Advancing Continuous Glucose Monitoring for Inpatient Clinical Decision Support: Individual Algorithmic Mean Absolute...

Continuous glucose monitoring (CGM) is widely used to monitor glucose levels in patients with diabetes and guide insulin dosing in outpatients. In inpatient care, special regulatory requirements necessitate CGM accuracy as a prerequisite for its integration into clinical decision support.

To meet the specific demands of in-hospital care, CGM accuracy was retrospectively evaluated in 226 patients using paired CGM and point-of-care glucose measurements, assessed via mean absolute relative difference (MARD), Clarke Error Grid (CEG) analysis, and a modified version of the U.S. Food and Drug Administration agreement rule. A dynamic, patient-specific algorithm incorporating time lag correction and linear modeling was developed to minimize MARD and applied in a second cohort of 24 patients within the clinical workflow.

Data analysis showed an initial MARD of 10.30%, with 99.02% of data points located in zones A and B of the CEG. The application of the patient-specific optimization algorithm improved the MARD by 4.33%. Evaluation of the patient-specific algorithm on the second inpatient cohort demonstrated a 5.58% reduction in intrapersonal MARD, indicating its potential applicability within clinical workflows.

Patient-specific algorithmic refinements of CGM data demonstrate the potential to adequately address the unique demands of inpatient diabetes care by reducing intrapersonal MARD, paving the way for the adoption of CGM systems into hospital environments.