Cleveland Clinic's $15B Revenue Cycle Has a 15% Denial Rate Problem — and Its CFO Just Called It 'Unsustainable'

The most operationally sophisticated revenue cycle in the country collected $15 billion last year, employs 3,800+ FTEs, and runs at a 2.92% cost to collect. And its CFO just said the system is broken — for both sides of the transaction.

The Numbers Behind the Problem

Cleveland Clinic's revenue cycle team processed more than $15 billion in U.S. net patient revenue in 2025 — one of the highest volumes of any health system in the country. With 3,800+ RCM FTEs and a cost to collect of 2.92%, the operation is leaner than most. By most benchmarks, this is what best-in-class looks like.

And yet: 15% of claims come in denied. That means for every $100 Cleveland Clinic bills, $15 is initially rejected and has to go back through the system via resubmissions, medical director reviews, peer-to-peer calls, and formal appeals.

The system's RCM team ultimately overturned 92% of those denials, bringing the net controllable loss to under 2%. So on paper, the outcome is acceptable. But the operational cost of getting there — the 3,800 FTEs, the embedded physician advisors, the 200 RPA bots — is the problem Tracy Peffley, Cleveland Clinic's system VP of revenue cycle management, put plainly in a recent Becker's podcast:

"Rising denial rates create unnecessary costs and administrative burden and create friction for providers, payers, clinicians and — most importantly — our patients. As an industry, we have an opportunity to move away from the reactive model that exists today toward a more collaborative one."

Why the CFO Used the Word "Unsustainable"

Dennis Laraway, Cleveland Clinic's EVP and CFO, made a pointed observation in his comments to Becker's that cuts to the core of the denial problem: both sides know the denials are wrong.

"It's not sustainable. And frankly, it's arguably unsustainable for the payers, too. If both sides are watching the same report card — and the end result is that 13% of those denials are ultimately overturned — then that doesn't seem like productive friction for anyone."

This is a significant statement coming from the CFO of a $15B health system. The math is simple: if 15% of claims are denied, and 92% of those denials are eventually overturned, then roughly 13% of total denials were wrong on arrival. Not gray area. Not ambiguous. Wrong — and both parties spent resources fighting over them.

Laraway isn't pointing fingers at payers. He's saying the system design itself is broken, and the fix isn't for providers to get better at fighting denials. It's to build something different.

The Fix: Embed Clinical Expertise Upstream

Cleveland Clinic's response isn't just technology. It's structural. The health system embedded two medical directors directly into revenue cycle operations — including a medical director of documentation excellence — plus a physician advisory program. The explicit goal: prevent denials before claims leave the building, not fight them afterward.

"When we look at incoming denials and peel back the layers — why are we receiving these secondary and tertiary denials? — it's really about taking a more integrated clinical-revenue cycle approach," Ms. Peffley said. "Leaning in with our clinical providers lets us better prevent those denials and creates a better experience for clinicians, allowing them to stay focused on patient care rather than worrying about prior authorizations or peer-to-peer reviews."

The clinical-RCM integration model addresses something most denial management programs miss: the root cause of the majority of denials isn't billing error, it's documentation. Coders can't capture what clinicians don't document. And appeals fail when the underlying clinical record doesn't support the level of care billed. Putting physicians inside the revenue cycle operation — not as consultants, but as embedded operators — closes that gap before it becomes a denial.

200 Bots and Still a Human in the Loop

AI is central to Cleveland Clinic's upstream strategy, but the approach is notably different from the headlines about automation replacing coders. The health system runs roughly 200 RPA bots within its revenue cycle, and recently created a dedicated technology and innovation domain within its operating model.

On inpatient coding specifically, Cleveland Clinic deployed a large language model to surface documentation and coding opportunities — with a human reviewer validating every result before any claim goes out the door.

"AI is augmenting the inpatient coder's work, using a large language model to surface potential opportunities, with a human in the loop 100% of the time validating those opportunities before anything goes out the door. We saw the same fear with computer-assisted coding and with RPA. People worried about losing their jobs. They didn't. AI in inpatient coding is no different. It's augmenting the great work our coders do and leveraging their critical thinking at the top of their license." — Tracy Peffley

This is a deliberate positioning. As other health systems — including CoxHealth's recent 53-role reduction — report removing coders from the workflow, Cleveland Clinic is investing in a model where AI surfaces opportunities and humans close them. The distinction matters: Cleveland Clinic has one of the highest case mixes in the country, which means the complexity of its inpatient coding is above average. The augmentation approach makes sense when the cases are genuinely hard.

The North Star: Near Real-Time Adjudication

Both Peffley and Laraway pointed to the same long-term target: a world where the claim adjudication process doesn't take weeks and doesn't require a team of hundreds to fight through. Laraway put the endgame in stark terms:

"If we can use AI productively to get behind the rules engines and create agreement — a contract between payer and provider — then 95% of claims should be auto pay. What we'll be arguing about with the payer is over what time period are you going to pay that claim."

Near real-time claim adjudication has been an industry aspiration for years. What's different now is that the technology infrastructure to support it — AI that can interpret policy, validate documentation, and auto-adjudicate routine claims — is actually within reach. The barrier isn't technical. It's the payer-provider adversarial dynamic that has built up over decades of denial gaming on both sides.

Laraway's framing is the right one: the goal isn't for providers to beat payer AI, or for payers to deploy better denial algorithms. The goal is to use AI to establish shared ground truth so that the 13% of denials that are clearly wrong never happen in the first place.

What This Means for the Rest of the Market

Cleveland Clinic is operating at scale most health systems will never reach. But the structural problems it's describing — a 15% incoming denial rate, an enormous human workforce needed to fight back to acceptable recovery rates, a reactive model that everyone agrees is broken — are not unique to a $15B system. They're universal.

The difference is that smaller systems don't have 3,800 FTEs to absorb the overhead. They don't have embedded medical directors in revenue cycle. They don't have 200 RPA bots. When a 200-bed community hospital faces the same 15% denial rate, the math looks very different — and the net controllable loss is often much worse than 2%.

The Cleveland Clinic story is important not because it's exceptional, but because it names the problem clearly, at a scale that makes the numbers undeniable, with executives willing to say publicly what most CFOs only say in private: the system is broken, the friction is productive for no one, and AI's real job isn't to help you win the denial fight — it's to make the fight unnecessary.

Source: Becker's Hospital Review, June 25, 2026. Quotes from Tracy Peffley, SVP Revenue Cycle Management, and Dennis Laraway, EVP & CFO, Cleveland Clinic.