Assessing Health Events Through the Lens of Context: A Case Study in Diabetes Management

Many health conditions such as diabetes, mental illness, and substance use require proper management in daily living to mitigate the occurrence of adverse events. However, it is a challenge to understand personalized factors that contribute to such events. The goal of this research is to enable improved management of chronic conditions by leveraging pervasive technology (i.e. mobile/wearable devices) for passive-sensing of the context surrounding adverse health events. Building on the premise that individuals engage in routine behaviors, this work aims to elucidate factors that positively or negatively influence wellbeing. The knowledge gained can inform evidence-based decisions for patients, caregivers, and researchers.

The proposed research will focus on the application domain of diabetes; however, we envision that the findings and tools developed will support behavioral interventions and improved management of related conditions. This study will first build a multisource bio-behavioral dataset for exploratory analysis of context surrounding adverse diabetes events. The dataset will include wearable device data from continuous glucose monitors and insulin pumps, biomarker data collected during routine clinical visits, and self-reported demographic and behavioral data. Next, our team will develop and evaluate automated algorithms to uncover personalized factors associated with different management outcomes. The research will be conducted through a collaboration between the computer science department at Dartmouth College and the endocrine team at Dartmouth Hitchcock Medical Center. This work will be a launchpad for event-triggered passive sensing of context to inform targeted interventions.