Reliability and accuracy of bedside capillary blood glucose measurement by glucometers compared to venous blood glucose in critically ill patients: A facility based cross-sectional study.

Background: The traditional reference standard, venous blood glucose requires venipuncture and laboratories usually return the test results after 60 min. Our aim was to determine the agreement accuracy of glucose (capillary) levels obtained by POC glucometers with glucose (venous) values by standard laboratory method (glucose-oxidase) and to assess whether and to what extent the glucometers perform uniformly well across the entire range of blood glucose values.

Methods: We compared the diagnostic accuracy of two-point of care glucometers with laboratory venous glucose, the reference standard using Bland-Altman plots and Clark error grid method to analyse the results.

Results: This study included a total of 110 patients (38[34 %] women; mean age 52.1 years (SD, 17.3); range 14-85 years. Fourteen patients (12 %) were known to have diabetes. The mean glucose value (glucometer 1) was 152.9 mg/dL (SD 83.1); range = 48-501 mg/dL; that by glucometer 2 was 152.2 mg/dL (SD 76.2); range = 30-458 mg/dL and by the laboratory was 148.6 mg/dL (SD 81.5); range = 52-480 mg/dL. Of the 110 subjects, 2(2 %) had blood glucose below 70 mg/dL; 85(77 %) between 70 and 180 mg/dL and 23(21 %) had blood glucose exceeding 180 mg/dL. The Bland-Altman plot showed a bias of 4 mg% (95%CI -9.8 to +1.1); and the limits of agreement were -63 and + 54 mg%. The area under the receiver operating characteristic curve for the two glucometers was 0.92 and 0.93 respectively.

Conclusions: Point of care glucose, measured by glucometers was in agreement with the venous glucose estimation. Both glucometers were equally accurate and performed uniformly well across the wide range of blood glucose values.