308

Abstract

Introduction: Patient-specific factors may explain conflicting evidence of optimal glucose control in intensive care unit (ICU) patients. Glucose variability appears to correlate with mortality, but this variability may be an indicator of disease severity, rather than an independent predictor of mortality. We sought to explore the contribution of glucose variability to mortality in ICU patients. Hypothesis: Coefficient of glucose variation is an independent predictor of mortality in ICU patients, after adjustment for disease severity and comorbidities. Methods: We queried the highly detailed electronic medical record of Intermountain Healthcare, for all uses of eProtocol-Insulin. eProtocol-Insulin is an explicit, electronic protocol that manages intravenous insulin in the ICU with previously reported high physician compliance and reproducibility. We queried use of e-Protocol-Insulin over the past 5 years from 8 hospitals, excluding patients with diabetic ketoacidosis and patients who were on the protocol for less than 24 hours. We used 30-day mortality as the outcome of interest. We performed backwards elimination multivariate logistic regression, with initial covariates of age, gender, coefficient of variation (standard deviation / mean glucose), Acute Physiology and Chronic Health Evaluation II score, occurrence of hypoglycemia < 60 mg/dL, and presence of each component of the Charlson comorbidity score. We used statistical interaction terms to model the effects of hypoglycemia on coefficient of variation. Results: We studied 3923 patients. The final model included: age, coefficient of variation, hypoglycemia, APACHE II, diabetes, myocardial infarction, hemiplegia, stroke, congestive heart failure, cancer, kidney disease, lung disease, liver disease, connective tissue disease. Coefficient of variation predicts 30-day mortality in the critically ill (OR 7.4, p < 0.001), even after adjustment for age, hypoglycemia, disease severity, and comorbidities. Conclusions: Coefficient of glucose variation is associated with mortality. This association is independent of disease severity, and independent of hypoglycemia. Future studies should incorporate coefficient of glucose variation as a parameter of interest.