The $262B Denial Problem: A Playbook for RCM Shops to Actually Work Their Appeal Pool

US hospitals faced $262 billion in initial claim denials in 2024. Less than one percent of them ever got appealed. Of the ones that did, forty-four percent were overturned. This piece is about the gap between those three numbers, and the workflow that closes it.

Every RCM shop I have ever audited has an unappealed denial pool worth a mid-six-figure annual number. Every one of them knows the pool exists. Very few of them work it. This is not a technology problem. It is a workflow, staffing, and measurement problem. AI is part of the solution, but it is the last part. The first three parts are cheaper, boring, and if you skip them, no amount of AI will save you.

1. The math nobody looks at

Multiply those numbers together for your own book. A mid-size RCM shop servicing 40 provider clients at 12,000 claims per month, at an industry-average 11% denial rate, denies about 1,320 claims a month. At an average commercial claim value of $137, that is roughly $181,000 in denied dollars per month.

If your team works 5% of those denials at a 40% overturn rate (typical for a small operation with no dedicated appeal team), you recover about $3,600 per month. If your team works 30% at the same overturn rate, you recover $21,700 per month. The gap between 5% and 30% is $18,100 per month, or $217,000 per year, on the same underlying denials. Nothing about the payer changed. Nothing about the claim changed. Only the workflow changed.

This is why "denial recovery" as a category is oversubscribed by consultants and undersubscribed by operators. The consultants know the math. The operators are too buried in this week's fresh claims to look at last month's aging.

2. Why almost nobody actually appeals

Four reasons, in order of frequency I encounter them in the field.

Bandwidth: the appeal always loses to the fresh claim

Every biller queue is triaged by dollar volume and recency. A $487 claim that just came in for the first time gets worked before a $487 denial that came back three weeks ago and needs an appeal. Structural reason: the fresh claim has 100% chance of being posted if worked; the denial appeal has maybe 40% chance of being overturned. The biller with a full queue mathematically maximizes their output by working fresh claims first. That is rational. It is also why the appeal pool accumulates.

The fix is not to ask billers to work appeals harder. The fix is to make the appeal a separate queue with a separate person owning it. Appeals-work-only, priced against recovery yield.

Timely filing pressure: the window is shorter than your aging

Payer appeal windows are shorter than most shops track. UnitedHealthcare commercial: 65 days. Cigna commercial: 65 days out-of-network, 180 in-network. Medicare Advantage: 60 days across the board (CMS uniform). Medicare FFS redetermination: 120 days.

If your average time-to-first-appeal from ERA posting is 30 days (industry benchmark: north of 45), and the payer allows 65 days total, you have 35 working days between the biller identifying the denial and the appeal being filed. That is achievable but tight. If you are at 45+ day time-to-first-appeal, you are already writing off half of the appealable UHC book before you touch it.

No ROI visibility: nobody measures the appealable pool

Ask an RCM shop owner "what is your appealable denial pool worth per month?" and you will get one of three answers: "I do not know" (most common), "not much" (spoken with confidence but not measured), or an actual number (rare, and always higher than they thought before they measured).

You cannot prioritize what you cannot see. The first month you actually measure the pool, the number is embarrassing enough to change behavior. This is why the measurement piece comes early in any real workflow build, not late.

Wrong staffing model: appeals are senior work assigned to junior staff

Appeals are complex. They require reading the denial reason code, cross-referencing the payer contract or coverage rule, pulling supporting documentation from the chart or clinical note, drafting a rebuttal, and knowing which payer rep or escalation path to send it to. This is senior-biller work.

Most shops assign appeals to their newest billers because the senior team is booked on posting and follow-up. Junior billers who cannot recognize a winnable appeal reason spend 45 minutes producing a rebuttal that a senior biller could produce in 10, and the file rate is lower. That teaches the shop "appeals do not work" and creates the cycle.

3. Which denials are actually worth appealing

Not every denial belongs in the appeal queue. A workflow that treats every denial equally wastes senior-biller time on unwinnable rebuttals. Three buckets, triaged same-day the ERA lands.

The high-overturn categories (CO-45, CO-50/11, CO-197) are where your workflow investment pays back the fastest. A shop that only appeals those three categories, at 30% appeal rate on each, will recover more than a shop that appeals everything equally at 10% each.

4. The recovery ceiling for your book

Every RCM shop has a specific ceiling on how much revenue can be recovered from appeals. It is not a benchmark or a rule of thumb; it is a math answer that depends on your claim volume, your denial rate, your average claim value, and your state's average overturn rate.

The formula:

If you want your specific number instead of the example above, use the RCM Pulse Report. It asks for your state, specialty, provider count, and denial rate; the appeal calculator on Step 3 returns your specific ceiling using your state's actual overturn rate from the CMS-0057-F transparency file.

