DiaData: An Integrated Large Dataset for Type 1 Diabetes and Hypoglycemia Research

Type 1 diabetes (T1D) is an incurable autoimmune disorder, which needs attentive monitoring to avoid high glucose variations. Affected cannot produce sufficient insulin and depend on external insulin injections. Multiple factors impact glucose levels, which can lead to dangerous side effects of hyperglycemia (≥ 180 mg/dL) and hypoglycemia (≤ 70 mg/dL). Data analysis can significantly enhance diabetes care by discovering individual trends and enabling tailored decision support. Particularly, machine learning (ML) approaches provide early alerts and predict glucose levels. However, the main limitation in diabetes research is the unavailability of large datasets. Therefore, this study systematically integrates 15 datasets to create a comprehensive database of 2510 subjects with glucose measurements recorded every 5 minutes. In total, 149 million measurements are included (Euglycemia (58.3%), Hyperglycemia (37.5%), and Hypoglycemia (4.2%)). Moreover, two sub-databases are extracted, including demographics or heart-rate data. The integrated dataset provides an equal distribution of sex and a variety of age levels. As a further contribution, data quality is assessed, revealing that missing values and data imbalance present a significant challenge. Thus, the application of ML models necessitates appropriate preprocessing methods.