The Consequence Library · How records are made and graded
After UnitedHealth's naviHealth subsidiary deployed the nH Predict algorithm to manage Medicare Advantage post-acute care, the skilled-nursing-facility denial rate rose from 1.4% in 2019 to 12.6% in 2022, a ninefold increase documented by the U.S. Senate Permanent Subcommittee on Investigations, while elderly patients were cut off from prescribed rehabilitation coverage.
Evidence caveat. The operational allegations, including the 1% adherence pressure and the ~90% appeal-reversal figure, come from the plaintiffs' complaint and are contested by UnitedHealth. They are reproduced here as allegations. The denial-rate change and the discovery order are documented by the Senate subcommittee and the court respectively. The court denied the request for the algorithm's source code; nothing here should be read as saying source code was ordered produced.
What happened
nH Predict, a predictive model trained on roughly 6 million patient records, projected recovery trajectories and length-of-stay targets. A class action filed on 14 November 2023 in the District of Minnesota (Estate of Gene B. Lokken v. UnitedHealth Group, No. 0:23-cv-03514) alleges that these projections became the effective coverage decision, that case managers were pressured to keep stays within 1% of the algorithm's prediction, and that roughly 90% of appealed denials were reversed while only about 0.2% of patients appealed. Those operational claims are allegations, not adjudicated facts, and UnitedHealth contests them. The confirmed record is separately significant: Gene Lokken's rehabilitation coverage was terminated after 19 days and his family paid approximately $150,000 out of pocket before his death; on 13 February 2025 a federal judge allowed breach-of-contract and good-faith claims to proceed; and on 9 March 2026 Magistrate Judge Shannon Elkins granted six of seven categories in the plaintiffs' motion to compel, ordering production of records on how nH Predict was built and used, and of internal AI Review Board materials, with an April 2026 production deadline. The court denied other requests, including the algorithm's source code and underlying medical guidelines. The Senate PSI's October 2024 "Refusal of Recovery" report, based on over 280,000 pages of internal documents, established the denial-rate shift.
Where control failed
Human review was part of the process. Plaintiffs allege that reviewers were pressured to remain close to model predictions, allegations that UnitedHealth contests. Separately, the Senate investigation documented the increase in denial rates. The record therefore distinguishes the existence of human review from the unresolved question of how much practical discretion reviewers exercised.
The authority question
The case raises whether formal reviewer authority was matched by practical discretion to depart from algorithmic recommendations. That question remains contested and should be read in the context of both the litigation and the Senate record.
What could be proven afterward
The decision records existed but were not available to the people affected. It took a class action and a 2026 discovery order to compel production of how the algorithm was built and used, and the Senate subcommittee needed over 280,000 pages of internal documents to establish the denial-rate change. At decision time, a patient received an outcome with no accessible record of how it was reached.
Control state, before and after
Before the consequence
Human review present in the process. No external, contemporaneous record of the human decision available to the patient or to a reviewer outside the company.
After the consequence
Senate PSI report October 2024. Claims allowed to proceed February 2025. Discovery order compelling AI Review Board materials March 2026.
Sources
- Primary: U.S. Senate Permanent Subcommittee on Investigations, "Refusal of Recovery", Oct 2024 · https://www.hsgac.senate.gov/subcommittees/investigations/
- Primary: Estate of Gene B. Lokken v. UnitedHealth Group, No. 0:23-cv-03514 (D. Minn.) · https://www.courtlistener.com/docket/68015377/estate-of-gene-b-lokken-v-unitedhealth-group-inc/
- Secondary: ArentFox Schiff, "Federal Court Orders Broad Discovery Against UHC in AI Coverage Denial Lawsuit" · https://www.afslaw.com/perspectives/alerts/federal-court-orders-broad-discovery-against-uhc-ai-coverage-denial-lawsuit
- Secondary: Becker's Payer Issues, "Judge orders UnitedHealth to hand over broad discovery in AI coverage denial case" · https://www.beckerspayer.com/legal/judge-orders-unitedhealth-to-hand-over-broad-discovery-in-ai-coverage-denial-case/
Record history
Published 16 August 2026. Load-bearing facts re-verified against the cited sources on 16 August 2026. Corrections and material changes are appended here with their dates. To report an error in this record, write to hello@regulayer.com with the record slug and the source you believe is authoritative.
This record describes what sources establish about a consequence and the control state around it. It separates confirmed fact from source-stated cause and from architectural analysis, and it makes no claim that any control or product would have prevented the outcome. Gap codes identify a failure class, not a remedy.