5. A six-step workflow that actually works

The workflow below is what a modern operation looks like. It is boring on purpose. The reason most shops do not run it is not that they do not know how; it is that they never carve out the operational discipline to run it consistently.

Step 1: Same-day ERA triage

The day the 835 lands, every denial gets routed into one of three buckets: appealable (send to appeal queue), correctable (send to correction queue for resubmission), not-appealable (write off, log the reason). This triage takes 30-60 seconds per denial with the right rules engine and payer-code-to-bucket mapping. Do not skip it and let the biller decide case by case; the biller will default to "I will get to it later."

Step 2: Appeal queue owned by a specific person

Appeals are their own queue, worked by a specific senior biller or a dedicated appeals coordinator. Not "someone will get to appeals when they have time." Assigned. Deadlined. Measured on recovery yield.

Step 3: AI-drafted rebuttal with human review

For each appealable denial in the queue, the workflow produces a first-draft rebuttal that pulls: the denial reason from the ERA, the relevant coverage rule from the payer's contract or LCD, the supporting documentation from the chart (or a note that documentation is missing), and the correct escalation path. The appeals owner reads the draft in about 2 minutes, corrects or approves, and files.

Step 4: Payer submission with tracking

File through the correct payer portal or clearinghouse channel. Log the appeal ID, the filing timestamp, and the expected response window. The tracking is critical for step 5.

Step 5: Response monitoring and second-level appeal

Every appeal gets a status check at day 30. If overturned, post the recovery. If denied at level one, evaluate for level-two appeal or external review. Some categories (medical necessity, high-dollar) are worth escalating; others are not.

Step 6: Root-cause feedback to submission

The most important step and the most commonly skipped. Every overturned appeal contains information about what should have been on the initial claim to avoid the denial in the first place. Missing auth on 30% of your CO-197 overturns? Fix the auth workflow upstream. CO-45 recoveries are all one payer with a specific fee-schedule carve-out? Update your contract-monitoring rules. The feedback loop from recovery to submission is where a real denial recovery operation quietly stops needing to work as many appeals.

6. The automation triggers that move the needle

Three specific automation triggers, in order of ROI.

Trigger 1: ERA parsing and bucket assignment

The single most valuable automation is parsing incoming 835 remittance files, extracting every denial with reason code and dollar value, and assigning each to the appropriate bucket per your triage rules. Removes the 30-60 seconds per denial of manual triage. On a 1,320-denial-per-month book, that is roughly 22 hours of monthly biller time returned.

Trigger 2: Documentation surfacing at appeal time

When the appeals owner opens a queued denial, the workflow pre-fetches the relevant chart notes, prior authorization records (if any), the specific coverage rule from the payer's LCD or contract, and prior appeals on similar denials from the same payer. Cuts appeal time from 20 minutes to 6 minutes on high-volume categories.

Trigger 3: Draft generation with template + citation

LLM plus a template library plus the correct payer citation produces a first-draft rebuttal in about 90 seconds. The appeals owner reads, adjusts the clinical language, adds any case-specific nuance, and files. This is the automation where governance matters most: never file without human review, especially on medical-necessity or clinical-appeal categories where the wrong language can hurt the shop's payer relationship.

7. Appeal templates by denial category

Every high-overturn category has a template shape that works. These are not the full templates (those are payer-specific and live in the workflow tool), but the structure.

CO-197 authorization denials

1. Payer info, patient info, DOS, claim number, appeal reference
2. Cite payer's own retroactive-auth policy (each payer has one; find it in the provider manual)
3. State the specific circumstance (emergency, unclear coverage at time of service, auth attempted but not obtained in system) with supporting clinical narrative
4. Include supporting documentation: chart note, admission notes, physician certification
5. Request retroactive auth grant and claim reprocessing

CO-45 charges exceed fee schedule

1. Payer info, patient info, DOS, claim number, appeal reference
2. State the correct contracted rate for this CPT and specialty at this location, citing the fee-schedule effective date
3. Attach the contract page showing the rate
4. Request reprocessing at contracted rate with retroactive adjustment

CO-50 / CO-11 medical necessity denials

1. Payer info, patient info, DOS, claim number, appeal reference
2. State the medical necessity: patient's clinical presentation, symptoms, prior treatment attempts, why this CPT was clinically indicated
3. Cite the payer's own coverage policy or LCD supporting the indication
4. Attach chart notes, physician orders, and any prior authorization documentation
5. Request reconsideration with clinical review

8. Payer-behavior tracking (why the same denial keeps recurring)

The pattern most shops miss: certain payers show consistent denial signatures that repeat every month. The same CPT gets CO-197 denied by the same payer for the same population. The same DRG gets downcoded to CO-45 by the same MAC. Nobody notices because each denial is worked (or written off) individually.

