BACKGROUND AND PURPOSE: The assessment of any health care intervention should consider both risks and benefits and take patients' preferences about these into account. The study reported in this paper aimed to elicit patient valuations of health states relevant to assessment of the prevention of stroke by warfarin anticoagulation therapy for patients with atrial fibrillation. METHODS: A sample of patients over the age of 60 years with atrial fibrillation from three family practices in North-East England was interviewed. Their health state values were elicited using the standard gamble method for general practitioner (GP)-managed warfarin treatment, hospital-managed warfarin treatment, major bleed, mild stroke and severe stroke. RESULTS: Of 180 patients, 69 (38%) agreed to participate, of whom 57 (83%) completed interviews. Median (mean) utility values were for GP-managed warfarin treatment 0.986 (0.948), hospital-managed warfarin treatment 0.984 (0.941), major bleed 0.880 (0.841), mild stroke 0.675 (0.641) and severe stroke 0 (0.189). There was wide variation in values between patients and the distributions were highly skewed. CONCLUSIONS: The results are of value in applying decision analysis to groups of patients. They should be used with caution in reaching decisions about appropriate treatment for individual patients, but may provide a starting point for necessary further exploration of those patients' individual preferences.