WISeR should pay its reviewers for accuracy, not savings. The Medicare model began in January 2026 and adds prior authorization for a set of services in traditional Medicare in 6 states. Private technology companies review the requests. They are paid a share of the spending their reviews avert.
The first year has been rough. Records released through a public records lawsuit show requests that waited weeks, and 1 that waited 83 days. CMS sent formal warnings to at least 4 of the 6 vendors in the first 3 months. In Washington, the vendor did not affirm 53% of the 6,096 requests it decided in the first quarter. CMS says the early problems have been fixed and that a non-affirmation is not a denial.
Delays can be fixed with staff and software. The payment design is harder to fix, because it shapes every close call. A way to pay reviewers for getting decisions right comes at the end.
Key takeaways
- The rollout was rough. CMS sent formal warnings to at least 4 of the 6 vendors in the first 3 months. A single request went unanswered for 83 days.
- First answers often changed. Washington’s vendor told CMS it affirmed 63.7% of requests on first review and 84.3% after resubmission.
- Pay follows averted spending. Vendors earn a share of the spending their reviews avert, reported at 25%. Poor quality scores cut that pay by about 10% at most.
- Appeals miss dropped requests. A request that ends without a service produces no claim to appeal. Yet a dropped request is one the vendor gets paid for.
- Outcome data remain unpublished. CMS has promised a 6-month report. It has not released appeal or overturn rates.
- Congress is split. The House Appropriations Committee voted unanimously in June to defund the model. A Senate vote to end it failed 46 to 50 in July.
- We think reviewers should be paid for accuracy. A fee per review, scaled by agreement with a blinded independent re-review, would remove the tilt toward no.
How WISeR works
CMS unveiled WISeR in June 2025 and launched it on January 1, 2026. The Wasteful and Inappropriate Service Reduction model is scheduled to run through 2031. Each of the 6 states has 1 vendor, and each vendor works with the state’s Medicare Administrative Contractor, according to the CMS provider fact sheet.
| State | Vendor | Medicare contractor |
|---|---|---|
| Arizona | Zyter | Noridian |
| New Jersey | Genzeon | Novitas |
| Ohio | Innovaccer | CGS |
| Oklahoma | Humata Health | Novitas |
| Texas | Cohere Health | Novitas |
| Washington | Virtix Health | Noridian |
The model covers 13 core services in its first year. Examples include electrical nerve stimulators and epidural steroid injections. It adds 2 skin substitute services in places with an active local coverage policy. Every service is governed by an existing Medicare coverage policy, so WISeR adds a review step rather than new criteria. Our January explainer described the design in detail.
A provider sends a request before the service or accepts a review of the claim before payment. The vendor’s software screens each request, and a clinician must review every non-affirmation. CMS’s FAQ says non-affirmations cannot be performed solely by technology. Standard decisions are typically issued within 3 days and expedited ones within 2.
What a non-affirmation means
A non-affirmation is not a denial in the legal sense. The provider can resubmit as many times as it likes and can ask for a peer-to-peer discussion. It can also deliver the service anyway and submit a claim. The patient must first receive a notice that Medicare may not pay. If the claim is then denied, the usual Medicare appeals apply.
That design is gentler than a denial on paper. In practice a non-affirmation pushes the provider to resubmit or to drop the service. Delivering it anyway puts the provider or the patient at financial risk.
How vendors are paid
Vendors are not paid per review. They are paid a share of the spending their reviews avert, calculated only from requests that did not end in a paid claim, according to the CMS fact sheet. They are paid once per beneficiary for a service, however many times a request is resubmitted. They are not paid for non-affirmations that are overturned on appeal, according to CMS’s payment methodology.
CMS has not published the share in its public documents. KUOW reported it at 25% of the cost of averted procedures. Pay can be reduced by up to 10% for slow or inaccurate decisions.
What the first year showed
Most of what is known comes from records CMS released to the Electronic Frontier Foundation, which sued for them under the Freedom of Information Act in March 2026. EFF published about 1,000 pages in September and more than 800 more in October.
Delays and warnings
The first release showed many requests waiting well past the 72-hour goal in the first months. A single request went unanswered for 83 days. EFF later noted that the turnaround figures cover the whole request pathway, including steps the vendors do not control.
