We searched the biochemistry laboratory computerised database for results of all biochemistry tests carried out during each patient's admission. The medical records of those with at least one plasma glucose value in the hyperglycaemic range (glucose =11.1 mmol/l and/or =7.0 mmol/l on casual and/or fasting measurements respectively) were reviewed by two observers using a standardised method and a cheek list for data collection. 707 (80.4%) patients had at least one plasma glucose value available. 100 patients (12.5%; 95% C.I. 10.3% to 14.9%) had at least 1 plasma glucose in the hyperglycaemic range. 56 of these patients were known diabetic patients and there were 3 patients, whose clinical records could not be traced. Out of the remaining 41 patients, 30 were male, their mean age was 68±12 years and the mean peak random glucose value was 13.4±1.8 mmol/l. There were 12 patients with two or more plasma glucose values in the hyperglycaemic range. A total of 8 of the 41 hyperglycemic patients died during their admission. Of the remaining 33 hyperglycemic patients who were discharged, 21 (63.6%; 95% C.I. 45% to 80%) were discharged without evidence in their clinical records that the elevated blood sugar had been noted. A significant proportion of hospital inpatients has abnormal plasma glucose concentrations consistent with a possible diagnosis of diabetes. There is a need to raise the level of awareness among health care professionals of this problem. We also need to develop structures within hospitals to detect previously undiagnosed diabetes. Routine, laboratory initiated follow-up of abnormal glucose results merits consideration.