Objectives: To assess the long term efficacy of and risks associated with computer aided oral anticoagulation for non-rheumatic atrial fibrillation (NRAF) in a district hospital setting. Design: Retrospective, age stratified, event driven clinical database analysis. Setting: District general hospital. Participants: 739 patients receiving warfarin for NRAF between 1996 and 2001. Patients were selected from an anticoagulation database through appropriate filter settings. Main outcome measures: Anticoagulation control (international normalised ratio (INR)) and hospitalisations for bleeding complications, thromboembolic events, and stroke. Results: Over 1484 patient-years, computer assisted anticoagulation was uncontrolled in 38.3% of patients (INR , 2.0 or . 3.0). No significant differences in INR control were observed with respect to patient age (, 65, 65-75, and . 75 years), although to achieve adequate control of anticoagulation, the frequency of testing increased significantly with age. Annual risks of bleeding complications, thromboembolism, and stroke were 0.76%, 0.35%, and 0.84%, respectively. No significant differences in these events were observed between the three age groups studied. Patients who had thromboembolic events and haemorrhagic complications were significantly more likely to have been under-anticoagulated (INR , 2.0) and over-anticoagulated (INR . 3.0), respectively, at the time of their clinical event.Conclusions: Computerised long term oral anticoagulation for NRAF in a community setting of elderly and diverse patients is safe and effective. Anticoagulation control, bleeding events, thromboembolic episodes, and stroke rates are directly comparable with those reported in major clinical trials. The authors therefore support the strategy of rate control with long term oral anticoagulation for NRAF in general clinical practice.