Reinforcement Learning-Based Adaptive Insulin Advisor for Individuals with Type 1 Diabetes Patients under Multiple...

Area of Science:

• Biomedical Engineering
• Endocrinology
• Digital Health

Background:

• Type 1 diabetes management often involves multiple daily injections (MDI) of insulin.
• Optimizing insulin dosing is crucial for glycemic control and preventing complications.
• Existing adaptive basal-bolus advisors (ABBA) require further refinement for real-world MDI use.

Purpose of the Study:

• To enhance the existing adaptive basal-bolus advisor (ABBA) for patients on MDI insulin therapy.
• To validate a novel approach combining self-monitoring of blood glucose (SMBG) with insulin injection devices.
• To assess the efficacy of the enhanced ABBA in improving glycemic control and safety.

Main Methods:

• Development and refinement of the adaptive basal-bolus advisor (ABBA) algorithm.
• Utilization of the DMMS.R simulator for in silico experiments.
• Testing the combined use of SMBG data and insulin pen usage within the simulator.

Main Results:

• The enhanced ABBA demonstrated superior performance compared to conventional methods.
• Increased time spent within the target blood glucose range was observed.
• Significant reduction in the incidence of both hyperglycaemic and hypoglycaemic events was achieved.

Conclusions:

• The enhanced ABBA offers a promising advancement for type 1 diabetes management in MDI patients.
• The combined use of SMBG and insulin pens with the ABBA improves glycemic control.
• This approach enhances patient safety by mitigating risks associated with blood glucose fluctuations.