The Interaction of Acute and Chronic Glycemia on the Relationship of Hyperglycemia, Hypoglycemia, and Glucose Variability to Mortality in the Critically Ill*

Objectives: To determine the relationship between preadmission glycemia, reflected by hemoglobin A1c level, glucose metrics, and mortality in critically ill patients. Design: Retrospective cohort investigation. Setting: University affiliated adult medical-surgical ICU. Patients: The investigation included 5,567 critically ill patients with four or more blood glucose tests and hemoglobin A1c level admitted between October 11, 2011 and November 30, 2019. The target blood glucose level was 90–120 mg/dL for patients admitted before September 14, 2014 ( n = 1,614) and 80–140 mg/dL or 110–160 mg/dL for patients with hemoglobin A1c less than 7% or greater than or equal to 7% ( n = 3,953), respectively, subsequently. Interventions: None. Measurements and Main Results: Patients were stratified by hemoglobin A1c: less than 6.5.( n = 4,406), 6.5–7.9% ( n = 711), and greater than or equal to 8.0% ( n = 450). Increasing hemoglobin A1c levels were associated with significant increases in mean glycemia, glucose variability, as measured by coefficient of variation, and hypoglycemia ( p for trend < 0.0001, < 0.0001, and 0.0010, respectively). Among patients with hemoglobin A1c less than 6.5%, mortality increased as mean glycemia increased; however, among patients with hemoglobin A1c greater than or equal to 8.0%, the opposite relationship was observed ( p for trend < 0.0001 and 0.0027, respectively). Increasing glucose variability was independently associated with increasing mortality only among patients with hemoglobin A1c less than 6.5%. Hypoglycemia was independently associated with higher mortality among patients with hemoglobin A1c less than 6.5% and 6.5–7.9% but not among those with hemoglobin A1c greater than or equal to 8.0%. Mean blood glucose 140–180 and greater than or equal to 180 mg/dL were independently associated with higher mortality among patients with hemoglobin A1c less than 6.5% ( p < 0.0001 for each). Among patients with hemoglobin A1c greater than or equal to 8.0% treated in the second era, mean blood glucose greater than or equal to 180 mg/dL was independently associated with decreased risk of mortality ( p = 0.0358). Conclusions: Preadmission glycemia, reflected by hemoglobin A1c obtained at the onset of ICU admission, has a significant effect on the relationship of ICU glycemia to mortality. The different responses to increasing mean glycemia support a personalized approach to glucose control practices in the ICU.