We also assessed whether the association between glycemic variability and mortality was independent of hypoglycemia and other patient attributes. Methods then Data collection Using Intermountain Healthcares electronic medical record, we performed a retrospective cohort analysis of all patients supported with eProtocol insulin from November 2006 to August 2012. Patients were drawn from 14 different ICUs from 8 different hospitals. These open ICUs included medical, surgical, and mixed patient populations, teaching and non teaching ICUs. Diagnostic categories regarding reason for admission were not assessed, although we excluded patients with diabetic ketoacidosis, as we believe they comprised a different patient experience than the typical ICU patient on intravenous insulin for blood glucose management.

Similarly, we also excluded patients supported with eProtocol insulin Inhibitors,Modulators,Libraries for 24 hours or with 10 blood glucose measurements. We included only a patients first ICU admission during the study period. We stratified the data by presence or absence of diabetes mellitus, determined by ICD 9 code. We calculated the acute physiology component of the acute physiology and chronic health evaluation II score to avoid collinearity in our regression model. We calculated the Charlson Comorbidity index using ICD 9 codes. Blood glucose management in Intermountain Healthcare ICUs includes standardized institutional processes. All ICUs used the OneTouch SureStep bedside glucose meter until 2010, when all facilities Inhibitors,Modulators,Libraries switched to the HemoCue glucose meter. All glucose meters were calibrated nightly.

All ICUs administered intravenous Inhibitors,Modulators,Libraries regular insulin using a smart pump in plastic tubing with a 1unit mL concentration. The Inhibitors,Modulators,Libraries time interval of blood glucose measurements was explicitly determined by eProtocol insulin, based on glucose stability. All study ICUs have a 2 1 patient nurse ratio. eProtocol insulin does not contain detailed nutrition rules but adjusts recommendations based on glucose calories above or below ? the estimated basal caloric requirement. Both the decision to initiate eProtocol insulin and the blood glucose target was determined by the individual practitioner at a target of either 95 mg dL, or at 115 mg dL. Shortly after the publication of NICE SUGAR, most physicians switched from 95 mg dL to 115 mg dL target. eProtocol insulin discontinues Inhibitors,Modulators,Libraries IV insulin and recommends concentrated IV dextrose when blood glucose is 60 mg dL. eProtocol insulin support is discontinued if patients are receiving bolus feeds. Statistical cause analysis Given the non parametric distribution of the data, we compared central tendencies between groups using Mann Whitney U tests, and compared proportions using Chi squared tests.