A new initiative at Dana-Farber leverages HDAI's AI platform embedded within the electronic health record (EHR) to improve serious illness conversations by extracting and summarizing patients' goals of care, both in inpatient and outpatient settings.

This platform includes predictive models to identify high-risk patients and generate rapid, comprehensive summaries, helping clinical teams streamline communication and better align care with patient wishes.

The initiative utilizes HDAI's HealthVision platform, which integrates large language model (LLM) technology directly into the EHR for quick, accurate synthesis of clinical narratives and patient risk stratification.

Addressing challenges of care complexity and data fragmentation in academic medical centers, the pilot project aims to demonstrate improved care quality, with plans to share results through conferences and publications in late 2025 and early 2026.

Led by experts in oncology, palliative care, and health informatics—Dr. Christopher Manz, Dr. Charlotta Lindvall, and Dr. Manuel Fanarjian—the study seeks to establish the effectiveness of AI-driven summaries in clinical practice.

The team’s focus is on using AI to enhance serious illness conversations, improve communication, and ensure care aligns more closely with patient preferences, leveraging their combined expertise in clinical and health informatics.

Building on prior research, the project demonstrates the feasibility of using LLMs to identify and summarize goals of care documentation in cancer patients, with results expected to be published in 2026.

Previous studies at Dana-Farber have shown promising potential for LLMs in advance care planning, laying the groundwork for this larger pilot to evaluate real-world clinical impact.

The recent pilot study will analyze the accuracy and feasibility of applying LLMs to communicate goals of care, with results anticipated in early 2026 that could set a new standard for AI in patient-centered oncology care.

This initiative aims to improve care quality by making patient preferences more accessible during hospital admissions, reducing unnecessary aggressive treatments near end-of-life, and enhancing communication between care teams.

HDAI emphasizes that this technology addresses key challenges such as care complexity, clinician workload, and fragmented data, ultimately aiming to improve care for patients with serious illnesses.

The program seeks to support more patient-centered care by providing real-time summaries of goals of care, which are often hidden in voluminous clinical notes, thereby enhancing communication and goal-concordant treatment.