Continuous Glucose Monitoring in Critically Ill Patients With COVID-19: Results of an Emergent Pilot Study

Abstract

Background:

Amidst the coronavirus disease 2019 (COVID-19) pandemic, continuous glucose monitoring (CGM) has emerged as an alternative for inpatient point-of-care blood glucose (POC-BG) monitoring. We performed a feasibility pilot study using CGM in critically ill patients with COVID-19 in the intensive care unit (ICU).

Methods:

Single-center, retrospective study of glucose monitoring in critically ill patients with COVID-19 on insulin therapy using Medtronic Guardian Connect and Dexcom G6 CGM systems. Primary outcomes were feasibility and accuracy for trending POC-BG. Secondary outcomes included reliability and nurse acceptance. Sensor glucose (SG) was used for trends between POC-BG with nursing guidance to reduce POC-BG frequency from one to two hours to four hours when the SG was in the target range. Mean absolute relative difference (MARD), Clarke error grids analysis (EGA), and Bland-Altman (B&A) plots were calculated for accuracy of paired SG and POC-BG measurements.

Results:

CGM devices were placed on 11 patients: Medtronic (n = 6) and Dexcom G6 (n = 5). Both systems were feasible and reliable with good nurse acceptance. To determine accuracy, 437 paired SG and POC-BG readings were analyzed. For Medtronic, the MARD was 13.1% with 100% of readings in zones A and B on Clarke EGA. For Dexcom, MARD was 11.1% with 98% of readings in zones A and B. B&A plots had a mean bias of −17.76 mg/dL (Medtronic) and −1.94 mg/dL (Dexcom), with wide 95% limits of agreement.

Conclusions:

During the COVID-19 pandemic, CGM is feasible in critically ill patients and has acceptable accuracy to identify trends and guide intermittent blood glucose monitoring with insulin therapy.

Keywords

continuous glucose monitoring COVID-19 critically ill hospital inpatient intensive care unit

Introduction

Coronavirus disease 2019 (COVID-19) is a life-threatening infection caused by severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2). Globally, there are 20,950 402 confirmed cases and 760,213 COVID-19-related deaths as of August 15, 2020.¹ In prior pandemics, such as severe acute respiratory syndrome,^2,3 Middle East respiratory syndrome,³ and influenza A (H1N1),³ diabetes was considered an independent risk factor. With COVID-19, the Centers for Disease Control and Prevention reports that diabetes is an underlying condition in 30% of US patients⁴ and is associated with threefold increased hospitalization risk⁵ and twofold risk of high severity,^6,7 poor outcome, and mortality.⁷ Because of numerous factors, including stress imposed by COVID-19 infection, critical illness, and frequent glucocorticoid use, patients with COVID-19 and diabetes can develop severe hyperglycemia or diabetic ketoacidosis, requiring intensive insulin therapy by the intravenous (IV) route. Optimal glucose control in hospitalized patients with COVID-19 is associated with markedly lower mortality compared with poor control of blood glucose.^6-8

A large, retrospective study in Wuhan, China, involving 7,337 patients noted lower all-cause mortality in patients whose blood glucose was maintained between 70 and 180 mg/dL. Increased mortality was seen in patients who were outside this range.⁶ A US retrospective study reported increased length of stay and mortality in patients with diabetes and hyperglycemia (>180 mg/dL), further emphasizing the importance of tight glycemic control.⁸

IV insulin infusion is the standard of care for hyperglycemia management in the intensive care unit (ICU); however, this protocol requires frequent blood glucose measurements every one to two hours. This presents a significant challenge amid the COVID-19 pandemic because of increased personal protective equipment (PPE) use and increased transmission risk to health care providers (HCP).To address these concerns, the US Food and Drug Administration (FDA) issued a policy in March 2020 to expand the availability and capability of noninvasive remote monitoring devices to reduce patient and HCP exposure during the current pandemic. This policy allowed the use of continuous glucose monitoring (CGM) systems in hospitals to assist with glucose monitoring in patients with COVID-19.^9,10 CGM use in the ICU provides a critical advantage of near-continuous blood glucose tracking, allowing real-time monitoring and out-of-range alarms, thereby reducing the need for frequent bedside blood glucose testing and decreasing exposure risks for HCP.

