Guidelines and standard frameworks for artificial intelligence in medicine: a systematic review

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Guidelines

Background:

• The integration of artificial intelligence (AI) into clinical settings necessitates robust guidelines.
• Evolving challenges in AI implementation require updated standard frameworks.
• This review focuses on evaluating the quality of existing AI in medicine guidelines.

Purpose of the Study:

• To evaluate the quality of current reporting guidelines for artificial intelligence (AI) in medicine.
• To summarize ethical frameworks, best practices, and recommendations for AI implementation.
• To identify areas for improvement in AI guideline development and applicability.

Main Methods:

• Utilized the Appraisal of Guidelines, Research, and Evaluation II (AGREE II) tool to assess guideline quality across six domains.
• Conducted a systematic search of two databases and manual searches, yielding 4975 initial studies.
• Selected eleven articles for data extraction based on predefined eligibility criteria.

Main Results:

• Guidelines generally demonstrated high quality in scope, purpose, and editorial independence.
• Significant variability was observed in the applicability and rigor of guideline development.
• Initiatives like TRIPOD+AI and CONSORT-AI showed high quality, particularly in stakeholder involvement, but applicability remains a challenge.
• Reproducibility, ethical, and environmental aspects of AI in medicine require further attention.

Conclusions:

• Highlights the need for integrated, comprehensive reporting guidelines adhering to FAIR principles (Findability, Accessibility, Interoperability, Reusability).
• Emphasizes the importance of transparency and open science for sustainable digital health research.
• Identifies current advantages, challenges, and limitations of AI in medicine reporting guidelines.