Prior authorization metrics should be published by service. By March 31, payers covered by CMS’s January 2024 interoperability and prior authorization rule, from Medicare Advantage organizations to issuers on the federal exchanges, must post their first annual prior authorization metrics. Each will list the items and services that require prior authorization and report approval and denial rates, along with decision times.
That is real progress. For most plans, it will be the first time any of these numbers are public. But every metric in the rule is aggregated across all items and services. A plan will report one approval rate for everything it reviews, from a routine MRI to a months-long course of home health care. That number cannot answer the questions patients and physicians actually ask.
A single number hides what matters
We showed last month how pooled metrics can hide differences between groups of enrollees in Medicare Advantage equity reports. The same problem applies to services. A plan that denies 2% of requests overall could deny 1% of imaging requests and 25% of requests for a particular therapy, and the public report would show only the 2%.
The aggregate rate also mixes 2 different things, what a plan chooses to review and how it decides. KFF’s analysis of 2024 Medicare Advantage data found that denial rates ranged from 4.2% at Elevance to 12.8% at UnitedHealthcare, while requests per enrollee ranged from 0.6 at Kaiser Foundation Health Plan to 3.0 at Elevance. A plan that requires prior authorization for many routine services will tend to report a high approval rate, because most of those requests are approved. A plan that reviews only a few high-risk services may report a lower one. Without service detail, a reader cannot tell a plan that decides harshly from one that simply reviews fewer things.
KFF also noted that the Medicare Advantage data CMS collects are aggregated to the contract level, which keeps researchers from seeing how prior authorization varies by service. The new public metrics will share that limitation.
Minnesota already does it
Service-level reporting is not hypothetical. Since 2022, Minnesota Statutes § 62M.18 has required health plan companies to post, by April 1 each year and for each commercial product, their prior authorization denials sorted by health care service. The reports also show whether each denial was appealed and how the appeal came out. Plans must report the reasons for denials as well, such as the patient not meeting criteria or the provider submitting incomplete information.
Other states have followed the federal model instead. Mississippi’s prior authorization statistics rule requires issuers to post, by March 31 each year, essentially the same aggregated metrics CMS requires.
The Minnesota approach shows that plans can produce service-level reports and keep publishing them year after year. The data exists in every plan’s utilization management system. The question is only which table gets published.
An approval rate for everything is an approval rate for nothing in particular.
What a useful report would include
A report that physicians and regulators could use would not need to be much longer than the federal one. It would need to be organized differently.
- Results by service category. Request counts and approval rates for each service or code group that requires prior authorization, starting with the categories that have the most requests and the most denials.
- Denial reasons. At least a split between denials for not meeting clinical criteria and denials for incomplete documentation. The first says something about the plan’s criteria. The second says something about the process.
- Appeals by service. The share of denials appealed and the share of appeals overturned, for each category. Across Medicare Advantage in 2024, KFF found that 11.5% of denials were appealed and 80.7% of appeals were at least partly overturned. A service with a high overturn rate is a service whose initial decisions deserve a closer look.
- Decision times by service. Averages and medians for each category, so that a slow process for one service is not hidden by fast decisions on others.
- Counts as well as percentages. Every rate should show the number behind it, with small cells suppressed rather than reported as 0% or 100%.
- A machine-readable file. A simple data file alongside the web page lets researchers and states compare plans without retyping tables.
A report like this also makes reform visible. In June 2025, health plans pledged to reduce the scope of claims subject to prior authorization by January 2026. Service-level reporting would show which services were removed and what happened to request volume and approval rates afterward. An aggregate rate could move for reasons no one could identify.
What we build for
We build tools health plans use in utilization review, and one of our rules is that every decision can be reconstructed. When each decision records the service it concerns and the reason for its outcome, a service-level report is a query, not a project. The same data that supports a single decision supports the public table.
We think plans should publish service-level metrics voluntarily, using the same definitions as the federal metrics so the service rows add up to the aggregate totals. The plans that do will learn things from their own reports. A service with thousands of requests and an approval rate near 100% is a strong candidate to stop reviewing, as we argued in our pieces on reviewing less and on gold carding. A service with a high overturn rate on appeal is a sign that something in the first review needs to change.
The federal metrics due on March 31 are a floor. The plans that treat them as a starting point, rather than a ceiling, will have an answer ready when someone asks what, exactly, all those approvals were for.