There have been several CGM trials in critically ill patients in the ICU, but only a few were randomized controlled trials.^11-15 Most trials focused on accuracy, but some addressed reliability, time to euglycemia, and mortality.^16,17 While there has been no convincing evidence that the use of CGM in critically ill patients leads to improved patient outcomes, this monitoring method reduces hypoglycemia^14,15 and is without major safety concerns.^16-18 Nonetheless, a consensus statement on CGM pointed to several challenges impeding its use in critically ill patients. Patient factors such as tissue perfusion, mechanical shear forces, biofilm deposits, and medications (eg, heparin, acetaminophen, dopamine) could cause erroneous readings. Other potential barriers include HCP ease of use, nurse acceptance, and require periodic calibration. However, the consensus panel agreed that special patient populations would benefit from CGM to detect real-time glucose trends, particularly ICU patients with high risk for glucose variability and hypoglycemia, those receiving glucocorticoids, and those with organ failure.¹⁸ These are all common factors for critically ill patients with COVID-19.

Anecdotal reports describe the increased use of CGM in many US hospitals during the COVID-19 pandemic. However, there is limited data regarding CGM in the ICU setting where patients are frequently managed with IV insulin, requiring a more frequent blood glucose measurement. Our aim was to evaluate the feasibility of using CGM in critically ill patients with COVID-19 for real-time sensor glucose (SG) trends with intermittent point-of-care blood glucose (POC-BG) testing to guide insulin therapy.

Methods

The study was approved by the Quality Department and Houston Methodist Institutional Review Board. Data for this retrospective project were obtained from electronic health records and bedside documentation.

Patients ≥18 years with a positive COVID-19 reverse transcription-polymerase chain reaction (RT-PCR) test admitted to the ICU with a history of diabetes or hyperglycemia (>300 mg/dL) and requiring IV or subcutaneous insulin therapy were included in this study. Patients were excluded if they: (1) had diabetic ketoacidosis/hyperosmolar hyperglycemic syndrome, pH < 7.0, or anasarca; (2) used >2 vasopressors or extracorporeal membranous oxygenation (ECMO); (3) were pregnant; or (4) required ≤48 hours of ICU care. Patients on dialysis or enteral/parenteral nutrition and those using potentially interfering medications such as acetaminophen, vitamin C, hydroxyurea, or albuterol were not excluded. All patients were enrolled and managed by the endocrinology and ICU teams, collaboratively.

A total of 11 patients were placed on one of two CGM devices, the Medtronic Guardian Connect or Dexcom G6. Individual patient accounts were created per manufacturer’s requirements. Sensor insertion on the upper arm or abdomen was performed by the nurse who was trained by the manufacturer’s instructions and monitored by an endocrinologist at the bedside. Bluetooth transmission and the manufacturer’s applications were used to connect sensors to either an iPad or iPhone outside the room to display real-time SG. Nurses were trained on application features, alarms/alerts, and calibrations (Medtronic only) and provided with references at the bedside. In the first 24 hours after sensor insertion, the SG was documented but not used for clinical management. All POC-BG values were obtained as per the usual practice using capillary, venous, or arterial blood samples, and measured at the bedside on a Roche Accu-Chek Inform II POC-BG monitor.

To ease the burden of hourly POC-BG for all patients with COVID 19, we modified our institution’s usual insulin infusion protocol from a target range of 140-180 mg/dL to a broader target of 100-200 mg/dL and allowing POC-BG up to every two hours. With CGM, POC-BG monitoring could be reduced to every four hours if the hourly SG was in range and without system alarms/alerts. If an alarm/alert was noted, POC-BG was performed immediately. If the POC-BG was above or below 100-200 mg/dl, then POC-BG testing increased to hourly until two consecutive readings were in range again. To address concerns about accuracy and patient safety, changes in insulin dosing were made using only POC-BG measurements. A log sheet was provided to document hourly SG, POC-BG, and administration of known potential sensor interferents per the device labels, including acetaminophen, vitamin C, hydroxyurea, and albuterol.¹⁹ CGM sensors were removed for computed tomography (CT) or magnetic resonance imaging (MRI) procedures.

