Objective: To test the efficacy of consultation designed to prevent anticoagulant-related bleeding. Design: Randomized, controlled trial. Setting: A large teaching hospital. Patients: A total of 101 patients at increased (> 15%) risk for major, in-hospital bleeding while starting long- term anticoagulant therapy who were identified using a validated prediction rule. Interventions: Fifty-five patients received usual care under the direction of the attending physician who had initiated anticoagulant therapy. Forty-six patients received guideline-based consultation in addition to usual care. Guideline-based consultation included individualized review of the risks and benefits of anticoagulant therapy and, on the basis of current practice guidelines, recommendations for daily management. Measurements: The main outcome was in-hospital bleeding, which was classified using a reliable, explicit index. Results: Major or minor bleeding occurred in 17 of 55 patients (31%) receiving usual care alone, compared with 6 of 46 patients (13%) receiving consultation in addition to usual care (P = 0.03). The protective efficacy of consultation was 58% (95% CI, 3% to 82%). Consultation was associated with similar reductions in the frequencies of major bleeding (from 13% to 4%) and minor bleeding (from 18% to 9%). Consultative recommendations had an 84% compliance rate and directly affected anticoagulant management: In the consult group, nonsteroidal anti- inflammatory agents were stopped in six patients (13%), and therapeutic ranges were achieved more often for activated partial thromboplastin times (52% compared with 45% in the usual care group, P = 0.08) and for prothrombin times (47% compared with 27% in the usual care group, P < 0.001). Nearly all housestaff and attending physicians (91%) for patients receiving consultation also reported that consultation improved housestaff learning. The consult group had a somewhat lower rate of thromboembolism in the 90 days after discharge (5% compared with 17%, P = 0.06). Death rates and mean lengths of stay were similar in the two groups. Conclusion: Guideline-based consultation was associated with reduction in the frequency of anticoagulant-related bleeding in patients at increased risk for major in-hospital bleeding.