CMS’s own records called the launch timeline extremely condensed, according to EFF. Innovaccer told CMS about a month before launch that it would go live without full functionality or full testing. It said it would affirm every request automatically until its system was ready, according to EFF. CMS kept the launch date.
The second release showed that CMS sent requests for information or corrective action to at least 4 vendors in the first 3 months. The letters cited delays and technical failures. They also cited poor communication with providers. Nextgov described the letters in detail.
- Innovaccer in Ohio. CMS asked for a corrective action plan in January. It followed with a second request in February, after finding that the vendor’s provider portal was not fully working by the deadline.
- Genzeon in New Jersey. A February letter cited a backlog of cases and duplicate tracking numbers.
- Zyter in Arizona. A February letter said determination letters were not going out on time.
- Virtix in Washington. A March letter cited slow decisions and questionable denials. Providers had also reported blank determination letters.
Providers felt the delays. The University of Washington’s medical system reported in April that procedures once approved within days were taking weeks, according to KUOW.
CMS says the picture has changed. A spokesperson told Nextgov that no corrective action plans are active and that all vendors are compliant. Virtix’s plan closed in August, and the company now reports average turnaround of about 1.2 days.
Affirmation rates
The clearest numbers come from Washington. Virtix did not affirm 53% of the 6,096 requests it decided in the first 3 months of 2026, according to KUOW. The company blamed provider misunderstanding and faxes sent to the wrong agency. It told CMS that 63.7% of requests were affirmed on first review and 84.3% after resubmission, according to Nextgov.
That gap deserves attention. About 1 in 5 Washington requests was turned down first and approved later. Each of those took at least another round. Each round took time from a clinic and from a patient. Some of the first answers were right, because the request lacked a needed document. Others may have been wrong. The records do not show which.
The comparison with private plans is imperfect but useful. Medicare Advantage insurers denied 7.7% of about 52.8 million prior authorization requests in 2024, according to KFF. WISeR’s services were chosen because CMS considers them prone to overuse, so higher rates are expected. A first-quarter rate above 50% is still far outside what plans report.
CMS urges caution with these figures. It told Nextgov that non-affirmation data do not equal coverage denials. That is true, and it is also the problem. The model’s most important outcome sits between a non-affirmation and a paid claim, and nobody has published it.
What CMS has not published
CMS has not published affirmation rates by vendor or service. It has not published appeal or overturn rates either. In June, 31 House members asked CMS for those figures by state, according to their letter. They also asked for the share of non-affirmed requests that ended in paid claims. That number would show how often a non-affirmation changed what care a patient received.
CMS told Nextgov that a fuller view of early results will come in a 6-month report. A senior official was more candid. CMS deputy administrator Kimberly Brandt said at an October conference that much about WISeR was not working, according to Nextgov.
How lawmakers and CMS responded
Congress has fought over WISeR all year. In May 2026, the Government Accountability Office found that the model counts as a rule under the Congressional Review Act. That opened the way for a vote to overturn it. The Senate rejected the resolution 46 to 50 on July 16. Our piece on Congress and prior authorization covered that vote.
The House took a different route. On June 9, the House Appropriations Committee voted unanimously to add a rider to the 2027 HHS spending bill, according to Becker’s ASC Review. The rider would bar funds for WISeR or any similar model that adds prior authorization to traditional Medicare, according to Polsinelli. It was the second year in a row that the committee voted to defund the model.
The rider has not become law. Congress funded the government into December with a stopgap signed in September, according to the National Council of Urban Indian Health. The stopgap runs out on December 11. The rider’s fate now rests on the final spending bill.
CMS adjusted the model in smaller ways. In April it published a correction that narrowed the code lists for several services, according to the Federal Register notice. In July it began an exemption program. A clinician who submits at least 10 requests and meets an affirmation threshold set by the vendor, no higher than 90%, is exempt from review for at least a year.
The next question is scope. Senator Elizabeth Warren wrote in September that CMS documents suggested plans to extend WISeR to oncology care and cardiac catheterization. HHS’s nominee for deputy secretary did not confirm those plans, according to Nextgov.
Why the payment design matters
Consider how a vendor earns money. An affirmation earns nothing because the service goes ahead and Medicare pays for it. A non-affirmation earns a share of the service’s cost if the request ends without a paid claim. A non-affirmation that is appealed and overturned earns nothing and may be recouped.
