Randomized trials have indicated that well-managed anticoagulation with warfarin could prevent more than half of the strokes related to atrial fibrillation. However, many patients with atrial fibrillation who are eligible for this therapy either do not receive it or are not maintained within an optimal prothrombin time-international normalized ratio (INR) range. We sought to determine whether an anticoagulation service within a managed care organization would be a feasible alternative for providing anticoagulation care. We performed a multi-site randomized trial in six large managed care organizations in the United States. Subjects were aged 65 years or older and had nonvalvular atrial fibrillation. At each site, physician practices were divided into two geographically defined practice clusters; each site was randomly assigned to have one intervention and one control cluster. The intervention cluster received an anticoagulation service that satisfied specifications for high-quality anticoagulation care and was coordinated through the managed care organization. Control clusters continued with their usual provider-based care. We measured the proportion of time that warfarin-treated patients in each of the clusters (intervention and control) were in the target range for the INR at baseline, and again during a follow-up period. Five of the six selected sites succeeded at developing an anticoagulation service. Patients in the intervention and control clusters had similar demographic characteristics, contraindications to warfarin, and risk factors for stroke. Among patients (n = 144 in the intervention clusters; n = 118 in the control clusters) for whom data were available during the baseline and follow-up periods, the changes in percentages of time in the target range were similar for those in the intervention clusters (baseline: 47.7%; follow-up: 55.6%) and in the control clusters (baseline: 49.1%; follow-up: 52.3%; intervention effect: 5%; 95% confidence interval: -5% to 14%; P = 0.32). Although it was feasible in a managed care organization to implement anticoagulation services that were tailored to local circumstances, provision of this service did not improve anticoagulation care compared with usual care. The effect of the anticoagulation service was limited by the utilization of the service, the degree to which the referring physician supports strict adherence to recommended target ranges for the INR, and the ability of the anticoagulation service to identify and to respond to out-of-range values promptly.