Modeling User Behavior in Context: A Systems-Level Approach to Mobile Health 198 views

Mobile devices such as smartphones and smartwatches have fundamentally shifted the healthcare landscape towards more individualized care. Advancements in passive sensing over the past decade have enabled consumer mobile devices to conduct unobtrusive, in-the-moment monitoring of objective features of health and wellbeing. Viewed together, these passively-sensed data comprise a system in which a user's context shapes their future health and behavior. In this dissertation, I take a systems-level approach to understanding health in context. To this end, I present our work on mobile sensing for symptom and medication adherence monitoring. I then present COMP-SCT, a novel framework for deriving personal, behavioral, and environmental features of user context at multiple time scales using Social Cognitive Theory. I apply COMP-SCT to two case studies, demonstrating its utility in using personal and behavioral factors such as mood, levels of engagement, and medication-taking behavior to predict mood and medication adherence among breast cancer patients. This work advances the state-of-the-art in translating raw, passively-sensed data into clinically-relevant insights for personal health monitoring in the wild.

PHD (Doctor of Philosophy)

mhealth; mobile health; wearable sensors; wearables; supervised learning; health behavior theory; machine learning

English

2022-12-11