The TRIPOD-LLM Statement: A Targeted Guideline For Reporting Large Language Models Use

Area of Science:

• Biomedical informatics
• Artificial intelligence in healthcare
• Scientific reporting standards

Background:

• Large Language Models (LLMs) are increasingly used in healthcare.
• Standardized reporting is crucial for LLM applications in medicine.
• Existing guidelines need adaptation for LLM-specific challenges.

Purpose of the Study:

• Introduce TRIPOD-LLM, reporting guidelines for LLMs in biomedical research.
• Extend the TRIPOD+AI statement to address unique LLM issues.
• Enhance transparency, reproducibility, and clinical utility of LLM studies.

Main Methods:

• Developed an extension of the TRIPOD+AI statement.
• Utilized an expedited Delphi process and expert consensus.
• Created a comprehensive checklist with a modular format.

Main Results:

• TRIPOD-LLM includes 19 main items and 50 subitems.
• A modular format with 14 main items and 32 subitems is applicable across LLM research designs.
• An interactive website is available for guideline completion and PDF generation.

Conclusions:

• TRIPOD-LLM provides essential guidelines for reporting LLM research in healthcare.
• Emphasizes transparency, human oversight, and task-specific performance.
• Aims to improve the quality and applicability of biomedical LLM research as a living document.