Monday, August 3, 2026
Are Hospitals Using AI to Deny Care
Well-supported — high-quality sources agree
14 sources · 13 support, 1 neutral
What we can confirm
Authoritative investigations, lawsuits, and policy reports show that large health insurers and Medicare Advantage plans have used AI tools and predictive algorithms in ways that contributed to denials or premature termination of post‑acute care and other services, sometimes overriding or constraining clinical judgment.[1][4][6][7][11][12][14] Multiple class‑action suits allege that UnitedHealth and its NaviHealth nH Predict model were used to cut off payment for skilled nursing and rehab stays despite physicians’ determinations of medical necessity, and a federal judge has ordered broad discovery in one such case.[1][6][9][11][14] A Senate Permanent Subcommittee on Investigations report and follow‑up letters criticize UnitedHealthcare, Humana, and CVS for using AI and other predictive technologies to deny or delay post‑acute care authorizations for vulnerable seniors in Medicare Advantage.[4][7][8] The AMA has surveyed physicians, finding widespread concern that health plans’ AI use is increasing prior‑authorization denials and overriding good medical judgment.[2] While insurers like UnitedHealth and Optum publicly deny that AI makes “final” adverse coverage decisions and characterize these systems as decision‑support rather than sole determinants of denial, credible reporting and congressional scrutiny indicate that AI‑driven tools materially shape when care is cut off or claims are denied, especially in Medicare Advantage post‑acute care.[3][4][7][9][11][14] Regulators such as CMS have issued guidance stating that AI predictions cannot be used alone as the basis for denying or downgrading services, implicitly acknowledging both current use and the risk of over‑reliance on such tools.[11][13] Overall, there is strong evidence that AI is already being used within payer workflows in ways that significantly contribute to care denials, though typically framed as supporting rather than formally replacing human decision‑makers.
Sources
Checked August 3, 2026 · we'll re-check as this develops
Community verdict
Log in to react to the verdict.
Not enough reactions yet to summarize community view.
💬Discussion