What did the HHS OIG report find about Medicare Advantage SNF denial overturn rates?

HHS OIG report OEI-09-24-00331 (September 2026) reviewed 19 Medicare Advantage Organizations and found that 95% of appealed SNF admission denials were overturned in favor of the enrollee. naviHealth, a UnitedHealth Group subsidiary that processed 50% of all SNF admission requests, denied 14% of requests — and 97% of those denials were overturned on appeal. Only 18% of SNF denials were ever appealed, meaning 82% of potentially reversible denials went uncontested and revenue was permanently abandoned.

What is in-flight claim intervention and how does it differ from traditional denial management?

In-flight claim intervention applies AI risk scoring and payer intelligence inside active revenue cycle workflows — during scheduling, registration, documentation, coding, and claim assembly — rather than waiting for denials to occur and then working them retroactively. Traditional denial management operates on a 30–90 day lag: a claim is submitted, denied, received, queued, and worked. In-flight intervention catches the same denial triggers before the claim ever leaves the building, preventing the denial entirely instead of recovering from it after the fact.

Why do 82% of SNF denials go uncontested despite the 97% overturn rate?

Staff bandwidth is the primary constraint. Hospitals spend $19.7 billion per year on denial appeals (HFMA September 2026), and even at that investment level, approximately half of all denials are never worked. SNF denials involve complex clinical documentation, multi-level appeal processes, and payer-specific requirements that demand specialized staff time. When billing teams are already overwhelmed with denial volume, lower-dollar or more complex denials — including many SNF authorizations — are deprioritized or abandoned entirely. The 82% uncontested rate represents pure lost revenue from a capacity problem, not a merit problem.

How does AI in-flight intervention prevent denials before they happen?

AI in-flight intervention works at five points in the revenue cycle workflow: (1) at scheduling, flagging authorization requirements and coverage gaps before the appointment is booked; (2) at registration, validating patient data completeness against known denial drivers; (3) during documentation, checking clinical records for medical necessity sufficiency; (4) at coding, validating CPT/ICD accuracy against payer-specific rules; and (5) at claim assembly, running pre-submission risk scoring against historical denial patterns. Each checkpoint catches a category of denial that traditional retrospective management would only discover 30–90 days after submission.

What is the $48 billion revenue loss from denials and how much is preventable?

Healthcare Finance News reported that final denials and uncompensated care contributed to $48 billion in net revenue loss across 2,300+ hospitals in 2025 — a 25% increase from the prior year. HFMA estimates that 15% of claims are initially denied and hospitals spend $19.7 billion per year on appeals, with only approximately half overturned. The OIG data showing 95–97% overturn rates on appealed SNF denials demonstrates that most denials are not based on clinical merit — they are process failures and administrative friction. AI in-flight intervention targets the preventable portion by catching denial triggers upstream, before claims are submitted.

97% of Medicare Advantage SNF Denials Get Overturned — Why In-Flight AI Intervention Is the Only Strategy That Works

A federal inspector just proved what every revenue cycle leader already suspected: most Medicare Advantage claim denials are wrong. HHS Office of Inspector General report OEI-09-24-00331, published in September 2026, reviewed 19 Medicare Advantage Organizations and found that 95% of appealed skilled nursing facility admission denials were overturned in favor of the patient. One contractor — naviHealth, a UnitedHealth Group subsidiary that processes 50% of all SNF admission requests across the MAOs reviewed — denied 14% of requests. 97% of those denials were overturned on appeal.

Read that again. Ninety-seven percent. Not a gray area. Not a close call. Nearly every denial that was challenged was reversed, meaning those denials should never have been issued in the first place.

But here is the number that should keep every hospital CFO up at night: only 18% of SNF denials were ever appealed. The other 82% — denials that had a 95–97% chance of being overturned — were accepted, written off, and absorbed as lost revenue. HFMA's September 2026 analysis confirms the macro picture: hospitals spend $19.7 billion per year on denial appeals, 15% of claims are initially denied, and only about half of denials that are worked get overturned.

The traditional response to data like this is to hire more denial analysts. Appeal faster. Work harder. But HFMA's September 2026 framework introduces a different concept entirely: in-flight intervention — the idea that payer intelligence and denial risk scoring must be applied inside active revenue cycle workflows, not after the rejection letter arrives.

82%
of Medicare Advantage SNF denials were never appealed — despite a 97% overturn rate when challenged (HHS OIG 2026)

The $48 Billion Problem: Why Retrospective Denial Management Is Structurally Broken

The traditional denial management workflow operates on a 30–90 day lag. A claim is submitted. Weeks later, a denial is received. It enters a work queue. An analyst reviews it, pulls documentation, writes an appeal letter, submits it, and waits again. The entire cycle — from initial service to final resolution — can take 120–180 days. During that time, the revenue sits in accounts receivable, consuming working capital and staff bandwidth.

