Objective: Our objective was to quantify the association between intensive care unit–acquired dysglycemia (hyperglycemia, hypoglycemia, and high variability) and in-hospital mortality.
Design: Retrospective, observational study.
Setting: eICU Research Institute participating hospitals with an active tele-ICU program between January 1, 2008, and September 30, 2010, representing 784,392 adult intensive care unit patients.
Patients: A total of 194,772 patients met inclusion criteria with an intensive care unit length of stay >48 hrs.
Measurements and Main Results: Acute Physiology and Chronic Health Evaluation IV standardized mortality ratios were calculated for dysglycemia present at admission and acquired in the intensive care unit. Intensive care unit–acquired dysglycemia was modeled using multivariable modified Poisson regression to account for confounding not incorporated in Acute Physiology and Chronic Health Evaluation. Dysglycemia severity was assessed by the relative risk of in-hospital mortality associated with the maximum, time-weighted average daily glucose; lowest glucose value throughout the intensive care unit stay; and quintiles of variability (coefficient of variation). The association of duration beyond thresholds of dysglycemia on mortality was also modeled. The adjusted relative risk (95% confidence interval) of mortality for the maximum intensive care unit average daily glucose was 1.13 (1.04–1.58), 1.43 (1.30–1.58), 1.63 (1.47–1.81), 1.76 (1.55–1.99), and 1.89 (1.62–2.19) for 110–150 mg/dL, 151–180 mg/dL, 180–240 mg/dL, 240–300 mg/dL, and >300 mg/dL, respectively, compared to patients whose highest average daily glucose was 80–110 mg/dL. The relative risk of mortality for the lowest glucose value was 1.67 (1.37–2.03), 1.53 (1.37–1.70), 1.12 (1.04–1.21), and 1.06 (1.01–1.11) for <20 mg/dL, 20–40 mg/dL, 40–60 mg/dL, and 60–80 mg/dL, respectively, compared to patients whose lowest value was 80–110 mg/dL. The relative risk of mortality increased with greater duration of hyperglycemia and with increased variability. The relative risk for the highest compared to lowest quintile of variability was 1.61 (1.47–1.78). The association of duration of hyperglycemia on mortality was more pronounced with more severe hyperglycemia.
Conclusions: The risk of mortality progressively increased with severity and duration of deviation from euglycemia and with increased variability. These data suggest that severe intensive care unit–acquired hyperglycemia, hypoglycemia, and variability are associated with similar risks of mortality.