Real-World Performance of Personal Continuous Glucose Monitors During Hospitalization

Objective: To assess the accuracy of patients' personal continuous glucose monitors (CGMs) worn in the hospital and accuracy correlation with laboratory values and vital signs and to compare CGM glucometric data with data from published literature and guidelines. Method: We enrolled adult patients with diabetes mellitus wearing outpatient-inserted CGMs at the time of hospital admission to a noncritical setting. CGM readings were paired within 5 min with point of care (POC) glucometers and laboratory (Lab) blood glucose levels. CGM accuracy was expressed using mean absolute relative difference (MARD) and Clarke Error Grids. CGM accuracy variation with labs and vital signs was analyzed with Spearman's correlation method. Results: For 188 hospitalizations, we analyzed 3316 CGM-POC pairs from 101 patients (56 with Dexcom® sensors and 45 with Abbott's FreeStyle Libre® sensors) and 771 CGM-Lab pairs for 97 patients. For CGM-POC pairs, MARD was 13.7%, 14.4%, and 11.8% for all sensors, Dexcom, and Libre sensors, respectively. MARD was 22.6% for POC glucose <70 mg/dL. For CGM-Lab pairs, MARD was 13.6%, 12.7%, and 15.4% for all sensors, Dexcom, and Libre sensors, respectively. Of CGM-POC pairs, 98.7% and 98.8% of CGM-Lab pairs were in zones A and B of Error Grid Analysis. There was no correlation between ARD and daily mean arterial blood pressure, hemoglobin level, glomerular filtration rate, and pulse oximetry. CGMs' time below range (TBR), time in range (TIR), and time above range were 2.2% (standard deviation [SD]: 4.7%), 58.7% (SD: 22.5%), and 39.9% (SD: 23.4%), respectively. The coefficient of variation was 31.2%. Conclusion: Except for hypoglycemia ranges, patients' personal CGMs had adequate accuracy for glucose monitoring in the hospital. Vital signs and Lab values did not interfere with CGM accuracy. The TBR and glucose variability were low, better than outpatient recommendations. TIR was in line with inpatient consensus guidelines, and "glucometrics" were comparable with reports for hospital inserted sensors.