Why WISeR providers should keep their own record while Congress argues

The week

Since January 2026, for a selected list of services, traditional Medicare in six states has required an extra step, and it involves software. Before that, a patient in Ohio whose surgeon recommended one of those procedures faced no such step: for decades, the surgeon could schedule it, perform it and bill for it.

The program is called the Wasteful and Inappropriate Service Reduction model, known as WISeR. It started in January 2026 and is scheduled to run through 2031 in Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington. For the selected services, providers in traditional Medicare must request prior authorization, meaning approval before the service is delivered, or face prepayment review, where the claim is examined before Medicare pays it. Prior authorization has long been familiar in private insurance and Medicare Advantage; for traditional Medicare it is new territory.

The reviews are carried out by technology companies using tools that include artificial intelligence. They are paid a share of the spending their reviews avert, which ties their income directly to what they decline. A licensed clinician must review any denial, so software alone cannot say no.

The model’s footing was contested all summer. On May 12, the Government Accountability Office, Congress’s independent auditor, concluded that the CMS notice implementing WISeR is a rule under the Congressional Review Act. That law gives Congress a fast-track way to overturn agency rules, and the finding opened the door to a vote. On June 9, the House Appropriations Committee adopted language barring fiscal 2027 funds for the model. On July 16, the Senate voted 46 to 50, falling short of taking up a resolution to overturn it.

So the week of July 27 found hospitals and physician groups in the six states in an uncomfortable middle. The model might be defunded later. It might run to 2031. In the meantime, it was deciding real claims for real patients every day. Providers entered August still bound by it, and still needing to manage it well.

For a hospital or physician group in the six states, the politics left an awkward position. The model might be defunded in a later spending bill, or it might run to 2031. Either way, it decides claims today. A revenue cycle, the chain of work that turns care delivered into payment received, cannot wait for Washington to settle the question.

The uncertainty comes from Congress, which had two levers this summer. One was the Congressional Review Act, which lets lawmakers overturn an agency rule with a simple majority in both chambers; GAO’s May finding made that route available, but the Senate vote on July 16 fell short. The other is the power of the purse: the House Appropriations Committee’s language would bar fiscal 2027 money for the model, but committee language only matters if it survives into a spending law. Neither lever has stopped the model so far, and neither has been put away.

The more useful response is for each organization to build its own record of how WISeR works in practice. That record does two jobs at once. It lets the organization manage the model day to day, and it gives the organization evidence to bring to its medical society, its hospital association, CMS or Congress. The record has three parts, which follow the life of a request: the decision, the documentation behind it, and what happens after a denial. First, what actually happens when software helps review a medical request.

How AI-assisted prior authorization works

Prior authorization is advance permission. Before delivering certain services, the provider sends the payer a request with supporting clinical information, and the payer decides whether the service meets its coverage criteria, the written conditions under which it will pay. For a spinal procedure, for example, criteria often ask whether the patient tried more conservative treatments first, what imaging showed and how long symptoms have lasted.

In WISeR, technology companies handle that review for traditional Medicare in the six states. Their tools, including artificial intelligence, read the submitted documents and compare them against the criteria. In practice, this kind of software looks for evidence in the record: a note that physical therapy was tried, an imaging report, a date range. Where it finds what the criteria require, the request can move quickly. Where it does not, the request is flagged, and because a licensed clinician must review any denial, a human makes the final call to refuse.

Two features of the design matter for providers. First, the reviewers are paid a share of the spending their reviews avert. Leaders should understand that incentive clearly, because it shapes how the model is debated and why the accuracy of denials attracts so much attention. Second, providers can choose between prior authorization and prepayment review for the selected services. Prior authorization gives an answer before care is delivered. Prepayment review lets the provider deliver the service first, then subjects the claim to review before payment, which shifts the financial risk onto the provider.

The software works like a very fast reader with a checklist. It is only as good as the documents it receives and the checklist it is given. If a key fact sits in a scanned referral letter nobody attached, the reader cannot find it, and the request may be denied for something that was true all along.

Part one: the decision record

The first part is a log of every WISeR request. It should capture the service, the ordering clinician, the site, the dates submitted and decided, the outcome, the stated reason for any denial, and whether the organization chose prior authorization or prepayment review. Most of these data exist already, spread across the authorization work queue, the scheduling system and the billing system. They are rarely joined, and for traditional Medicare they may not be tracked at all, because these services never needed authorization before.

