IA_PSPA_34 · Patient Safety in the Use of Artificial Intelligence

Subcategory: Patient Safety and Practice Assessment

Status: One of three activities CMS added for the 2026 performance year. CMS removed eight and modified seven, bringing the total inventory to 104.

Why this matters to our clients: anesthesia practices are adopting AI-assisted tooling across preoperative risk stratification, intraoperative decision support, documentation, and case review. Any practice that has stood up governance around those tools (validation before deployment, monitoring for drift or bias, a defined human-in-the-loop escalation path, an incident process when a tool produces a wrong or unsafe output) is likely already performing the substance of this activity and simply has not been documenting it as an Improvement Activity.

Practices doing AI governance work can review this activity against their actual processes and documentation. Attestation requires performing the specified work for the required period; using an AI tool alone does not establish eligibility.

What to document. CMS has not published a detailed validation-documentation list for this activity in the same format as the older inventory entries, so confirm requirements at https://qpp.cms.gov/mips/improvement-activities before attesting. Based on how CMS has framed AI safety expectations elsewhere, a defensible file would include: an inventory of the AI tools in clinical or documentation use, with each tool's intended use and risk classification; evidence of validation before deployment; a monitoring process for outputs that could influence a clinical decision; a defined human-in-the-loop review and escalation path; and a record of any incident where a tool produced an incorrect or unsafe output, with the response.