Objective: To investigate the nature and importance of blood glucose abnormalities in an unselected heart failure (HF) population. Design: Cohort study. Setting: Urban University hospital. Patients: All index emergency HF admissions to one University hospital during the year 2000 were studied. Results: 454 consecutive index admissions had blood chemistry, diabetic status and follow-up information recorded. 390 (86%) patients had an echocardiogram, of whom 117 (30%) had preserved left ventricular systolic function and 110 (24%) had diabetes. Sixty (13%) patients had abnormal glucose tolerance (8.0-10.99 mmol/l), and 284 (63%) patients had a normal admission blood glucose (<8 mmol/l). 51 (11.2%) patients died in hospital. After adjustment for other prognostic attributes, abnormal glucose tolerance (Cox hazard ratio HR, 95% CI: 5.920, 1.03 to 34.00; p = 0.046) but not diabetes (HR 3.46, 0.75 to 16.02; p = 0.112) predicted in-hospital mortality. During follow-up (median 812 (range 632-978) days), 104 (36.6%), 30 (50.0%) and 55 (50%) patients with a normal admission blood glucose concentration, abnormal glucose tolerance and diabetes, respectively, died (log rank test p = 0.0037, adjusted p = 0.075). Compared with patients with normal admission blood glucose, abnormal glucose tolerance (adjusted HR: 1.41 (0.92 to 2.16); p = 0.12) and diabetes (adjusted HR: 2.02 (1.41 to 2.88); p = 0.0001) predicted mortality. Considering glucose on admission as a continuous covariate, a 2 mmol/l increase was associated with a HR of 1.08 (1.03 to 1.13), p = 0.0010, which after adjustment for the above covariates became 1.08 (1.03 to 1.13), p = 0.0023. Conclusions: Admission blood glucose concentration and diabetes are prognostically important in HF and could help target some patients for more intensive therapy.