If you tag every denial with (payer, CPT/DRG, denial code, month), then aggregate monthly, three patterns will jump out within 60 days:

1. Recurring auth failures on specific CPTs from specific payers. Fix upstream: get the auth for those CPTs before the claim goes out.
2. Recurring rate carve-outs on specific procedures. Fix by updating contract-compliance monitoring to flag the specific carve-out at bill time.
3. Payer-specific silent rule changes. A payer starts denying something they used to pay. The pattern appears in the aggregate before anyone notices in individual claims. Catching it in month one saves 60+ days of denied claims.

The tag-and-aggregate workflow is boring and takes 5 minutes a month. Most shops do not run it. The ones that do stop needing to work as many appeals over time because the upstream fixes retire recurring denials at source.

9. Six metrics to instrument

1. Appeal rate. Percent of denials appealed each month. Target: 25-40%.
2. Overturn rate. Percent of appealed denials overturned. Target: 40-55%.
3. Dollars recovered per appeal. Total recovered / total appealed. Rising is good.
4. Time-to-first-appeal. Days from ERA posting to first appeal filed. Target: under 15 days.
5. Appealed-and-lost pool. Denials appealed but overturned zero. Leading indicator that triage is letting through unwinnable categories.
6. Cost-per-recovered-dollar. Total appeal-workflow cost / total dollars recovered. Should trend down as automation matures.

Six metrics. Every month. Every one of them will move if the workflow is real. If any three go sideways or backwards, the workflow needs adjustment before more investment goes into it.

10. Denial management software vs a custom AI biller

Everything above describes a workflow. The question every operator asks next is which tool actually runs it. There are two categories in the market right now and they solve different problems.

Denial management software

Category leaders here are Waystar, Availity, Change Healthcare, and Experian Health. What denial management software does well: aggregate the 835 remittances, categorize the denial codes, dashboard the pool, and give your team a queue to work. What it does not do: draft the appeal, look up the payer-specific rebuttal citations, attach the supporting documentation, or track filing deadlines and escalate the aged ones. Your team still does the actual work. The software is the workbench, not the worker.

Denial management software is priced per user or per claim volume, ranges from $3 to $25 per provider per month for the light tier up to enterprise contracts north of $100K per year at hospital scale, and is a good fit when your team already has denial-work capacity and just needs better tooling.

Denial management services

The outsourced version. You hand denials to a firm that has an offshore or nearshore team of billers who work the appeals for you and share the recovery. Recovery rates are variable, timely-filing risk moves to the vendor, and you give up institutional knowledge of your denial patterns. This is a good fit when you cannot hire billers at all and need throughput without headcount.

A custom AI biller

The third category is what our AI Biller is. Instead of giving your team a better workbench or renting an outside team, we build a custom worker that lives inside your PM environment and does the work. The AI biller reads every denial as it lands, classifies it, drafts the appeal with the right payer-specific language, attaches the supporting documentation from the chart, and queues it for a human coder to review and submit. It monitors filing deadlines and escalates before they expire. It watches for payer rule changes and flags them before your claims start denying against the new rule.

It is delivered in 30 days, sits inside your workflow rather than pulling your team into a new one, and costs less than one biller per quarter. If it does not outperform your current process on three agreed metrics inside 30 days, the fee is refunded.

The right fit depends on where your bottleneck is. Software helps if you have capacity and need better tooling. Services help if you cannot hire. A custom AI biller helps if the problem is that your team is good but there just are not enough of them to work the pool, which is the situation almost every RCM shop is actually in.

11. Where to start Monday morning

Five steps. Skip any of them and the rest do not work.

1. Pull 90 days of denials from your top account. Export the 835 files. Aggregate by denial reason code. Rank by dollar volume. Now you have baseline.
2. Score the appealable pool. Apply the three-bucket triage from section 3. What percent of your top three denial categories is appealable? At what overturn rate would recovery pay back?
3. Assign one person to appeals for one week. Take a senior biller off their regular queue for one week. Have them work the top-three-category appealable pool. Track appeals filed, hours spent, dollars recovered. Now you have a real productivity number for your operation.
4. Pick one automation to build first. Almost always ERA parsing + bucket assignment. Small build, large ROI, unlocks the rest of the workflow.
5. Set the 90-day success bar in writing. "By day 90, appeal rate moves from X to Y on our top three denial categories. Overturn rate stays at Z or better. Recovery increases by $N." Write it down. Review at day 90. Do not move the goalpost.

12. Frequently asked questions