Every denials team has a version of the same Friday afternoon. There are 340 open appeals sitting in the work queue, three FTEs to work them, and a controller asking why days in A/R crept up again this month. Somebody picks the top of the list, usually sorted by denial date, sometimes by dollar amount, and starts working down it. By Monday, the team has burned twelve hours on appeals for claims that were never going to get overturned, while a $38,000 inpatient denial with a strong clinical argument sat untouched three rows down.

That's not a staffing problem. It's a prioritization problem, and it costs most denial management programs more than they realize.

Why "biggest dollar amount first" isn't a strategy

Sorting the appeal queue by claim value feels like prioritization, but it's really just triage by size. It ignores the second variable that determines whether an appeal is worth the labor: the probability that you'll win it.

A $12,000 denial for lack of prior authorization on an elective procedure your payer contract explicitly requires pre-cert for is, frankly, a lost cause. A $1,800 denial coded as a bundling edit that your team has overturned 70% of the time in the last two quarters is a much better use of a biller's afternoon, even though it's a tenth of the dollar amount.

Revenue cycle teams that sort purely by size end up spending their best hours on their worst odds. The fix isn't more staff. It's a scoring model that weighs both variables together.

The two variables that matter

Revenue impact is the easy half. It's the dollar amount at stake: the expected reimbursement if the appeal succeeds, net of any contractual adjustment. For claims with partial payment already posted, use the shortfall, not the billed charge.

Denial probability is where most organizations either skip the math entirely or get it backwards. Here, it's more precise to talk about appeal win probability, meaning the likelihood that a specific denial, given its payer, denial reason (CARC/RARC code), documentation status, and claim age, will be overturned if you appeal it. Some teams call this a "denial recovery score" or "appeal viability score." Whatever the label, it's the number that turns a flat worklist into a ranked one.

Multiply the two, and you get an expected value for every open appeal:

Priority Score = Revenue Impact × Appeal Win Probability

Take that same Friday queue. The $38,000 inpatient denial has thin documentation and a denial reason your team has historically overturned about 12% of the time. That's an expected value of roughly $4,560. A $9,000 authorization denial sitting further down the list, with an 65% historical overturn rate for that payer and reason code, works out to about $5,850. Under a dollar-only sort, the $38,000 claim wins every time. Under an expected-value sort, the $9,000 claim gets worked first, and the bigger denial gets routed for the extra clinical support it needs instead of sitting in a general queue.

Building the win-probability score without guessing

The revenue side of this equation is usually sitting in your claims data already. The probability side takes more work, but it doesn't require a data science team. Most mid-size RCM departments can build a workable version from six inputs they already have:

  • Denial reason code (CARC/RARC). Some categories, like timely filing, duplicate claim, or non-covered service, have close to zero overturn potential unless there's a documented payer error. Others, like medical necessity or bundling edits, overturn at meaningfully higher rates when the clinical documentation supports the claim.

  • Payer. Overturn rates for the same denial reason can swing 20 to 30 points between payers. Medicare Advantage plans and ACA marketplace plans, which industry benchmarking from Kodiak Solutions and Experian Health has shown carry some of the highest initial denial rates in 2026, don't behave the same way on appeal as traditional Medicare or a stable commercial contract.

  • Claim age relative to appeal deadline. A denial that's 80 days into a 90-day appeal window with missing documentation isn't the same bet as one filed the week the denial posted.

  • Documentation completeness. Whether the clinical notes, itemized bill, and medical necessity criteria are already assembled, or whether someone has to chase a physician for an addendum first.

  • Historical overturn rate for this denial type, at this payer, in your own data. This is the single highest-value field, and it's usually just sitting in your appeals tracking spreadsheet or your clearinghouse's denial module, unanalyzed.

  • Appeal level. First-level appeals generally clear at meaningfully higher rates than second-level or external review, simply because the easier wins get resolved earlier in the process.

A logistic regression on eighteen months of your own appeal outcomes, run in Excel or a basic BI tool, will get most departments 80% of the way to a usable win-probability score. Machine learning isn't a prerequisite. The AI-enabled platforms on the market do this at scale and refresh the model continuously, which matters more once your appeal volume gets into the thousands, but a mid-size hospital or specialty group can build version one with a spreadsheet and a few hours of analyst time.

The four-quadrant view: what to do with the score

Once every open appeal has a revenue-impact number and a win-probability number, plotting them on a simple 2x2 tells the team exactly where to spend their day.

High dollar, high probability claims should be worked today. This is the obvious bucket, but it's shocking how often high-probability, high-dollar claims sit in queue behind older, lower-value ones simply because of FIFO habits. These go to your most experienced appeal writers, with the shortest possible turnaround.

High dollar, low probability claims are worth appealing, but not with a form letter. This is where you attach a peer-to-peer review, pull in a physician advisor, or escalate to a payer relations contact before the deadline. If the win rate is low because the documentation is weak, this quadrant is also your best early-warning signal for a front-end fix, usually a coding or authorization gap upstream that's costing you the same denial over and over.

Low dollar, high probability claims are ideal candidates for standardized appeal letters, junior staff, or automation. The per-claim value doesn't justify a senior biller's time, but the aggregate recovery across a hundred small claims can rival a handful of large ones.

Low dollar, low probability claims should be written off deliberately, not by neglect. Every denials team has claims that will cost more in labor to appeal than they'll ever recover. The mistake isn't writing these off. It's writing them off by default because nobody sorted the queue, rather than as a documented decision with a policy behind it.

