Summary
Requires developers and deployers to identify foreseeable clinical decision-support bias risks, developers to mitigate risks and provide documentation, and deployers to monitor and mitigate biased impacts.
Healthcare Implications
Healthcare organizations must evaluate bias risks and monitor covered clinical AI. Vendor documentation is due upon request or initial sale, whichever comes first, and material updates; a limited internal-development documentation exception applies.
Operational Implications
- Developers and deployers must reasonably identify foreseeable bias risks; developers must reasonably mitigate them and deployers take proportionate mitigation steps.
- Deployers must regularly monitor identified clinical decision-support systems for biased impacts; no fixed monitoring interval is specified.