Association of Hypoglycemia, Hyperglycemia, and Glucose Variability With Morbidity and Death in the Pediatric Intensive Care Unit

Abstract

OBJECTIVE. We evaluated retrospectively plasma glucose levels and the degree of hypoglycemia, hyperglycemia, and glucose variability in a PICU and then assessed their association with hospital length of stay and mortality rates. METHODS. Electronic medical records at the Packard Children's Hospital at Stanford University were reviewed retrospectively for all PICU admissions between March 1, 2003, and March 31, 2004. Patients with a known diagnosis of diabetes mellitus were excluded. The prevalence of hyperglycemia was defined with cutoff values of 110, 150, and 200 mg/dL. Hypoglycemia was defined as ≤65 mg/dL. Glucose variability was assessed with a calculated glucose variability index. RESULTS. In 13 months, 1094 eligible admissions generated 18865 glucose values (median: 107 mg/dL; range: 13–1839 mg/dL). Patients in the highest maximal glucose quintile had a significantly longer median PICU length of stay, compared with those in the lowest quintile (7.5 days vs 1 day). Mortality rates increased as patients' maximal glucose levels increased, reaching 15.2% among patients with the greatest degree of hyperglycemia. Hypoglycemia was also prevalent, with 18.6% of patients (182 of 980 patients) having minimal glucose levels of ≤65 mg/dL. There was an increased median PICU length of stay (9.5 days vs 1 day) associated with glucose values in the lowest minimal quintile, compared with those in the highest quintile. Hypoglycemia was correlated with mortality rates; 16.5% of patients with glucose levels of ≤65 mg/dL died. Glucose variability also was associated with increased length of stay and mortality rates. In multivariate logistic regression analyses, glucose variability, taken with hyperglycemia and hypoglycemia, showed the strongest association with mortality rates. CONCLUSIONS. Hyperglycemia and hypoglycemia were prevalent in the PICU. Hypoglycemia, hyperglycemia, and, in particular, increased glucose variability were associated with increased morbidity (length of stay) and mortality rates.