Statistical Analysis

Baseline characteristics were summarized for each CGM system. Data were presented as median, (interquartile range [IQR]: 25th-75th percentiles) for continuous variables and as number and percentage for categorical variables. Fisher’s exact test (for categorical variables) and the Mann-Whitney test (for continuous variables) were used to compare patients between the CGM systems. The POC-BG was used as the reference glucose in all comparative data analyses. Mean absolute relative difference (MARD) was calculated as the absolute difference between time-matched measurement and reference, divided by reference value, multiplied by 100. MARD was compared between SG and POC-BG overall and in three clinically pertinent POC-BG ranges: <70 mg/dL, 70-180 mg/dl, and >180 mg/dL. The Wilcoxon signed-rank test was used to compare SG and POC-BG values. Concordance correlation coefficient for agreement by Lin’s method^20,21 was used to measure both precision (Pearson’s correlation coefficient) and point accuracy. Point accuracy is a measure of how far a line of perfect concordance deviates from a 45-degree angle through the origin with a value of one equaling no deviation. Agreement was also illustrated using the Bland-Altman (B&A) plot²² and the Clarke error grid analysis (EGA).²³ Statistical significance was defined as two-tailed, P < .05. All analyses were performed with STATA version 16 (StataCorp LLC, College Station, TX) and R version 3.6.3 (R Foundation for Statistical Computing, Vienna, Austria).

Results

A total of 437 paired SG and POC-BG readings were obtained from 11 critically ill patients with COVID-19 on intravenous and/or subcutaneous insulin therapy (Medtronic. n = 6 and Dexcom. n = 5). SG and POC-BG were paired when within 0-30 minutes of each other. This interval was necessary to accommodate the delays in nursing documentation in the bedside log between completing the bundled tasks in the patient’s room and subsequently documenting the SG reading located outside the room. Baseline characteristics are described in Table 1. Eight patients had type 2 diabetes, one patient had type 1 diabetes, one had post-transplant diabetes, and one had prediabetes. Median A1C at admission was 7.9% (IQR: 7.10%-10.20%). Use of insulin and noninsulin diabetes medications prior to admission was similar between the CGM devices. The median initial Sequential Organ Failure Assessment (SOFA) score (a calculation of the severity of organ dysfunction for critically ill patients used to predict mortality ranging 0-24 where a score of ≤9 predicts mortality of up to 33% and ≥11 of 95% mortality)²⁴ was eight for the entire cohort and considered moderate but not statistically significant. One patient was on ECMO, but not at the time of sensor use. Comorbidities of obesity, hypertension, coronary artery disease, cancer, and organ transplantation were similar between CGM devices. The Medtronic group had higher median ventilator days (37.5 vs 14.0, P = .034). ICU and hospital length of stay did not differ. One patient died during hospitalization.

Table 1.

Baseline Characteristics.

	Total (n = 11)	Dexcom (n = 5)	Medtronic (n = 6)	P value
Age (years)	63.00 (36.00-72.00)	58.00 (54.00-63.00)	68.50 (36.00-72.00)	.36
Male	5 (45.45)	3 (60.00)	2 (33.33)	.57
Race/ethnicity
African-American	2 (18.18)	0 (0.00)	2 (33.33)	.061
Asian	2 (18.18)	0 (0.00)	2 (33.33)
Caucasian	7 (63.64)	5 (100.00)	2 (33.33)
Body mass index (kg/m²)	33.02 (24.80-43.18)	42.90 (35.79-43.18)	26.64 (23.60-33.02)	0.14
Hypertension	10 (90.91)	4 (80.00)	6 (100.00)	0.45
Diabetes mellitus
Type 1	1 (9.09)	1 (20)	0 (0)	1.0
Type 2	8 (72.73)	4 (80)	4 (66.67)
Post-transplant diabetes mellitus	1 (9.09)	0 (0)	1 (16.67)
Prediabetes	1 (9.09)	0 (0)	1 (16.67)
Coronary artery disease	2 (18.18)	1 (20.00)	1 (16.67)	1.00
Cancer	2 (18.18)	0 (0.00)	2 (33.33)	.45
Organ transplant	2 (18.18)	1 (20.00)	1 (16.67)	1.00
Smoker	1 (9.09)	0 (0.00)	1 (16.67)	1.00
Preadmission insulin	5 (45.45)	3 (60.00)	2 (33.33)	.57
Preadmission noninsulin diabetes medications	6 (54.55)	3 (60.00)	3 (50.00)	1.00
Hgb A1c (%)	7.90 (7.10-10.20)	8.40 (7.90-9.90)	7.65 (7.10-10.20)	.58
Preadmission ACEI or ARB therapy	7 (63.64)	4 (80.00)	3 (50.00)	.55
# of ventilator days	28.00 (14.00-45.00)	14.00 (12.00-17.00)	37.50 (29.00-47.00)	.034
ECMO	1 (9.09)	1 (20.00)	0 (0.00)	.45
ICU LOS (days)	27.00 (16.00-30.00)	18.00 (16.00-27.00)	29.00 (25.00-45.00)	.17
Hospital LOS (days)	30.00 (28.00-38.00)	30.00 (30.00-31.00)	33.50 (25.00-48.00)	.85
SOFA score	8 (5-9)	7 (5-8)	8.5 (5-11)	.52

