Temporal data analytics for diabetes monitoring with applications on pediatric and adult patients

The research activity described in this thesis has been conducted within the Laboratory for Biomedical Informatics “Mario Stefanelli” of the University of Pavia, Italy, from October 2018 to November 2021. It was motivated by the need of a diabetes management application that allowed the integration of patient-generated health data (PGHD) from different wearable sensors, providing temporal data analytics functionalities to gain deeper insights in the data and to enhance critical events prediction. Diabetes mellitus (DM) is a life-long condition that continues to rise in prevalence across the globe, representing one of the fastest growing health challenges of the last decades. DM management is mainly focused on maintain near-normal glycemic values for reducing the risk of long-term life-threatening complications, without causing substantial falls in circulating glucose. Self-monitoring of blood glucose (SMBG) and continuous glucose monitoring (CGM) systems are essential to achieve the goal of a safe and prolonged glycemic control, especially in subjects on insulin therapies. Anyway, it is important to consider also other factors for providing a complete glycemic profile contextualized within the day. The Advanced Intelligent Distant – Glucose Monitoring (AID-GM) platform, developed at the Biomedical Informatics Laboratory of the University of Pavia, consents the integration of PGHD from multiple sources, such as CGM systems, personal fitness trackers (PFTs), and self-reported daily diaries. Therefore, patients and healthcare providers were allowed to share and visualize CGM measurements integrated by lifestyle, sleep, and HR information. In this context, the research activity is focused on the development of novel and state of the art temporal data analytics strategies based on multivariate PGHD to discover new insights in the collected data. Moreover, the long-term remote monitoring has been supported by the implementation of the proposed temporal data analytics functionalities in the AID-GM platform, turning the time-series raw data into relevant clinical information and making available to clinicians a complete overview of each patient’s conditions as a decision support in their healthcare tasks.