The scale of the problem is staggering. Healthcare Finance News reported that final denials and uncompensated care contributed to $48 billion in net revenue loss across 2,300+ hospitals in 2025 — a 25% increase from the prior year. The Experian 2025 State of Claims report found that 68% of revenue cycle leaders identified inaccurate or incomplete patient data at intake as the primary denial driver.

This is the fundamental structural failure of retrospective denial management: the errors that cause denials happen at scheduling, registration, and documentation — but the detection happens 30–90 days after the claim is submitted. By the time the denial arrives, the clinical encounter is over, the provider has moved on, the documentation window has closed, and the cost of recovery is 10× the cost of prevention.

"Seeing a problem sooner only helps if someone can do something about it. A denial should leave the process better than it found it — trace denials back to the workflow and data that caused them." — HFMA September 2026, "From Reactive Denial Management to Upstream Prevention"

The OIG data crystallizes why this matters. Those 97% of overturned naviHealth SNF denials were not clinically justified. They were process failures — authorization workflows that flagged admissions incorrectly, criteria applied inconsistently, documentation requirements enforced selectively. If AI had been scoring those authorization requests in real time, inside the workflow, the denial triggers would have been identified and addressed before the denial was ever issued.

naviHealth and the Contractor Problem: When One Gatekeeper Controls Half the Market

The OIG report contains a detail that reshapes how providers should think about payer risk. naviHealth — a single UnitedHealth Group subsidiary — processed 50% of all SNF admission requests across the 19 MAOs reviewed. It denied 14% of those requests, compared to 11% for MAOs processing internally and 9% for other third-party contractors.

The concentration risk is obvious: half of all SNF authorization decisions flow through one entity that denies at a higher rate than the market average and gets overturned 97% of the time when challenged. But only 18% of enrollees (or their facilities) challenge those denials.

For providers, this creates a specific and actionable problem. When you submit an SNF admission request and it routes through naviHealth, the historical data says there is a 14% chance it will be denied — and a 97% chance that denial is wrong. The rational response is not to accept the denial and write off the revenue. The rational response is to build a system that anticipates the denial before it arrives and either prevents the trigger condition or auto-generates the appeal within hours of receipt.

Nursing home residents face even worse odds. The OIG found that nursing home residents were denied SNF admissions at 40% — compared to 11% for other enrollees — raising concerns about systematic access barriers for the most vulnerable Medicare Advantage population.

50%
of all SNF admission requests processed by a single contractor (naviHealth) — which denied at 14% and was overturned 97% of the time

In-Flight Intervention: The AI Strategy That Catches Denials Before They Happen

HFMA's September 2026 framework introduces the concept of moving from "retrospective visibility to in-flight intervention." The distinction is architectural, not incremental. Retrospective visibility means you can see denials after they happen — dashboards, reports, trend analysis. In-flight intervention means AI is embedded inside the workflow, scoring denial risk at every checkpoint, and either preventing the trigger condition or routing exceptions to human staff before the claim is submitted.

Here is what in-flight intervention looks like at each stage of the revenue cycle:

1. At Scheduling: Authorization Risk Scoring

When a patient is scheduled for a procedure or admission, AI cross-references the procedure code, payer, plan, and patient history against known authorization requirements. If the patient's Medicare Advantage plan routes through naviHealth for SNF authorizations, and the historical denial rate for that combination is 14%, the system flags the authorization as high-risk and initiates proactive documentation strengthening — before the authorization request is submitted.

2. At Registration: Data Completeness Validation

68% of revenue cycle leaders identified inaccurate or incomplete patient data at intake as the primary denial driver (Experian 2025). AI validates registration data against known denial patterns in real time. Missing subscriber IDs, mismatched demographic information, stale coverage data, and coordination of benefits gaps are caught at registration — not discovered as a denial reason code 60 days later.

3. During Documentation: Medical Necessity Sufficiency

For services requiring prior authorization or medical necessity documentation, AI scores the clinical record against payer-specific criteria during the encounter — not after. If a skilled nursing admission request requires specific functional status documentation that the current record lacks, the system alerts the clinical team while the patient is still being evaluated. The documentation gap is closed in minutes, not argued about in an appeal letter three months later.

4. At Coding: Payer-Specific Rule Validation

HFMA's framework emphasizes that "actual payer behavior does not always mirror the written rule." AI maintains a continuously updated model of actual payer adjudication behavior — not just published guidelines — and validates codes against that behavioral model before submission. A modifier that is technically correct but consistently denied by a specific payer is flagged for review, not submitted and denied.

5. At Claim Assembly: Pre-Submission Risk Scoring

Before any claim leaves the building, AI runs a final risk score against the combined denial probability of the payer, plan, procedure, diagnosis, provider, and facility combination. Claims above the risk threshold are routed for human review. Claims below the threshold are submitted with high confidence. The goal is not zero denials — it is zero preventable denials.

82% Never Appealed: The Revenue Your Billing Company Abandons

The OIG finding that 82% of SNF denials were never appealed is not a compliance statistic. It is a revenue catastrophe expressed as a percentage.

