Americans. Variations found among counties are described in this article. A total of 183,174 hospitalized patients age 65 years in 1991-1993 with newly diagnosed colorectal carcinoma who resided in one of 480 large counties were identified in Medicare files. The county-level truncated age, race, and gender adjusted incidence rates for the population age 65 years, the proportion of patients with a code indicating distant metastases, and the 2-year case-fatality rates were determined. Corresponding rates from the SEER database were compared. The median truncated adjusted 3-year incidence rate was 870 per 100,000 (Quartile 1-Quartile 3, 779-955), with almost twofold differences among counties even after the exclusion of outliers. The median proportion of patients with codes indicating distant metastases was 23.4% (range, 10.2-46.9%; Quartile 1-Quartile 3, 20.8-25.8), and the average 2-year case-fatality rate was 39.2% (range, 26.5-51.4%; Quartile 1-Quartile 3, 37.0-41.6). Medicare files tended to underestimate the truncated incidence rate according to SEER, but among counties the two sets of rates were closely correlated (r = 0.94, P < 0.0001). Medicare files are a potential alternative source of national data for the study of colorectal carcinoma incidence among the elderly at the county level. The data also suggest significant variations among counties in colorectal carcinoma incidence, stage, and mortality that could be used in public health initiatives.