Abstract
OBJECTIVE: To validate that, using patient demographics and other risk factors readily obtained from computerized databases, one can predict a priori the risk for developing a nonsteroidal antiinflammatory drug (NSAID) associated gastrointestinal (GI) bleed prior to exposing patients to therapy. METHODS: We conducted a retrospective cohort analysis using computer-stored information from a large group-model health maintenance organization. All patients who received one or more prescriptions for a single NSAID over a 9 month period were eligible. Historical and risk factor data was obtained for age, sex, prior GI bleeds, use of GI medications, prednisone use, and use of disease modifying antirheumatic drugs (DMARD). We tested a model (eSCORE) that is based on a previous risk stratification method. The primary outcome was a hospital admission for a GI bleed (GI event). RESULTS: A total of 303,211 NSAID patient-users met eligibility requirements. Serious GI events occurred in 302 patients, for a rate of 0.68% (0.68 events per 100 patient-years' exposure). All the risk factors except DMARD use were associated with a significant increase in the GI event rate. Higher eSCORE points were associated with increased GI event rates. CONCLUSION: The study supports the concept that the rate of GI events can be predicted by a defined set of easily assessed patient criteria using the eSCORE. Stratification of patients by risk score can guide the physician to appropriate therapeutic options, with the potential of protecting patients at greatest risk for a GI event.
