OBJECTIVE: To examine the effect of mail-based interventions on appropriate prescribing of COX-2 inhibitors for patients at increased risk for adverse gastrointestinal (GI) events. METHODS: The 600 highest-volume prescribers of COX-2 inhibitors were identified from a health plan database from October to December 2001, and randomly assigned to one of three intervention groups: 1) High intervention (profile incorporating probability of GI bleed); 2) Medium intervention (pharmacy claim information only; no probability modeling); and 3) No intervention. Intervention physicians received monthly profiles for 6 months (May 2002 through October 2002). Risk factors used to calculate the probability of GI bleed included: history of ulcer, arthritis, GI bleed, concurrent use of anticoagulants or oral steroids, and age greater than 65 years. Appropriate usage metrics were changes in mean proportion of patients with each risk factor and percentage of COX-2 inhibitor prescriptions out of all anti-inflammatory medications. Proportions were compared between intervention groups. Statistical analyses were performed using multivariate Generalized Estimating Equations (GEE). RESULTS: Baseline proportions of COX-2 prescriptions for high intervention, medium intervention, and no intervention groups were 0.602 ± 0.192, 0.603 ± 0.202, and 0.589 ± 0.202, respectively; 5-month post-intervention proportions were 0.587 ± 0.269, 0.588 ± 0.253, and 0.619 ± 0.255, respectively; adjusted GEE coefficients were -0.048 (p=0.001), -0.044 (p=0.002), for the high and medium intervention groups, respectively. Proportion of patients with a history of ulcer, GI bleed, and arthritis also increased by 33%, 41%, and 26%, respectively, across all groups. However, differences between groups were not significant. CONCLUSIONS: Preliminary analysis has shown a positive change in prescribing patterns. First, there was a decrease in the proportion of patients receiving COX-2 inhibitors. Second, in patients receiving COX-2 inhibitors, the proportion of patients with select risk factors (history of ulcer, GI bleed, and arthritis) increased.