What this looks like in a real workflow

The scoring model only pays off if it changes how the queue gets worked, not just how it gets reported to finance.

A few things that make the difference between a spreadsheet exercise and a real operational shift:

  1. Re-score weekly, not once. Win probability changes as your own overturn history accumulates and as payers shift behavior. 2026 has already seen a wave of payers moving to AI-driven adjudication that changes which denial reasons are winnable, so a static model goes stale within a quarter.

  2. Assign by quadrant, not by biller availability. Route high-dollar, low-probability claims to your most senior appeal writers or a physician advisor pool specifically, rather than whoever's queue is shortest.

  3. Set different SLAs per quadrant. High-value, high-probability claims should have a 24 to 48 hour turnaround target. Low-dollar, high-probability claims can sit in a batch that runs weekly.

  4. Feed the outcomes back into the score. Every resolved appeal, won or lost, is a new data point. Teams that treat this as a one-time build rather than a living model watch their scoring accuracy decay within two or three quarters.

  5. Report expected value recovered, not just dollars appealed. This reframes denial management conversations with finance leadership around what was recoverable, not just what was contested, and it makes it much easier to justify headcount or a platform investment with a real ROI case.

Common mistakes worth naming directly

Treating all "medical necessity" denials the same. This is the widest category and the one with the most internal variance. A medical necessity denial on an imaging study with a clear clinical indication in the chart behaves nothing like one where the documentation genuinely doesn't support the code billed.

Building the score around billed charges instead of expected net reimbursement. Contractual write-offs mean the real revenue impact is often 40 to 60% lower than the charge on the claim. Scoring on gross charges systematically overweights certain service lines.

Ignoring the labor cost side of the equation entirely. Industry estimates put the cost to rework a single denied claim somewhere between $25 and $180 depending on complexity, worth factoring in as a rough floor below which an appeal isn't worth pursuing regardless of win probability.

Never closing the loop with the front end. A prioritization model tells you where the money is stuck. It doesn't fix why it got stuck in the first place. The highest-performing denial teams route their high-dollar, low-probability quadrant findings straight back to coding, authorization, and registration as a standing agenda item, not just an appeals problem.

The KPIs that show this is working

A few numbers tend to move within one to two quarters of implementing a real prioritization model, and they're worth tracking separately from your overall denial rate:

  • Expected value capture rate: dollars recovered against the total expected value the model projected, a cleaner efficiency measure than raw dollars appealed.

  • Appeal win rate by quadrant: confirms the model is predictive, not just intuitive.

  • Average days from denial to appeal submission, broken out for the high-priority quadrant specifically.

  • Cost per dollar recovered, factoring in labor hours against recovered revenue.

  • Denial recurrence rate for the CARC codes flagged as high-dollar/low-probability, which should trend down if the upstream feedback loop is working.

Most of these are extractable from whatever appeals tracking tool or clearinghouse denial module your team already has, once someone decides they're worth pulling.

Frequently Asked Questions

What's the difference between denial probability and appeal win probability? In casual use, revenue cycle teams often say "denial probability" to mean two different things: the likelihood a claim gets denied in the first place, and the likelihood a denied claim gets overturned on appeal. For prioritization purposes, the number that matters is the second one. This article uses "appeal win probability" specifically to avoid that ambiguity.

How much claims history do we need before this scoring model is reliable? Twelve to eighteen months of appeal outcomes is usually enough to get directionally useful overturn rates by payer and denial reason, assuming you have at least a few hundred resolved appeals per major category. Below that volume, blend your own data with published industry benchmarks for the same CARC code and payer type until your sample size catches up.

Do we need a denial management platform to do this, or can we build it in-house?

You can build a first version in a spreadsheet using historical overturn rates you already have, and many mid-size organizations do exactly that. A platform becomes worth the cost once appeal volume is high enough that manual re-scoring becomes a bottleneck, or when you want the model to update automatically as new outcomes come in.

Should low-probability, high-dollar denials always be appealed anyway?

Not automatically. If the underlying issue is a hard contractual exclusion or a documented payer policy you can't overcome, appealing is usually wasted labor. If the low probability is driven by weak documentation rather than the merits of the case, it's often worth the extra effort of a peer-to-peer review or physician advisor input before deciding to write it off.

How often should the priority score be recalculated? Weekly is a reasonable cadence for most teams. Payer behavior, staffing, and your own overturn history all shift quickly enough that a model built once and left alone will be noticeably stale within a quarter.

What's a realistic first KPI to report to leadership after implementing this? Expected value capture rate tends to land best with finance leaders, since it ties directly to the dollars the model predicted were recoverable versus what was collected, rather than just the volume of appeals filed.

Does QWay Healthcare only handle appeals, or does it also address the root causes of denials?

Both. QWay's denials management services cover the appeals workflow itself, while the broader AI-governed RCM platform is built to catch denial-driving errors upstream, before claims are ever submitted. That upstream and downstream pairing matters because, as this article covers, a prioritization model tells you where revenue is stuck, but closing the loop with coding, authorization, and registration is what actually reduces the volume of denials reaching the appeals queue in the first place.

Where to go from here

Sorting appeals by expected value instead of raw dollar amount takes most teams two to three quarters to fully operationalize, and it doesn't require adding headcount or buying a new platform. It won't fix a broken front-end process that's generating avoidable denials in the first place. For that, take a look at our practical framework for preventing denials before submission. But a working priority score will make sure the appeals team's limited hours are pointed at the claims most likely to turn into cash.

External References