Readability of Orthopaedic Patient Educational Material: An artificial intelligence application

Area of Science:

• Medical Education
• Health Informatics
• Artificial Intelligence in Healthcare

Background:

• Patient educational materials in orthopaedic trauma often exceed appropriate reading levels.
• Improving readability is crucial for patient comprehension and adherence to treatment.

Purpose of the Study:

• To evaluate the efficacy of artificial intelligence (AI) in simplifying orthopaedic trauma patient educational materials.
• To assess AI's ability to reduce reading levels to a patient-appropriate standard (6th grade) while preserving accuracy.

Main Methods:

• 35 orthopaedic patient education articles from Level 1 Trauma Centers were analyzed for reading level.
• Articles were rewritten using four iterations of Generative Pre-Trained Transformer (GPT) AI models with a specific prompt.
• Readability, accuracy, cost, and token reduction were assessed for multiple AI models.

Main Results:

• All GPT AI models successfully lowered the mean reading grade level of the articles.
• GPT-4o and GPT-4o-Mini demonstrated the highest effectiveness and consistency.
• GPT-4o-Mini presented the optimal balance of effectiveness, cost ($0.61 for 35 articles), and information accuracy.

Conclusions:

• AI is a viable tool for enhancing the readability of patient educational materials in orthopaedic trauma.
• GPT-4o-Mini is recommended as the most efficient AI model for this application, considering performance and cost.