Use of Continuous Glucose Monitor in Critically Ill COVID-19 Patients Requiring Insulin Infusion

An Observational Study

Eileen R. Faulds; Andrew Boutsicaris; Lyndsey Sumner; Laureen Jones; Molly McNett; Keaton S. Smetana; Casey C. May; Elizabeth Buschur; Matthew C Exline; Matthew D. Ringel; Kathleen Dungan


J Clin Endocrinol Metab. 2021;106(10):e4007-e4016. 

In This Article


Data from 19 patients were included in this analysis. CGM data were not downloaded for 1 patient because of inadvertent disposal of the transmitter; however, this patient was included in the analysis of baseline characteristics and EHR data including POC frequency. In 1 patient, the sensor displayed falsely low (40–55 mg/dL) readings over a period of 3.5 hours after initial sensor insertion and warmup despite multiple calibrations, necessitating the need for sensor replacement. CGM values from this sensor were not used clinically and were excluded from analyses. Baseline characteristics are shown in Table 1. The mean age was 58 years, 79% were male, and 47% were White. Most (84%) had preexisting type 2 diabetes, 89% had hypertension, and 78% had a smoking history. Preadmission therapies included metformin (58%) and multiple injections of insulin per day in 32%. The mean body mass index was 32 kg/m2. The mean hemoglobin A1c was 8.5%, and 22% had evidence of diabetic ketoacidosis or hyperosmolar nonketotic state at admission. More than one-half of patients had an estimated glomerular filtration rate of less than 60 mL/min/1.723 m2.

Characteristics at the time of sensor placement are shown in Table 2. The mean SOFA score was 5.8 ± 2.3, 42% were receiving dialysis, 89% of patients required mechanical ventilation, and 37% were receiving vasopressors.

Details of CGM use, insulin requirements, and frequency of POC BG measurements are shown in Table 3. The mean duration of IV insulin was 103 ± 58 hours and the mean daily dose of IV insulin on sensor day 1 was 136 ± 102 units per day. The CGM was placed by the critical care clinical nurse specialist in 8 patients, nurse manager in 4 patients, and consult team in 5 patients (the remaining 2 sensor placements were not recorded). Twelve patients used 1 sensor, whereas 7 patients had a second sensor placed. The mean duration of the first sensor was 191 hours (approximately 8 days). The sensor was replaced or removed before 10 days because of accidental dislodgement (n = 1), placement error (n = 1), accuracy concern (n = 1), and hospital discharge or patient expiration (n = 3).

The median time to sensor validation was 137 minutes (IQR 114–206). Nonadjunctive CGM glucose was used prematurely after only 1 validated sensor-meter pair, rather than the required 2 sensor-meter pairs in 2 patients. After excluding 1 sensor described previously, the sensor was validated in 17 patients within 24 hours, and the remaining patient just over 24 hours. In 9 patients, the sensor achieved validation threshold, but the nurse continued to check POC BG values for a range of 1 to 21 hours after validation. In patients requiring IV insulin, the median number of POC values was 7 (IQR 6, 16) on sensor day 1 with a significant reduction in number of POC BG values from day 1 to day 2 (P = 0.009), but not from day 2 to day 3 (P = 0.17). The reduction in POC testing represented a 71% reduction in frequency compared with the study institution's standard of care (24 POC measures/day). However, the median number of sensor values recorded in the EHR was 22 (IQR 18, 24) times per day and declined only slightly over time. The CGM value was assumed to be used for dosing titration if there was no POC BG value within 1 hour of measurement. The median number of titration values was 11.5 (IQR 0, 15) on day 1 and increased steadily thereafter. On day 1, 41% of all IV insulin titrations were performed nonadjunctively using CGM glucose; by days 2 through 5, nonadjunctive titration using CGM had risen to 63%. CGM glucose <100 mg/dL, triggering a low threshold alarm, was associated with more frequent POC monitoring (r s = 0.2216, P < 0.001). There were no reports of sensor site issues, such as bleeding, bruising, erythema, or infection identified from chart review.

Glucose control is described in Table 4. The mean sensor glucose was 170 ± 25 mg/dL on day 1 and 165 ± 18 mg/dL on subsequent days. The percent TIR (70–180 mg/dL) was 64 ± 23% on day 1 and 72 ± 16% on days 2 through 7 of sensor wear. The percent time spent >250 mg/dL was 9.6 ± 9.7% on day 1 and 6.8 ± 8.5% on days 2 through 7. The mean percent time < 70 mg/dL was 1.5% on day 1 and 0.16% thereafter. One patient had sensor glucose <54 mg/dL on day 1 (3% of time), whereas 2 patients had measurable sensor glucose <54 mg/dL on days 2 through 7 (0.3% and 1%). Interestingly, of the CGM values <70 mg/dL that had a paired POC value within 5 minutes, none of the POC values were also <70 mg/dL. Only 1 patient (with a history of type 1 diabetes, end-stage renal disease, and history of severe hypoglycemia) had measured POC blood glucose <54 mg/dL.

The overall MARD (adjusted for number of observations per patient) between the sensor and meter glucose was 13.9 ± 7.8 (median 11.9, IQR 3.3–29.4) on day 1 and 13.5 ± 8.1 (median 10.6, IQR 9.0–15.0) on days 2 through 7. The overall MAD between sensor and meter glucose was 25.0 ± 16.1 (median 19.7, IQR 14.6–26.9) on day 1 and 23.2 ± 15.5 (median 18.4, IQR 12.5–29.0) on days 2 through 7 (Table 5). Univariate linear mixed effect models examining associations between patient clinical condition and MARD appear in Table 6. Models show a negative association between body mass index and MARD (coefficient = -0.291, P = 0.007). Additionally, the presence of insulin as a home medication was associated with a 3.7% decrease in MARD (P = 0.022). Overall, there were 37 POC values < 100 mg/dL, for which 23 (62%) paired CGM values were within 20 mg/dL of the POC value and 634 POC values ≥ 100 mg/dL, for which 191 (70%) paired CGM values were within 20% (Table 7). There did not appear to be significant difference on day 1 of sensor wear compared with subsequent days. Finally, there were 55 CGM values with paired POC values that were <100 mg/dL, 15 (27%) of which were confirmed by POC BG.