Medical Billing Denial Benchmarks 2026

This is the whole ledger. Each row states the figure exactly as its source states it, the population it was measured on, the denominator, whether it is a measurement or a target, and where we read it. Nothing appears elsewhere on this page that is not in this table or explicitly labelled synthetic.

How to read this table without making the standard mistake. The rows are not comparable to each other. The hospital rows describe institutional claims; the physician-practice rows describe professional claims; the payer-side rows describe what an insurer did with everything it received, across every provider type. A practice that compares its own denial rate to the 19% marketplace figure, or to a hospital figure, will reach a wrong conclusion in whichever direction the mismatch runs. Compare like to like, or compare to your own prior period — which is usually the more useful comparison anyway.

What is not here, and why. HFMA's MAP Keys are the standard revenue-cycle KPI definitions, and the page publishes equations and data-source guidance without target values. So there is no HFMA denial target, no HFMA clean-claim target and no HFMA top-quartile figure to cite, and any page attributing one to HFMA has misattributed it. MGMA's fuller practice medians sit inside licensed DataDive products and are not free to quote. We state those blanks rather than filling them.

On denominators. The single technical point that decides whether two denial rates can be compared at all is what sits underneath. A denial rate should be computed on claims that reached adjudication, because a claim rejected before adjudication never got a payment determination and has no appeal rights. That matters practically: a practice whose rejection rate is climbing shrinks its own denial-rate denominator and can watch its denial rate improve while its collections fall. Our clean claim rate page separates the four metrics this is routinely confused with.

Specialty mix is one of the largest drivers of denial-rate variance, because each specialty has its own dominant denial pattern — different code families, different payer behaviours, different documentation requirements.

What this table is. It is an operational map of the patterns our own teams work most often, by specialty. It is not a benchmark table and carries no percentages, because we could not find a public source that publishes denial rates by physician specialty with a stated denominator and population. Treat each row as a place to look first in your own data, not as a claim about how often it happens nationally.

The pattern across specialties is structural rather than statistical: the specialty's procedure mix decides which denial family dominates. Procedural specialties — cardiology, orthopedics, dermatology, urgent care — cluster on bundling and modifier issues. Endoscopy-driven specialties show the same profile, with multi-procedure colonoscopy sessions a standing NCCI exposure, covered on our gastroenterology billing services page. Visit-based specialties — mental health, family practice, internal medicine, pediatrics — cluster on coding-level scrutiny, modifier 25 and missed add-ons, including the G2211 longitudinal-care exposure that is the recurring primary care billing issue. Therapy specialties cluster on time-discipline and concurrent-billing rules.

Use this as a hypothesis list. The only authoritative statement about your specialty's denial mix is the one you compute from your own remittances.

We have removed the frequency claim that used to head this section. The widely repeated "about 20 CARC codes account for roughly 80 percent of denial volume" is not traceable to a published study with a stated population and denominator, and we could not source one. Payers do not publish code-level adjudication frequencies, so there is no national CARC frequency distribution to cite — and inventing one to make a page look authoritative is the exact failure this page exists to avoid.

What we can say is narrower and more useful: the codes below are the ones that recur in physician-practice remittances, and knowing what operational category each sits in is what lets you route a denial worklist. The meanings in the middle column are our plain-language restatements, not X12's published descriptions — those are copyrighted and must be read from the X12 list itself. Every code below was Current on the X12 list read 17 September 2026.

A counting warning that changes reported denial rates materially. Not every CARC line is a denial. X12's own FAQ for the code list defines "adjusted" as meaning the charge amount differs from the payment amount, and lists denial, zero payment, partial payment, reduced payment, penalty applied, additional payment and supplemental payment as examples. Rows 1, 2, 3 and 45 above are ordinary cost-sharing and contractual adjustments, not denials. A denial rate computed by counting every CARC line is inflated by routine write-offs, and that single classification error accounts for a good share of the wildly divergent denial rates practices report to each other.

The operational implication is the same one that runs through this whole page: classify at ingestion — contractual adjustment, patient cost-sharing, or denial — then categorise only the third group and route each category to the team that owns the prevention step. Our CARC and RARC reference covers the group-code and remark-code reading in full.

> This is an illustrative example demonstrating MedPrecision's reporting methodology. It is not presented as the result of an actual client engagement. Every practice-level figure below is synthetic. No patient, practice, or payer data appears anywhere on this page. Benchmark comparators are real and cited.

Example Practice A — the same illustrative practice used across all of our sample deliverables, so the artifacts can be read end to end: a four-physician, two-APP family medicine group in Ohio, roughly 2,050 encounters and $412,000 in gross charges per month. Reporting period: June 2026.

The benchmarks above describe the industry. This is what the same analysis looks like applied to a single practice for a single month — the artifact, not the average.

