Adherence to TRIPOD+AI guideline: an updated reporting assessment tool

Area of Science:

• Healthcare research methodology
• Medical informatics
• Artificial intelligence in medicine

Background:

• Incomplete research reporting leads to waste and limits utility.
• Reporting guidelines like TRIPOD improve healthcare research reporting.
• TRIPOD was updated to TRIPOD+AI to incorporate AI/ML advancements.

Purpose of the Study:

• To develop an updated adherence tool for the TRIPOD+AI reporting guideline.
• To provide a uniform method for assessing reporting completeness in prediction model studies using AI/ML.

Main Methods:

• The original TRIPOD adherence tool was updated and expanded based on the TRIPOD+AI guideline.
• Adherence elements and scoring rules were refined through expert discussion and pilot testing.
• Separate tools were created for model development, evaluation, and combined studies.

Main Results:

• The updated TRIPOD+AI adherence tool includes 37 main items and 136 adherence elements.
• Automated scoring rules are incorporated into the tool.
• Specialized tools were developed for different study types (development, evaluation, combined).

Conclusions:

• A uniform approach to assessing TRIPOD+AI adherence facilitates cross-study comparisons.
• This tool aids in monitoring reporting quality over time.
• It incentivizes primary study authors to improve reporting compliance.