Due Date Was April 29. UHC Asked for More Time.

Introduction

A federal judge told UnitedHealth to hand over the nH Predict documents by April 29. On April 29, UnitedHealth asked for more time, and the court said yes. The case has since slid to December 2027. In between, UHC made five prior authorization announcements, not one of them touching the Medicare Advantage post-acute care at the center of the lawsuit, and the Massachusetts attorney general sued for $100 million the week after. The documents are still sealed.

The Docket Is the Story

Back in April I wrote about the deadline itself: a magistrate judge had ordered UnitedHealth to produce internal records for its AI claim-denial tool, and the clock was running. That piece asked what the documents would show. This one answers a different question, which is why we still don't know.

Here's the timeline, pulled straight from the CourtListener docket (No. 0:23-cv-03514, D. Minn.). On March 9, Magistrate Judge Shannon Elkins granted most of the plaintiffs' motion to compel. On March 30, UHC asked for more time, and on March 31 the court reset the deadline to April 29. On April 29 itself, UHC filed a second motion for an extension, this one unopposed. The court granted it May 4. Then on May 15 both sides jointly moved to amend the entire schedule, and on May 21 the court signed off: fact discovery now due March 11, 2027, trial readiness pushed to December 6, 2027. The old trial-ready date had been September 2026. In one order, the case slid more than a year.

That's the spine of the thing. Every extension was either filed by UHC or filed jointly. None of it is illegal, none of it is even unusual for complex litigation. What makes it worth your attention is what was happening in public at the exact same time. Between the March 9 order and the documents actually changing hands, UnitedHealth ran a months-long campaign telling everyone how much it cares about cutting red tape. That's the gap I keep circling back to. In the court record the company is managing a clock; in the press releases it's reforming. Both are true, and they cover the same eight weeks.

What the Reform Announcements Carefully Avoided

Count them. On April 20, UHC exempted rural hospitals from prior authorization. On April 24, it championed industry-wide electronic prior authorization standards. On May 5 it pledged to eliminate 30% of its prior authorization requirements by year-end. On May 29 it cut roughly two-thirds of pediatric prior authorization rules. And on March 31, the same day the court extended its production deadline, it posted prior authorization metrics on its website for the first time, claiming only 2% of medical services need prior authorization and 92% of requests clear in under 24 hours.

Read that list against the lawsuit. The Lokken case is about Medicare Advantage post-acute care: skilled nursing, rehab, the care an elderly patient needs after a hospital stay. The 30% reduction applies to commercial and Medicaid plans, the pediatric cuts cover kids, and the rural exemptions cover rural hospitals. None of it touches the population the algorithm allegedly burned. You can run a real reform program and still route every headline around the one product line a federal court is making you explain.

The Numbers the Court Pointed At

The reason any of this matters comes down to a denial rate that moved. According to the Senate Permanent Subcommittee on Investigations report from October 2024, which Judge Elkins cited directly in her discovery order, UHC's post-acute care denial rate more than doubled after it deployed naviHealth and nH Predict in 2019, going from 8.7% in 2019 to 22.7% in 2022. Skilled nursing denials over the same window climbed from 1.4% to 12.6%. Those are the established findings the court used to justify expanding discovery back to January 2017, two years before the tool went live.

Then there are the plaintiffs' allegations, which are sharper and not yet proven. The lawsuit claims nH Predict had a roughly 90% error rate, measured by how often denials got reversed on appeal, and that only about 0.2% of affected policyholders ever appealed. If both numbers hold up, the business logic writes itself: deny by algorithm, bank on almost nobody fighting back, and reverse the handful who do. UHC hasn't conceded either figure, and the court denied the plaintiffs' request for the source code itself, so the algorithm's internal logic stays a black box.

Which makes the contradiction in Hemsley's testimony land harder. On January 22, 2026, CEO Stephen Hemsley told the House Ways & Means Committee that UHC's denial rate runs "less than 2%." His own testimony PDF never mentions nH Predict, naviHealth, post-acute care, or the Lokken case. When Rep. Robin Kelly pressed him on whether AI was still improperly denying claims, he couldn't say. The court record the company is fighting to keep sealed puts post-acute denials at 22.7%. The documents that would reconcile a 2% headline with that 22.7% reality were due April 29, the same day his company asked for more time.

Who Benefits From the Clock

UnitedHealth's shareholders and executives, and the math is in the filings. The surviving claim in Lokken is breach of the implied covenant of good faith and fair dealing, the bad-faith doctrine that can turn a wrong denial into punitive damages. If the sealed documents show UHC knew its denial methodology produced systematic errors and kept deploying it, that exposure balloons. Every quarter the case doesn't resolve is a quarter that risk stays unpriced.

Look at how Hemsley gets paid. Per the 2026 proxy statement, his package is a $1 million salary plus a $60 million stock option grant that cliff-vests in May 2028, with a holding requirement added in February 2026 that bars him from selling the net shares before May 2030. None of his compensation depends on claim approval or denial-reversal rates; it rides entirely on the stock price between now and 2030. A litigation timeline that runs to December 2027 at the earliest is a timeline where the share price gets years to recover without a punitive-damage verdict baked into it. Meanwhile Q1 2026 came in strong: $7.23 adjusted EPS, a medical benefit ratio that improved to 83.9%, full-year guidance raised above $18.25 a share. A company posting numbers like that can outwait almost anyone.

There's a quieter beneficiary too. CMS had the same Senate denial-rate data the court cited, and on May 1, 2026 it fined UHC $48,869 for a contract administration issue, a rounding error against a company guiding to more than $439 billion in revenue. The same agency is now building its WISeR model in six states using the exact vendor-incentive structure the Senate documented inside UHC's denials: pay the contractor a share of the spending it reduces. The discovery process is doing the oversight work the regulator declined to do.

The Pattern Isn't One Courtroom

On May 29, Massachusetts Attorney General Andrea Campbell sued UnitedHealthcare for at least $100 million, accusing it of defrauding the state's MassHealth Senior Care Options program from 2015 through 2025. The mechanics differ from Lokken, but the structure rhymes. Campbell's complaint alleges UHC coded members as sicker than they were to pull higher payments, and that its own internal reviews flagged improper classifications starting in 2018 and 2019. The company, per the AG, never disclosed those findings and never repaid the state. Campbell's words: UHC "knowingly violated these obligations by manipulating health assessments to increase its profits." UHC calls the suit meritless.

That 2018-2019 detail is the one that sticks with me. It's the same shape as the question at the heart of Lokken, which is whether UHC knew the algorithm was producing bad denials and kept running it anyway. In Massachusetts, the allegation is that the company caught its own overcharging years before a regulator did and said nothing. If that's how UHC handles what its internal reviews turn up in Boston, it's fair to ask what its sealed documents are hiding in Minnesota.

The Bottom Line

The March 9 order wasn't vague about what UHC has to turn over. The court specifically required the naviHealth acquisition documents and the projected cost savings UHC expected from those post-acute claims, plus the performance reviews and compensation data for the medical directors and care coordinators who denied coverage to the 300 putative class members. That's the heart of it: did the financial model assume those denials, and did the people signing off on them get rewarded for the volume? Those are the records still behind the protective order, and a December 2027 trial date means they could stay there for another eighteen months.

So if UHC ever denied your post-acute or skilled nursing claim, the file that would tell you whether an algorithm made that call, and whether the people who approved it were paid to, exists. It's been produced to the plaintiffs and sealed from everyone else. The next time someone asks UnitedHealth to explain the gap between a 2% denial rate and a 22.7% one, will the answer come from those documents, or from another press release about a population the lawsuit was never about?