AI assists clinicians in responding to patient messages at Stanford Medicine

Key Points:

  • Large language models reduce healthcare providers’ workload and burnout
  • Integrating generative AI with clinicians can improve cognitive burden
  • AI-generated drafts improve clinician experience despite not saving time


Stanford Medicine researchers have discovered that integrating large language models into health care workflows can assist in drafting responses to patient portal messages. These AI-generated drafts, reviewed by clinicians before being shared, address clinical inquiries like symptoms of a cold or medication side effects. A reduction in clerical burden and burnout among clinicians was reported in a study of these AI-generated responses.


The introduction of the large language model GPT in late 2022 sparked excitement within the medical field about the potential of AI. Patricia Garcia, MD, highlighted the model’s capabilities in creating language and its utility for tasks like writing messages and clinical notes. This integration of generative AI with a “human in the loop” system marks an early step in demonstrating how AI can support health care providers.


The study, detailed in a paper published in JAMA Network Open, was led by Garcia and Stephen Ma, MD, PhD, with support from Stanford’s Department of Medicine and Technology and Digital Solutions team. Christopher Sharp, MD, emphasized the importance of rigorously evaluating the real-world safety and usefulness of AI in health care.


Stanford Medicine has been proactive in incorporating AI tools while prioritizing patient safety and privacy through the RAISE Health initiative. The research team integrated a large language model into electronic health records, ensuring compliance with HIPAA regulations. Clinicians received draft responses generated by the AI model within seconds upon patient message receipt, streamlining the response process.


After a pilot period, clinicians reported that AI-generated drafts alleviated cognitive burden and improved feelings of work exhaustion, despite not saving time. Garcia emphasized the positive impact of this tool on burnout relief and its potential for broader applicability as it evolves. Plans are underway to expand the use of this AI tool to additional clinicians at Stanford Health Care in the future.


By effectively deploying AI tools like large language models, Stanford Medicine is paving the way for innovative solutions to alleviate burden and enhance efficiency in health care settings. The research serves as a practical demonstration of integrating AI technology seamlessly into clinical workflows to support health care teams collectively responding to patient messages. As this technology continues to evolve, its impact on reducing burnout and improving workflow efficiency is expected to grow.



Prompt Engineering Guides



©2024 The Horizon