Internal validation of procedure codes on Medicare claims for digital mammograms and computer-aided detection.

BACKGROUND Billing data, such as Medicare claims, are a potential data source for evaluative studies of new digital mammography technologies, such as digital mammography and computer-aided detection (CAD). OBJECTIVE To assess the internal validity of procedure codes on Medicare claims for screening mammography interpreted with either digital mammography or CAD. STUDY DESIGN We assessed agreement of procedure coding for digital versus film mammography and CAD use versus nonuse for mammograms with claims appearing in two independent Medicare claims files (the Carrier Claims and Outpatient files). SUBJECTS AND SETTING Women enrolled in fee-for-service Medicare who received screening mammography in 2001 to 2003 within Surveillance, Epidemiology, and End Results regions, representing 25% of the U.S. population (n = 35,642 women who received 57,632 total mammograms). MEASURES Coding agreement beyond chance (kappa), overall, and by year. RESULTS Overall coding agreement was excellent for both digital versus film mammography [kappa = 0.81 (95% confidence interval, 0.80-0.83)] and for CAD use versus nonuse [kappa = 0.82 (95% confidence interval, 0.81-0.83)]. However, in 2001, agreement was only moderate (kappa = 0.51 for digital versus film and kappa = 0.50 for CAD use versus nonuse) but increased to higher levels in 2002 and 2003 (kappa > 0.80 for both technologies). CONCLUSIONS High internal agreement in procedure codes for digital mammography technologies is consistent with accurate coding on Medicare claims after 2001, although external validation remains desirable.