Data are presented as median (25th, 75th percentiles) for continuous variables, and n (%) for categorical variables. Fisher’s exact test (categorical variables) and Mann-Whitney test (continuous variables) are used to compare subjects between Dexcom and Medtronic. ACEI, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blockers; ECMO, extracorporeal membranous oxygenation; HgbA1c, glycated hemoglobin; ICU, intensive care unit; LOS, length of stay; SOFA, Sequential Organ Failure Assessment.

Feasibility was assessed by the ease of sensor insertion, successful remote display of real-time SG data, utility and ease of app use, extrinsic reasons for data gaps, need for early sensor removal, and complications from sensor placement. Both systems required a tedious initial setup to create individual accounts on the manufacturer’s cloud-based platforms. However, sensor insertion was done easily within five minutes and display set up within ten minutes. The Medtronic sensor required more steps and at least twice daily calibrations, whereas the Dexcom insertion was simpler with no calibrations. The applications for each system differed in the information displayed, with a key advantage of the Medtronic application which allows concurrent entry of POC-BG and superimposes both POC-BG and SG readings on the display (Figure 1).²⁵ One patient was noted to have minimal bleeding at the insertion site after sensor removal.

Figure 1.

Real-time sensor data displayed outside the patient room. Upper left: Dexcom G6 on an iPhone. Upper right: Medtronic Guardian Connect on iPad. Bottom: Medtronic display of blood glucose (red teardrops) superimposed on sensor glucose (blue line graph).

Reliability was evaluated qualitatively by documenting data gaps in either transmission to the display device or early sensor failure which required removal/replacement. Both systems were successful in meeting our initial aim of transmitting real-time SG data to the remote display device and allowing the trending of POC-BG and SG values (Figure 2). Dexcom patients did not experience any data gaps or sensor failure. In the Medtronic group, one display lost communication with the sensor, and two patients required early sensor replacement. Sensor data were interrupted or removed for one patient during CT and MRI. One patient, in whom blood pressure cuff placement was adjacent to the sensor, showed severe deviations in SG that were corrected immediately after changing the cuff placement.

Figure 2.

Examples of a patient on each of the continuous glucose monitoring (CGM) systems with paired sensor glucose and blood glucose measurements graphed over time to demonstrate superimposed trending.

Accuracy was assessed using several methods: MARD, Clarke EGA, and B&A plots. The overall MARD in all glucose ranges was 13.1% for Medtronic and 11.1% for Dexcom (P = .13), with a lower MARD noted for both when POC-BG was >180 mg/dL (12.6% and 9.5%, respectively). There were insufficient data points to accurately assess MARD for POC-BG <70 mg/dL. Although no acceptable MARD has been defined for these specific CGM systems in critically ill patients, Wollersheim et al suggest that CGM with a MARD of <14% is acceptable and a MARD >18% represents poor accuracy.^26,27 By this threshold, the overall MARD for both systems was acceptable, but Dexcom performed better than Medtronic. The concordance correlation coefficients were 0.79 and 0.89, and Pearson’s correlations coefficients and point accuracies were 0.84 and greater in both systems, with Dexcom having better performance than Medtronic (Table 2).

Table 2.

