OBJECTIVE: To evaluate the effectiveness of a program to improve hypertension screening practices in primary care. DESIGN: Retrospective quasi-experimental study. SETTING: Two hospital-based family medicine centres (FMCs). PATIENTS: In the study FMC, two study groups of randomly selected adult patients: 425 who visited the FMC before implementation of the screening improvement program (from Apr. 1, 1983, to Mar. 31, 1984) and 418 who visited it afterward (from Apr. 1, 1986, to Mar. 31, 1987). These patients were matched with 392 and 442 control patients respectively seen during the same time frames at the second FMC. INTERVENTIONS: Educational sessions for physicians to standardize blood pressure measurement and knowledge of the recommendations from the Canadian Hypertension Society on hypertension screening and diagnosis, and specific operational incentives to improve hypertension screening, including a reference guide placed in each physician's office, a coloured form for recording blood pressure measurements placed in every patient's chart and a follow-up and recall card file. MAIN OUTCOME MEASURE: Frequency of blood pressure measurements recorded in patient charts. RESULTS: The hypertension screening rate was 60% per year in the study group before program implementation and 79% in the study group afterward; the corresponding rates in the two control groups were 72% and 59% (p < 0.0001). Patients were more likely to be screened if they visited the physician for a periodic health examination than for other problems (e.g., psychosocial or dermatologic) and if they had a scheduled appointment rather than no appointment. Physician characteristics that were positive predictors of screening were low age, female sex and payment on a salary basis. CONCLUSION: Physician education and incentives are effective in improving hypertension screening practices in hospital-based FMCs without incurring additional costs or other use of resources. Further evaluation of such a program should be undertaken in other primary care settings.