Consider the math for a single health system. If a hospital generates 1,000 SNF admission requests through Medicare Advantage plans in a year, and the average denial rate across MAOs is 12%, that produces 120 denials. If only 18% are appealed, that means 22 appeals are filed and 98 denials are accepted without challenge. At a 95% overturn rate, 93 of those 98 uncontested denials would have been overturned if someone had filed the appeal.

At an average SNF admission value of $15,000–$25,000, those 93 abandoned denials represent $1.4 million to $2.3 million in recoverable revenue — from a single facility, in a single year, for a single service type. Scale that across a multi-hospital system with thousands of MA-covered SNF admissions, and the abandoned revenue reaches eight figures.

The reason these denials go uncontested is not that billing teams do not care. It is that they are already buried in denial volume across every service line. AI changes the capacity equation in two ways simultaneously: it prevents a large percentage of denials from occurring in the first place (in-flight intervention), and it auto-generates appeal packages for denials that do occur (zero-delay response). The combination eliminates the triage problem that forces billing teams to abandon winnable appeals.

Metric	Retrospective Denial Management	AI In-Flight Intervention
When denial risk is detected	30–90 days after claim submission	During scheduling, registration, documentation, coding, and claim assembly
Prevention capability	None — operates after denial is issued	Catches denial triggers before claim is submitted
Appeal generation time	Days to weeks after denial received	Same-day automated appeal with supporting documentation
Denial coverage rate	18% appealed (OIG SNF data)	100% — every denial scored, routed, and either prevented or appealed
Staff allocation	Analysts work denials retroactively	Staff handle exceptions flagged by AI in real time
Revenue impact	82% of recoverable denials abandoned	Preventable denials eliminated; remaining denials auto-appealed

From Dashboard to Workflow: Why AI Must Live Inside the Revenue Cycle

Most revenue cycle AI products marketed in 2026 are dashboards with predictive labels. They can tell you which denials are likely, what the top denial reason codes are, and which payers have the highest denial rates. This is retrospective visibility — the exact approach HFMA's framework says is insufficient.

The difference between visibility and intervention is the difference between a weather forecast and an umbrella. Knowing it will rain does not keep you dry. Knowing a claim has a 40% denial probability does not prevent the denial — unless the system can act on that probability inside the workflow, before submission.

In-flight intervention requires AI that is architecturally embedded in the revenue cycle workflow, not layered on top of it as a reporting tool. The system must have:

Real-time payer intelligence: Not published guidelines, but continuously updated models of actual payer adjudication behavior. HFMA notes that "actual payer behavior does not always mirror the written rule" — the system must learn from outcomes, not just documentation.
Workflow integration at every checkpoint: Scheduling, registration, documentation, coding, and claim assembly each represent a denial prevention opportunity. A system that only scores claims at submission misses five earlier intervention points.
Exception routing to human staff: In-flight intervention does not replace human judgment. It eliminates the volume problem — the hundreds of routine verifications and risk scores that consume staff time — and routes only genuine exceptions that require clinical or administrative decision-making.
Feedback loops that improve the process: Every denial that occurs despite in-flight intervention must feed back into the risk model. As HFMA states, "a denial should leave the process better than it found it." The system traces denials to the workflow step where prevention failed and adjusts.

What the OIG Recommends — And Why Providers Cannot Wait

The OIG report recommends that CMS "take action to address breakdowns in initial reviews" that produce the extremely high overturn rate. CMS has not explicitly concurred or nonconcurred with the recommendations; an update is expected by December 7, 2026.

But providers cannot afford to wait for regulatory intervention. Even if CMS acts — even if naviHealth's authorization practices are reformed — the structural problem remains. MA plans expanded prior authorization requirements 37% since 2022. Payers are deploying their own AI to accelerate denials. The US Senate Permanent Subcommittee on Investigations confirmed that payers use AI to deny claims at rates 16× higher than previous benchmarks.

The OIG data does not describe a problem that regulation will fix. It describes an asymmetry that only provider-side AI can close. Payers have industrialized the denial process. Providers must industrialize prevention.

"Nearly half of healthcare executives identify revenue cycle as their top area for IT investment." — HFMA September 2026

The practices and health systems that deploy in-flight AI intervention now — before the regulatory cycle plays out, before the next denial wave arrives — are the ones that will stop losing revenue to a process that the federal government just confirmed is wrong 97% of the time.

Stop Treating Symptoms. Prevent the Disease.

The OIG report is a landmark. It puts a federal number on what providers have experienced for years: Medicare Advantage denials are frequently unjustified, disproportionately harm vulnerable populations, and represent billions in lost revenue that is never recovered because the appeal process overwhelms staff capacity.

But the most important insight is not in the OIG report. It is in HFMA's framing: the shift from retrospective visibility to in-flight intervention. The answer to a 97% overturn rate is not more appeal analysts. It is a system that prevents the denial from being issued in the first place — by embedding payer intelligence, risk scoring, and exception routing inside every step of the revenue cycle workflow.

Retrospective denial management is treating symptoms. In-flight AI intervention prevents the disease. The data is clear. The architecture exists. The only question is when you deploy it.