10 Real-World Glucose Meter Accuracy, ICU vs Non-ICU. (11th January 2018)

• Record Type: Journal Article
• Authors: Zhang, Ray; Scott, Mitchell G
• Is Part Of: American journal of clinical pathology. Volume 149(2018)Supplement 1
• Page Start: S168
• Page End: S168
• Publication Date: 2018-01-11
• Subjects: Diagnosis, Laboratory -- Periodicals; Pathology -- Periodicals; 616.07
• DOI: 10.1093/ajcp/aqx149.379
• Languages: English
• ISSNs: 0002-9173

Abstract

Abstract: In 2014, the Food and Drug Administration (FDA) proposed relegating the use of glucose meters on critically ill patients as off-label practice due to accuracy concerns. Such use defaults to high-complexity testing per Clinical Laboratory Improvement Amendments regulations, according to the Centers for Medicare and Medicaid Services (CMS). However, most hospital intensive care units (ICUs) in the US currently depend on personnel who do not meet high-complexity testing educational requirements to operate meters for glycemic control (de facto standard of care for critically ill patients). As replacing all meter operators with high-complexity qualified personnel is practically unachievable, ICUs in the United States currently face a serious logistical impasse. To investigate whether real-world accuracy of modern glucose meters is worse in ICU patients vs general inpatients, we examined all glucose tests performed at our 1, 300-bed academic hospital over a three-year period. In total, 1, 171, 007 glucose meter results from seven ICUs and 30 general inpatient units were studied. Results having a coincidental main-lab glucose test collected (via PPID) within 1 hour of meter collection were identified. Meter tests were performed using Roche Accu-Chek Inform II instruments, while main-lab tests were performed by the Roche Hitachi Modular prior to April 5, 2016 or the Roche Cobas c502/c702 subsequently. Blood samples received for testing more than a half-hour after collection were removed from analysis to minimize in vitro glycolysis artifacts, as were results >550 mg/dL (linear range of the meter). Paired meter/main-lab results were divided by patient location into "ICU" or "non-ICU" and by time between meter and main-lab collections. Meter accuracy was evaluated using the Parkes consensus error grid as well as current FDA criteria for acceptable performance in split-sample studies for manufacturers seeking hospital-use clearance. When collection times were within 1 minute, the percent of meter tests passing the FDA-98% criterion was 95% for ICU and 92% for non-ICU. The percent passing the FDA-95% criterion was 89% and 85%, respectively. ICU location was associated with higher meter accuracy by both criteria (χ 2 P < 10 –15 ). No meter result fell within consensus error grid zones D or E. Within five minutes, the percent of meter tests passing the FDA-98% criterion was 93% for ICU and 90% for non-ICU. The percent passing the FDA-95% criterion was 86% and 84%, respectively. ICU location was again associated with higher meter accuracy by both criteria ( P < 10 –15 ). As time between meter and main-lab collection increased, the percent of meter results passing either FDA criteria decreased but ICU meter accuracy remained consistently higher than non-ICU meter accuracy (Wilcoxon signed rank P < 10 –10 ). Glucose meters were no less accurate (in fact, statistically more accurate) in ICUs than in non-ICUs. While the current FDA and CMS stances are well intentioned, disrupting glucose meter workflow in ICUs due to the inability to meet high-complexity requirements appears more likely to do harm than good.