MARD, Concordance Correlation Coefficient, Pearson’s Correlation Coefficient, and Point Accuracy for Agreement Between Sensor Glucose and POC Blood Glucose (Reference) for Two Systems.

	Number of measurements (n)	MARD (%)	Concordance correlation coefficient	Pearson’s correlation coefficient	Point accuracy
Medtronic
Overall	238	13.10 ± 10.96	0.79	0.84	0.94
POC blood glucose
<70	0
70-180	114	13.62 ± 11.05	0.65	0.69	0.94
>180	124	12.62 ± 10.90	0.58	0.71	0.81
Dexcom
Overall	199	11.09 ± 8.42	0.89	0.90	0.99
POC blood glucose
<70	8	18.12 ± 11.96	−0.06	−0.81	0.07
70-180	92	12.19 ± 8.63	0.75	0.75	0.99
>180	99	9.49 ± 7.50	0.69	0.72	0.96

Data were mean ± SD.

MARD, mean absolute relative difference; POC, point-of-care.

The Clarke EGA was used to demonstrate the clinical accuracy of both systems. For Medtronic, 74.4% of readings were in zone A, and 100% were within zones A and B (Figure 3). For Dexcom, 86.4% of readings were in zone A, 98.0% were within zones A and B, and 2.0% were in zone D (Figure 4). Both systems performed well, with >98% of readings in the acceptable zones of A and B.

Figure 3.

Clarke error grid analysis—Medtronic.

Figure 4.

Clarke error grid—Dexcom.

The B&A plots also illustrate the agreement between SG and POC-BG. For Medtronic, the mean bias was −17.76 mg/dL with 95% limits of agreement between −72.91 mg/dL and +37.40 mg/dL, whereas for Dexcom, the mean bias was −1.94 mg/dL (P < .001), with 95% limits of agreement between −48.74 mg/dL and +44.86 mg/dL (Figures 5 and 6).

Figure 5.

Bland-Altman plot—Medtronic.

Figure 6.

Bland-Altman plot—Dexcom.

To evaluate the role of CGM in reducing HCP exposure, workload, and PPE use, the number of POC-BG measurements performed was analyzed in eight of the 11 patients who had CGM for 72 hours. The first 24 hours (day 0) were used as the baseline without SG, and POC-BG frequency was as usual care. Day 0 was compared with days 1 and 2 when SG was used to reduce POC-BG testing. While guidelines were provided on when to reduce POC-BG testing, there was nonuniformity in nursing compliance, often with excess POC-BG. Despite this, we saw a 33.11% overall reduction by day 2 (P = .023) when compared with day 0 (Table 3).

Table 3.

Number of POC-BG Glucose Measurements for Each Day After Sensor Insertion.

	Day 0	Day 1	Day 2	% reduction from day 0 to day 1	% reduction from day 0 to day 2
Overall	14.18 ± 5.60	12.36 ± 4.82	8.75 ± 3.92^a,b	6.7 ± 29.7	33.11 ± 29.03^a
Dexcom	13.4 ± 5.41	10.8 ± 3.96	6.67 ± 0.58	9.62 ± 38.83	52.2 ± 11.09
Medtronic	14.83 ± 6.18	13.76 ± 5.43	10 ± 4.64	4.33 ± 23.30	21.65 ± 31.24

Data were presented as mean ± SD. Day 0, first 24 hours after sensor insertion, with usual care frequency of POC-BG; days 1 and 2, guidance to decrease POC-BG if SG in target glucose range.

P = .023 compared with day 0.

P = .031 compared with day 1.

POC-BG, point-of-care blood glucose.

We did not perform a formal nursing survey, but both systems were generally well accepted. On daily clinical rounds, nurses did not cite sensor insertion or maintenance as burdensome. The ability to monitor hypoglycemia and hyperglycemia trends between POC-BG testing was frequently viewed as a benefit, even more than the reduced POC-BG testing. A few nurses remained skeptical about the absolute value differences between SG and POC-BG and continued with the usual frequency of POC-BG monitoring, but there were frequent requests to expand CGM to more patients.

