OBJECTIVE: To quantify use of health-care resources for management of severe hypoglycaemia. METHODS: We defined severe hypoglycaemia as an episode of coma or seizure requiring assistance from NHS personnel. We excluded any episodes managed exclusively by persons outside the health services. We included all patients with Type 1 or Type 2 diabetes registered with DARTS (Diabetes Audit and Research in Tayside Scotland) and receiving treatment with insulin or oral hypoglycaemic drugs from June 1, 1997 to May 31, 1998. We identified episodes from the records of three acute-care hospital emergency rooms, eight minor injury units in primary-care hospitals and from Tayside Ambulance Service records. RESULTS: Over 150,000 records were searched to identify 295 people with one or more episodes, of which 39 were excluded because they were not residents of Tayside and another 12 because their identity could not be reliably confirmed. The incidence of major hypoglycaemia per 100 patient years was 12.0 (95% CI 8.2-13.7) in Type 1 diabetics, 6.8 (CI 5.2-8.9) in Type 2 diabetics taking insulin and 0.8 (CI 0.6-1.0) in Type 2 diabetics on oral hypoglycaemics. Fifty two patients required inpatient treatment. This compares with an incidence of 2.3 per 100 patient years for patients treated with insulin in the UKPDS trial and 5.4 in the conventional arm of the DCCT trial. The total health-care costs were £115,541 of which 64% was for inpatient stay, 25% for ambulance and 11% for emergency room. CONCLUSIONS: The incidence of hypoglycaemia was three to five fold higher than previously reported in clinical trials, and extrapolation of our costs suggests that hypoglycaemia could cost the UK NHS as much as £16.2m per year.