Algorithm-Enabled, Personalized Glucose Management for Type 1 Diabetes at the Population Scale: Prospective...

Area of Science:

• Endocrinology
• Medical Informatics
• Diabetes Technology

Background:

• Continuous glucose monitors (CGMs) are standard care for type 1 diabetes (T1D), recommended by the American Diabetes Association.
• Limited open-source, hardware-agnostic tools exist for analyzing CGM data at a population level for clinical use.
• Clinicians need efficient methods to manage large volumes of CGM data for timely patient intervention.

Purpose of the Study:

• To develop an open-source, hardware-agnostic tool (TIDE) for analyzing CGM data to identify patients needing intervention.
• To minimize the number of patients requiring review by clinicians (physicians, certified diabetes educators) through automated flagging.
• To facilitate asynchronous communication and proactive patient management via electronic medical records.

Main Methods:

• Developed Timely Interventions for Diabetes Excellence (TIDE) using consensus guidelines to analyze CGM data.
• TIDE generates generic and personalized flags for deteriorating glucose control (e.g., mean glucose, glucose increase).
• Conducted a 7-week prospective study in a pediatric T1D clinic and simulated workload on 8 external datasets.

Main Results:

• TIDE achieved 99% sensitivity in identifying patients appropriate for contact.
• The tool reduced the number of patients requiring clinician review by 42.6%.
• Analysis of 8 external datasets (1365 patients) showed low rates of flags per patient per review period.

Conclusions:

• TIDE is an effective open-source tool for large-scale, personalized CGM data analysis in clinical settings.
• It enables sensitive and efficient identification of patients with deteriorating glucose control, reducing clinician workload.
• TIDE supports telemedicine-based T1D care, enhancing guideline-based population health management.