In two hospitals an identical diabetes teaching and treatment programme (in-patient, Monday to Friday, group teaching) was set up. Seventy-eight consecutive, conventionally treated Type 1 diabetic patients (duration of diabetes 10 +/- 6 years), referred during a certain period, were reinvestigated after 1 year, and again (for assessment of metabolic control only) 22 months after the teaching and treatment programme. Initially, mean glycosylated haemoglobin was 2.6%, after one year 1.0%, and after 22 months 1.5% above the upper limit of the normal range (p less than 0.001). Hospital admissions were reduced from a mean of 10 to a median of 1 day per patient per year (p less than 0.001). The long-term quality of diabetes care achieved by the diabetes teaching and treatment programme was unrelated to intelligence quotient, diabetes duration, or diabetes-related knowledge. Patients with normal levels of glycosylated haemoglobin on follow-up (33% of all patients) had particularly good compliance rates, and significantly lower initial values of glycosylated haemoglobin than patients with glycosylated haemoglobin levels greater than or equal to 10%. The data indicate that the diabetes teaching and treatment programme resulted in a substantial long-term improvement of metabolic control and a striking reduction of hospital admissions. The study substantiates the feasibility of applying this teaching and treatment programme on a large scale to other hospitals, so as to improve the quality of diabetes care and decrease health care costs.