The Role of Artificial Intelligence Model Documentation in Translational Science: Scoping Review

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Health Services Research

Background:

• Artificial intelligence (AI) and machine learning (ML) hold significant promise for healthcare but face challenges in clinical adoption.
• Gaps in transparent documentation for medical modeling software (MMS) hinder the translation of AI/ML tools from research to practice.
• Lack of standardized reporting impedes evidence generation for implementing AI/ML-based clinical decision support.

Purpose of the Study:

• To scope current documentation practices for AI- and ML-based MMS.
• To identify the role of documentation in translating ethical and explainable MMS into clinical workflows.
• To elucidate barriers and facilitators for AI/ML MMS translation.

Main Methods:

• A scoping review was conducted following PRISMA-ScR guidelines.
• Searches of PubMed and reference lists identified publications on AI/ML MMS documentation, ethics, and explainability.
• A 2-stage screening process and data extraction focused on documentation barriers, standards, and recommendations.

Main Results:

• Twenty-one of 115 retrieved papers met inclusion criteria, focusing on ethics and explainability in MMS documentation.
• Key challenges identified include bias, accountability, governance, and explainability.
• Recommendations emphasize proactive evaluation, multidisciplinary collaboration, transparency, traceability, and guiding standards.

Conclusions:

• Addressing documentation-related barriers (bias, accountability, governance, explainability) is critical for MMS clinical translation.
• Transparent, strategic documentation aligned with translational and regulatory science can reduce barriers.
• Improved documentation practices will facilitate the adoption of AI/ML-based MMS in clinical settings.