PURPOSE: The international Surviving Sepsis Campaign (SSC) management bundle calls for maintaining blood glucose levels >70 and <150 mg/dL. Participating institutions submit median blood glucose levels for analysis in the SSC database. We hypothesized that median blood glucose may not be the most accurate descriptor of glucose control in critically ill patients.
METHODS: We evaluated every blood glucose measurement for patients admitted to our MICU in January, 2007. We excluded from analysis patients admitted for IL-2 or epoprostanol infusion or for diabetic ketoacidosis or non-ketotic hyperglycemia. We compared median with mean blood glucose and with time spent in the control range. For determining time spent within parameters we dichotomized the measurements as ‘in’ or ‘out’ and considered the level to remain in one state until a measurement demonstrated a change. Our target was >90% time within parameters, using a standard glucose control protocol.
RESULTS: 82 patients were admitted; 64 patients and 2,504 glucose measurements were analyzed, 38 men (59.4%) and 26 women (40.6%). Age was 53.8±16.2 years (mean±SD). The principal diagnosis was respiratory failure (45.3%). The median glucose level was 125 mg/dL (interquartile range 104–152 mg/dL). The mean glucose was 135±53 mg/dL. Glucose values were crudely normally distributed, but skewness was 2.31 and kurtosis 9.67; tests of normality failed. 17 patients (26.6%) were in control ≥90% of time; the individual mean and median values for these patients were all within parameters. 50 patients had individual means within parameters, but only 16 of them were in parameters >90% of time (32.0%). 56 had medians within parameters, 16 in parameters >90% of time (28.6%).
CONCLUSION: Glucose values are not normally distributed and are skewed toward higher levels. Both median and mean glucose levels overestimate tight glucose control in the MICU population.
CLINICAL IMPLICATIONS: Successful quality improvement requires accurate and meaningful data. Better descriptors of glucose control could refine the SSC overall survival data. Per cent of time in the control range may be a better indicator of overall success of glucose control measures.
DISCLOSURE: Steven Simpson, No Financial Disclosure Information; No Product/Research Disclosure Information