The Denial Numbers Are Public. Read the Overturn Rate.

Payer prior-authorization metrics are public now. The most important number is still missing: how much of your own denied revenue was worth appealing.

Our position is simple. Public payer data should be treated as an external signal, not a complete insurer scorecard. A hospital has to combine that signal with its own denied dollars, appeal rate, overturn rate, cost to appeal, and cash recovered before it can decide where intervention will produce value.

This is not a revenue-cycle reporting exercise. It is a capital-allocation decision about where scarce staff time, automation, and executive attention can return cash without weakening clinical accountability.

The first public reports changed the conversation

Under CMS-0057-F, impacted Medicare Advantage, Medicaid, CHIP, and federally facilitated Marketplace payers must now publish annual prior-authorization metrics for medical items and services other than drugs. The required information includes approval and denial percentages, standard requests approved after appeal, extensions, and average and median decision times. The first reports, covering calendar year 2025, were due by March 31, 2026.

KFF collected reports from 14 major insurers representing approximately 71 million enrollees across the market segments it examined. Its August analysis found that insurers denied 12 percent of standard requests in Medicare Advantage, 14 percent in Medicaid managed care, and 18 percent in the federally facilitated ACA Marketplace.

The market differences are not a league table. Populations, service mixes, benefit designs, and prior-authorization requirements vary. The useful signal is that denial performance is now visible enough for hospital leaders to ask sharper questions than, “Are denials up?”

The overturn signal deserves attention, not a headline

KFF found that 67 percent of appealed standard denials in its 2025 Medicare Advantage analysis were overturned. The corresponding figures were 47 percent in Medicaid managed care and 43 percent in the federally facilitated Marketplace.

Those percentages do not show how many total denials were appealed. A separate KFF analysis of 2024 Medicare Advantage data does. Of approximately 4.1 million denied prior-authorization requests, only 11.5 percent were appealed. Among the denials that were appealed, 80.7 percent were overturned. The years and reporting methods differ, so the figures should not be blended into a single benchmark. Together, they support one conclusion: an initial denial is not always a durable decision.

The HHS Office of Inspector General found a more concentrated version of the pattern. In its review of skilled-nursing-facility admission requests across 19 Medicare Advantage organizations during June 2024, 18 percent of denials were appealed and 95 percent of those appeals were overturned. That is not an all-service national rate. It is still a serious signal about initial-review quality, contractor oversight, and the unexamined denials that never reached appeal.

Public data is not a payer scorecard

The new transparency has limits. Payers are not uniformly required to publish numeric counts. The reports do not consistently break results out by service, denial reason, or geography. CMS recommends a reporting template, but does not currently require every payer to use it. Case mix, service mix, reporting choices, and gold-card programs can all distort a raw comparison.

That means a payer with a higher aggregate denial rate is not automatically behaving worse, and a payer with a lower rate is not automatically performing better. A percentage without its denominator, scope, and root cause is a clue, not a verdict.

The reporting gap should change the hospital's response. Do not wait for a perfect national dataset, and do not turn an imperfect one into a simplistic ranking. Use the external numbers to identify where your own internal analysis should begin.

Build the missing denial-yield view

Most denial dashboards count cases, dollars, or aging. Fewer connect the entire economic chain:

• Initial denied allowed amount
• Appeal-eligible denied amount
• Appeal rate by payer, service line, and reason
• Overturn probability for comparable cases
• Net cash recovered after a favorable decision
• Staff and vendor cost per appeal
• Days from denial to decision and from decision to cash

The operating question is not, “How many appeals can we automate?” It is, “Which eligible denials have the highest expected net yield, and what caused them?”

A high overturn rate may reflect poor initial payer decisions. It may also reflect missing documentation that is corrected during appeal. Those are different problems. The first belongs in payer performance and contracting discussions. The second belongs in order entry, documentation, authorization, and revenue-cycle workflow redesign.

Put AI in the triage layer, with accountable review

AI can help assemble documentation, classify denial reasons, match cases to payer rules, and rank appeal candidates. It should make the work queue more intelligent. It should not make clinical accountability disappear.

KFF identified nine states with enacted laws addressing AI in prior authorization or claims review as of April 28, 2026. The requirements differ, but common themes include licensed human review, attention to individual clinical circumstances, disclosure, and periodic assessment of AI outcomes. Hospitals should know which rules apply to their own utilization-management tools and to the payers they challenge.

The practical control is straightforward: record which model or automation touched the case, preserve the supporting evidence, route medical-necessity judgment to a qualified reviewer, and retain the final decision owner.

Read the vendor claim like a CFO

Janus Health reports that Carle Health reduced authorization-related outpatient write-offs by 20 percent, generated a $2.4 million annual net revenue impact, saved time equivalent to 6.4 full-time employees, and achieved touchless auto-approval for 63 percent of in-scope authorizations.

That is a stronger case study than an anonymous percentage because it names the health system and reports financial, labor, and workflow outcomes. It is still vendor-published evidence. Before applying it to another hospital, a buyer needs the baseline write-off amount, measurement period, included payers and services, total in-scope volume, implementation cost, recurring cost, attribution method, and evidence that the result persisted.

The larger lesson is not whether 20 percent is good. It is whether your business case can show the full bridge from automation activity to avoided write-offs and collected cash.

Five actions for this week's leadership meeting

1. Choose one payer and one service line with material denied dollars.
2. Separate appeal-eligible denials from valid or uneconomic cases.
3. Measure appeal rate, overturn probability, cost, cycle time, and cash recovered.
4. Trace each high-value denial to payer behavior or an internal root cause.
5. Assign human decision ownership for every AI-assisted medical-necessity workflow.

The public reports have made denial performance visible. They have not made it decision-ready. The hospital that connects external payer signals to internal denial economics will know which cases to pursue, which workflows to repair, and which payer conversations deserve executive attention.

What denial measure can your executive team not produce today? Reply and tell us. We read every response.