Bias and equity assessment
The mapping
| Framework | Where this control sits |
|---|---|
| Joint Commission RUAIH | Focus area 3 — Risk and bias reduction |
| CHAI governance playbooks | Playbook 5 — Risk and impact assessments |
| NIST AI RMF | MEASURE |
| HTI-1 | HTI-1 source attributes include information on fairness and validity |
The artifact: Local bias assessment
Who signs it: The quality or equity officer, jointly with the AI committee
What an assessor actually asks for
Performance broken out by the subgroups you actually serve, on your own population, with a date. Vendor-reported fairness on someone else’s population is an input, not an answer.
Why the mapping is not obvious
The September 2025 guidance asks both that tools were bias-tested during development and that you regularly monitor and audit deployed tools. Most programmes do the first by asking the vendor and never do the second. The second is the one RUAIH’s monitoring area will also look for, which is why this control appears in two places.
The most common failure
Bias work done once, at selection, and never repeated. A model fair on the population you had at go-live can drift as your population, referral patterns or documentation habits change. Bias assessment belongs on the monitoring calendar, not in the procurement file.
Where this sits in the whole map
This is one control in the RUAIH ↔ CHAI ↔ NIST crosswalk. The artifact itself is specified at Local bias assessment.
Written and reviewed by Neel Chauhan, MD MBA, physician-executive and founder of the Healthcare AI Institute. Last reviewed 2026-07-30.
Generated from data/crosswalk.yaml, where the mapping and the commentary for each control are authored individually. Reviewed on each framework revision.
The Institute accepts no vendor sponsorship, holds no vendor equity and takes no referral fees.