Impact of Patient-Driven Meal Management Errors on Type 1 Diabetes Control: A Sensitivity Analysis

Type 1 Diabetes (T1D) is characterized by an impairment of glycaemia regulation, due to a lack of endogenous insulin production. Consistent advancements in T1D treatment have been possible thanks to Artificial Pancreas (AP), a technology based on a closed-loop control algorithm that regulates insulin infusion such to achieve the most efficient glycaemia control. Current commercial AP systems require patient to estimate the carbohydrates content in the ingested meals and to announce it to the system, in order to promptly counteract the postprandial glucose raise. However, various inaccuracies might affect the announced meal: i) the estimate might be inaccurate (carb counting error), ii) a meal could be announced with a delay or iii) the patient might forget to announce it (unannounced meal). Understanding the impact of these inaccuracies on the performances of an AP could help healthcare providers and patients in prioritizing the most impactful factors. To this aim, in this work we conduct sensitivity analysis using the UVa/Padova Simulator, a well established simulator of T1D physiology, and considering the Pavia/Padova AP, that is a Model Predictive Control based control algorithm.