OBJECTIVES: We determined rates of gastrointestinal (GI) complications among patients receiving nonsteroidal anti-inflammatory drugs (NSAIDS), according to risk of GI events and adherence to Veterans Affairs (VA) NSAID prescribing guidelines; and calculated the associated costs of hospitalizations from GI events by risk and adherence to guidelines. METHODS: In November 2001, we identified 7625 patients treated with NSAIDS in the New Mexico VA Health Care System. Using VA prescribing guidelines we assessed each patient's risk for GI events (low, moderate, high) and whether their treatment adhered to VA prescribing guidelines. We then reviewed patient records for GI hospitalizations and diagnoses within the following 2 years. Our data included demographic information, hospitalizations, prescription medications, and diagnoses. Costs were based upon 2002 Medicare reimbursement values. We compared outcomes and costs by risk level and adherence to guidelines, using chi square analyses for categorical data and t-tests for costs. RESULTS: Patients at moderate and high risk (n = 2,288) had more (p <0.001) GI hospitalizations (1.86%), compared to patients at low risk (n = 5337, 0.83%) with an odds ratio of 2.24, 95% confidence interval 1.47-3.41. Adherence to guidelines was not associated with fewer hospitalizations (odds ratio 1.41, confidence interval 0.67-2.96). Among patients at moderate-risk who were hospitalized, non-adherence to criteria (n = 26) was associated with higher (p = 0.027) mean costs ($5,709 ± 2991) compared to those adherent to criteria (n = 9, $4,037 ± 1248). Additional hospitalization costs due to non-adherence totaled $112,099. CONCLUSION: The VA guidelines used to assess risk of GI complications from NSAIDS was related to rates of GI hospitalizations over a 2-year follow-up period. Rates of hospitalizations were not affected by adherence to guidelines. However, mean hospitalization costs were significantly lower among patients at moderate risk who were prescribed according to guidelines. Limitations are that the study is observational and that costs are limited to GI hospitalizations.