Chronobiologically-informed features from CGM data provide unique information for XGBoost prediction of longer-term glycemic dysregulation in 8,000 individuals with type-2 diabetes

Type 2 Diabetes causes dysregulation of blood glucose, which leads to long-term, multi-tissue damage. Continuous glucose monitoring devices are commercially available and used to track glucose at high temporal resolution so that individuals can make informed decisions about their metabolic health. Algorithms processing these continuous data have also been developed that can predict glycemic excursion in the near future. These data might also support prediction of glycemic stability over longer time horizons. In this work, we leverage longitudinal Dexcom continuous glucose monitoring data to test the hypothesis that additional information about glycemic stability comes from chronobiologically-informed features. We develop a computationally efficient multi-timescale complexity index, and find that inclusion of time-of-day complexity features increases the performance of an out-of-the-box XGBoost model in predicting the change in glucose across days. These findings support the use of chronobiologically-inspired and explainable features to improve glucose prediction algorithms with relatively long time-horizons.Summary: Diabetes mellitus (DM), one of the most common conditions in the world, is a chronic metabolic disease that causes high blood sugar levels. Elevated blood sugar levels can lead to secondary conditions such as heart disease, high blood pressure, and kidney disease. It is therefore important to monitor sugar levels in the blood in order to allow individuals to decide how to best control their diabetes. Noninvasive continuous glucose monitors (CGMs) allow for the monitoring of blood sugar every few minutes instead of the historical self-administered “finger prick” technique. Estimates from CGMs have often be used to predict when someone’s blood sugar levels may get too high or too low in the nearby future – often within an hour; however, longer-term dysregulation can reflect an individual’s overall blood sugar stability. In this study we instead use CGM estimates to predict blood sugar dysregulation on the scale of days, instead of the nearby future, by incorporating information related to the body’s ability to self-regulate blood sugar levels across time. We then use machine learning to show that this additional information is better at predicting longer-term dysregulation than typical methods in statistics.