To conduct a scoping review to identify studies validating commercially available AI-based products, assess demographic reporting, and provide specific recommendations for improving such reporting.
Key Findings:
Limited evidence exists on subgroup bias in commercial medical AI, raising concerns about its impact on patient care.
Few studies report demographic subgroup performance metrics, indicating a need for better practices.
Current regulations do not mandate peer-reviewed evidence for AI product approval, which could lead to unchecked biases.
Interpretation:
The review highlights a significant gap in the reporting of demographic subgroup performance in AI-based radiology products, which may contribute to algorithmic bias and affect patient outcomes.
Limitations:
Limited data availability on commercial medical AI products, which restricts comprehensive analysis.
No regulatory requirements for comprehensive demographic reporting, hindering accountability.
Conclusion:
The findings suggest a need for improved reporting practices to facilitate future meta-analyses of algorithmic bias.
