Medicare built a program that pays private companies a share of the money they save by refusing care, then handed those companies 6.4 million people. In the first three months, the government’s own files show it had to cite four of the six of them for failing at the job. This is not a pilot anymore. It is a live experiment on the sickest, oldest, least mobile population in the country, run by vendors whose software was still being finished while the clock ran.
The program is called WISeR, for Wasteful and Inappropriate Service Reduction. CMS says it exists to screen out low-value care in original Medicare, and it runs in Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington. Six private companies process the requests, one per state, using artificial intelligence alongside human reviewers. It launched on January 1, 2026, less than seven months after it was announced, and it is scheduled to keep going until the end of 2031. Thirteen categories of procedures and devices require permission first: epidural steroid injections, skin substitutes, incontinence devices, kyphoplasty for spinal fractures.
Traditional Medicare never had this layer. Privatized Medicare Advantage did. The people inside the six pilot states now carry the version without the appeals staff, the member services line, or the brand name.
Four of Six — The Vendors the Government Had to Cite in Its Own First Quarter
The paper trail comes from a Freedom of Information Act lawsuit the Electronic Frontier Foundation filed in March. EFF published roughly 1,000 pages in September, and reporting this week filled in the vendor-level detail. CMS sent a request for information or a corrective action plan to four of the six contractors inside the program’s opening months.
Genzeon, covering New Jersey, drew a February 6 letter over a backlog of coverage determinations and the accidental creation of duplicate tracking numbers. Zyter, covering Arizona, was written up on February 27 for missing determination letters and for months of confusion between Medicare Part A and Part B claims — a distinction that decides which coverage rules apply at all. Virtix, covering Washington, was cited on March 17 after providers reported slow determinations, denials that looked wrong, and determination letters that arrived blank.
Innovaccer, covering Ohio, got the first corrective action request of all, on January 7, six days into the model.
A pilot this young should not be generating corrective action plans about its own launch. Corrective action plans are what you write about a program that has been running for years and rotted.
83 Days — The Answer That Never Came
CMS tells the public that WISeR vendors should resolve prior authorization requests within 72 hours, and its own performance specifications treat three calendar days for a standard request and two for an expedited one as the standard being measured. Internal status reports from the program’s first months show a significant share of requests ran far past that.
One request sat unanswered for 83 days. A single prior authorization, 83 days, no determination, logged in a CMS status report and buried on page 234 of a document dump nobody outside a courtroom would ever have read. For a skin substitute or a spinal procedure, 83 days is not a delay. It is a decision made by silence.
The 72-hour promise was written by the same agency that was receiving the status reports showing it broken, week after week, in its own program.
5,944 — Two Companies That Said No More Often Than Yes
Volume tells you what the timeliness numbers cannot. Two of the six vendors together denied 5,944 prior authorization requests in the first three months of 2026. Virtix, the Washington contractor, denied more requests than it approved. That is not a screening function. That is a wall with a form on it.
Compare the money. Vendors can earn up to 20 percent of the program savings their denials and modifications generate. When their quality scores fall — for lateness, for inaccuracy — their payment drops by 5 to 10 percent. One penalty is four times the size of the other, and it points in the opposite direction. A company loses more by approving care than by getting it wrong.
The incentive is not hidden in a footnote. It is the design. The brake is a rounding error next to the accelerator.
Innovaccer — Ohio’s Screening Was Switched Off on Day One
The most damning record is the one that shows the technology was not the thing making decisions in Ohio at all. Innovaccer told CMS in October and December 2025 letters that its provider portal would not be ready for a January launch. It warned the agency the schedule was unrealistic, and it warned early enough that a delay would have cost nothing but a calendar quarter.
CMS launched on time anyway. Innovaccer’s answer was to auto-affirm every request that arrived — approve everything, sight unseen — until the software existed. The agency sent a second corrective action request in February after concluding the first had been ignored, then a formal performance notice of non-compliance.
Read that again. The AI screen that justified the entire experiment was not running in Ohio while the experiment was running in Ohio. And the fix for a broken gatekeeper was to remove the gate.
Three People Crying at a Bedside — What the Providers Wrote
The administrative language stops being administrative in the provider feedback forms. An Ohio clinician, answering the March 2026 survey about Innovaccer, described watching three patients weep at the bedside while waiting on authorization for kyphoplasty and vertebral augmentation, in deep pain and unable to get an answer. A second wrote that patients were calling offices in tears because pain procedures had stalled for weeks with no guidance. A third described cases open for more than six weeks without the tracking number needed to move a surgery forward, while patients were cancelled off the schedule.
Senator Maria Cantwell’s April 2026 letter, drawing on hospital association data, put a harder number on the drift: procedures that took two weeks before WISeR now take four to eight. The patient does not experience that as a utilization statistic. They experience it as a second month of not being able to stand up straight.
46 to 50 — The Vote That Left 6.4 Million People Inside the Experiment
Congress saw it coming and could not stop it. The Government Accountability Office determined in May 2026 that the WISeR model notice is a rule under the Congressional Review Act, which meant it should have gone to Congress before it took effect. Democrats used that opening in July. The resolution to terminate the program failed 46 to 50, on a party-line vote.
The House has moved further since. Its Appropriations Committee advanced a bipartisan amendment to suspend WISeR funding, and Energy and Commerce unanimously advanced a bill to end it. None of it is law. Senator Elizabeth Warren wrote in September that documents suggest CMS wants to add oncology care and cardiac catheterization to the list — the services where a bad week is not a bad week.
Kimberly Brandt, the CMS deputy administrator, said the quiet part at a conference last month: a lot about WISeR is not working. Asked what would fix it, she said she did not have an answer.
The agency running the program has no repair plan and no intention of pausing, and the vendors are paid by the denial. Somewhere in Ohio, a scheduler is on hold with a company that has already been cited twice.
Sources: EFF, STAT, Tech Times.