A Perspective on Harmonizing Diabetes Management Datasets

Abstract

Diabetes management datasets are often compiled from various sensors and devices, including diabetes technology, activity trackers, and other health-related equipment, resulting in heterogeneous data formats. Despite the abundance of available data, inconsistencies in dataset formats and data-sharing practices limit the ability to build on prior work and compare results across studies. Standardizing data-sharing formats can improve consistency, facilitate dataset consolidation, and reduce the data processing burden for researchers. This article explores the current state of data-sharing practices in diabetes management research and proposes guidelines for harmonizing datasets using a unified time-aligned tabular format. We demonstrate the application of these guidelines on three widely used datasets and highlight key challenges in achieving data harmonization. We call on the broader research community to develop and adopt detailed recommendations for standardized data-sharing practices.