Wearable AI in clinical decisions

Consumer wearables capable of rendering prediagnoses, such as for atrial fibrillation (AF), have the potential to influence downstream clinical decision-making. However, postmarket surveillance has been limited by the absence of codified indicators of wearable use in electronic health records (EHRs). In this talk, we demonstrate a novel approach to address this gap using a weak supervision-based framework to identify wearable-driven AF prediagnoses from clinical notes. Leveraging data programming via labeling functions and the Snorkel framework, we constructed a labeler model to probabilistically annotate notes and fine-tuned a Clinical-Longformer classifier. Using this approach, we were able to study the clinical characteristics of individuals with AF prediagnoses, their subsequent healthcare use, and clinical outcomes in a real-world cohort. This presentation demonstrates the feasibility and utility of EHR-based surveillance for consumer health technologies.

Ben Viggiano is a second year PhD student at Stanford University studying Biomedical Informatics and Computer Science advised by Euan Ashley and Chris Ré. He is primarily interested in developing methods to utilize foundation models to advance research methodology in biology, medicine, and healthcare.

Krishna Pundi is a clinical cardiac electrophysiologist and instructor at the Palo Alto VA Medical Center and Stanford University. His research interest is in combining traditional epidemiologic approaches of understanding disease with novel methods of data acquisition to define clinical, demographic, and arrhythmia morphology features that predict cardiovascular events and death.

This presentation will explore the transformative potential of wearable AI technologies in clinical care, highlighting how these intelligent systems may enhance patient safety and support improved decision-making across diverse healthcare settings. Drawing on recent advancements in digital biomarkers and AI-enabled wearables, we will examine practical applications and discuss the challenges and opportunities for integrating these innovations into routine clinical practice.

Arjun Mahajan is a medical student at Harvard Medical School with a focus on clinical informatics and autoimmune diseases. His work focuses on applying machine learning and multimodal large language models to improve diagnosis, clinical decision-making, and patient safety, with support from national grants from the American College of Rheumatology and Amazon, amongst others. Previously, he conducted research at Kaiser Permanente and helped develop wearable technologies in autism to address impairing motor stereotypy – work awarded by the White House and Department of Defense. Arjun was also recently appointed to the editorial team at Nature Digital Medicine (News & Views). He completed his undergraduate degree at UC Berkeley and subsequently obtained an MS in Data Science.

Dylan Powell is an Assistant Professor (Lecturer) in Public Health & Innovation at the University of Stirling within the faculty of Health Sciences & Sport.

Excited by the value of technology, data and innovation in health care and life sciences to optimise care and wellbeing for patients and society. Prior to joining the University, Dylan worked for Deloitte Healthcare and Life Sciences practice and Knowledge Sharing lead for the UK Deloitte Clinical Network. Dylan’s professional background as a clinician (Physiotherapist) and researcher has spanned the NHS, academia, professional sport, and within industry. His PhD in Computer Science explored the use of wearables and digital biomarkers in neurological conditions including Sports-Related Concussion.

Dylan was appointed to the editorial team at Nature Digital Medicine (News & Views) in September 2023.