Top reasons medical claims get denied

Most practices treat denials as a billing department issue. The data says otherwise. Optum's Revenue Cycle Denials Index, built on 124 million hospital claim remits across more than 1,400 US hospitals, found that 44% of all denials originate in front-end revenue cycle processes — registration, eligibility and authorization. That share has risen steadily from 34% in 2019–20.

This matters because front-end denials are the cheapest to prevent and the most expensive to ignore. A coding error is caught by a coder. An eligibility error is caught by nobody until the remittance arrives weeks later, by which point the care has been delivered and the cost is sunk.

Here is how denials distribute across categories, from the same Optum dataset covering 2023:

Within the registration and eligibility category, coordination of benefits accounts for half of all denials. Patients carry coverage the practice does not know about, or coverage order changes and nobody re-checks. Benefit maximums account for another 27%, and plan coverage issues 17%.

The common thread is that eligibility is verified once, at intake, and then assumed stable. It is not stable. Coverage terminates, employers change carriers, and Medicaid managed care assignments shift. Verifying before each visit rather than once per patient removes an entire denial category.

Premier Inc. measured the average cost of contesting a single denied claim at $57.23 in 2023 — a 31% increase in one year. Across US providers, that adds up to $25.7 billion spent annually adjudicating claims with payers, of which roughly $18 billion is potentially unnecessary.

The more striking figure is what happens when practices do appeal: approximately 70% of denied claims are ultimately overturned and paid. The revenue is usually collectible. The obstacle is capacity, not merit.

If your denial rate is above 10% — and 41% of providers now report exactly that, up from 30% in 2022 — the highest-leverage work is not in the billing office.

Knowing the distribution is only useful if it changes what you do on Monday. The three largest categories have different owners and different fixes.

A rejection and a denial are frequently conflated, and the difference determines where the claim actually is.

A rejection happens before adjudication: the clearinghouse or payer refuses the claim for a format or data problem, so it never enters the payer's system. It can be corrected and resubmitted without an appeal, and critically, it does not appear in any aging report built from payer data.

A denial happens after adjudication: the payer processed the claim and decided not to pay. It carries a CARC code, it is visible in remittance, and changing the outcome requires an appeal.

The dangerous one is the rejection, because it exists in your system as submitted and in the payer's system not at all. It ages silently until timely filing closes — which is how CO-29 denials are manufactured months before they appear.

A claim adjustment reason code says what the payer did. The remittance advice remark code that accompanies it usually says why, and it is the field most often ignored because it sits further right on the remittance.

CO-16 is the clearest example. On its own it means “claim lacks information or has submission errors,” which is not actionable — it could be almost anything. The attached remark code names the missing element: a specific identifier, an absent attachment, a required field left blank. Working CO-16 from the reason code alone produces guesswork and repeat denials; working it from the remark code produces a fix.

The same applies to CO-97, where the service is bundled into another already adjudicated. The reason code tells you it was bundled; only the remark and the paired claim tell you what it was bundled into, and therefore whether a modifier would have been appropriate or whether the bundling was correct.

Denials are usually worked one claim at a time because that is how they arrive. It is the least efficient possible order, because denials are not independent events — they arrive in clusters produced by a single upstream cause.

One payer changing an authorization requirement generates dozens of denials that look individually unrelated. A fee schedule loaded incorrectly generates a steady trickle of underpayments across months. A provider whose credentialing lapsed generates every denial on their panel at once.

Sorting the month’s denials by payer, by code, by provider and by CPT takes a few minutes in any reporting tool and answers a different question than the claim-level queue does. The queue asks which claim to fix next. The clusters ask which cause to remove, and removing one cause typically prevents more future denials than a week of individual rework recovers.

Not every denial is worth the labour. Premier Inc. puts the average administrative cost of fighting a single denied claim at $57.23, which means a denial on a claim worth less than roughly that amount costs more to pursue than it returns — unless working it removes a cause that will recur.

That caveat is the important half. A $40 denial is not worth appealing on its own economics, but if it is the twelfth instance of the same fee schedule error, the fix is worth far more than the twelve claims combined. The decision is therefore not about the individual balance but about whether the denial is isolated or representative.

A workable rule: appeal anything above the cost-to-work threshold on its own merit, and for anything below it, look at whether the same code and payer appear more than a handful of times that month. If they do, fix the cause and resubmit the batch. If they do not, write it off and spend the time on the clustered categories instead.

The practices that reduce denials rather than merely working them do the same unglamorous thing every month. It is not sophisticated and it does not take long.