The best utilization program is not the one that reviews the most requests. Picture a utilization review nurse’s queue on a Monday. Say it holds 120 requests, and history says 110 of them will be approved as submitted. She will still open every one and check it against criteria. For most of those cases, the review changes nothing.
That is the quiet cost at the center of prior authorization. A plan’s reviewers are among its scarcest clinical resources, and most of their hours go to confirming that care already headed for approval can proceed. The fix is not to review faster. It is to review less, and to spend the hours saved on the cases where a clinician’s judgment actually changes what happens to a patient.
Most reviews end in yes
The national numbers make the point. Medicare Advantage insurers received nearly 53 million prior authorization requests in 2024 and denied 7.7% of them, according to a KFF analysis of plan-reported data. Only 11.5% of those denials were appealed.
Read those figures from the reviewer’s chair. Out of every 100 requests, about 92 end in approval. Of the 8 that are denied, fewer than 1 is appealed, and when it is, the denial usually falls. Most of the review effort in the system goes to requests that would have been approved anyway, and a meaningful share of the denials it produces do not survive a second look.
The same analysis shows how much of this is a design choice. In 2024, Kaiser Foundation Health Plan received about 0.6 requests per enrollee, while Elevance Health received about 3.0. Denial rates ranged from 4.2% at Elevance to 12.8% at UnitedHealth Group. Patients do not differ that much from one insurer to the next. Review programs do.
Measure review by its yield
Plans usually judge a utilization program by what it denies and what it saves. We think the better measure is yield, meaning the share of reviews in which a reviewer’s judgment changes the outcome. A denial that holds on appeal counts, and so does an approval redirected to a more appropriate site of care. An approval of a complete, routine request does not.
The arithmetic is simple enough to run on any service. Say a plan reviews 4,000 requests a year for a common imaging study. It approves 3,960 as submitted and denies 40, and 30 of those denials are reversed after a peer-to-peer call or an appeal. That leaves 10 reviews out of 4,000 that changed the outcome. Every other review cost a nurse’s time and a patient’s wait, and produced the answer the request arrived with.
Measured that way, many reviews have a yield close to zero. A service approved on first submission nearly every time, whose rare denials are mostly reversed, is not being managed. It is being delayed. The new federal decision clocks make that wait more visible without making it shorter for the provider who has to submit the request.
Yield also exposes the other failure. Denials are not always right. In a 2022 review of Medicare Advantage denials, the HHS inspector general found that 13% of prior authorization denials met Medicare coverage rules and would likely have been approved under original Medicare. A team buried in routine approvals has less attention left for the hard cases where those mistakes happen.
A review that almost always ends in yes is not oversight. It is a waiting room.
Spend reviewer time where judgment changes outcomes
Retiring low-yield reviews frees capacity for the work prior authorization is supposed to do. That work is concentrated in a smaller set of decisions.
- Genuine clinical ambiguity, where reasonable clinicians disagree and the criteria turn on details in the record.
- High-cost services with wide variation in practice, where an unnecessary procedure carries real risk for the patient as well as real cost.
- Site-of-care decisions, such as whether a procedure belongs in a hospital outpatient department or an ambulatory surgery center.
- Long or complex stays, where concurrent review can catch a discharge that is ready to happen.
Retiring a review does not mean giving up oversight. A plan can still sample a small share of claims for a retired service after the fact. It can also bring review back for a specific provider whose ordering pattern shifts. Those checks cost a fraction of reviewing every request in advance, and they keep the burden on the outliers rather than on every clinician who orders the service appropriately.
The industry has committed to moving in this direction. In June 2025, health plans pledged to show reductions in the scope of claims subject to prior authorization by January 1, 2026, among other commitments. The public lists of services requiring authorization, which Medicare Advantage and Medicaid plans must post by March 31, will show which reviews were actually retired.
A plan’s UM committee is the natural place to make these calls. It already reviews every prior authorization policy each year. Adding yield to that review turns an annual compliance check into an annual decision about where reviewer time goes.
What we build for
We build tools that health plans use in utilization review, so we see this tradeoff from inside the queue. Our view is that software should make it easy to retire a review, not just to perform one.
That means showing each plan its yield by service, so its committee can see which reviews earn their cost. It also means approving complete, routine requests quickly, because a machine can say yes. Only a person can say no, so the cases that need judgment should reach a clinician with the criteria and the record facts already in front of her. Keeping the evidence for each decision lets a committee that retires a review see later whether utilization moved, and decide again with data rather than instinct.
The nurse with 120 requests on a Monday does not need a faster way to approve 110 of them. She needs those 110 not to be in her queue, so the 10 that remain get the attention they deserve.