Osteoporosis is a major public health problem becoming increasingly important as our population ages. Approximately one in two women will experience an osteoporotic fracture in their lifetime. Additionally, millions have low bone mass that increases fracture risk. OBJECTIVE: To develop a methodology, in a managed care environment, that identifies members at high-risk for osteoporotic fractures. METHODS: This was a retrospective database study. The study population was from a Midwest health plan of 1.38 million lives. The database consisted of all pharmaceutical, professional, and institutional claims linked at the patient level. The study population was identified by the following criteria: continuously enrolled from 7/1/95 to 3/31/99, complete coverage, individuals older than age 45 for women and age 65 for men with fracture of interest, specific fracture by site; hip (820.0 to 821.39), vertebrae (806 to 806.91), and forearm (813.0 to 813.9x), past history of fracture and specific site by V codes (13.5 and 15.5), chronic users of steroids (? 180 days exposure over a period of 24 months), history of transplant (by CPT-4 and V codes), a diagnosis code of 733.0x., current history (in study period) of ICD-9 codes for pathologic fractures 733.1 to 733.19, a family history of musculoskeletal disease (V code 17.8), and individuals with bone density studies (by CPT-4 codes). RESULTS: Over 120 million claims were screened to stratify the population (prior to 1997) into risk bands. We used 2,000 patients with current osteoporotic fractures to develop and to verify a predictive model of fracture risk. CONCLUSIONS: A claims based model identified high-risk patients for fractures. Application of the model to 1999 data generated a list of patients for intervention.