Glycemic control for patients with type 1 diabetes on multiple daily injections: A customized decision support system based on patients adherence

Type 1 Diabetes is a chronic condition characterized by the pancreas’s inability to produce insulin, requiring lifelong insulin therapy to manage blood glucose levels. While machine learning algorithms have been widely applied in the context of this chronic condition, few studies report practical results on glycemic control based on these predictions, particularly in the context of multiple daily injections therapy. This study aims to evaluate the clinical benefits of a machine learning-driven decision support system designed to assist patients on multiple daily injections therapy in taking corrective actions to meet standard glycemic targets. The system also offers patients the flexibility to customize their treatment plan according to their adherence levels. The proposed approach was validated through in silico simulations that modeled realistic scenarios, including physical activity. The system comprises four distinct operational modes, all of which outperformed standard therapy, especially in reducing time spent in hypoglycemia, without compromising mean glucose levels. This study provides a novel contribution by demonstrating how to effectively apply machine learning-based predictions to tailor patients’ therapy.