Utility of an Artificial Intelligence Language Model for Post-Operative Patient Instructions Following Facial Trauma

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Surgical Patient Education

Background:

• Effective post-operative instructions (POIs) are crucial for patient recovery and outcomes following facial trauma.
• Current POIs are often derived from electronic medical record (EMR) templates or general internet searches, which may vary in quality and patient-friendliness.

Purpose of the Study:

• To compare the utility of AI-generated POIs from ChatGPT against traditional EMR templates and Google search results for facial trauma procedures.
• To assess the understandability and actionability of these different sources of patient information.

Main Methods:

• POIs for four common facial trauma procedures were generated using ChatGPT, sourced from EMRs, and retrieved via Google search.
• A panel of surgeons evaluated the POIs using the Patient Education Materials Assessment Tool (PEMAT-P), focusing on understandability and actionability.

Main Results:

• All evaluated POIs demonstrated a wide range of understandability (55%-94%) and actionability (33%-94%) scores.
• ChatGPT-generated POIs achieved high understandability scores (82%-88%) and moderate actionability scores (60%-80%).
• Procedure-specific scores for ChatGPT POIs ranged from 50% to 75%.

Conclusions:

• ChatGPT presents a promising tool to augment the provision of post-operative instructions for facial trauma patients.
• Tailoring AI-generated instructions requires detailed and patient-specific prompts to ensure accuracy, clarity, and practical usability.