926-P: Implementation of a Novel AutoBolus Feature in an Automated Insulin Delivery (AID) System

Abstract

AID systems aim to reduce mental burden experienced by people with diabetes by automatically adjusting insulin delivery in response to real-time glucose levels; however, systems still require pre-meal boluses for optimal outcomes. A system providing an automatic bolus response (‘AutoBolus’) to rising glucose levels (e.g., after meals) lessens user burden and reduces hyperglycemic events by compensating for missed or under-estimated pre-meal boluses. We propose a novel method to deliver an AutoBolus based on rising glucose levels via a machine learning based meal detection algorithm. AutoBolus meal detection is paired with a two-part novel insulin delivery scheme with an immediate safe upfront delivery, followed by temporarily tuning AID algorithm parameters to progressively respond to the post-prandial glucose response by delivering additional insulin. If a decreasing glucose condition is detected, the AID algorithm parameters revert to baseline. In-silico simulation demonstrates AutoBolus increases time in euglycemia by as much as 11% in adolescents, 8% in adults, and 6% in children for missed boluses for three meals in a 24-hour period. Our results show that the proposed AutoBolus insulin delivery partially compensates for missed pre-meal boluses increasing time in euglycemia, reducing user burden. Figure. Illustrative diagram showing the two-part novel insulin delivery scheme upon meal detection. Disclosure R.Narayanaswami: None. Y.Zheng: Employee; Insulet Corporation. W.J.Whiteley: Employee; Insulet Corporation. M.Sevil: None. S.Salavati: Employee; Insulet Corporation.