In order to assess the outcome of hospitalized diabetic patients in an urban health district the notes of a cohort identified from a survey of all inpatients on a single day in spring 1990 were reviewed. One hundred and ten cases were reviewed (8.4% of all inpatients); median age 73 years (range 26-99), 59 female. Fifty-five percent were medical patients (general or geriatric) and 16% were general surgical. Six remained inpatients after 6 months. Sixteen died, of whom 10 had macrovascular disease. Median length of stay was 22 days (2-300), significantly above the district average in all specialties (< 10 days). Of 15 patients with foot problems, 5 died and 3 had major amputations. Only 23% of all patients had documented evidence of screening for diabetic complications. The discharge diagnoses failed to acknowledge diabetes in 54 cases (including 10 deaths). Only 10% had formal advice from the diabetes team and subsequent audit revealed that metabolic management was commonly suboptimal in non-physician units. These data suggest that inpatient diabetes is costly and carries a high mortality. The incidence is substantially underestimated by conventional episode statistics. The evidence from this cohort of diabetic inpatients suggests that improved communication and recognition of the importance of diabetes could usefully contribute to the quality of care achieved.