Alleged lawsuit targets UnitedHealth prior‑authorization AI 'nH Predict' after social posts claim errors
- A real lawsuit already exists — and it is not new today. Medicare Advantage members sued UnitedHealth and NaviHealth in November 2023 over denials tied to nH Predict. - The headline number in circulation is from the complaint: it says nH Predict had a 90% error rate, while just 0.2% of denied claims were appealed. - What changed more recently is procedural. In February 2025, a judge let the case move forward, as AI-driven prior authorization faces wider Medicare scrutiny.
Health insurance prior authorization is the thing here — the system that decides whether an insurer will pay for care before or during treatment. The stakes are simple. If that system gets an answer wrong, a patient can lose rehab days, face a surprise bill, or get pushed out of care early. The gap is that social posts are talking like a brand-new lawsuit just landed against UnitedHealth’s “nH Predict” tool. But turns out the core lawsuit is older, and the real news is that it has survived an early effort to kill it. ### Is there actually a lawsuit? Yes. A class action was filed on November 14, 2023, against UnitedHealth Group and its subsidiary NaviHealth in federal court in Minnesota. The plaintiffs are Medicare Advantage beneficiaries and families who say the companies used an algorithm called nH Predict to cut off post-acute rehab coverage too early. That means the social posts are pointing at a real case — but not a fresh filing from May 2026. (statnews.com) ### What is nH Predict supposed to do? nH Predict is described in the lawsuit and follow-up reporting as a tool that estimates how long a patient should need care in settings like nursing homes or rehab facilities after a hospital stay. The allegation is not just that the tool existed. It is that UnitedHealth pushed clinicians to stay very close to the algorithm’s predicted length of stay when making payment decisions. UnitedHealth has denied that the tool itself makes final coverage decisions and says coverage is based on Medicare rules and plan terms. (statnews.com) ### Where did the “90% error rate” claim come from? From the complaint itself. The suit says nH Predict had a roughly 90% error rate, based on denials later reversed in internal appeals or by administrative law judges. That number is an allegation, not a court finding. It matters because it is the sharpest claim in the case, and it is the piece now getting flattened into social-media shorthand. (statnews.com) ### What about the “80% reversed on appeal” stat? That number is real in a broader Medicare Advantage sense, but it is not specific proof about nH Predict. KFF’s latest analysis of CMS data says 80.7% of prior-authorization denials that were appealed in Medicare Advantage were partially or fully overturned in 2024. The catch is that only a small share of denials get appealed at all. So “most appealed denials get reversed” does not mean “most denials are wrong,” but it does tell you the system has a serious accuracy or process problem. (statnews.com) ### What changed in the case? The big procedural turn came on February 13, 2025. A judge ruled the lawsuit could move forward, rejecting UnitedHealth’s argument that patients had to finish Medicare’s long appeals process before suing. That matters because discovery — the phase where plaintiffs can seek internal documents — may reopen. In plain English, the case moved from “might get tossed” to “still alive, and potentially more revealing.” (kff.org) ### Why is this flaring up again now? Because automated prior authorization is under broader pressure. CMS launched the WISeR model on January 1, 2026, to use AI and machine learning with human review in traditional Medicare for selected services in six states. Providers and hospitals say the rollout has been rocky, with reports of approval times stretching from days or two weeks to 15–20 days or even 4–8 weeks in some Washington cases. That wider fight makes every insurer-AI story feel newly urgent. (statnews.com) ### So what should you take from the posts? Basically — they are mixing a real case, an old filing date, and broader industry stats into one explosive claim. The lawsuit is real. The 90% figure is an allegation from the complaint. The 80%-plus reversal figure is a broader Medicare Advantage appeal statistic. And the most concrete recent development is that the UnitedHealth case is still moving, not that it suddenly appeared this week. (cms.gov) ### Bottom line The story is less “a surprise new lawsuit hit UnitedHealth today” and more “an existing AI-denial lawsuit is gaining staying power while the whole prior-authorization system comes under heavier scrutiny.” If the case gets deeper into discovery, that is when the most important facts — how nH Predict was used, and how often it overrode clinical judgment in practice — could get much clearer. (statnews.com 1) (statnews.com 2)
