OBJECTIVE : Using routine HbA1c measurement to determine the prevalence of diabetes mellitus (known and previously unrecognized) and their hospital outcomes among hematology and oncology inpatients. METHODS : This was a prospective, observational study. Routine automated HbA1c testing was performed in all hematology and oncology inpatients aged ≥54 years at a tertiary hospital, July 2013-January 2015. The outcome measures were: (i) prevalence of known and previously unrecognized diabetes, and (ii) hospital outcomes: length-of-stay (LOS), intensive-care-unit (ICU) admission, 30-day/18-month readmission, and 18-month mortality. RESULTS : Over the 18-month study period, 1076 inpatients aged ≥54 years were admitted to hematology (n = 298) and oncology (n = 778) units: 21% had known diabetes and 7% had previously unrecognized diabetes. Patients with known diabetes had a longer LOS (IRR: 1.18, 95%CI: 1.02-1.37, p = 0.03), compared to those without diabetes, adjusting for age, hemoglobin level, estimated-glomerular-filtration-rate, admission specialty unit, Charlson's comorbidity index score, and glucocorticoid exposure. No significant differences were observed in ICU admission, 30-day/18-month readmission, and 18-month mortality among patients with known, previously unrecognized and no diabetes (p ≥ 0.05). CONCLUSIONS : Approximately one in five hematology or oncology inpatients aged ≥54 years had known diabetes, and one in fourteen had previously unrecognized diabetes. Those with known diabetes had a longer hospital stay. Routine HbA1c measurement is can be useful for identifying previously unrecognized diabetes, particularly among patients with high glucocorticoid exposure. Further study is required to determine cost-effectiveness in screening for unrecognized diabetes and optimal management of these patients.