MetaboNet: The Largest Publicly Available Consolidated Data Set for Type 1 Diabetes Management

Area of Science:

• Biomedical Informatics
• Data Science in Healthcare
• Endocrinology Research

Background:

• Type 1 diabetes (T1D) algorithm development is hindered by fragmented and non-standardized data.
• Existing T1D datasets have structural differences, complicating data integration and algorithmic comparability.
• A unified data resource is needed to advance T1D management algorithm research.

Purpose of the Study:

• To establish a unified and accessible data resource for type 1 diabetes (T1D) algorithm development.
• To overcome limitations posed by fragmented and non-standardized T1D management data.
• To facilitate improved comparability and generalizability of T1D algorithms.

Main Methods:

• Harmonized multiple public type 1 diabetes (T1D) datasets into a single resource named MetaboNet.
• Included datasets with continuous glucose monitoring (CGM) and insulin pump data.
• Retained auxiliary data like carbohydrate intake and physical activity where available.

Main Results:

• The MetaboNet dataset contains 3135 subjects and 1228 patient-years of CGM and insulin data.
• It is substantially larger than existing standalone benchmark datasets.
• A public subset is available for download, with a restricted subset accessible via application.

Conclusions:

• A harmonized public dataset for T1D research (MetaboNet) has been created.
• Access pathways for both unrestricted and DUA-governed data components are detailed.
• The dataset's breadth supports more generalizable algorithmic performance evaluations in T1D research.