Clinical AI & Wellness

How often do Doctors discuss wellness with their patients?

Dr. Zeke Emanuel’s recently-published book, Eat Your Ice Cream, summarizes a wide body of research regarding wellness behaviors and practices. He mentions well-known wellness activities like exercise and diet, but also draws attention to lesser-discussed wellness concepts like social connectivity and cognitive engagement.

Based on this research, we set out to quantify how often wellness and preventive healthcare are discussed during clinical encounters.

In collaboration with Abridge, a healthcare AI company specializing in ambient scribes for doctor-patient encounters, I developed a large language model (LLM)-based natural language processing pipeline to identify and classify discussions of nine wellness domains across a dataset of up to 50,000 de-identified primary care transcripts.

In this work, I operationalized a theory-driven annotation codebook, translating concepts from social science, behavioral research, and clinical medicine into standardized, reproducible labeling criteria for both human annotators and LLMs. This work combines interdisciplinary research, clinical NLP, and AI-assisted annotation to enable large-scale, systematic analysis of patient-provider conversations.