Artificial Intelligence-Driven Adaptation of Pediatric Traumatic Brain Injury Case Descriptions for Family...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Pediatric Traumatology

Background:

• Complex medical information, such as pediatric traumatic brain injury (TBI) case descriptions, presents communication challenges.
• Large language models (LLMs) offer potential solutions for simplifying complex medical texts for diverse audiences.

Purpose of the Study:

• To evaluate GPT-4o's ability to generate age- and education-appropriate versions of pediatric TBI case descriptions.
• To assess if LLM-generated adaptations maintain clinical accuracy and appropriate emotional tone.

Main Methods:

• Five pediatric TBI case descriptions were adapted for four distinct audience scenarios.
• Text complexity was quantified using Flesch-Kincaid (FKS), Gunning Fog, and SMOG indices.
• Clinical experts rated the fidelity and emotional appropriateness of the adapted texts on a 3-point scale.

Main Results:

• Original texts exhibited high complexity (FKS 18.2-20.5), suitable for college-educated adults.
• Adaptations for parents with high school education were often over-simplified (FKS 4.75-7.1).
• Versions for 12-year-olds were well-matched (FKS ~5-6), while 8-year-old versions (FKS 4.0-6.8) showed reduced fidelity despite appropriate emotional tone.

Conclusions:

• LLMs can assist clinicians in drafting patient-friendly explanations of complex medical cases.
• Careful review and tailoring by clinicians are crucial to ensure accuracy and appropriateness for specific audiences.
• LLM-generated content requires validation to meet specific educational and clinical needs.