Discussion

With the FDA’s allowance to expand CGM use to the inpatient setting, we have a unique opportunity to learn the benefits and limitations of this technology in real-world practice. While a few other hospitals have employed CGM technology emergently in critically ill patients with COVID-19 as we did, there is still very limited data available. Our aim was not to replace POC-BG entirely but to use SG trends between POC-BG for optimal glycemic control while reducing POC-BG frequency, HCP exposure, and PPE use with fewer POC-BG when SG was in range. In this respect, the pilot study was successful and demonstrated adequate CGM feasibility, reliability, and accuracy for real-time glucose trending in highly complex patients. We demonstrated a 33.11% reduction in POC-BG when CGM was used to guide monitoring. Overall, Dexcom provided a more favorable experience and better accuracy without the need for daily calibrations compared with Medtronic. Although a formal nursing survey was not completed, acceptance was high with easy incorporation into the workflow.

While it was rewarding to adapt the CGM systems in this emergent setting, a more promising outcome was the accuracy demonstrated by MARD and Clarke EGA. The MARD of 11.1% in Dexcom and 13.1% in Medtronic, along with 98% and 100% of Clarke EGA readings, respectively, in zones A and B, suggest that the current sensor technology has overcome many of the issues that caused it to fail accuracy expectations in prior clinical trials of critically ill patients. Additionally, these accuracy results were in patients with a moderately high mean SOFA score and included conditions that cause significant glucose variability including hypoxia, acidosis, fever, corticosteroid therapy, vasopressor therapy, dialytic therapy, and enteral and parenteral nutrition. We did not restrict the use of medications known to interfere with sensor accuracy such as acetaminophen with Medtronic and hydroxyurea with Dexcom. Many studies would have excluded these factors, but we aimed to include a realistic population of patients generally managed in our ICU, in real-world care. In CGM outpatient studies, MARD is reported at 9.0% for Medtronic with two calibrations and 9.9% for Dexcom G6.^28-30 While there are no guidelines for a threshold MARD that is considered acceptable in critically ill patients, one recent letter published about Dexcom G6 use in noncritically ill COVID-19 inpatients reported a MARD of 9.77% using capillary blood glucose as the reference.³¹ In the complicated, critically ill patient with COVID-19, a higher MARD would be expected than in outpatient studies or inpatient studies in noncritically ill patients. Given this context, the accuracy of the results we found was above our expectations. While they may not yet meet the high standards required for FDA approval in ICU patients, they are sufficient for our main objective in this pandemic, which is to monitor glucose trends, reduce HCP exposure, and reduce PPE use.

Another area of controversy in CGM trials has been the reference glucose source. It is well established that different methods used to measure blood glucose result in significant differences which is a valid concern, especially in clinical trials of glycemic management. It is also recognized that the sampling site (venous, capillary, or arterial) and measurement method (blood gas analyzer, point-of-care blood glucose monitor, central laboratory) can introduce errors.³² Additionally, there are limited manufacturers of POC-BG monitors that have an FDA indication for use in critically ill patients.³³ However, many hospitals, including our own, do not use such a manufacturer and instead have performed internal validations and modified practices to minimize potential sources of error in this patient population. The routine clinical practice uses glucose monitors with a variety of sampling sites, including venous, arterial, or capillary blood. In this regard, it is most practical to compare SG to the usual practice; therefore, we chose POC-BG as our reference.

It was clear from our study that the current CGM systems need further modifications for universal use. Our project utilized training and supervision by an endocrinologist daily, which is not a readily available resource, especially on a large scale. Another critical advancement, which we did not have on an emergent basis, is incorporation into electronic health records for ongoing quality review.

Limitations of our study include small sample size and unrandomized assignment of patients to each group. Patients were assigned to a CGM system based on the inclusion/exclusion criteria and sensor availability. To simplify nurse training, we designated each system to different ICUs, one surgical and the other medical, but both for only patients with COVID-19. This could have introduced bias from the broader clinical practices of intensive care management, especially those that affect glucose control as mentioned previously. We were unable to evaluate a benefit in patient outcomes or cost-effectiveness, which is the main priority for technology adoption.

If we are to achieve the desired glycemic control shown to benefit critically ill patients with COVID-19 as well as protect the HCP’s risk of transmission and workload while preserving PPE, CGM has significant value. Our findings support the urgent need for randomized controlled trials using CGM in this complex patient population, especially given the significant ICU resource constraints during the COVID-19 pandemic.

