The objective was to determine the main parameters taken into account for the decision of antithrombotic treatment of atrial fibrillation (AF) by vitamin K antagonist or aspirin. This was a prospective clinical study of four clinical services of geriatric medicine. Two hundred and nine inpatients, 84.7 +/- 7 years (women 60.8%), with chronic AF were included. The patients were distributed into two groups (anticoagulant or aspirin) according to medical decision. All the decision criteria for treatment were recorded: cardiopathy, conditions of life, clinical examination (nutrition and autonomy, mini-mental state examination (MMSE), walking evaluation, comorbidity), subjective evaluation of risk of falls and glomerular filtration rate. The thromboembolic risk and the bleeding risk, evaluated subjectively for each patient, were compared with two scores of thrombo-embolic risk and bleeding risk. The evolution of the patients was recorded after 3 months. Student's t-test and chi-squared tests were used for statistical analysis. One hundred and two patients (48.8%) received anticoagulant and 107 patients received aspirin. Patients in the aspirin group were significantly older (86.5 +/- 6.5 vs. 82.9 +/- 7.1 years), with more frequent social isolation, higher systolic blood pressure, and had more important subjective bleeding risk and risk of falls. Patients in the anticoagulant group had significantly more valvulopathies and a more important subjective thromboembolic risk. Thrombo-phlebitis antecedents, dementia, denutrition and walking alterations were only slightly more frequent in patients in the aspirin group. Physicians underestimated thromboembolic risk (one-third of patients) and they overestimated bleeding risk (half of the patients). After 3 months, the two groups did not significantly differ for death, bleeding or ischaemic events. In common practice, the decision of antithrombotic treatment for AF should take into account not only cardiovascular but also geriatric criteria.