The Most Common Medical Billing Denial Codes (CARC/RARC) and How to Fix Them

By Clayton Severson, Principal Consultant · Published July 6, 2026

The denial codes you will see most are CO-16 (missing or invalid information), CO-197 (no prior authorization), CO-11 (diagnosis inconsistent with the procedure), CO-50 (not medically necessary), and CO-29 (timely filing expired). Most are preventable before the claim ever goes out — the durable fix is upstream at registration, eligibility, and coding, not in the appeal.

What are CARC and RARC codes?

When a payer pays a claim for less than the billed amount — or nothing at all — it explains the adjustment with standardized codes on the electronic remittance (the 835) and the explanation of benefits. Two code sets do the work, and they travel together.

The codes are maintained nationally and updated a few times a year, so the exact list shifts at the edges — but the denials that drive the bulk of lost revenue are remarkably stable from payer to payer.

What are the most common denial codes?

The table below lists the denial codes that recur across specialties and payers, the root cause behind each, and where in the revenue cycle it actually gets fixed. All are CO (contractual obligation) group codes — the denials a practice can act on.

Code	What it means	Typical root cause	Prevent / fix at
CO-16	Claim lacks information or has a submission error	Missing or invalid field — member ID, NPI, diagnosis (a RARC names which)	Registration & claim scrubber
CO-197	Precertification / authorization absent	Service performed without the required prior auth	Pre-service authorization
CO-11	Diagnosis is inconsistent with the procedure	Dx-to-CPT linkage error or wrong code	Coding
CO-50	Not deemed a medical necessity by the payer	Diagnosis does not support the service under payer policy	Coding & documentation
CO-29	The time limit for filing has expired	Claim submitted past the payer's filing deadline	Billing workflow & A/R follow-up
CO-18	Exact duplicate claim or service	Resubmission or a system double-send	Billing system / clearinghouse
CO-97	Benefit is included in another, already-adjudicated service	Bundling / NCCI edit / missing modifier	Coding
CO-22	May be covered by another payer (coordination of benefits)	Primary vs. secondary not verified	Eligibility & COB check
CO-109	Not covered by this payer — send to the correct one	Claim routed to the wrong payer or plan	Eligibility & registration
CO-27	Expenses incurred after coverage terminated	Eligibility not re-checked at the date of service	Eligibility verification
CO-4	Procedure inconsistent with the modifier, or a required modifier is missing	Modifier error or omission	Coding
CO-B7	Provider not eligible / certified to be paid for this service on this date	Enrollment or credentialing gap with the payer	Provider enrollment & credentialing

Two codes people often file under "denials" do not belong here. CO-45 (charge exceeds the fee schedule or contracted amount) is the single most frequent adjustment on most remittances, but it is a contractual write-off, not a denial — the difference between your billed charge and the negotiated rate. And PR-1, PR-2, and PR-3 (deductible, coinsurance, copay) are patient-responsibility balances, not payer denials. Chasing those as if they were denials is one of the quiet ways a denial team loses hours it does not have.

How do you read a denial code on an 835 or EOB?

Read it in three parts: group code, CARC, then RARC. A line that reads CO-16 with remark M76 means a contractual denial for missing information (CO-16), and the RARC (M76) specifies the culprit — a missing, incomplete, or invalid diagnosis. The CARC tells you the category; the RARC tells you exactly what to fix. A denial without a useful RARC is a signal to call the payer rather than guess, because resubmitting blind usually earns the same denial a second time.

Which denials are preventable?

Most of them. Look back at the "prevent / fix at" column: registration, eligibility, authorization, and coding account for nearly every high-volume denial in the table. Those are all front-end steps that happen before the claim is submitted. Genuinely clinical denials — a true medical-necessity dispute that needs a physician's documentation — are the minority. That is the single most useful thing to internalize about denials: the codes look like a back-office billing problem, but the causes are overwhelmingly upstream, at the front desk and in the coding queue.

How do you fix and prevent the most common denials?

Prevention and rework are two different disciplines, and a healthy revenue cycle runs both:

• Verify eligibility and benefits at scheduling — real-time eligibility catches terminated coverage (CO-27), wrong-payer routing (CO-109), and COB issues (CO-22) before the visit, not after the denial.
• Automate prior authorization — flag services that require auth at booking so CO-197 never happens; it is close to 100% preventable and rarely appealable after the fact.
• Scrub claims before submission — an edit engine that checks required fields, NPIs, modifiers, and NCCI conflicts kills most CO-16, CO-4, and CO-97 denials at the door.
• Track timely-filing clocks — a worklist sorted by days-to-deadline stops CO-29, which is the most infuriating denial because the money was earned and simply left on the table.
• Work denials by reason code, not by claim — a queue grouped by CARC lets you fix the pattern (and the upstream cause) instead of re-solving the same denial one claim at a time.
• Close the loop to the root cause — every worked denial should feed a trend so a recurring CO-197 from one payer becomes an authorization rule, not a permanent tax on your A/R.

That last two points are where denial management with AI earns its keep: classifying raw payer codes into workable buckets, routing each to the right owner, and surfacing the reason-code trends a human staring at one claim at a time will never see.

Where denial management actually breaks down

In the billing operations I have worked in, the denials themselves are rarely the hard part — the workflow around them is. Two failures recur. The first is that denials get worked reactively, one claim at a time, so the same CO-16 or CO-197 shows up every week and nobody ever fixes the upstream gap that produces it; the queue never shrinks because the faucet is still running. The second is classification: payers return the same underlying problem under a thicket of slightly different codes and free-text remarks, and without a crosswalk that collapses them into a handful of workable categories, the team cannot even see which denials are worth chasing. Fixing those two things — pattern over one-off, and a clean reason-code taxonomy over raw payer output — usually recovers more revenue than any individual appeal ever will.

Frequently asked questions