Background: Diabetes diagnosis is delayed 4–7 years and 50% are undiagnosed. Forty percent of hospitalized patients with any blood glucose level (BGL) ≥10 mmol/L have diabetes 3 months post-discharge, yet less than 5% are detected in hospital. We review identification of, and responses to, hyperglycaemia in inpatients at a teaching hospital. Methods: The world's largest retrospective review of medical records for inpatients with venous BGL ≥11.1 mmol/L without known diabetes over 12 months (2005–2006). The primary outcome was recognition of hyperglycaemia; secondary outcomes were treatment and documentation of follow up. Logistic regression was performed with variables including BGL, admitting team, length of stay and endocrine team review. Results: Of 10 973 people screened, 162 were eligible. The median age was 58 years and BGL 13.3 mmol/L, with increased mortality and length of stay. Hyperglycaemia was noted as definitely in 26%, maybe in 24% and definitely not in 50%. Forty percent of patients were treated in hospital and 19% on discharge. Follow up was documented for 24%. A higher BGL and review by the endocrine team were strongly associated with clinical recognition on uni- and multivariate analyses. However, where an endocrine review was sought for non-hyperglycaemia reasons, similar rates of non-recognition occurred. Conclusion: Despite evidence for improved inpatient outcomes when treated, and high short-term progression to frank diabetes, inpatient hyperglycaemia remains frequently missed. In-hospital recognition is cheap, and vital for the implementation of activities to improve outcomes and prevent progression and complications. Changes to systems for checking pathology results, medical officer education and inpatient screening guidelines are indicated.