A joined log turns anecdotes into measures. Instead of a surgeon telling the board that “approvals are taking forever,” the organization can show turnaround by service and by month, so a slipping reviewer is visible early. Turnaround is not a theoretical concern. By early July, CMS had directed Virtix Health, the reviewer for Washington, to submit a corrective action plan after it missed required response times. A provider with its own timestamps could see that delay in its schedule before CMS acted, and could document how many procedures it pushed back.

The log also shows whether outcomes differ by clinician or site, which separates documentation habits from reviewer behavior. If one surgeon’s requests are denied far more often than colleagues’ for the same procedure, the cause is probably in how that surgeon’s notes are written, not in the reviewer. If every site sees the same rise in denials in the same month, the cause is more likely on the reviewer’s side.

The log can also track the gap between first-pass and final decisions. Early Texas figures, as reported by the Washington Post, showed roughly 62 percent of requests approved on first submission and about 84 percent after physician review. That gap is the space where staff time, delayed care and frustration live. A provider that can set its own numbers beside published ones is in a stronger position than one relying on press reports.

Part two: the documentation record

The second part is a copy of exactly what was submitted with each request. Technology-assisted review reads the record against coverage criteria. When a request is denied for insufficient documentation, the question is whether the clinical facts were missing or only hard to find. Those problems need different fixes. Missing facts are a clinical or ordering problem. Hard-to-find facts are a packaging problem, and packaging problems are cheap to solve once they are visible.

Tagging each denial by the criterion cited lets clinical documentation teams find patterns. A spine practice may find that conservative therapy history sits in referral letters that are never attached. A pain clinic may find that imaging results are summarized in notes but not included. Each pattern can be fixed with a template change, an order-entry prompt or a checklist at scheduling. That is cheaper than appealing the same denial repeatedly.

Where services are ordered in the EHR, the electronic health record where clinicians write notes and place orders, a pre-submission check for required elements catches the problem at source.

Scope changes matter here too. In April, CMS delayed two of the listed services, deep brain stimulation and a lumbar decompression procedure, to allow more operational readiness. A documentation library keyed to each service’s current status stops staff from sending requests that are no longer required, or skipping ones that are. Both errors cost money: the first wastes staff time, the second risks nonpayment.

This record also feeds the wider argument about AI in utilization review, the industry term for checking whether care is necessary and covered. If reviews track the published criteria, a provider’s files will show it. If they do not, specific examples tied to cited criteria will carry more weight with CMS and lawmakers than general complaints.

Part three: the record after a denial

The third part covers what happens next. A denied request can be resubmitted with more documentation. A provider can also deliver the service and bill it, accepting prepayment review and the normal Medicare claims and appeals process. Each path has a cost. Resubmission takes staff time and delays care. Proceeding without an affirmed request risks nonpayment and adds work for patient financial services, the team that handles billing and patient balances.

Tracking every denial to final payment shows which path works for which service. The log should record resubmissions and their results, claims sent through prepayment review, appeal levels reached, final payment and the staff hours spent. Over a few months, the overturn rate, the share of denials later reversed, becomes one of the most telling measures available. A high rate suggests first-pass reviews are missing something. It is also the figure most likely to matter in an evaluation or an oversight hearing.

Beneficiary effects belong in this record as well. Patients who cancel, travel to another state or switch to a Medicare Advantage plan leave a trace in scheduling and registration data. Counting them is harder than counting dollars, and it is the concern critics of the model raise most often. A patient who simply gives up on a procedure does not appear in any denial statistic, which is exactly why a provider should try to count them.

One health system’s first months

Suppose a health system with three hospitals and a large orthopedic group operates in one of the six states. In the model’s first weeks, its surgeons complain about delays and denials, and the complaints reach the board as stories. The chief financial officer cannot say how many requests have been filed, how many denied, or what the delays are costing.

The system assigns its utilization management director to own the problem. Within a month, she joins the authorization queue, the scheduling system and the billing system into a single log. The first report shows that most denials for one spinal procedure cite the same criterion, prior conservative treatment, and that in most of those cases the treatment did happen but was recorded in a physical therapist’s notes that were never attached. A change to the order form, prompting the surgeon’s office to attach therapy records, cuts those denials sharply over the following weeks.

The log also shows that turnaround for one service has started to lengthen. The director brings dated examples to the Medicare Administrative Contractor, the regional company that processes Medicare claims, and to the state hospital association. When the board next asks about WISeR, it gets numbers rather than stories, and the system’s representatives can speak in public forums with evidence of their own.

Four ways the model reaches past billing

WISeR is more than a billing headache for the finance team, and leaders should see it in at least four dimensions.

