Data-driven carbohydrate counting accuracy monitoring : a personalized approach

Abstract

Accurate carbohydrate counting is crucial for type 1 diabetes mellitus patients on intensive insulin therapy to get on-target blood glucose values. So, it is fundamental to assess their ability to estimate meals’ carbohydrate content and, if needed, recommend carbohydrate counting training. In this context, we propose a personalized data-driven approach to monitor the patients’ ability to estimate the carbohydrate content of meals. The proposed approach uses personalized data to compute a safe range for the carbohydrate counting error according to the characteristics of each patient and adjust this interval to the patient’s daily routines and food habits. Initially, the proposed method uses the insulin-to-carbohydrate ratio, the insulin sensitivity factor, the blood glucose limits, and the blood glucose target to compute a safe interval for the carbohydrate counting error, so the patient could train to reach this goal. Then, the app uses collected daily life data (i.e., blood glucose, meals carbohydrates content, and insulin bolus) to adjust the initial safe interval for the carbohydrate counting error according to the patient’s needs. Preliminary assessment using the FDA-approved University of Virginia (UVA)/Padova Type 1 Diabetes Simulator shows the potential of the proposed approach to help type 1 diabetes patients being aware of their needs for carbohydrate counting education and how accurate they should be to achieve suitable blood glucose levels. Therefore, this tool has the potential to be a great asset to healthcare professionals and patients, improving the carbohydrate counting learning outcomes and leading to better glycemic control.