Insulin protocols are increasingly used in ICUs to achieve desired glucose ranges while minimizing practice variation and potentially, harm to patients. Glucose levels may vary, however, with different testing methods, and these differences may be clinically significant. Point of care testing with glucometers is now widely performed, however, the adequacy of this method in critically ill patients is largely unknown. The purpose of this study is to evaluate the variations in recommendations given by a computerized insulin protocol to glucose levels obtained by different testing methods.
Patients admitted to the ICU with shock (SBP < 90 mm Hg despite adequate volume resuscitation or requiring vasopressor therapy) were evaluated. When indicated, an arterial blood sample was obtained and glucose was measured using a glucometer (ACCU-CHEK Comfort Curve, Roche), as well as in the clinical laboratory (“gold standard”) using the colorimetric plasma glucose analyzer (VITRIOS). A simulation was subsequently run using this patient abstracted data on our computerized insulin protocol. The computerized protocol has been implemented in our ICU since January of 2003 and targets a glucose level of 80 to 110 mg/dl. The protocol uses the current blood glucose level as well as the rate of decline in glucose to determine the insulin infusion rate.
Simulations were run on glucose results from 21 patients (61 total samples). The protocol treatment recommendation obtained for each glucose level determined by the glucometer was compared to the recommendation obtained for the corresponding glucose level determined by the clinical lab. See table for the results. Two out of the 21 patients in our simulation had a difference of greater than 1 unit/hour of insulin infusion rate, between the recommendations.
Blood glucose determinations made with a glucometer can result in patients receiving higher doses of insulin, per our simulation.
While our study shows statistical significance in variations in insulin dosage, if a glucometer is used, the clinical significance is uncertain.
ResultsProtocol recommendationDifferencePGlucometer-basedClinical lab-basedTotal insulin (units)Mean15.1612.692.470.0025SD12.8411.763.19Range0 to 42.180 to 41.22-0.8 to 13.41Insulin rate (units/hour)Mean2.091.740.310.0017SD1.521.430.44Range0 to 4.820 to 4.71-0.12 to 1.55
Srinivas Chakravarthy, Grant monies (from sources other than industry) This study was supported by an award from The CHEST Foundation of the American College of Chest Physicians and Ortho Biotech Products, LP.; Grant monies (from industry related sources) The glucometer, chemistry strips and reagents were provided by Roche.