Operations and Monitoring of Predictive Health AI Solutions (CTA-2136)

This standard outlines the methodology and requirements of post-market operations and
monitoring related to the accuracy of predictive health AI solutions. This standard does not
identify real-world metric triggers by which the model might require re-training.
Additionally, this standard does not address:
• Generative health AI solutions and large language models.
• Workflow AI models like scribes.
• Natural Language Processing to review unstructured data in Electronic Medical Records
(EMR) and transform it into structured data.
• How the initial health AI algorithms are trained.