The medical algorithmic audit

AI Generated Summary*

A viewpoint proposing an audit framework for medical artificial intelligence, adapted from the SMACTR internal auditing method, which moves from scoping through mapping and artifact collection to testing and reflection. It argues that such systems can fail unpredictably, through spurious correlations and hidden stratification, so errors need proactive study. Suggested testing includes examining errors, checking subgroups, and adversarial probing, while failure modes and effects analysis ranks risks. In an illustrative hip fracture audit, the error rate was 2.5% overall but 50% in hips with abnormal bones or joints. Mitigation options are outlined for developers and clinical users, and auditing is framed as a shared responsibility.

*This summary was generated by AI and is published unedited. Oku does not alter these summaries. It may contain errors or omissions and is provided for general informational purposes only. Oku does not guarantee its accuracy, completeness, or reliability. For authoritative information, please refer to the original, peer-reviewed article.