Denial Rate

Denial rate is a core RCM health metric. A rising rate usually means more dollars stuck in AR, more rework, and higher write-offs.

Every behavioral health operator should know two things about denial rate: how it is calculated in your shop, and which denial reasons are driving it. Without those two pieces, month-over-month trends are easy to misread.

Denial vs rejection

A denial is a claim the payer has received and formally adjudicated, then decided not to pay in full. Denials appear on 835 remits with CARC and often RARC codes, such as CO-197 for authorization issues or CO-16 for missing information.

A rejection is a claim that never made it into the payer's adjudication system. Rejections usually come from your clearinghouse or the payer gateway for format or basic data problems, such as invalid member ID, missing NPI, or ICD version mismatch. Rejections typically arrive as 277CA or clearinghouse messages, not as 835 denials.

You should define whether your "denial rate" includes only true payer denials, or also clearinghouse rejections. Many teams track these separately: a claim-rejection rate for front-end scrub issues, and a payer-denial rate for back-end issues.

Count vs dollars

Denial rate can be claim-count based or dollar-based.

• Claim-count based: number of denied claims divided by total submitted claims. This is useful to measure operational friction for staff, since each denied claim usually means extra touches.
• Dollar-based: denied allowed amount or denied billed charges divided by total allowed or billed charges. This is useful to understand financial impact, since ten small denied claims are not equal to one large denied partial-hospitalization claim.

Both views are valid. The key is to label which version you are using on dashboards and reports. Many executives care about the dollar impact first, while billing leads care about claim counts to size staffing.

Break it down by reason

A single topline denial rate is not actionable by itself. The value comes from breaking it down in a way that maps to how you can fix the problem.

At a minimum, segment denial rate by:

• Payer or line of business (commercial, Medicaid, Medicare Advantage, EAP, carve-out behavioral health plan)
• Denial category, such as: eligibility, authorization, medical necessity, non-covered service, coding or modifiers, coordination of benefits, timely filing
• Service type or level of care, such as: residential, PHP, IOP, outpatient therapy, MAT, psychiatry

For example, you might see an 18 percent denial rate overall, but 35 percent for one Medicaid MCO on residential claims. When you break that 35 percent down by CARC, you discover most are CO-197 concurrent auth denials after day 21. That tells you exactly where to dig: authorization workflows and clinical documentation for continued-stay reviews.

A denial-rate report is only as useful as its grouping rules. Invest time in mapping CARC and RARC codes into clear operational buckets. This is where an AI denials agent that reads every 835 and applies consistent categories can save your analysts many hours and keep the groupings stable over time.

Beyond the main formula, a few variations are worth knowing. Each answers a slightly different question.

Imagine you submit 1,000 claims in June across all behavioral health programs.

By the end of August, your reporting shows:

• 1,000 total claims in the June submission cohort
• 180 of those claims received at least one denial at some point
• The 835s show that $90,000 of allowed amounts were initially denied out of $600,000 total allowed for June

If you use the cohort-based, claim-count formula:

Denial rate = (180 denied claims ÷ 1,000 total claims) × 100 = 18 percent.

If you look at the dollar-based view:

Dollar-based denial rate = ($90,000 denied ÷ $600,000 total allowed) × 100 = 15 percent.

When you drill into the 180 denied claims by CARC category, you find:

• 80 claims denied with CO-197 for authorization exceed days or units
• 50 claims denied with CO-16 for missing or invalid data
• 30 claims denied with CO-22 for coordination of benefits
• 20 claims denied for other reasons

Now the topline 18 percent tells a story. More than 40 percent of denied claims are auth related, which likely points to concurrent review breakdowns for residential and PHP, and 28 percent are basic data quality issues you could reduce with better eligibility and demographic checks up front.

• Counting clearinghouse rejections as denials without labeling them, which makes denial rate look worse but hides whether the issue lives in billing or in registration and eligibility.
• Calculating denial rate by dollars in one report and by claim counts in another, then comparing them as if they were the same metric, which leads leaders to chase false trends.
• Only looking at initial denials and ignoring how many are overturned on appeal, so a payer with aggressive but reversible CO-197 denials looks just as bad as a payer with hard, non-appealable medical-necessity denials.
• Failing to separate carve-out behavioral health plans from medical plans, so routing denials for plans like Optum BH or Magellan are buried inside generic commercial denial statistics.
• Mixing lines of business, such as Medicaid and commercial, into one denial rate, which hides that Medicaid MCOs may drive most of the denials for residential or IOP stays.

Behavioral health denial rates often run higher than primary care or surgical specialties. The mix of carve-out behavioral health plans, heavy prior-authorization rules, and long-stay services creates more opportunities for payers to deny part or all of a claim.

