Objectives: To determine whether relative hyperglycemia was associated with in-hospital mortality in critically ill patients independent of other prognostic variables and whether this association is affected by background glycemia.
Design: Prospective observational study.
Setting: Mixed medical-surgical ICU in a metropolitan teaching hospital.
Patients: From 2,617 admissions to ICU between January 27, 2016, and March 30, 2017, 1,262 consecutive patients who met inclusion and exclusion criteria were studied.
Interventions: Glycosylated hemoglobin was used to estimate average glucose concentration over the prior 3 months. Glucose concentration on ICU admission was divided by estimated average glucose concentration to calculate the stress hyperglycemia ratio, an index of relative glycemia. Risk of death score was calculated using data submitted to the Australia and New Zealand Intensive Care Society.
Measurements and main results: In this study, there were 186 deaths (14.7%). Admission glucose was significantly associated with mortality in univariate analysis (odds ratio = 1.08 per mmol/L glucose increment; p < 0.001) but not after adjustment for risk of death score (odds ratio = 1.01; p = 0.338). In contrast, stress hyperglycemia ratio was significantly associated with mortality both in univariate analysis (odds ratio = 1.09 per 0.1 stress hyperglycemia ratio increment; p < 0.001) and after adjustment for risk of death score (odds ratio = 1.03; p = 0.014). Unlike admission glucose concentration, stress hyperglycemia ratio was significantly associated with mortality in patients with glycosylated hemoglobin less than 6.5% (odds ratio = 1.08 per 0.1 stress hyperglycemia ratio increment; p < 0.001) and glycosylated hemoglobin greater than or equal to 6.5% (48 mmol/mol) (odds ratio = 1.08 per 0.1 stress hyperglycemia ratio increment; p = 0.005).
Conclusions: Unlike absolute hyperglycemia, relative hyperglycemia, as assessed by the stress hyperglycemia ratio, independently predicts in-hospital mortality in critically ill patients across the glycemic spectrum. Future studies should investigate whether using measures of relative hyperglycemia to determine individualized glycemic treatment targets improves outcomes in ICU.