On paper, appeals keep the system honest. A wrong no gets appealed and overturned. The vendor then loses the payment and takes a lower quality score. That check works only when there is something to appeal. If a clinic drops a non-affirmed service, no claim exists. No appeal can follow, and the vendor still gets paid for the request.
Appeals are rare even when they are possible. Medicare Advantage enrollees and providers appealed only 11.5% of denials in 2024, and 80.7% of those appeals succeeded, according to KFF. In an earlier review, HHS’s inspector general found that plans overturned 75% of their own denials on appeal from 2014 to 2016, according to OIG. In a system like that, most wrong answers are likely never corrected.
The quality adjustment is small next to the reward. A vendor that loses up to 10% of its pay for poor scores still earns the rest on every request that ends without a claim. EFF’s review of the model’s data reporting guide found reductions of only 5% to 10%. A federal memo in the records acknowledged that the payment design could encourage denials, according to KUOW.
Pay a reviewer for savings and every close call leans toward no.
What the recovery auditors taught
Medicare has tried paying reviewers this way before. Its Recovery Audit Contractors are paid contingency fees on the improper payments they find. After years of provider objections, CMS announced contract changes in December 2014, according to Becker’s Hospital Review.
Under those changes, recovery auditors receive no contingency fee until the second level of appeal is exhausted, according to CMS’s list of program improvements. They must keep their overturn rate at the first level of appeal below 10%. They must also maintain an accuracy rate of at least 95%.
WISeR repeats the contingency model with a weaker check. A recovery auditor’s finding always produces a claim adjustment that a provider can appeal. A WISeR non-affirmation often produces nothing to appeal at all.
Medicare also shows that review can work without a savings share. Traditional Medicare has required prior authorization for some hospital outpatient services since July 2020, under 42 CFR § 419.83. The contractors that run those reviews are not paid a share of the spending they avert.
Pay reviewers for accuracy
Our July piece argued for tying vendor pay to accuracy as measured by appeals. The first year shows why appeals are not enough. Accuracy has to be measured on a sample of all decisions, including the ones nobody contested.
We think CMS should change what it buys. It should pay for correct decisions, not for averted spending.
- A fee per completed review. The fee would cover the cost of a careful review and would be the same for a yes as for a no.
- An accuracy multiplier. Each quarter independent clinicians would re-review a random sample of each vendor’s affirmations and non-affirmations. They would not see the vendor’s answer or its software’s output. The vendor’s fee would rise or fall with how often it agreed with them.
- A request followed to its end. For every non-affirmation, CMS would record how the request ended. Requests that ended with no service would be sampled more heavily, because those are the cases appeals never reach.
- Public scorecards. CMS would publish first-pass and final affirmation rates by vendor and service each quarter. It would also publish turnaround times and accuracy scores.
| Feature | WISeR today | Pay for accuracy |
|---|---|---|
| What the vendor is paid for | A share of spending averted | A fee for each completed review |
| What raises pay | Requests that end without a paid claim | Decisions that match an independent re-review |
| Cost of a wrong no | Lost pay if overturned on appeal, plus a quality cut of up to 10% | A lower accuracy score on every sampled case, even without an appeal |
| Cost of a wrong yes | The share of savings the vendor gave up | A lower accuracy score |
| Dropped requests | Counted as savings | Tracked and sampled for re-review |
| Public data | Not yet published | Quarterly scorecards by vendor and service |
A wrong yes would still cost the vendor. That matters, because the model exists to catch services that do not meet Medicare’s criteria. Paying for accuracy keeps that goal and removes the tilt.
The savings question does not go away. It moves to where it belongs. CMS has contracted for an evaluation of the model, according to EFF’s records. That evaluation can compare spending in WISeR states with spending elsewhere. Savings should be measured there, at the level of the model, rather than in each vendor’s invoice.
How accuracy would be measured
The standard already exists. Every WISeR service is governed by a national or local coverage policy, so re-reviewers would apply the same written criteria the vendors apply.
The re-review would have to be blind. A reviewer who sees the vendor’s answer tends to agree with it. A reviewer who sees a software score tends to anchor on it. Our piece on auditing payer AI argued that software should be checked the way a human reviewer is checked.
Disagreements would go to a second reviewer. Some cases are close, and reasonable clinicians will differ. A vendor should not lose pay on a case that 2 independent reviewers split on. The multiplier should reflect clear disagreements, not honest close calls.
