628-P: Accuracy of Continuous Glucose Monitoring (CGM) for Inpatient Diabetes Management

In the hospitalized patient with diabetes (DM), adequate glucose control requires frequent measurement of point-of-care (POC) blood glucose. Use of inpatient continuous glucose monitoring (CGM) could reduce workload on the staff, increase ease and frequency of glucose assessment, provide continuous glucose trends, and decrease nurse exposure to infectious patients, but data on accuracy of CGM in the hospital setting is limited. We conducted an observational study of 44 adult patients with DM who were admitted to our hospital non-intensive care units from July-October 2020. After placement of Abbott 14-day Freestyle Libre CGM sensors, patients continued to receive standard POC glucose measurement using the Accu-Chek device. Insulin dosing was based on POC data. 44% percent of participants were female, and 69% had type 2 DM. Median weight was 100.1 kg (IQR 89.0 - 109.5) and median age was 56.6 years (IQR 47.9 - 63.5). Relative to POC glucose measurement, the CGM mean absolute relative difference (MARD), %15/15, and %20/20, three standard measures of CGM accuracy, were 21.4%, 32.6%, and 47.9% respectively. CGM paralleled POC glucose, but CGM readings were consistently lower compared to POC by 20.9% (IQR 12.4 - 28.6). The variation did not change with POC glucose levels or sensor wear-time, though most data (82%) were from the first three days of sensor wear. On multiple regression analysis, participant age, sex, weight, DM type, baseline renal function, and admission HbA1c did not contribute to variation in individual’s MARDs. In conclusion, we report the accuracy of the Freestyle Libre CGM relative to POC measurements in hospitalized patients. CGM was lower than POC at all glucose levels, and accuracy did not change with other clinical variables measured. Future studies evaluating newer generations of CGMs relative to standardized plasma glucose measurements will further determine the safety and efficacy of using these new technologies for insulin dosing and glucose alerts in hospitalized patients. Disclosure J. J. Wright: None. A. Williams: None. R. Weaver: None. J. M. Williams: None. S. Bao: None.