Denials by CARC code

127 of the 1,940 claims that reached adjudication were denied on first pass — a 6.5% payer denial rate. Each row carries the disposition decision, because a denial count without a disposition is a statistic rather than a worklist.

Percentages are rounded to one decimal, so the column sums to 99.9% rather than 100%.

Disposition summary — what actually happens to the money

We publish the write-off line. $2,285 in this month is not coming back, and 5 of those claims died on timely filing — a process failure, not a payer decision. A denial analysis that reports only recoverable dollars is a sales document.

How this month compares to public data

• The practice's 6.5% denial rate above is a claim count measured on claims that reached adjudication. AAFP's 5–10% guidance is computed on a different basis — denied claim dollars over submitted claim dollars — so read it as an order-of-magnitude sanity check here, not a like-for-like comparison.
• For context on the payer side rather than the practice side: KFF found HealthCare.gov insurers denied 19% of in-network claims in 2024, with individual insurer rates ranging from 3% to 36%. That is a different denominator — marketplace plans, all claim types — and should not be read as a practice target.
• KFF also reports that of denied in-network Marketplace claims in 2024, consumers appealed less than 1%, and insurers upheld 66% of the appeals they received. The lesson for a practice is the opposite of discouraging: almost nobody appeals, so the appeal channel is systematically under-used.
• On the prior-authorization side, KFF found Medicare Advantage insurers denied 4.1 million requests (7.7%) in 2024, only 11.5% of those denials were appealed — and 80.7% of appealed denials were partially or fully overturned. The 19 CO-197 authorization denials above are worth appealing for exactly that reason.

The root-cause reading

Grouping by CARC answers what was denied. Grouping by originating step answers why, and only the second one generates a fix:

1. Front-end capture — 47 claims ($10,330). CO-16, CO-27 and CO-22 all originate at registration and eligibility verification, not in billing. This is the largest single root cause and the cheapest to fix.
2. Coding and modifiers — 42 claims ($8,510). CO-4, CO-11 and CO-97 point at coding and scrubber-rule gaps.
3. Authorization — 19 claims ($7,410). The highest dollar-per-claim group at $390, handled by prior authorization.
4. Process failure — 5 claims ($1,275). CO-29 timely filing. Unrecoverable, and the only category that is entirely self-inflicted.

The single highest-value action is not the largest dollar bucket — it is the front-end capture group, because those 47 denials recur every month until the registration workflow changes.

The same reason code can require a different appeal at different payers, which is why payer-specific templates outperform generic denial management. What follows is an operational map of behaviours our teams encounter, organised by payer.

Read the caveat before the list. These are patterns, not published statistics. We have removed the numeric audit thresholds that previously appeared here — percentages of visits, session counts, unit caps — because we could not tie any of them to a retrievable payer policy document, and a threshold quoted to the percentage point is exactly the kind of figure a reader will act on. Payer medical and reimbursement policies change without notice and vary by plan and by state, so verify the current policy for your plan and state before you build a rule on any row below.

UnitedHealthcare and Optum

• Scrutiny of modifier 25 on same-day E/M plus procedure, most visibly in dermatology.
• Documentation review on multi-stage Mohs surgery cases.
• Functional outcome measures required at intervals for continued physical therapy authorization, with continued-treatment denials where outcome data is missing.
• Bilateral procedure reductions applied differently from Medicare in some orthopedic cases.
• Component-based vaccine administration expected for patients under 19, with denials where the wrong series is billed for the patient's age.

Aetna

• Session-length documentation for 90837 challenged on records request, with recoupment exposure where start and stop times are not documented.
• Supervision and interpretation components of a stress test bundled when both are billed under the same NPI on the same date.
• Add-on psychotherapy codes not reimbursed for some non-physician behavioral health provider types, which limits add-on billing by provider credential rather than by service.
• Fracture-care codes bundled with a same-day E/M.

Cigna

• Retrospective chart review of established-patient E/M levelling.
• Echocardiography downcoded from complete to limited where the documentation does not list every element of a complete study.
• Family therapy benefit limits applied per plan, with couples therapy not always counted as a distinct benefit.
• Combined unit limits applied to therapeutic exercise plus manual therapy on some plans.
• Multi-biopsy claims held pending a records request.

BCBS plans

• Joint-replacement authorization denials where conservative-treatment documentation is not included in the request.
• Phototherapy diagnosis pairing required, with documentation of failed topical therapy.
• Group therapy reimbursement varying by state, and bundled with individual treatment codes when billed on the same day.

Medicare fee-for-service

• The 90-day global surgical period bundles related follow-up care into the surgical fee; modifier 24 for an unrelated E/M or 58 for a staged procedure is required on separately billable post-operative encounters.
• NCCI Procedure-to-Procedure edits apply automatically, denying separately billable components without a distinct-service modifier.
• Telehealth place-of-service selection — the patient's home versus another location — drives a reimbursement differential. The rules here have changed repeatedly since the public health emergency and are the most time-sensitive item on this page; check the current year's Physician Fee Schedule position rather than any guide, including ours. Our telehealth medical billing services page covers the current place-of-service treatment.

