The objective was to describe, by means of a retrospective study of three years' routine District information (financial years 1991/92 to 1993/94), the in- and out-patient activity for patients with diabetes, and compare this with the non-diabetic population. The clinical resource usage by patients with diabetes relative to those without was estimated by (a) their relative probability of admission by specialty, (b) attendance rates at out-patient clinics, (c) primary diagnosis, and (d) operations and procedures. The setting was a District Health Authority with a population of 408 000. All in-patient and out-patient records were reclassified as attributable to a patient with or without diabetes by cross-referral to routine records of patients with identified diabetes from separate hospital databases. The main outcome measures were: (a) relative frequency, and crude and age-specific relative probability of admission by specialty, ICD9 primary diagnosis, and OPCS4 primary operation and procedure, and (b) out-patient attendance rates by specialty. Patients with diabetes were responsible for 5.5% of admissions and 6.4% of out-patient attendances. However, because of increased length of stay, patients with diabetes occupy 9.4% of bed days. The relative risk of admission for diabetes related complications was: coronary heart disease 11.8 (95% CI = 11.4-12.3), cerebrovascular disease 11.8 (10.8-12.8), neuropathy and peripheral vascular disease 15.6 (13.6-17.9), eye complications 10.4 (9.3-11.7), and renal disease 14.7 (12.6-17.3). Recognised diabetes related vascular (9.3-11.7), and renal disease 14.7 (12.6-17.3). Recognised diabetes related vascular complications accounted for at least 23% of admissions of patients with diabetes. The relative risk of admission for diagnoses and procedures not known to be related to diabetes were similar for non-diabetic and diabetic patients. The pattern of out-patient activity mirrored that of the relative probability of admission. It is concluded that previous estimates of the proportion of NHS resources used for the treatment of patients with diabetes had been significantly underestimated. Patients with diabetes were found to occupy 1 in 10 non-obstetric, non-psychiatric beds. Many of these admissions were for diagnoses and procedures that are known to be related to diabetes.