Here’s a number that should keep hospital CFOs up at night: between 35% and 65% of denied claims are never reworked. Not appealed. Not resubmitted. Just abandoned.
That’s permanent revenue destruction. Not a collection problem. A capacity problem wrapped in a technology gap.
The denial rate at U.S. hospitals hit 11.81% in 2024, up from 11.5% the year before, with private payer denials averaging closer to 15%. For a health system with $500M in annual net patient revenue, that’s $75M in claims hitting a wall before any recovery effort begins. And for more than half of those claims, no one ever tries again.
This post breaks down what’s actually driving the denial management crisis, why manual processes can’t solve it, and what the math looks like when technology handles what staff can’t.
Why denial rates keep climbing
Payers are getting more sophisticated, not less. Commercial insurers now deploy AI to deny claims at scale. UnitedHealth’s nH Predict algorithm has been cited in litigation as generating a 90% error rate on appealed denials. Cigna’s PxDx system reportedly enabled physicians to deny 300,000+ claims in a two-month span, at an average of 1.2 seconds per review.
The result: medical necessity denials surged 60% in professional settings and 123% in inpatient settings in 2025 alone (per HFMA analysis). Coding-related denials climbed 125% (per ICD10monitor). Requests for additional information increased alongside every other denial category.
Providers were not built to fight algorithms with billing staff.
DENIAL LANDSCAPE AT A GLANCE
Initial denial rate (2024): 11.81% across 2,100+ hospitals
Private payer average: ~15% | Medicare Advantage: 15.7% | Medicaid: 16.7%
Medical necessity denials (inpatient) up 123% in 2025 vs. 2024
Claims never reworked: 35-65% of total denials
Annual provider spend on claims adjudication: $19.7B-$25.7B
Administrative cost per denied claim: $43.84 average; $63.76 for commercial
The cost is bigger than most finance teams realize
Providers spend an estimated $19.7 to $25.7 billion annually adjudicating denied claims. The per-claim rework cost averages $43.84 for all payers, hitting $63.76 for commercial claims (per Premier analysis).
Twenty-two percent of healthcare leaders report annual denial losses exceeding $500,000. One in 10 reports losses above $2 million annually.
But those numbers only capture the claims that actually get worked. The bigger number is the revenue that never gets counted because it was written off before anyone tried.
An organization losing 5% of net patient revenue to denial-related waste, on a $200M revenue base, is watching $10M evaporate annually. Most of it never shows up as a discrete line item on any report. It just shows up as margin that isn’t there.
Why the manual model breaks
Revenue cycle teams are handling more volume with fewer people. Over 60% of revenue cycle departments report staffing vacancies exceeding 15%. The average annual overhead for a full-time coder is $215,000.
Manual denial management requires a specialist to read each remittance advice, determine root cause, research the payer’s specific requirements, draft a response, and track the filing. For claims under $5,000, the labor cost to work the denial often exceeds the potential recovery. So those claims get written off.
That’s a rational decision by an overwhelmed billing team. It’s also $10 million a year in abandoned revenue for a mid-size system.
The timely filing problem compounds it. Most payers require resubmission within 90-180 days. When staff backlogs push claims past that window, the revenue is gone permanently. No appeal is possible. The deadline problem isn’t a process failure; it’s an arithmetic one. There aren’t enough hours.
What technology changes
The question isn’t whether to automate denial management. The question is which parts of the workflow automation can actually own versus which parts still need human judgment.
The answer turns out to be most of it. Roughly 80-85% of denials are systemic: wrong code, missing auth, wrong payer ID, routine missing information. These follow repeatable patterns that machine learning classifies reliably.
The remaining 15-20% involve genuine clinical complexity, payer negotiation, or unusual circumstances that warrant human expertise. That’s where specialists should spend their time.
AI-driven denial management platforms like PULSE, which TSI developed for healthcare RCM operations, do the following: ingest remittance advice, classify denial type and root cause, prioritize by financial impact and overturn probability, draft the appeal using generative AI against payer-specific templates, and surface it to a human reviewer who approves or modifies in minutes instead of hours.
