Objective:

To provide a scoping review of the applications of natural language processing (NLP) and large language models (LLMs) in sports injury assessment and rehabilitation decision-making.

Approach:

• Methodology: Followed the Joanna Briggs Institute (JBI) scoping review framework and PRISMA-ScR guidelines to search databases and included studies utilizing NLP or LLMs.
• Data Extraction: Two researchers independently screened studies and extracted data, resolving disagreements through discussion or arbitration.
• Quality Assessment: Study quality was assessed using the MINimum Information for Medical AI Reporting (MINIMAR) criteria.

Key Findings:

• 27 studies were included, primarily from the United States (37.0%).
• The primary research types included algorithm evaluation and benchmarking.
• Healthcare professionals were the primary target population.
• Text data sources mainly consisted of simulated/synthetic question-answering scenarios.
• Overall compliance with MINIMAR report quality assessment was 61.70%.

Interpretation:

LLMs have potential in sports injury assessment and rehabilitation decision-making but face challenges in readability, hallucination control, and methodological reporting.

Limitations:

• Significant shortcomings in readability and hallucination control.
• Lack of patient perspective in existing studies.
• Geographic coverage is limited.

Conclusion:

Future efforts should focus on improving model reliability and bridging geographic gaps in research.