Evaluating the Completeness of the SEER-Medicare Linked Database for Oral and Pharyngeal Cancer.

Oral and pharyngeal cancer is a persistent oral health problem. Baseline and trend data to measure progress are lacking. Our long-term goal is to create an algorithm using Medicare claims to identify oral and pharyngeal cancer cases among those ages 65 and older. The goal of this project was to assess the completeness of the SEER-Medicare linked database for identifying incident oral and pharyngeal cancer cases. We compared incidence rates from the "gold-standard" SEER limited-use database to those from the SEER-Medicare linked database using a quasi-likelihood extension of Poisson regression, allowing for over-dispersion. Adjustment for age, sex, race and ethnicity, and interaction terms between these explanatory variables with data source were used to assess the completeness of the SEER-Medicare linked database among these subgroups. Approximately 6.4% of the cases were missing from the SEER-Medicare linked database. The completeness varied by race and ethnicity (p=0.066). Future development of an algorithm to identify oral and pharyngeal cancer cases using Medicare claims alone can potentially identify over 93% of the cases; however, Hispanic, non-Hispanic black, and non-Hispanic other race and ethnicity subgroups will be less likely than non-Hispanic whites to be identified in such future algorithms.