Two denial categories tend to dominate. First, routing denials when services that belong to a behavioral health carve-out get billed to the medical plan. For example, a residential SUD claim sent to the Blue Cross medical plan instead of the affiliated behavioral health administrator. These show up as eligibility or coverage denials, even if the member has full benefits. If you do not separate carve-out routing denials in your denial-rate reporting, the root cause looks like a generic eligibility problem instead of a payer-mapping issue that scheduling and billing can fix.

Second, authorization denials in the middle of long episodes. Residential, PHP, and IOP claims are often billed per diem, with authorization granted in small chunks. The first 10 or 14 days might pay clean, then days 15 to 28 deny with CO-197 because the concurrent auth was not secured or documented. Topline denial rate will rise sharply for those levels of care, even though the early days look fine. Break denial rate out by prior-authorization status and day-of-stay so you can see exactly where episodes fall off auth.

Behavioral health operators should also pay attention to denial rate by payer-program, such as Medicaid FFS, each Medicaid MCO, and each commercial plan's behavioral health administrator. A single MCO can account for most of your auth denials on one level of care. Targeted work with that payer and tightening internal concurrent-review workflows can move denial rate by several points, which usually translates into hundreds of thousands of dollars per year for larger programs.

AI can help with denial rate by handling the high-volume, detailed work of reading every 835, classifying CARC and RARC combinations, and updating denial-rate dashboards in near real time. Instead of analysts pulling ad hoc reports from the PM system, an AI agent can maintain consistent groupings for eligibility, authorization, COB, coding, and medical-necessity denials, and surface which payer or program is driving the highest rates.

Supabill's denials agent ingests remits across payers, maps each denial to stable categories, and tracks denial rate at the claim, line, and dollar level. The same shared rules engine can hold payer-specific quirks, such as how a Medicaid MCO codes concurrent-auth denials for PHP versus IOP, so you get cleaner behavioral-health-specific views. Humans still own the hard parts: deciding which denial patterns to prioritize, negotiating with payers, refining clinical documentation, and crafting appeals that match medical-policy language. AI can do the grunt work of counting, categorizing, and flagging anomalies so your teams spend time on fixes, not on spreadsheets.

Should denial rate include partial denials or only fully denied claims?

Denial rate should include both full and partial denials, but they need to be labeled clearly. A full denial means no lines on the claim were paid. A partial denial means at least one line or day was paid and at least one was denied. If you only count fully denied claims, you will miss a large share of residential, PHP, and IOP issues where part of the stay is denied for lack of authorization or medical necessity. A practical approach is to track: claim-count denial rate, line-level denial rate, and a dollar-based denial rate. That combination tells you how many claims are affected, how many units are at risk, and how much money is involved. Source

What is the difference between first-pass denial rate and overall denial rate?

First-pass denial rate looks only at the initial payer adjudication. It asks: out of claims the payer touched for the first time, how many were denied or partially denied on that first decision. Overall denial rate, sometimes called lifecycle denial rate, considers all denials those claims ever experienced until they are finally resolved or written off. For behavioral health, first-pass denial rate shows how well front-end processes like eligibility checks, benefits verification, and prior authorization are working. Overall denial rate shows how many claims keep getting stuck over time, including concurrent auth and medical-necessity issues that surface mid-episode. Source

How often should denial rate be reviewed for a behavioral health program?

Monthly is the minimum for leadership, but weekly review is typical at the billing-operations level, especially for high-risk programs like residential and PHP. Weekly reports allow you to spot spikes in authorization or routing denials before a whole month's worth of claims stack up. If your payers remit frequently via 835, an automated agent can update denial rate daily, and the team can set alerts for sudden jumps by payer, level of care, or denial category. Source

How should re-billed or corrected claims be handled in denial rate calculations?

The cleanest method is cohort based: tie denial rate to the original claim from a given service period, and treat all corrected claims and resubmissions as activity on that same claim. The cohort still counts as one claim that experienced a denial, even if you send three corrected claims. This avoids inflating both the numerator and denominator with resubmissions. Separately, you can report overturn rate or recovery rate, which measures how many denied dollars or claims were eventually paid after appeals. Source

How do coordination of benefits denials affect denial rate in behavioral health?

Coordination of benefits denials, such as CO-22 or CO-109, often inflate denial rate in behavioral health because clients frequently have overlapping coverage: commercial plus Medicaid, or primary medical plus a separate behavioral health plan. These are not usually clinical problems. They reflect missing or outdated insurance information, incorrect primary payer selection, or failure to update coverage during long treatment episodes. If you track COB denials separately in your denial-rate reporting, you can target front-desk, intake, and verification workflows to reduce them, instead of treating them like payer hostility or clinical denials. Source

• X12 — Claim Adjustment Reason Codes
• KFF — Claims denials research