Admission plasma glucose and diabetes mellitus in elderly admissions to hospital.

Over 6 months, all admissions to three geriatric wards were studied to define an admission plasma glucose level (APG) that identified previously undiagnosed diabetes mellitus. Subjects with APG> or =7.0 mmol l(-1) had a modified oral glucose tolerance test (OGTT) when well before discharge if their dose of steroid and/or thiazide was constant, and they were neither terminally ill nor dead; excluded were 1 subject on reducing steroid doses, and 9 moribund admissions without APG. If the first 2 h OGTT result was > or =11.1 mmol l(-1), a second OGTT was performed 6 weeks later to fulfil 1985 WHO criteria. Subjects with APG<7.0 mmol l(-1) did not have OGTT. Seventy had a previous diagnosis of diabetes; scrutiny of records and OGTT refuted the diagnosis in 5, who were excluded from further analysis. Diabetes was only commonly found among those with APG> or =8.0 mmol l(-1), and the proportion was small until APG> or =13 mmol l(-1), although even then only 47% (95% CI 21-73%) had diabetes. Fourteen of 28 subjects with initial OGTT results suggesting diabetes were not diabetic on retesting. Inpatient mortality was higher if APG> or =7.0 (Odds ratio 2.82; CI 1.63-4.89) or the subject had known diabetes (Odds ratio 2.43; CI 1.15-4.97) compared to APG<7; there was no age or sex difference between these three groups. We conclude that, unless overtly diabetic, diagnosis of diabetes in elderly medical admissions needs later confirmation.