The numbers matter here. TSI’s PULSE platform achieves an 86% denial overturn rate and reduces the time to write an appeal by up to 26 days. Payment per appeal increases 43%. The minimum economically workable claim drops from $5,000 to $500, meaning previously abandoned claims become recoverable.
PULSE PERFORMANCE (TSI Healthcare RCM data)
86% denial overturn rate
43% increase in payment per appeal
50% reduction in time-to-appeal
Up to 26 days faster appeal writing
Minimum workable claim: $5,000 reduced to $500
~47% higher approval likelihood
That last point is worth sitting with. When the floor drops from $5K to $500, claims that were previously certain write-offs become worth pursuing. At scale, that population is large. It’s also where many health systems discover their biggest recoverable opportunity.
The 14% problem
Only 14% of healthcare providers currently use AI to reduce denials, despite 67% believing it could help.
The gap between belief and adoption tells you something about the perceived barrier. Most revenue cycle leaders don’t doubt that the technology works. They’re uncertain about integration complexity, staff change management, and whether the ROI justifies the disruption.
Of providers who have adopted AI for denial management, 69% report reduced denials and/or increased resubmission success. That’s a high adoption satisfaction rate for any healthcare technology category.
The question for CFOs and VP Revenue Cycle leaders isn’t whether denial management AI works. It’s whether the cost of not adopting it, in margin erosion and permanent write-offs, is higher than the cost of change.
For most systems processing more than $100M in annual net patient revenue, the math is straightforward.
What to look for in a denial management partner
Not all denial management solutions are equivalent. Before evaluating platforms, it’s worth defining what you’re actually measuring:
- Overturn rate: What percentage of appealed denials are reversed? The industry baseline is around 35-50%. Anything above 70% indicates a strong appeal writing methodology.
- Time-to-appeal: How long from denial to submitted appeal? Manual processes typically run 15-30+ days. Automated platforms should close this under 5 days.
- Claim floor: What’s the minimum claim size the platform addresses economically? If the floor stays at $5K, you’re still abandoning a significant population.
- Root cause capture: Does the platform capture what’s working by payer and denial type over time, building institutional knowledge? Or does it start fresh on every claim?
- Human review layer: Is there a structured workflow for clinical escalations, or does everything go through the same automation path?
A useful follow-on read: our breakdown of how PULSE specifically handles the denial workflow, including how SAGA AI drafts appeals and how the 86% overturn rate is generated. That’s next in this series.
Frequently asked questions
What percentage of denied claims can be recovered?
Industry data suggests 50-90% of denied claims are recoverable if worked within the timely filing window. The key variable is whether your team has the capacity to work them. Most don’t, which is why the average abandonment rate runs 35-65%.
What causes most healthcare claim denials?
The largest categories are prior authorization issues, missing or incorrect patient information, coding errors, medical necessity disputes, and duplicate claim submissions. The first 3 categories are largely preventable with better front-end workflows; the medical necessity category is where AI has the most impact on appeals.
How does AI improve denial management outcomes?
AI improves outcomes by automating the classification and triage step (which 80-85% of denials require), generating appeal drafts against payer-specific templates, and tracking what works by payer over time. The result is faster throughput, higher appeal quality, and the ability to work smaller claims that would be uneconomical to process manually.
Is denial management outsourcing worth it?
For most health systems, yes, if the vendor has demonstrated appeal outcomes and EHR integration capability. The benchmark numbers to ask for: overturn rate, time-to-appeal, and minimum workable claim threshold. Those three metrics separate capable partners from aspirational ones.
What's the ROI of denial management technology?
It varies by organization, but the typical calculation starts with the abandonment rate: what percentage of your denied claims are currently written off without being worked? If that’s 40% of a $10M denial pool, you’re leaving $4M on the table annually. Most denial management technology pays for itself in the first year if the overturn rate is meaningfully above your current baseline.
TSI’s Healthcare RCM practice specializes in denial management, EBO, and bad debt recovery for mid-to-large health systems. PULSE, TSI’s denial management platform, achieves an 86% overturn rate with 43% higher payment per appeal. For a conversation about what denial management looks like at your organization, contact TSI Healthcare RCM.