Several recent randomized clinical trials of anticoagulation in atrial fibrillation have demonstrated significant reduction in stroke rates with a small incidence of bleeding complications. The objective of this study was to determine whether the recommendations resulting from these trials have been implemented into routine practice, and if the anticoagulation control, therapeutic efficacy, and low complication rates achieved in the trials have been matched in community practice. We analyzed the anticoagulation practices and outcomes obtained for patients in atrial fibrillation at a large staff model health maintenance organization (HMO). We reviewed the medical records of all patients in atrial fibrillation as of April 1990. We compared demographic characteristics and clinical risk factors between HMO patients and those in the clinical trials. We also compared anticoagulation monitoring, adequacy of anticoagulation control, and clinical outcomes at the HMO with those achieved in the clinical trials. Of 238 HMO patients in atrial fibrillation, 198 were without contraindications and therefore eligible for anticoagulation. Of these, 168 were offered anticoagulation (84.8%) and 156 were receiving anticoagulation therapy (78.8% of those eligible). The HMO patients had a greater prevalence of comorbidities than those in the clinical trials. The routine monitoring interval at the HMO was estimated at between 36.3 and 40.9 days (compared with 21 to 28 days reported in the clinical trials). The prothrombin time ratios at the HMO were in the target range on 50% of days compared with 68% of days in the clinical trials. The annual stroke and major bleeding rates in the HMO patients (1.3% and 0.6%, respectively) were not significantly different from the rates in the clinical trials (1.3% and 1.1%, respectively). The annual minor bleeding rate of 13.6% at the HMO was greater than the 7.8% to 8.4% rates in the two trials with better anticoagulation control (Boston Area Anticoagulation Trial for Atrial Fibrillation and Stroke Prevention in Atrial Fibrillation Study) but was not significantly different than the rates of 12.7% and 13.7% of the two trials with poorer anticoagulation control (Canadian Atrial Fibrillation Anticoagulation Study and Stroke Prevention in Nonrheumatic Atrial Fibrillation Study). Anticoagulation practices in this community setting appear to be good in that a large majority of patients were receiving anticoagulation therapy, and there were few major adverse outcomes. However, this study illustrates two common problems in attempting to apply the results of randomized clinical trials to routine practice: (1) differences between community patient populations and those on which the conclusions of clinical trials are based, and (2) less successful application of therapeutic interventions in settings other than that of a controlled clinical trial. The greater prevalence of comorbidities in the HMO patient population appears to convey a greater overall risk of thromboembolism and bleeding complications than in the clinical trials. In addition, the suboptimal anticoagulation control achieved at the HMO may increase the risks and decrease the potential benefits compared with those achieved in the clinical trials. Thus, the efficacy demonstrated in the clinical trials of anticoagulation in atrial fibrillation may not be directly translated into effectiveness in practice.