Challenges of estimating health service utilization for osteoarthritis patients on a population level

OBJECTIVE: To examine how estimates of osteoarthritis (OA) related health service utilization and medical care charges vary based on how the population of patients is defined, we compared a large cohort of patients identified through an administrative OA diagnosis relative to a subgroup of patients in whom this diagnosis had been validated through medical record review. METHODS: We identified all members (\u3e or = 18 years of age) of a Massachusetts group model health maintenance organization (HMO) with documentation of at least one health care encounter associated with an OA diagnosis during the period 1994-96 (n = 10,740). From this population we randomly selected 700 subjects. Trained nurse reviewers abstracted relevant clinical, laboratory, and radiologic data from their medical records. Physician reviewers evaluated the abstracted information and rated the evidence for the presence of OA according to 3 levels (definite, possible, and unlikely). All persons rated by the physician reviewers as having definite OA were included in the validated subgroup (n = 442). Health service utilization and medical care charges were assessed in all persons with an administrative OA diagnosis who were not randomly sampled (n = 10,040) and the validated subgroup (n = 442) across the following domains: (1) ambulatory encounters associated with an OA diagnosis, (2) relevant radiographic studies, (3) relevant surgical procedures, and (4) relevant medication use. RESULTS: Those in the validated subgroup had higher rates of ambulatory OA associated health care encounters, radiographic studies, surgical procedures, and analgesic and/or antiinflammatory medication dispensings. Patients in the validated subgroup were significantly more likely to be in the highest quartile for total one year charges for the care of OA. CONCLUSION: Estimates of health service utilization are substantially higher for populations of patients in whom a diagnosis of OA has been validated through medical record review, as compared with unvalidated populations identified solely through diagnoses contained in administrative records. Thus using health service utilization estimates based on an unvalidated sample may lead to an inaccurate estimate when extrapolated to the overall population of patients with OA