Our goal was to determine the association between random admission hyperglycemia and new diagnosis of diabetes after discharge in patients hospitalized with pneumonia.

Clinical data, including the Pneumonia Severity Index, were prospectively collected on all 2124 patients without diabetes admitted with pneumonia to 6 hospitals in Edmonton, Alberta, Canada. Admission glucose was classified as: normal (4.0-6.0 mmol/L, reference group) versus mild (6.1-7.7 mmol/L), moderate (7.8-11.0 mmol/L), and severe (11.1-20.0 mmol/L) stress hyperglycemia. New diagnosis of diabetes over 5 years was ascertained using well-validated criteria within linked administrative databases. Multivariable Cox models were used, and sensitivity, specificity, and likelihood ratios were calculated.

Mean age was 68 years; 1091 (51%) were male, and 1418 (67%) had stress hyperglycemia. Over 5 years, 194 (14%) with stress hyperglycemia were diagnosed with diabetes. Compared with the 45 of 706 (6%) incidences of diabetes in normal glycemia patients (4.0-6.0 mmol/L), a strong graded increase in risk of new diabetes existed with increasing hyperglycemia: mild (59 of 841 [7%]; adjusted hazard ratio [aHR] 1.09; 95% confidence interval [CI], 0.74-1.61) versus moderate (86 of 473 [18%]; aHR 2.99; 95% CI, 2.07-4.31) versus severe (49 of 104 [47%]; aHR 11.43; 95% CI, 7.50-17.42). Among moderate-to-severe hyperglycemia (≥7.8 mmol/L) patients, the sensitivity, specificity, and positive and negative likelihood ratios for new diabetes were 57%, 77%, 2.1, and 0.6, respectively, with a number-needed-to-evaluate of 5 to detect one new case of diabetes.

Moderate-to-severe random hyperglycemia in pneumonia patients admitted to the hospital is strongly associated with new diagnosis of diabetes. Opportunistic evaluation for diabetes may be warranted in this group.