Reliability of Inpatient CGM: Comparison to Standard of Care.

Abstract

Optimal inpatient glycemic management targets a blood glucose (BG) of 140-180 mg/dL and is an important safety measure for hospitalized patients with hyperglycemia. Traditional barriers to appropriate insulin administration include incorrect timing of prandial insulin administration, failure to administer basal insulin to persons with insulin deficiency/type 1 diabetes mellitus (DM), and inaccurate insulin dosing or timing resulting in hypoglycemia. Given the ongoing rapid assimilation of technology to manage our patients with DM, we investigated the use of continuous glucose monitoring (CGM) in the inpatient setting as a potential solution to traditional barriers to optimal hyperglycemia management for inpatient care. In this study, we evaluated the efficacy of use of inpatient CGM for insulin dosing in comparison with current standard of care and whether CGM could aid in minimizing hypoglycemic events. This study evaluated the use of Abbott professional (blinded) Freestyle Libre CGMs in participants treated with basal bolus insulin administered with subcutaneous insulin (basal bolus therapy [BBT]: n = 20) or on intravenous insulin (IVI) infusions (n =16) compared with standard point of care (POC) BG measurements. All participants on IVI were admitted with a diagnosis of diabetic ketoacidosis (DKA). The CGM data was not available in real time. Sensors were removed at the time of discharge and data uploaded to Libre View. Continuous BG data were aggregated for each subject and matched to POC BG or lab chemistry values within five minutes. The POC BG results were assessed for comparability (CGM vs standard BG testing). Data were further analyzed for clinical decision-making for correction insulin. The overall mean absolute relative difference including both IVI and BBT groups was 22.3% (SD, 9.0), with a median of 20.0%. By group, the IVI arm mean was 19.6% (SD, 9.4), with a median of 16.0%; for BBT, the arm mean was 24.6% (SD, 8.1), with a median 23.4%. Using the Wilcoxon two-sample test, the means were not different (P = .10), whereas the medians were (P = .015). The CGM consistently reported lower glucose values than POC BG in the majority of paired values (BBT arm mean difference = 44.8 mg/dL, IVI mean difference = 19.7 mg/dL). Glucose results were in agreement for the group 83% of the time with Bland-Altman Plot of Difference versus the mean of all glucometric data. Analysis of correction dose insulin using either CGM or POC BG values resulted in a negligible difference in calculated insulin dose recommended in those receiving subcutaneous insulin. Corrective doses were based on weight and insulin sensitivity (type 1 vs type 2 DM). Participants initially on IVI were included in a data set of BBT once IVI therapy ceased and basal bolus insulin regimen was started. The data of all basal bolus therapy participants with 1142 paired values of CGM versus POC glucose were used. The dosing difference was less for CGM than POC BG in the majority of paired values, and there was an absolute difference in dose of insulin of only 1.34 units. In the IVI group with 300 paired values of CGM versus POC glucose, there was an absolute difference in dose of insulin of only 0.74 units. About a third of the patients studied in the BBT arm experienced a hypoglycemic event with POC BG <70 mg/dL. If used in real time, CGM would have identified a hypoglycemic event for our patients on average 3 hours and 34 minutes before it was detected by standard POC BG. Two participants incurred severe nocturnal hypoglycemia during the study with POC BG <54 mg/dL with hypoglycemia detected on CGM up to 3 hours and 42 minutes before POC testing. These results suggest that the use of inpatient CGM arrives at similar correction insulin dosing. The routine use of CGM for inpatients would consistently underestimate the BG compared with POC BG and could aid in minimizing and predicting hypoglycemia in the hospital setting. Our data support that the model of adoption of real-time inpatient CGM technology is anticipated to have significant impact in the clinical setting in efforts to maintain adequate glycemic control targeting BG 140-180 mg/dL while minimizing the frequency of hypoglycemic events.