Summary#
The Protecting Patients from Automated Denials Act is an introduced bill that would amend title XVIII of the Social Security Act. It would set conditions for Medicare Advantage plans to deny prior-authorization requests based on artificial intelligence (AI); its stated goal is to establish requirements for that use. The bill is not law, and the materials do not include the current law needed to verify what rules already apply.
- For plan years starting on or after January 1, 2027, a plan using AI output as a basis for a denial would have to get advance review and approval from a qualified physician reviewer, under the plan medical director’s clinical direction.
- The physician would have to attest that they used independent medical judgment, that AI did not generate or dictate the denial, and that any software used to prepare it served only administrative purposes.
- The plan would have to give the requesting provider the attestation and opportunities to speak with the reviewer and discuss the proposed clinical basis for the denial. It would also have to tell the provider and the patient that AI was used and give them the reviewer’s National Provider Identifier number.
- The plan would have to keep records about the AI use, review, and information provided for at least 10 years, and report AI-based denials to the Secretary every 90 days.
- The Secretary could audit and inspect related AI use. The bill would also require the Secretary to issue a final rule within one year after enactment.
What it means for you#
- Medicare Advantage patients: If a plan denies your prior-authorization request based on AI output, it would have to tell you AI was used and identify the physician reviewer. The bill does not give you the provider’s attestation or a direct opportunity to speak with the reviewer.
- Providers who submit requests: For an AI-based denial, the plan would have to provide you the reviewer’s signed attestation and an opportunity to communicate with the reviewer and discuss the proposed clinical basis.
- Medicare Advantage plans: Plans that impose prior-authorization requirements would have to report any denials based on AI output, even though the review, disclosure, and record rules apply to AI-based denials. The bill does not require plans to use AI or prior authorization.
- The Secretary: The Secretary would have to issue a final rule and would have the power, but not a duty, to audit or inspect plans’ related AI use.
- Patients generally: The bill does not change which services are covered, create a new appeal process, or ban AI use in prior authorization.
Money#
No publicly available information on the bill’s estimated costs or savings.
- Plans would have to arrange qualified physician review, provide information, keep records for at least 10 years, and submit reports. The bill provides no cost estimate.
- The bill includes no appropriation, tax, fee, or fine.
What is unclear#
- The bill does not set a deadline or detailed process for the provider’s discussion with the reviewer, or explain what happens if the provider cannot reach the reviewer.
- It does not specify consequences for a plan that fails to meet the requirements.
- The bill leaves unclear how to judge whether a reviewer’s specialty “correlates” with the requesting provider’s specialty.
- The bill’s AI definition is broad. The final rule may clarify how it applies to particular tools.
- The supplied material does not include the amended law, so the bill’s relationship to current prior-authorization rules is not verified.
Case for#
- A possible argument for the bill is that a physician’s independent medical judgment could serve as a check on AI output before an AI-based denial is issued.
- Giving providers the attestation and a chance to discuss the proposed clinical basis could help them understand and challenge a denial.
- Long-term records, recurring reports, and audit authority could make it easier for the Secretary to examine how plans use AI.
Case against#
- One concern is that the bill requires physician review and documentation but does not specify consequences for noncompliance.
- The bill does not explain how quickly providers must get to speak with reviewers, or how that opportunity would work in practice.
- The staffing, recordkeeping, and reporting duties could impose costs on plans, but the available material provides no estimate.
- The required reports and disclosures may not by themselves show whether AI-based denials were medically appropriate; the bill does not set a standard for that assessment.