INTRODUCTION: Estimates of osteoporosis (OP) prevalence based on bone mineral density testing and fracture occurrence may be imprecise for small demographic groups. Medicare data are a useful supplemental source of information on OP. METHODS: We studied people ages > or = 65 years covered by Medicare 2005. Cases of presumed OP were beneficiaries with physician services or inpatient claims for OP or for an associated fracture (hip, distal forearm, spine) in 1999-2005. RESULTS: Among 911,327 beneficiaries with 6 or 7 years of Medicare coverage, the overall prevalence of OP and associated fractures was 29.7%. Prevalence was four times higher for women than men, increased with age, and was two times higher for whites, Hispanic Americans, and Asian Americans than African Americans. Among people with OP-associated fracture claims, the proportion with an OP diagnosis was 49.7% overall (women, 57.1%; men, 21.9%) and was lower for men than women and for African Americans than other ethnic groups. CONCLUSIONS: The low proportion of beneficiaries who had an OP-associated fracture and also had an OP diagnosis, particularly among men and African American women, suggests suboptimal recognition and management of OP. Study limitations included lack of validation of our definition of OP and potential misclassification of race/ethnicity.