Medicare Advantage

This is the one payer category with genuinely published denial data, and it sits on the payer side rather than the provider side. KFF's analysis of CMS data found that Medicare Advantage insurers fully or partially denied 4.1 million prior authorization requests in 2024 — 7.7% of all requests, against 6.4% in 2023 — that just 11.5% of denied requests were appealed, and that 80.7% of appealed denials were partially or fully overturned. Separately, OIG found on a stratified random sample of 250 prior authorization denials and 250 payment denials from 15 of the largest Medicare Advantage organizations that 13 percent of denied prior authorization requests met Medicare coverage rules, and 18 percent of denied payment requests met Medicare coverage rules and the organization's own billing rules.

Those two findings point the same way and are the strongest argument on this page for appealing: a low appeal rate against a high overturn rate means the appeal channel is systematically under-used, and an audit found a material share of denials should not have been issued at all.

Prior authorization is the denial category with the most published data, because Medicare Advantage reporting and federal audit work have produced figures with stated populations and samples. It is also the category where federal rules are actively changing what payers must do.

What is measured. Two sources, both payer-side or audit-side rather than provider-side:

• KFF's analysis of CMS data found that Medicare Advantage insurers fully or partially denied 4.1 million prior authorization requests in 2024, a somewhat larger share — 7.7% — of all requests than in 2023, when the figure was 6.4%. Of denied requests, just 11.5% were appealed, and 80.7% of appealed denials were partially or fully overturned.
• OIG reviewed a stratified random sample of 250 prior authorization denials and 250 payment denials issued by 15 of the largest Medicare Advantage organizations during 1-7 June 2019, and found that 13 percent of the denied prior authorization requests met Medicare coverage rules and 18 percent of the denied payment requests met Medicare coverage rules and the organizations' own billing rules. Note the denominator: denials sampled, not requests submitted.

What we removed. This section previously led with a physician-survey figure on prior-authorization-related adverse events, attributed to a named annual survey. We could not retrieve the underlying report to verify the figure or its sample this session, so we have removed it rather than relay it. If you have seen that number cited elsewhere, treat it as unverified until you can read the source yourself.

What changes in 2026 and 2027. The CMS Interoperability and Prior Authorization Final Rule (CMS-0057-F) applies to Medicare Advantage organizations, state Medicaid and CHIP fee-for-service programmes, Medicaid managed care plans, CHIP managed care entities, and Qualified Health Plan issuers on the federally facilitated exchanges. CMS states the rule requires prior authorization decisions within 72 hours for expedited requests and seven calendar days for standard requests, with those timeframes beginning to apply from 1 January 2026 for impacted payers other than exchange QHP issuers. Payers must publicly report certain prior authorization metrics annually on their websites, with initial reporting due by 31 March 2026, and most of the API requirements must be implemented by 1 January 2027.

The operational consequence is concrete rather than rhetorical: from 2026 onward, impacted payers are publishing their own prior-authorization approval, denial and appeal metrics. That is the first time provider-facing teams will be able to compare payers on authorization behaviour using each payer's own published numbers rather than anecdote — and it is a better source than any industry average, because it is specific to the payers you actually bill. Build the annual check of those postings into your payer-management calendar.

On the provider side, the corresponding adjudication code is CARC 197 — authorization absent — and it remains one of the most operationally preventable denial categories, because the prevention step happens at scheduling rather than at billing.

There is no published top-quartile denial rate for physician practices that we could verify, so this page does not print one. What exists is one freely published target, from a body that publishes it as its own guidance rather than as a measurement:

AAFP, verbatim: "A 5% to 10% denial rate is the industry average; keeping the denial rate below 5% is more desirable."

Read that as what it is — practice-management guidance with no population, sample or data year attached. AAFP does publish its denominator, and it is not the one most practices report on: the dollar amount of claims denied over the dollar amount of claims submitted, rather than a claim count on claims that reached adjudication. It is a useful managerial anchor and it is not a measured benchmark, and the difference matters when you put it in a board pack. AAFP publishes two related targets on the same page, also as guidance: the adjusted collection rate should be 95% at minimum with an average of 95% to 99%, and days in A/R should stay below 50 days at minimum with 30 to 40 days preferable. AAFP publishes no clean-claim-rate target, and neither does anyone else we could find for free — see our clean claim rate page for that blank stated in full.

HFMA's MAP Keys publish the standard KPI definitions and equations without target values, so there is no HFMA denial target to cite. Any "HFMA top-quartile denial rate" you encounter is misattributed.

