Pilot study of the accuracy of bedside glucometry in the intensive care unit

Objective. To evaluate the accuracy of bedside glucometry among critically ill patients. Design: Prospective audit, with sequential specimen collection. Setting: Thirty-two-bed cardiovascular, neurosurgical, and medical-surgical intensive care unit in a single Canadian center. Patients. Ten critically ill adults, who were sequentially followed during their intensive care unit stay. Eight had diabetes mellitus, and three were in shock. Measurements and Main Results, Repeat arterial blood samples were obtained by the attending nurse, who withdrew a portion of the sample and performed reflectance glucometry at the bedside. The remainder was immediately sent i n a vacuum-sealed plasma separation tube to the hospital laboratory, and an alyzed using a conventional plasma glucose analyzer by a laboratory technol ogist. Sequential samples were taken at intervals of at least 12 hrs of one other. A total of 105 arterial glucose pairs were obtained. There was a si gnificant correlation between the laboratory and glucometry determined gluc ose concentrations (Intraclass correlation coefficient = 0.86, p < .0001). The overall average laboratory-glucometry glucose difference was -0.04 mmol /L (95% confidence interval [CI] -2.3-2.2 mmol/L). Five out of 105 values ( 4.8%) lay beyond these confidence bounds. The square of the Pearson correla tion coefficient (?) between the mean glucose level and the laboratory-gluc ometry glucose difference was not significant (0.01, 95% CI 0.005-0.04; p = .22), suggesting the absence of any trend between rising glucose concentrat ion and the laboratory-glucometry difference. Conclusions., Bedside glucose testing of arterial whole blood samples may b e an accurate alternative to laboratory plasma glucose measurement among cr itically ill adults, within approximately 2.3 mmol/L of certainty. Because previous studies have suggested that this bedside technique may be prone to a moderate degree of error among patients in shock as well as those with a n abnormal blood pH or hematocrit larger studies are needed to confirm our findings.