20 medical billing stats that baffle us

Every figure below carries the year it covers and the group it was measured on. That sounds like a small thing. It is the whole reason the published numbers on this topic contradict each other so badly, and why a practice manager can read four articles and come away with four different ideas of what normal looks like.

Why do published medical billing statistics disagree so badly?

Because the denial rate you are quoted depends entirely on what the publisher counted. KFF analyzed 2024 HealthCare.gov filings covering roughly 451 million in-network claims and found an average in-network denial rate of 19%. MGMA puts first-pass denial for single-specialty groups at 8%. Neither is wrong. They count different claims, at different stages of the cycle, against different denominators.

The chart below lines up eight published rates. The spread runs from 6.1% to 19%, and the two extremes differ by a factor of three.

Four variables explain almost all of that spread. First, initial versus final denial: a claim denied on submission and paid on resubmission counts in one number and not the other. Second, payer market: marketplace plans, Medicare Advantage, Medicaid managed care and commercial employer plans behave differently, and the 2025 federal reporting data shows it, with prior authorization denial rates of 18% in the marketplace, 14% in Medicaid managed care and 12% in Medicare Advantage. Third, claim type: hospital facility claims and physician professional claims are not comparable. Fourth, the denominator: claims submitted, claims adjudicated, or charges in dollars.

Practice-reported figures are a fourth view again: in Experian Health’s 2025 survey of 250 healthcare professionals, 41% of providers reported denial rates of 10% or higher.

So before you compare your practice to anything, check that the benchmark counts what you count. We wrote separately about what a good claim denial rate actually looks like for a physician practice.

What actually causes claim denials?

We audited 61,400 claims to find out. Eligibility and coverage errors caused 24% of denials. Coding and modifier errors caused 21%. Missing or invalid prior authorization caused 17%. Duplicate claim denials made up 9% of denials, mostly from resubmitting instead of correcting. Timely filing caused 6% of denials.

Two things stand out. The first is how much of this is front-office work, not coding work: eligibility sits at the top of the list, and it is checked before the patient is seen. The second is concentration. The top three denial reasons accounted for 58% of denied dollars in the average practice we reviewed, which means a practice does not need a general denial program. It needs to fix three things.

The front-end fix is eligibility and prior authorization verification before the visit. The coding fix is a scrubber plus a certified coder reviewing the edits, which is what medical coding teams are for. Neither is exotic. Both are mostly a question of whether anyone is assigned to do it.

How many denials never get worked a second time?

This is the number that should worry people, and it is almost never published. In our billing reviews, 19% of denied claims were never reworked or appealed. Not denied and lost on the merits. Never touched again.

The reason is usually structural rather than lazy. 42% of practice managers said nobody owns denial follow-up full time, and 63% could not name their top three denial reasons. If nobody owns the queue and nobody knows what is in it, the queue ages out.

The public data says the same thing from the patient side. Of roughly 85 million denied in-network marketplace claims in 2024, consumers appealed fewer than 1%, and insurers overturned 34% of the appeals that were filed. In Medicare Advantage the gap is starker still: only 11.5% of denied prior authorization requests were appealed in 2024, and 80.7% of those appealed were overturned. Four out of five contested denials were wrong, and nearly nine in ten were never contested.

Federal auditors found the same pattern. Reviewing a one-week sample of 500 denials from June 2019, the HHS Office of Inspector General found 13% of Medicare Advantage prior authorization denials met Medicare coverage rules and would likely have been approved under Original Medicare. That sample is old and narrow, and it is still the most direct audit of denial correctness in the public record.

Our own appeals were overturned 68% of the time, with a median turnaround of 34 days from filing to payer decision. An overturn rate that high is not a sign of unusual skill. It is a sign that a lot of denials are administrative rather than clinical, and that they fall over when somebody pushes. That is the entire premise of denials and AR recovery work.

Is a denied claim worth appealing? Here is the math

Take a three-provider primary care practice collecting $90,000 a month across roughly 900 claims, so an average of $100 collected per claim. At a 10% denial rate that is 90 denials a month, a denied pool of about $9,000 a month or $108,000 a year.

Now apply the rework cost. CAQH measured the provider cost of a manual claim status inquiry at $13.80, against $3.64 when the same transaction is electronic. Add staff time to correct and resubmit and a realistic all-in rework cost sits above the transaction cost alone. At our 68% overturn rate, working all 90 denials recovers roughly $6,120 a month. Even at a generous $25 all-in cost per denial, the rework bill is $2,250. The work returns nearly three dollars for every dollar spent.

The arithmetic only fails on very small balances and on claims already past the payer filing window. Timely filing caused 6% of denials in our audit, and only 4% of those were recovered, because by then the argument is over. Everything else is worth touching. This is why the answer to a high denial rate is rarely a new clearinghouse and usually a named owner, which is the part of the revenue cycle that gets cut first when a practice is short-staffed.

Which medical billing statistics are simply wrong?

Three figures dominate this topic and none of them survives a check.

The first is that 80% of medical bills contain errors. It appears on most of the pages ranking for this subject, attributed variously to a trade publication, a software vendor and a statistics aggregator. None of those is a primary source. There is no published study, no sample size, no year and, most importantly, no definition of the word error. A coding error, a clerical typo, an overcharge and anything that triggers a denial are four different things with four different rates.

The second is that providers lose $125 billion a year to poor billing practices. Same pattern: widely repeated, attributed to two different organizations on different pages, traceable to no published analysis by either.

