Large-Scale Data Analysis for Glucose Variability Outcomes with Open-Source Automated Insulin Delivery Systems.

Area of Science:

• Endocrinology
• Biomedical Engineering
• Data Science

Background:

• Open-source automated insulin delivery (AID) systems integrate continuous glucose monitors (CGM), insulin pumps, and algorithms for diabetes management.
• Community adoption of AID, like OpenAPS, necessitates research into glucose-related outcomes, particularly glucose variability (GV).
• Existing studies often lack in-depth GV analysis, despite its importance in diabetes management.

Purpose of the Study:

• To review the evolution of open-source AID and clinically approved GV metrics.
• To evaluate glucose variability (GV) outcomes using large-scale data analytics from the OpenAPS Data Commons dataset.
• To identify lessons learned and future research questions from analyzing complex diabetes data.

Main Methods:

• Utilized the OpenAPS Data Commons dataset (n=122), comprising over 46,070 days of data and 10 million CGM data points.
• Performed data cleaning and applied methods for measuring glucose variability (GV).
• Conducted data analysis based on individual self-reported demographics.

Main Results:

• Evaluated glucose variability (GV) outcomes using large-scale data analytics.
• Described data cleaning processes and GV measurement methods.
• Presented results of data analysis, including demographic correlations.

Conclusions:

• Affirmed the efficacy of open-source AID in managing diabetes, consistent with previous findings.
• Highlighted the value of large-scale datasets like OpenAPS Data Commons for diabetes research.
• Identified key lessons and emerging research questions for future AID innovation.