Conclusions

Our experience with the emergent adoption of CGM systems in critically ill patients with COVID-19 in the ICU proved to be feasible, reliable, and accurate for real-time glucose trending and as an adjunct to POC-BG to reduce HCP exposure and PPE use during this pandemic. However, further modifications of the technology are needed to readily incorporate CGM use into routine ICU clinical protocols. Larger randomized controlled trials are needed to validate our findings and to demonstrate patient outcome benefits and/or cost-effectiveness.

Footnotes

Acknowledgements

The authors acknowledge Robert Phillips, MD, PhD, for his leadership and support of this study; Abhishek Kansara, MD as a clinical team member, Anisha Gupte and Melissa Whipple for study coordination; Samridhi Syal for assistance in data collection. We especially thank our ICU intensivists and staff for their collaboration and dedication: Deepa Gotur, MD, Daniela Moran, MD, Michael Sirimaturos, PharmD, and the ICU nursing staff.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: study was supported by the Division of Diabetes and Endocrinology, Department of Medicine, Houston Methodist Research Institute and Houston Methodist System Quality Department. Sensor equipment and technical support was provided by Dexcom and Medtronic but they had no role in study design, data collection, or analysis.

ORCID iDs

Archana R. Sadhu

Tariq Nisar

References

John Hopkins University & Medicine. COVID-19 dashboard by the Center for Systems Science and Engineering (CCSE) at John Hopkins University (JHU). Coronavirus Resource Center. 2020. https://coronavirus.jhu.edu/map.html. Accessed August 15, 2020.

Zhao

, et al. Genomic characterization and epidemiology of 2019 novel coronavirus: implications for virus origins and receptor binding. Lancet. 2020;395(10224):565-574.

Singh

Gupta

Ghosh

Misrab

Diabetes in COVID-19: prevalence, pathophysiology, prognosis and practical considerations. Diabetes Metab Syndr. 2020;14(4):303-310.

Stokes ES. Coronavirus disease 2019 case surveillance: United States, January 22-May 30, 2020. Centers for Disease Prevention. June 15, 2020. https://www.cdc.gov/mmwr/volumes/69/wr/mm6924e2.htm. Accessed August 15, 2020.

Centers for Disease Control and Prevention. COVID-19 associated hospitalization related to underlying medical conditions. Centers for Disease Control and Prevention. August 10, 2020. https://www.cdc.gov/coronavirus/2019-ncov/covid-data/investigations-discovery/hospitalization-underlying-medical-conditions.html. Accessed August 15, 2020.

Zhu

She

Cheng

, et al. Association of blood glucose control and outcomes in patients with COVID-19 and pre-existing type 2 diabetes. Cell Metab. 2020;31(6):1068-1077.e3.

Kumar

Arora

Sharma

, et al. Is diabetes mellitus associated with mortality and severity of COVID-19? A meta-analysis. Diabetes Metab Syndr. 2020;14(4):535-545.

Bode

Garrett

Messler

, et al. Glycemic characteristics and clinical outcomes of COVID-19 patients hospitalized in the United States [published correction appears in J Diabetes Sci Technol. 2020 Jun 10;1932296820932678]. J Diabetes Sci Technol. 2020;14(4):813-821.

U.S. Department of Health and Human Services, Food and Drug Administration, Center for Devices and Radiological Health. Enforcement policy for non-invasive remote monitoring devices used to support patient monitoring during the coronavirus disease 2019 (COVID-19) public health emergency: guidance for industry and Food and Drug Administration staff (March 2020), FDA-20202-D-1138. https://www.fda.gov/regulatory-information/search-fda-guidance-documents. Accessed September 1, 2020.

10.

Dexcom. Fact sheet for healthcare providers: use of Dexcom continuous glucose monitoring systems during the COVID-19 pandemic. https://www.dexcom.com/hospitalfacts. Accessed September 1, 2020.

11.

De Block

Gios

Verheyen

, et al. Randomized evaluation of glycemic control in the medical intensive care unit using real-time continuous glucose monitoring (REGIMEN trial). Diabetes Technol Ther. 2015;17(12):889-898.

12.

