Beth Israel Deaconess Medical Center (BIDMC) Receives National Library of Medicine Award to Advance Patient-Centered AI

Boston, MA | August 26, 2025

Beth Israel Deaconess Medical Center (BIDMC) received an award from the National Library of Medicine (NLM) to develop an AI tool that improves patients' access to and understanding of personalized medical information. The $1.5 million 4-year study is led by Sigall K. Bell, MD, Director of Patient Safety and Discovery at Open Notes Labs and Associate Professor of Medicine at Harvard Medical School, and Adam Rodman, MD, MPH, Director of AI Programs at Harvard Medical School. The researchers will co-design and test, together with patients, “HealthCoach,” a large language model (LLM)-driven chatbot available in English and Spanish.

“Research shows that when patients read their medical notes, they better understand their care, identify safety issues, and take a more active role in managing their health,” said Dr. Bell. “But many patients remain excluded due to language and literacy barriers. An AI tool can help close these gaps if it is designed with patients to meet their specific needs.” The team envisions that HealthCoach will translate and simplify notes based on patient priorities. They will also embed a previously developed patient-clinician alignment tool called “OurDX” into the chatbot, to help patients prepare for their next primary care visit conversationally with HealthCoach.

About HealthCoach

  • HealthCoach will help patients of all literacy levels:

  • Learn from their open medical notes

  • Prepare for primary care visits

  • Co-create a “pre-note” that reflects their values, priorities, and questions

By bringing patient perspectives and unique knowledge directly into the visit pre-note, HealthCoach aims to strengthen patient safety, engagement, and shared decision-making.

Study Highlights

  • Co-design HealthCoach with diverse patient partners in two U.S. states

  • Use de-identified patient notes and patient-clinician alignment tool data to iteratively refine the chatbot

  • Develop novel patient-centered evaluation tools for AI-generated notes

  • Test chatbot-generated notes in 2 languages using 10 domains (for example: accuracy, clarity, and usefulness), and both patient- and clinician-raters

  • Conduct a live trial with English- and Spanish-preferring patients preparing for upcoming primary care visits

The project will produce freely available resources, including the HealthCoach chatbot and educational materials, to promote broad, equitable access to AI-enabled healthcare tools. More about this study can be found on the NIH RePORTER website.