Why 41% of Healthcare Claims Get Denied, and How AI Is Changing the Math

A friend of mine, a family practice doc in Ohio, called me last month. She was furious. She'd spent 22 minutes on the phone with a payer rep trying to overturn a denial for a diabetic patient's continuous glucose monitor. The reason for denial? A missing modifier on the claim form. Not a medical judgment call. Not a question of necessity. A modifier.

She got it overturned. Most providers don't even bother.

And that's the whole problem in a nutshell.

The Numbers Are Getting Worse, Not Better

Let's start with the stat that should make every healthcare executive lose sleep: 41% of providers in 2025 say their denial rates now exceed 10%. That's up from 30% just three years ago. Initial claim denials hit 11.8% in 2024, climbing from 10.2%. And if you think that 1.6-percentage-point jump sounds small, do the math on millions of claims.

Hospitals spent $19.7 billion in 2022 just trying to overturn denied claims. Not treating patients. Not upgrading equipment. Not hiring nurses. Fighting paperwork.

Here's what really gets me: 77% of those denials stem from paperwork errors and plan design issues. Not medical judgment. We're not talking about doctors ordering unnecessary procedures. We're talking about missing authorizations, coding mistakes, and eligibility snafus. Bureaucratic friction, plain and simple.

Who's Denying the Most?

Not all payers are created equal here. Oscar Health leads the pack with a 25.3% denial rate. Molina isn't far behind at 22%. Medicare sits at 8.4%, which sounds almost reasonable until you remember the sheer volume of Medicare claims flowing through the system daily. Medicaid? A brutal 16.7%.

I think the disparity alone tells you something important. Denial rates aren't some fixed cost of doing business. They're a choice payers make about how aggressively to gate claims.

The Appeal Paradox Nobody Talks About

This is the part that genuinely baffles me. Fewer than 0.2% of denied claims are appealed internally. Read that again. Less than two-tenths of one percent.

But here's the kicker: 54.3% of denied claims that are appealed get overturned.

So more than half the time, the denial was wrong. Or at least arguable. And almost nobody fights back. Why? Because the cost of appealing often exceeds what you'd recover on any single claim. The system is essentially designed to make giving up the rational economic choice. That's not great.

The Human Cost

And let's not pretend this is just a balance-sheet problem. Sixty percent of consumers said a claim denial delayed their care. Half of those people said their condition got worse while they waited. People are sicker because of administrative friction.

In my view, that ought to be the headline every time we talk about denials. Not the revenue impact. The patient sitting at home wondering if they can afford to get the treatment their doctor already said they need.

Where the Pain Is Growing Fastest

Two trends are accelerating the damage.

Medical necessity denials are exploding. They've risen 70%, with the average cost per denial now hitting $450. These are the hardest to overturn because they require clinical documentation, peer-to-peer reviews, and sometimes external appeals. They eat staff time like nothing else.

Telehealth denials surged 84%. The pandemic pushed everyone into virtual care. Payers initially played along. And now, surprise, they're clawing it back through denials. Providers who built telehealth programs expecting stable reimbursement are getting hammered.

Enter AI. But Let's Be Honest About Where We Are

Here's where I'm supposed to tell you that artificial intelligence is going to fix everything. I won't. But the early data is genuinely encouraging.

Among providers already using AI to tackle denials, 69% say it's reduced their denial rates. That's a strong signal. The catch? Only 14% of providers are actually using AI for this purpose right now. The vast majority of health systems are still running denial management the old way: spreadsheets, manual audits, overworked billing staff chasing down codes at 4 p.m. on a Friday.

What AI Actually Does Well

• Pre-submission scrubbing: AI tools can flag likely denials before a claim goes out the door, catching missing modifiers, authorization gaps, and eligibility mismatches in real time.
• Pattern recognition: Machine learning models identify which payers deny which claim types and why, so staff can preemptively address known triggers.
• Appeal prioritization: Given that 54.3% of appeals succeed, AI can rank denied claims by likelihood of overturn and expected recovery value. That means teams chase the right ones first.
• Documentation support: Natural language processing helps clinicians produce notes that meet payer requirements the first time around, reducing those medical necessity denials.

My guess is the 14% adoption figure will double within 18 months. Revenue cycle leaders are under enormous pressure, and the ROI math on AI-driven denial prevention is pretty hard to argue with when you're staring at an 11.8% denial rate.

Why Hasn't Everyone Adopted This Already?

Fair question. A few reasons.

First, integration is messy. Most health systems run on legacy billing platforms that don't play nicely with modern AI tools. Ripping out and replacing your RCM stack is a multi-year project nobody wants to sign up for.

Second, trust. Revenue cycle teams have seen plenty of silver bullet vendors come and go. They're skeptical. And honestly, they probably should be. Not every AI product on the market actually delivers.

Third, staffing. You need people who understand both the clinical documentation side and the technology to set up these systems properly. That talent pool is thin.

What Smart Organizations Are Doing Right Now

The health systems I've watched get this right tend to follow a similar playbook:

• They start with denial analytics before denial prevention. That means understanding their specific payer-by-payer, code-by-code denial patterns before deploying any AI.
• They focus on pre-submission interventions, not post-denial appeals. Stopping a denial from happening costs a fraction of overturning one.
• They treat AI as a tool for their existing staff, not a replacement. The best results come from billing specialists augmented by machine learning, not from trying to automate humans out of the loop.
• They measure ruthlessly. Tracking denial rates by payer, by service line, by coder, and by week. If the AI isn't moving the needle, they adjust fast.

The Bigger Question

Look, we can fine-tune denial management all day long. AI will absolutely help providers claw back revenue and reduce the administrative tax on clinical staff. That's real, and it matters.

But here's what I keep coming back to: we've built a $19.7 billion-a-year industry around fighting over whether to pay for care that doctors already ordered. And the patients caught in the middle, the 60% who say denials delayed their treatment, they're not a line item. They're people.

So yes, adopt the AI. Fix the workflows. Hire the analysts. Scrub the claims. Do all of it.

But maybe also ask: is a system where providers need machine learning just to get paid for legitimate services actually the system we want?