The samples would need to be large enough to mean something by vendor and by service. That costs money. It also costs far less than a year of wrong answers that no one counts.
Where Redlake stands
Redlake builds tools that health plans use in utilization review. A machine can say yes, but only a person can say no. Every recommendation our tools make shows its reasons, and every decision can be reconstructed. WISeR already shares the first rule. The others would make it stronger.
Publish how often clinicians overrule the software. WISeR requires a clinician to review every non-affirmation. The rate at which those clinicians reverse the software shows whether that review is real. A rate near zero would be a warning.
Put the reason in every non-affirmation. A non-affirmation should name the criterion that was not met and the document that was missing. Blank determination letters are the opposite of that standard. Our piece on denial letters argued that a denial should read like a clinical note.
Keep the decision reconstructable. The record should show what the software recommended and who decided. It should also show which version of the criteria the decision relied on, as our piece on denial audit trails described.
Pay for accuracy. A reviewer should be paid the same for a correct yes as for a correct no. Savings belong in the model’s evaluation, not in the reviewer’s pay.
Exempt the clinicians who earn it. The exemption program is a good start. Our review of state gold card laws found that fewer than 4% of Texas physicians qualified in that program’s first years. The threshold decides whether an exemption matters.
Publish the outcomes. Affirmation rates mean little without overturn rates and without the share of non-affirmed requests that ended with no care at all.
What to watch in the rest of 2026
- The 6-month report. Watch whether CMS publishes appeal and overturn rates along with affirmation rates.
- The December spending deadline. Watch whether the House rider survives in a final spending bill.
- Year 2 services. Watch for changes to the service list before January 2027. A move into cancer care or cardiac catheterization would widen the model.
- Exemptions. Watch how many clinicians earn an exemption and where vendors set their thresholds.
- More records. EFF’s lawsuit continues, and CMS is still producing documents.
- Payment terms. Watch whether CMS changes how vendors are paid for the model’s second year.
About this analysis
Method. The sources are CMS’s public WISeR documents and the records EFF obtained through its lawsuit, as described by EFF and by news reports through November 2, 2026. The model’s payment design is compared with the Recovery Audit Program and with Medicare Advantage appeal data. The pay-for-accuracy design is our proposal.
Data notes. Vendor-level figures come from records and company statements reported by KUOW and Nextgov, not from CMS publications. Turnaround figures in the records cover the whole request pathway. The 25% share was reported by KUOW and does not appear in CMS’s public documents. Medicare Advantage figures cover a different program and different services.
Sources. CMS WISeR provider fact sheet and FAQ (2026). CMS WISeR payment methodology. CMS WISeR exemption program fact sheet (July 2026). Federal Register correction notice CMS-5056-CN (April 2026). EFF (September 2026). EFF (October 2026). KUOW (October 2026). Nextgov (November 2026). Letter from 31 House members to CMS (June 2026). Becker’s ASC Review (June 2026). Polsinelli (June 2026). National Council of Urban Indian Health (September 2026). KFF (January 2026). HHS-OIG report OEI-09-16-00410 (September 2018). CMS Recovery Audit Program improvements (2016).
Frequently asked questions
What is the WISeR model?
WISeR is a CMS Innovation Center model that adds prior authorization for selected services in traditional Medicare. It runs in 6 states from 2026 through 2031, with a technology vendor reviewing requests in each state.
Which services does the model cover?
The first year covers 13 core services. Examples include electrical nerve stimulators and epidural steroid injections. Skin substitute services are added where a local coverage policy is in place.
Is a non-affirmation a denial?
Not formally. The provider can resubmit or deliver the service and submit a claim. If Medicare then denies the claim, the usual appeal rights apply.
How are the vendors paid?
They receive a share of the spending their reviews avert. KUOW reported the share at 25%. Pay is reduced for poor quality scores and withheld for non-affirmations overturned on appeal.
Can patients appeal a decision?
Patients can appeal a denied claim through Medicare’s usual process. A non-affirmation by itself is not an appealable payment decision.
Does software make the decisions?
Software screens requests, but CMS requires a clinician to review every non-affirmation. CMS has not published how often those clinicians disagree with the software.
Will the model be defunded?
The House Appropriations Committee voted to bar its funding in June 2026. That rider has not become law, and the model continues while Congress works on a final spending bill.