Accuracy of point-of-care blood glucose measurements in critically ill patients in shock.

Abstract

A widely used method in monitoring glycemic status of ICU patients is point-of-care (POC) monitoring devices. A possible limitation to this method is altered peripheral blood flow in patients in shock, which may result in over/underestimations of their true glycemic status. This study aims to determine the accuracy of blood glucose measurements with a POC meter compared to laboratory methods in critically ill patients in shock. POC blood glucose was measured with a glucose-1-dehydrogenase-based reflectometric meter. The reference method was venous plasma glucose measured by a clinical chemistry analyzer (glucose oxidase-based). Outcomes assessed were concordance to ISO 15197:2003 minimum accuracy criteria for glucose meters, bias in glucose measurements obtained by the 2 methods using Bland-Altman analysis, and clinical accuracy through modified error grid analysis. A total of 186 paired glucose measurements were obtained. ISO 2003 accuracy criteria were met in 95.7% and 79.8% of POC glucose values in the normotensive and hypotensive group, respectively. Mean bias for the normotensive group was -12.4 mg/dL, while mean bias in the hypotensive group was -34.9 mg/dL. POC glucose measurements within the target zone for clinical accuracy were 90.2% and 79.8% for the normotensive and hypotensive group, respectively. POC blood glucose measurements were significantly less accurate in the hypotensive subgroup of ICU patients compared to the normotensive group. We recommend a lower threshold in confirming POC blood glucose with a central laboratory method if clinically incompatible. In light of recently updated accuracy standards, we also recommend alternative methods of glucose monitoring for the ICU population as a whole regardless of blood pressure status.