Boom

Sechterberger

Rijkenberg

, et al. Insulin treatment guided by subcutaneous continuous glucose monitoring compared to frequent point-of-care measurement in critically ill patients: a randomized controlled trial. Crit Care. 2014;18(4):453.

13.

Leelarathna

English

Thabit

, et al. Feasibility of fully automated closed-loop glucose control using continuous subcutaneous glucose measurements in critical illness: a randomized controlled trial. Crit Care. 2013;17(4):R159.

14.

Kopecký

Mráz

Bláha

, et al. The use of continuous glucose monitoring combined with computer-based eMPC algorithm for tight glucose control in cardiosurgical ICU [published online ahead of print February 20, 2013]. Biomed Res Int. doi:10.1155/2013/186439

15.

Holzinger

Warszawska

Kitzberger

, et al. Real-time continuous glucose monitoring in critically ill patients: a prospective randomized trial. Diabetes Care. 2010;33(3):467-472.

16.

Umpierrez

Klonoff

DC.

Diabetes technology update: use of insulin pumps and continuous glucose monitoring in the hospital. Diabetes Care. 2018;41(8):1579-1589.

17.

van Steen

Rijkenberg

Limpens

van der Voort

Hermanides

DeVries

JH.

The clinical benefits and accuracy of continuous glucose monitoring systems in critically ill patients—a systematic scoping review. Sensors (Basel). 2017;17(1):146.

18.

Wallia

Umpierrez

Rushakoff

, et al. Consensus statement on inpatient use of continuous glucose monitoring. J Diabetes Sci Technol. 2017;11(5):1036-1044.

19.

Basu

Slama

Nicholson

, et al. Continuous glucose monitor interference with commonly prescribed medications: a pilot study. J Diabetes Sci Technol. 2017;11(5):936-941.

20.

Lin

LI.

A concordance correlation coefficient to evaluate reproducibility. Biometrics. 1989;45(1):255-268.

21.

Lin

LI-K

. A note on the concordance correlation coefficient. Biometrics. 2000;56:324-325.

22.

Bland

Altman

DG.

Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476):307-310.

23.

Clarke

Cox

Gonder-Frederick

Carter

Pohl

SL.

Evaluating clinical accuracy of systems for self-monitoring of blood glucose. Diabetes Care. 1987;10(5):622-628.

24.

Ferreria

Bota

Bross

Mélot

Vincent

JL.

Serial evaluation of the SOFA score to predict outcome in critically ill patients. JAMA. 2001;286(14):1754-1758.

25.

Sadhu

AR.

Battlefield endocrinology. J Diabetes Sci Technol. 2020;14(4):776-777.

26.

Wollersheim

Engelhardt

Pachulla

, et al. Accuracy, reliability, feasibility and nurse acceptance of a subcutaneous continuous glucose management system in critically ill patients: a prospective clinical trial. Ann Intensive Care. 2016;6(1):70.

27.

Krinsley

Chase

Gunst

, et al. Continuous glucose monitoring in the ICU: clinical considerations and consensus. Crit Care. 2017;21(1):197.

28.

Welsh

Gao

Derdzinski

, et al. Accuracy, utilization, and effectiveness comparisons of different continuous glucose monitoring systems. Diabetes Technol Ther. 2019;21(3):128-132.

29.

Bailey

Chang

Christiansen

Clinical accuracy of a continuous glucose monitoring system with an advanced algorithm. J Diabetes Sci Technol. 2015;9(2):209-214.

30.

Christiansen

Garg

Brazg

, et al. Accuracy of a fourth-generation subcutaneous continuous glucose sensor. Diabetes Technol Ther. 2017;19(8):446-456.

31.

Reutrakul

Genco

Salinas

, et al. Feasibility of the inpatient continuous glucose monitoring during the COVID-19 pandemic: early experience. Diabetes Care. 2020;43(10):e137-e138.

32.

Finfer

Wernerman

Preiser

, et al. Clinical review: consensus recommendations on measurement of blood glucose and reporting glycemic control in critically ill adults. Crit Care. 2013;17(3):229.

33.

Klonoff

Umpierrez

Rice

MJ.

A milestone in point of care capillary blood glucose monitoring of critically ill hospitalized patients. J Diabetes Sci Technol. 2018;12(6):1095-1100.