Abstract Objective: To develop a working model with which prescribing behaviour among general practitioners might be influenced. Design: Intervention based on feedback on prescribing rates and problem-oriented educational outreach visits, using educational material and local opinion leaders. Randomised study with three parallel intervention groups of general practitioners, which also served as controls for each other. The pharmacotherapeutic fields chosen were hypertension, peptic ulcer/dyspepsia and depression. Prescription data were retrieved from the electronic patient records for periods of 1 year before and after the intervention. Setting: Six health care centres and three continuing medical education groups in Stockholm. Subjects: Forty general practitioners. Main outcome measures: Drug prescribing rates and patterns before and after the intervention. Results: In the hypertension field, desired trends in fractional prescribing (favouring diuretics and beta blocking agents) were recorded, with a significant (P<0.05) effect on prescriptions for agents acting on the renin–angiotensin system, despite a pre-existing prescribing behaviour already much in line with the goals. In the peptic ulcer/dyspepsia field, desired trends were recorded for both types of therapies addressed. The fractional prescribing rates for proton-pump inhibitors decreased from 61.0% to 52.6% in the intervention arm and increased from 68.1% to 76.0% in the control arm (not significant due to low power). The depression group focused on better general attention to the disease and only minor changes were registered. Conclusion: Feedback of individual prescribing rates, combined with problem-oriented educational outreach visits, is a promising model for the improvement of prescribing behaviour. Data from the electronic patient record were feasible for feedback on prescribing rates.