What You Should Know:
– A new study by researchers at Mass General Brigham suggests large language models (LLMs), a type of artificial intelligence (AI), could be a helpful tool for doctors to streamline communication with patients. However, the study also highlights the importance of human oversight to ensure patient safety.
– The findings, published in The Lancet Digital Health, emphasize the need for a measured approach to LLM implementation.
The Burden of Physician Communication
Physicians today face increasing administrative tasks, including responding to patient portal messages. This can contribute to burnout and hinder patient care. The study investigated the potential of AI to assist doctors in drafting replies to patient messages.
AI Generates Draft Responses for Review
Researchers used a powerful LLM called GPT-4 to create draft responses to 100 hypothetical patient questions about cancer. Radiation oncologists then reviewed and edited these AI-generated responses.
Promising Results, But Caution Needed
The study found that:
- Doctors perceived AI assistance as efficient.
- Over 80% of AI-generated responses were deemed safe by doctors.
- Nearly 60% of these responses required no further editing before sending to patients.
However, the study also identified potential risks:
- A small percentage (7.1%) of unedited AI responses could mislead patients and pose health risks.
- In rare cases (0.6%), these responses could even delay essential medical care.
Importance of Maintaining Human Oversight
Interestingly, physicians often retained the educational content generated by the AI when editing responses. While this highlights the potential benefit of AI-generated patient education, the study emphasizes the importance of human oversight to mitigate safety risks.
Mass General Brigham Committment to Responsible AI
Mass General Brigham is committed to responsible AI development and implementation. They are currently conducting a pilot program integrating AI message drafting into their electronic health records system. Future research will focus on patient perception of AI-generated communication and potential bias in AI algorithms.
“Generative AI has the potential to provide a ‘best of both worlds’ scenario of reducing burden on the clinician and better educating the patient in the process,” said corresponding author Danielle Bitterman, MD, a faculty member in the Artificial Intelligence in Medicine (AIM) Program at Mass General Brigham and a physician in the Department of Radiation Oncology at Brigham and Women’s Hospital. “However, based on our team’s experience working with LLMs, we have concerns about the potential risks associated with integrating LLMs into messaging systems. With LLM-integration into EHRs becoming increasingly common, our goal in this study was to identify relevant benefits and shortcomings.”