Feasibility of Automated Precharting using GPT-4 in New Specialty Referrals

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Workflow Optimization

Background:

• Precharting is crucial for efficient specialty clinic visits.
• Manual precharting is time-consuming and resource-intensive.
• Automating precharting can potentially improve clinic workflow and patient care.

Purpose of the Study:

• To evaluate the feasibility of using GPT-4 for automating precharting for new otolaryngology referrals.
• To describe design strategies for developing an automated precharting tool.
• To assess the accuracy and user feedback of the GPT-4 based precharting utility.

Main Methods:

• Development of a GPT-4 based precharting tool for otolaryngology new patient referrals.
• Iterative testing involving prompt design and token limit management.
• Retrospective testing for accuracy against physician consensus (95.0% agreement).
• Prospective pilot study in a real-world clinical setting.

Main Results:

• The automated precharting tool achieved 95.0% agreement with physician consensus in retrospective testing.
• Prospective pilot showed favorable user feedback on summary quality.
• Users expressed high intention to use summaries but lower perceived time savings.

Conclusions:

• Automated pre-charting using large language models like GPT-4 is feasible for specialty referrals.
• The developed design features can guide future vendor solutions for clinical chart summarization.
• Further optimization is needed to enhance perceived time savings in clinical practice.