Enhancing Preoperative Orthopaedic Communication: A Comparative Analysis of Large Language Model- and...

Area of Science:

• Orthopaedic surgery
• Medical communication
• Artificial intelligence in healthcare

Background:

• Effective communication is crucial for informed consent in orthopaedics.
• Healthcare systems face workforce and time constraints impacting patient dialogue.
• Large language models (LLMs) offer potential solutions for streamlining patient-clinician correspondence.

Purpose of the Study:

• To assess the feasibility of LLM-assisted correspondence in orthopaedic practice.
• To evaluate LLM-generated letters for understandability, readability, and complication profile inclusion.
• To compare LLM outputs against human clinician-generated letters.

Main Methods:

• Six common orthopaedic operations were selected.
• Standardized prompts were used to generate letters via four LLMs (OpenAI o1, DeepSeek, Gemini, Copilot) and two human clinicians.
• Understandability (PEMAT-P), readability (Flesch-Kincaid, Gunning Fog, SMOG), and complication inclusion were measured.

Main Results:

• LLM-generated letters demonstrated superior understandability and readability compared to clinician letters.
• OpenAI o1 and Gemini showed high complication profile compliance.
• LLM text was written at a 7th-8th grade reading level, simpler than clinician letters.

Conclusions:

• LLMs can outperform traditional clinician correspondence in clarity and completeness.
• Integrating LLM workflows can reduce administrative burden and improve patient decision-making.
• Further research is needed for ethical and safe clinical deployment of LLMs.