Automation of insulin bolus dose calculation in type 1 diabetes: a feasibility study

Background Diabetes care in type 1 diabetes has remained a challenge over time. Ability to count carbohydrates in every meal and varying insulin dose according to individual insulin/carbohydrate ratio allows people with type 1 DM to have wider food choices with better glycemic control. Difficulties in carbohydrate counting may largely be solved by use of technology. Methods This work was done at the endocrine unit of a superspeciality centre involved in care of people with type 1 diabetes. The process of development of software and its preliminary application and results from its use in clinical care in a small group of interested patients is presented in this manuscript. Carbohydrate counting tool for Indian foods was developed, and subsequently, bolus dose calculation was automated by using the reinforcement algorithm in an android app platform named "T1-Life". Data on app usability and acceptability is documented in this pilot study report. Results Five patients completed 3 months of this app usage. Among five people with combined usage over 115 patient weeks, a total of 2661 insulin dose predictions were made. This translates to 3.31 patient-initiated bolus dose predictions per day. Of the total bolus dose predictions made, 82% were accepted by the participants. With usage of app, time in range (70-180 mg/dl) increased by an average of 16.67% in children who used CGMS in their first week as well as last week of observation. Conclusion T1-life, integrated carbohydrate counting and reinforcement-based insulin bolus dose prediction system, has good patient usability and acceptability.