Money. Every denial that is later overturned still costs staff time. Every procedure delayed is revenue pushed into a later month, and sometimes lost if the patient goes elsewhere. Every service delivered under prepayment review and then denied is a cost the organization absorbs. A standard financial report shows none of these clearly unless someone builds the record to show them.

Patients and clinicians. Delays affect patients who are often older and in pain. Surgeons and specialists who spend time on paperwork instead of care grow frustrated, and frustration has its own costs in morale and retention. A record that shows where delays come from lets leaders respond with something more useful than sympathy.

Compliance. Prior authorization rules and appeal processes have deadlines and formal requirements. An organization that misses them loses rights it would otherwise have. A single owner and a consistent log reduce that risk, and the compliance office should see the monthly report for that reason.

Reputation and voice. WISeR is a live political question. Hospital associations, medical societies, CMS and members of Congress are all looking for evidence about how it works. An organization that can say, precisely, how many requests it filed, how long they took and how many denials were overturned will be listened to. One that can only offer impressions will be one voice among many. The evidence also protects the organization if its own practices are questioned, because it shows what was submitted and when.

There is a wider reason too. AI-assisted review of medical necessity is not confined to this one model. The habits an organization builds now, joining its data, tagging denials and following them through to payment, will serve it wherever similar tools appear.

One owner, one log

The three parts work best under one owner, usually a utilization management or revenue cycle lead. That person reports monthly to clinical leadership and to the compliance office, and keeps a short version ready for the state hospital association. That means:

• One joined log, not three disconnected systems.
• Denials tagged by the criterion cited, reviewed regularly by clinical documentation staff.
• Every denial followed through to final payment, with staff hours counted.
• A documentation library that reflects which services currently require review.
• An attempt, however imperfect, to count patients who cancel, travel or switch coverage.

The failures are understandable. Some organizations treat WISeR as temporary and refuse to invest in tracking, then find themselves with no evidence a year later. Some respond to denials one at a time, appealing the same documentation gap again and again instead of fixing the template. Some assume every denial is the reviewer’s fault, which may be true sometimes but leaves fixable internal problems in place. And some let the political debate set their agenda, waiting for Congress rather than managing the model that exists.

One practical caution applies when the record leaves the building. The log contains patient information, so the version shared with an association, a regulator or a legislative office should be summarized or de-identified, with privacy and compliance staff signing off on what goes out. The goal is to share the pattern, not the patient.

Organizations should confirm current resubmission and appeal rules with their Medicare Administrative Contractor and with counsel, because program details have changed during the year.

What the record should be able to show

1. How many WISeR requests have we submitted, by service and site, and how many were approved on first submission?
2. What is our average turnaround by service and month, and is any reviewer slipping?
3. For each denial, which coverage criterion was cited, and what are the three most common gaps in our documentation?
4. When we resubmit or appeal, how often is the denial overturned, and how many staff hours does it take?
5. How many patients have cancelled, gone out of state or switched to Medicare Advantage after a denial or delay?

With hindsight

WISeR kept running. The House committee’s funding bar has not become law. Fiscal 2027 began under a continuing resolution, a stopgap spending law, that runs into December and generally carries forward 2026 terms, so the fight moves to the full-year bill.

CMS has adjusted the model during its first year, delaying two services and pressing a reviewer on turnaround. That shows the details remain open even if the model survives. Which services are covered, how fast reviewers must respond and how CMS enforces those standards are all live questions, and providers with evidence are better placed to influence the answers.

We have not seen a formal CMS evaluation, and conclusions about accuracy or effects on care would be premature. That restraint matters. Supporters and critics of AI-assisted review both have strong views, and early anecdotes prove nothing either way. Public numbers so far come from press reports and individual reviewers rather than from CMS.

That makes provider-side measurement more valuable, not less. When official data are thin, the organizations that can describe their own experience precisely, with dates, services, cited criteria and outcomes, carry unusual weight in the conversation. They can also run their own operations better, because they know where their documentation falls short and where the reviewer is the bottleneck.

The continuing resolution also means uncertainty will last at least into the winter. Budgets, staffing plans and payer strategies for 2027 have to be built without knowing whether the model will be funded, which is another reason to rely on what the organization can measure for itself.

The questions reach beyond the six states. AI-assisted review of medical necessity is spreading, and the questions WISeR raises about turnaround, accuracy, documentation and patient access will come up elsewhere. Organizations that log decisions, documentation and outcomes consistently can describe their own experience with evidence, whatever Congress decides.