Why Rising Denial Rates Are an Operating Model Problem

By Angela Troccoli, Head of Marketing, Revecore 

Denial rates hit 11.8% industry-wide in 2024, and net revenue leakage from denials grew 25% year-over-year in 2025, reaching $48.4 billion. Leakage is growing faster than the denial rate itself, which means the cost of resolving each denial is climbing even as the number of denials climbs too. Hospitals spent close to $18 billion overturning denials in 2025 alone, much of it on claims that were eventually paid. 

In the past, organizations would respond to numbers like these by hiring more denial specialists, buying a new denials tool, or adding a layer of review before claims go out. Those moves can help at the margins, but the underlying structure is usually what's driving the numbers in the first place: patient access, coding, billing, and collections operating as disconnected functions, each with its own metrics and its own leadership, each optimizing locally. Denials, underpayments, and complex claims show up as separate line items because separate teams track them, even when the root cause is shared. 

The Real Problem Isn't Staffing or Software

HFMA's February 2026 survey of finance and revenue cycle leaders puts a number on how widespread this gap is. Only about 7% of respondents described their teams as "very prepared" for where revenue cycle operations are heading, with another 44% landing at "somewhat prepared." That points to a gap in operating model readiness more than a gap in staffing or software. A recent HFMA analysis put it plainly: intelligent revenue cycle strategy starts with process standardization, data governance, and aligned metrics across teams, and technology gets layered on after that foundation exists, not before. Adding automation to an already fragmented process tends to speed up the fragmentation rather than fix it. 

Why Revenue Cycle Silos Create More Denials

McKinsey's Michael Peterson has made a similar point about where health systems tend to start: back-end functions like AR follow-up, underpayment management, and denials are usually the first place hospitals apply AI, because the work is labor-intensive and rule-governed. That's also exactly the work that's hardest to fix with a tool alone if governance and workflow design haven't caught up first. 

What Leading Health Systems Are Doing Differently

BJC HealthCare is one system that made the structural move rather than the tooling move. After operating 24 hospitals across Illinois, Missouri, and Kansas as largely independent facilities, BJC consolidated patient access, middle revenue cycle, revenue management, and business operations under a single shared services model starting in 2019, with a tiered huddle structure that surfaces problems daily instead of quarterly. Harold Mueller, BJC's Chief Revenue Officer, described the results in a recent conversation with our CEO, Noah Breslow: avoidable write-offs down almost $9 million in 2025, with another $15 to $20 million targeted for 2026, tied to a task force built around a single owner rather than a distributed committee. 

In practice, this kind of shift tends to rest on a few consistent mechanics, whichever system is making it. A single point of contact gets embedded at each facility, someone who sits between hospital operations and revenue cycle leadership, tracks trends in the data, and runs a recurring cadence with local leadership rather than waiting for issues to surface on their own. Workflows, reporting, and policy get standardized across every facility instead of left to local preference, which tends to be where cost-to-collect improvement actually comes from. And the cadence itself usually splits in two: cash collections get watched daily, while a broader set of KPIs, typically anchored on AR and cost-to-collect, gets reviewed monthly so short-term noise doesn't bury the underlying trend. At BJC, that structure took the form of facility-based revenue cycle liaisons, standardized workflows under the shared services model, and a board that tracks technology efficiency alongside the underlying financial results. 

Technology Works Best After the Operating Model Changes

The technology BJC used mattered less than the sequence. The operating model came first, starting in 2018 and 2019, well before the analytics existed to make the work easier. The systems and vendor relationships got layered on afterward, and worked better because a structure already existed for them to plug into. 

Preparing Revenue Cycle Operations for the Next Wave of AI-Driven Denials

Denial rates show no sign of leveling off. Payer scrutiny is increasing, prior authorization requirements are expanding, and AI-driven claims review is scaling faster than most provider organizations can respond to manually. The systems best positioned to absorb that pressure over the next few years will likely be measured less by which denials software they bought and more by whether patient access, billing, and collections were ever built to talk to each other in the first place.