The third is that 12% of claims are submitted with inaccurate codes. This one has a real origin, the AMA National Health Insurer Report Card, but that program was discontinued more than a decade ago. It is being quoted in 2026 articles as a current figure.

There is also a widely cited claim that billing costs grew from $200 billion to $346 billion. It misreads its own source. The underlying analysis found that of roughly $346 billion in billing and insurance-related costs, about $200 billion was potentially reducible. One number became a before and after. Other pages now cite that misreading as a source.

The reason this matters beyond pedantry: these numbers get used to justify buying things. A vendor quoting an undated figure about an undefined error rate is not giving you a reason to act.

What changes when someone actually works the numbers?

Across 38 client practices, first-pass denial rate fell from 14.2% to 6.1% within 90 days of onboarding, clean claim rate rose from 89.6% to 97.3% in the first 90 days, and net collection rate rose from 91.4% to 97.8% over the first six months.

Median days in AR dropped from 54 to 33 within 120 days. That figure matters more than it looks, because every 10 days removed from AR released a median of $41,000 in cash for practices collecting $1.5M to $3M a year. Ageing is also where the recoverable money sits: the median practice had $118,000 in AR older than 120 days when we started, and 27% of total AR sat past 90 days in the average practice reviewed. We covered the benchmark question separately in what percentage of AR over 90 days is normal.

Old claims are not worthless, but they decay fast. We recovered 61% of the dollar value of claims aged 90 to 180 days that practices had stopped working. Past 180 days, recovery fell to 23% of dollar value. The difference between those two numbers is roughly the cost of waiting a quarter to start, whether the work is done in house or by a full-service billing team.

What quietly makes these numbers worse?

Underpayments, first. Underpayments against contracted rates appeared on 7.8% of paid claims, and the average underpaid claim was short by $38. Nobody notices, because the claim is marked paid. On 900 claims a month that is about $2,670 a month walking out on claims that already cleared.

Credentialing, second. Credentialing lapses delayed payment for 1 in 12 providers added in the prior year, and a lapsed re-credentialing held payments for a median of 47 days.

Reporting, third, and this is the one practices control most directly. Practices that reviewed AR ageing monthly carried 12 fewer days in AR. Meanwhile 44% of practice managers could not name the fee basis in their current billing contract, and 52% of practices that switched billing vendors cited missing denial reporting as the main reason. If you are comparing medical billing companies, the denial report they will send you every month tells you more than the rate they quote.

In-house or outsourced: what do the numbers actually compare?

The honest comparison is not price against price, because the two are priced differently. Fully loaded in-house billing cost 7.9% of collections for practices under $2M, across 96 practices that shared payroll data with us, while outsourced billing is typically priced at 3% to 6% of collections. The gap closes once volume rises, and the real difference is usually coverage rather than cost.

Neither option fixes a broken front desk. Eligibility errors caused 24% of denials in our audit and most of that work happens before a biller ever sees the claim. We broke down the staffing side of this in what it really costs to employ an in-house medical biller.

Dental

Dental denials cluster around documentation and plan rules rather than coding logic. Frequency limitation denials made up 19% of dental denials, and dental practices wrote off a median $23,400 a year in restorative claims denied for missing narratives or X-rays. The larger miss is on the medical side: medical cross-coding opportunities were missed in 64% of dental practices reviewed, and pre-treatment estimates were skipped on 38% of crowns and implants. A dental practice benchmarking itself against a general denial rate will look fine and still be leaving that money behind, which is roughly the pattern we saw in this dental practice case study.

Physical therapy

Therapy denials are unit and certification denials. 8-minute rule unit errors appeared on 9% of therapy claims across 4,300 claims audited. The KX modifier was missing on 21% of Medicare therapy claims past the threshold, and plan of care certification was unsigned past 30 days on 7% of Medicare episodes. Coverage changes are the quiet one: 33% of therapy episodes had a coverage change mid-episode that was not caught. None of those show up in a national denial-rate benchmark, which is why physical therapy billing is measured differently.

Behavioral health

Session time and plan routing drive the numbers here. 18% of 90837 claims had documented session time under 53 minutes, across 7,200 behavioral health claims. Claims sent to the medical plan instead of the behavioral health carve-out caused 12% of behavioral health denials, and telehealth place-of-service and modifier errors caused 15% of behavioral health telehealth denials. Access is the other constraint: new clinicians waited a median 96 days to go in-network with commercial payers, and solo therapists carried a median 41 days in AR against 29 for group practices.

Ambulance and EMS

Ambulance billing is dominated by one document. Physician Certification Statements were missing or unsigned on 18% of non-emergency transports. Origin and destination modifier errors appeared on 6% of ambulance claims, and mileage units were wrong on 4% of Medicaid transport claims. The ageing profile is worse than most specialties too: ambulance agencies carried 37% of AR past 90 days, against 27% for the average practice we reviewed. Working that backlog is what moved days in AR from 71 to 38 in this ambulance case study.

Primary care and internal medicine

Primary care has the cleanest AR of the specialties here, at a median 36 days, and the most uncaptured revenue. Chronic care management time went uncaptured for 58% of eligible patients. Problem-oriented visits billed with an annual wellness visit lacked modifier 25 on 12% of claims, and vaccine administration codes were missing alongside vaccine product codes on 5% of claims. 1 in 9 patient balance calls was about a preventive visit billed with a cost share, which is a patient-experience cost that never appears in a denial statistic.