What we can describe without a number is the set of operational disciplines that distinguish practices whose denial rate falls from practices whose denial rate is merely reported. These are stated as practice, not as evidence of an outcome:

1. Pre-submission scrubbing with current edit tables. Refresh NCCI Procedure-to-Procedure edits, MUE limits and payer-specific rules on a calendar. Stale tables produce both false-positive scrubs and missed denials.
2. Real-time eligibility verification before service. 270/271 transactions confirming coverage, cost-sharing, deductible accumulation and behavioral-health carve-out routing. Eligibility denials are the most preventable category and the cheapest to fix.
3. Authorization tracking integrated with scheduling. No service performed without a confirmed authorization where payer policy requires one.
4. A denial worklist categorised at ingestion, with payer-specific routing — and with contractual adjustments and patient cost-sharing separated out before categorisation, so the worklist is denials rather than everything.
5. A same-week feedback loop to the front-end teams. Eligibility items to the front desk, authorization items to the authorization team, bundling items to coding. Without the loop, the worklist never shrinks.
6. Modifier discipline created at the encounter, not at appeal. Documentation supporting a distinct-service or unrelated-E/M modifier has to exist before the claim goes out; retrofitting it at appeal time is what makes bundling denials unwinnable.
7. A timely-filing aging report run daily, with a same-day charge-entry service level behind it.

On the economics of appealing. The only cost-of-appeal figure we could verify is Premier's, and its population is hospitals, health systems and post-acute providers, not physician practices: an average cost incurred fighting a denial of $57.23 per claim on 2023 claims, against $43.84 the year before, alongside the finding that approximately 70 percent of denials are overturned and paid. Do not import that per-claim cost into a physician-practice model — compute your own from biller minutes per appeal and your loaded rate. The directionally useful part is the shape: a high overturn rate against a real per-claim cost is what makes prevention worth more than appeal, and it is an argument that survives without a precise number.

What we removed from this section. A cost-per-appeal range attributed to HFMA, a figure attributed to Advisory Board, a claim that roughly 65 percent of denied claims are never appealed industry-wide, and a set of appeal overturn rates by denial category. None could be tied to a retrievable source with a stated population, and several were attributed to bodies that do not publish them.

The rule this page follows. Every quantitative claim either carries a population, a denominator and a link to a source we retrieved on 17 September 2026, or it is labelled synthetic, or it is not on the page. Where a figure could not be tied to a retrievable source with a stated population, we removed it rather than softening it into a range — because a softened unsourced figure is still an unsourced figure, and it is harder for a reader to catch.

The sources used, with their limits:

The figures we removed from this page, and why. Listed explicitly so a reader who remembers seeing them here knows they were withdrawn rather than relocated:

• A 9-to-12 percent industry first-pass denial range attributed to MGMA DataDive and HFMA. Neither publishes it; we could not source it.
• An HFMA top-quartile denial target below 5 percent, and an HFMA clean-claim target of 95 to 98 percent. HFMA's MAP Keys publish definitions only, with no target values.
• An MGMA median net collection rate near 93 percent and an HFMA top-quartile net collection target of 96 percent or above.
• An MGMA median of around 40 days in A/R and an HFMA best-practice target under 35 days.
• "Approximately 20 CARC codes account for around 80 percent of denial volume," attributed to MGMA, HFMA and AAPC published guidance. AAPC does not publish it, MGMA does not publish it, HFMA does not publish it, and no payer publishes code-level frequency data at all.
• A claim that approximately 65 percent of denied claims are never appealed industry-wide, attributed to HFMA.
• A cost-per-appeal range of roughly $25 to $118 per claim, attributed to HFMA and to Advisory Board.
• Appeal overturn rates of 40-50 percent industry-average, 65-75 percent best-practice, and category-specific ranges of 50-70, 55-65 and 70-80 percent.
• A statement that best-in-class multi-specialty groups run 3-4 percent.
• A physician-survey figure on prior-authorization-related serious adverse events. The underlying report could not be retrieved this session to verify the figure or its sample.
• A Medicare Advantage denial-rate range of approximately 5 percent to over 18 percent by insurer, attributed to KFF. KFF's published Medicare Advantage figures are prior-authorization determinations, which are now cited directly above instead.
• Numeric payer audit thresholds in the payer-behaviour section — percentages of visits, session counts, unit caps and biopsy counts. These read as payer policy but were not tied to a retrievable policy document.

On the specialty and payer sections. Those describe operational patterns our teams work, not measured frequencies, and they are labelled as such in place. A pattern is a place to look in your own data; it is not a claim about how often something happens nationally. Where a section previously cited our own specialty pages as the source for a numeric claim, that circular attribution has been removed — a figure sourced to another page of this site is not sourced.

The one comparison that is always valid. Your own prior period, on a fixed definition. Every external figure on this page carries a population that is probably not yours; your last quarter does not.