The study evaluates whether data concerning drug prescriptions available from the National Health System (NHS) can be used to provide an estimate of the prevalence and pattern of treatment of diabetes in a well defined health district in southern Italy. In Italy virtually all drug prescriptions are obtained through the NHS. For a period of three consecutive months all prescriptions of insulin, hypoglycemic agents and strips for blood and urine testing were monitored in a well defined area near Naples. 2958 cases were thus identified giving a prevalence of drug treated diabetes mellitus of 2.01%; prevalence was higher in females than in males (2.68 vs 1.35%) and increased with age from 0.05% in the age group below 9 years to 13.67 in the age group over 70 years. This case-finding procedure was validated by cross-check with independent sources of cases. To evaluate the sensitivity of the method a list of 820 "known" cases of drug treated diabetes mellitus was obtained from a random sample of local general practitioners (GPs) and diabetic clinics: 73.5% of these cases were also detected through the analysis of drug prescriptions. To evaluate the probability of misclassification a random subsample of 602 cases identified through prescriptions was submitted to the GPs working in the study area, for 517 the diagnosis of diabetes was confirmed, thus yielding a positive predictive value of 85.9%. After correction for sensitivity and probability of misclassification the prevalence of drug treated diabetes in our population was 2.52%. The pattern of prescriptions in this population is also given.(ABSTRACT TRUNCATED AT 250 WORDS)