Measuring antibiotic prescribing practices among ambulatory physicians: accuracy of administrative claims data

To assess the accuracy of administrative claims data for measuring antibiotic prescribing behavior, we conducted a stratified randomized medical record review of office visits by children with pharyngitis, and adults with acute bronchitis, to primary care physicians in Colorado in 1998. The diagnoses of pharyngitis (n = 422) and acute bronchitis (n = 497) based on administrative data were verified in 83% and 79%, respectively, of medical records. The sensitivity, specificity, and positive predictive value of administrative data in identifying antibiotic treatment for pharyngitis was 68%, 91%, and 90%, respectively, and for bronchitis was 79%, 84% and 98%, respectively. The sensitivity, specificity, and negative predictive value of administrative data in identifying group A streptococcal test ordering for pharyngitis was 71%, 86%, and 30%, respectively. Absence of testing in administrative data (when present in the medical record) was more frequent among visits to physicians associated with a capitated health plan. We conclude that administrative claims data are accurate sources for measuring and profiling antibiotic prescribing practices in ambulatory practice, although they underestimate actual antibiotic treatment decisions by individual physicians. Measuring and profiling antibiotic prescribing behavior in relation to group A streptococcal test utilization may overestimate inappropriate antibiotic treatment by physicians enrolled in capitated contracts.