Can I bill the patient for a code 18?

No. Code 18 carries a rule saying to use it only with group code OA, except where a state workers' compensation regulation requires CO. Neither of those puts money on the patient. A duplicate is a bookkeeping event, not a coverage decision, and the patient did nothing to cause it. If your line says PR-18, that is outside the code's own rule and worth a call.

Should the group code be CO or OA?

Both are defensible, which is unusual. The code list limits code 18 to group code OA, with one exception for state workers' compensation rules that require CO. The industry table of approved combinations disagrees with its own reprint of that note and approves CO on every entry for this code. Payers use OA, CO and PI in practice. The money answer is the same for all three: not the patient's.

What should I do first with a duplicate denial?

Find the original claim, not the denied one. Most duplicate denials mean the first claim was already processed, and a good share of them mean it already paid. If it paid, the work is posting the payment and closing the second claim. Sending the claim again is the one thing that guarantees another duplicate denial.

Is a duplicate denial worth appealing?

Usually not, and on Medicare whether you even may depends on the kind of claim. The manual says an exact duplicate denial on a physician or supplier claim may be appealed, that institutional exact duplicates are rejected with no appeal rights, and that DMEPOS exact duplicates may not be appealed. A suspect duplicate, meaning one a person reviewed rather than one the system matched outright, may be appealed by anyone. Even where the right exists, the useful work is checking whether the original paid.

RemitBench/Denial codes/CO-18

CO-18 denial code

Code 18 means the payer already has this line. It is not a decision about the service, and most of the time the first claim has already paid.

What it means: The payer matched this claim or this service line to one it already has. It stopped there rather than paying for the same care twice.
Who owes the money: Nobody owes it The code's own text says to use it only with OA, apart from one workers' compensation exception that calls for CO. Neither reaches the patient. A duplicate is bookkeeping, and nothing about it should ever land on a statement.
Is appealing worth it: Rarely, and on Medicare whether you even may depends on what kind of claim it was. Go and find the original claim first. That is where the answer is, and often the payment.

Do not send the claim again. This is the reflex that turns one wasted cycle into three. A third copy matches the same original and denies the same way. One Medicare contractor puts it as a rule: a new claim is only appropriate when there is no claim on file at all, or the claim was unprocessable. The same notice warns that rebilling claims already paid or denied without a valid reason can get a practice referred to a program integrity contractor.

Look at the original before you decide anything. A duplicate denial tells you a first claim exists. It does not tell you what happened to it. Paid, denied and still processing are three completely different afternoons.

What a duplicate means to a payer's computer

No human decided this. A payer's system compares each incoming line against lines it already holds, and it looks at a small set of fields.

Medicare publishes its list, and it is worth reading once. On a claim from a physician, practitioner or other supplier, an exact duplicate is a claim or line matching another one on every one of these:

The patient's Medicare identifier.
The provider number.
The from date of service.
The through date of service.
The type of service.
The procedure code.
The place of service.
The billed amount.

When all eight match, the line stops. That is the whole test. It cannot see that you did the procedure twice, that two dentists in your office each did one, or that the second one was on the other side of the mouth. It sees the same eight fields twice.

Hospital and other institutional claims use a slightly different list, with the type of bill and the total charges in place of some of the above. Commercial payers do not publish theirs, but they are built on the same idea.

Calling will not get this overturned. CMS says these edits are hard coded and that contractors cannot override or bypass them. The only thing that changes the answer is a claim that no longer matches, which means the fix is on your side of the wire.

This is why the code splits into two very different problems. Either you really did send the same thing twice, which needs no work at all, or the payer's test caught two separate services that happen to look alike. Those need opposite responses, and the code is identical.

Why the payer sent it

There are a lot of these. One Medicare contractor publishes a monthly table of its top denials for its Part B jurisdiction, and duplicates sit at the top of it, with about 79,500 in Ohio and 21,400 in Kentucky. Worth one caveat: in Kentucky alone, medical necessity denials are actually higher. So "the number one denial" is true of that contractor's combined numbers rather than everywhere.

The real duplicates

These are the majority, and they are the good news, because there is nothing to fix.

The claim was sent twice, usually because nobody could see a payment yet and someone resent it.
It went out on paper and electronically.
Your clearinghouse or your software sent a batch twice.
A corrected claim was sent as a plain new claim, so the payer read it as the same claim arriving again.
Both your office and a billing service worked the same account.
The claim was resent while the first one was still processing. Payers are often quiet for a fortnight, and a resend during that window is the most common cause on this list.

The false duplicates

These are the ones worth your time. The service really did happen more than once, and the claim did not say so clearly enough for the payer's test.

The same procedure was genuinely done twice on the same day.
A procedure was done on both sides, and the claim did not say which side.
A test was repeated the same day to check a result.
Two providers in the same group, with the same specialty, each saw the patient. Many payers treat one group and one specialty as one provider.
The patient came back later the same day for something unrelated.
The original claim was replaced or adjusted, and the replacement was not marked as one.

Exact duplicates and suspect duplicates

Medicare splits this into two kinds, and they behave completely differently. Knowing which one you have tells you whether a person ever looked at your claim.

Kind	What happens to it
Exact	Every field on the list matches. The system denies or rejects it automatically. No human sees it, and the contractor cannot switch the edit off.
Suspect	The fields are close but not identical. The claim is held and a person at the contractor reviews it and decides whether to pay. The criteria for these are not published.

Kind

What happens to it

Exact

Every field on the list matches. The system denies or rejects it automatically. No human sees it, and the contractor cannot switch the edit off.

Suspect

The fields are close but not identical. The claim is held and a person at the contractor reviews it and decides whether to pay. The criteria for these are not published.

Whether you can appeal depends on who you are

This is the part nobody tells you, and it is written plainly in Medicare's own manual. The same exact duplicate gets three different answers depending on the kind of claim.

Kind of claim	What Medicare does, and whether you may appeal
Physician	Denied, and the manual says such denials may be appealed. This covers practitioners and most other suppliers.
Institutional	Rejected, and the manual says these do not have appeal rights. A rejection was never a decision, so there is nothing to appeal.
DMEPOS	Denied, and the manual says these may not be appealed. Same word as the physician row, opposite answer.
Suspect	The manual says a duplicate denial based on anything other than the exact duplicate criteria may be appealed. So a suspect duplicate is appealable whoever you are.

Kind of claim

What Medicare does, and whether you may appeal

Physician

Denied, and the manual says such denials may be appealed. This covers practitioners and most other suppliers.

Institutional

Rejected, and the manual says these do not have appeal rights. A rejection was never a decision, so there is nothing to appeal.

DMEPOS

Denied, and the manual says these may not be appealed. Same word as the physician row, opposite answer.

Suspect

The manual says a duplicate denial based on anything other than the exact duplicate criteria may be appealed. So a suspect duplicate is appealable whoever you are.

Read that table again before you file anything. "Duplicate denials cannot be appealed" is repeated everywhere and it is not true across the board. On a physician claim you do have appeal rights. Whether using them is a good idea is a separate question, and usually the answer is still no, because the payer is right.

The one sentence that ties it together

Medicare's appeals chapter says it more usefully than the duplicate chapter does: duplicate items and services are not given appeal rights, unless what you are appealing is whether the service was in fact a duplicate.

That is the test to apply. "It is a duplicate, please pay it anyway" is not an appeal that exists. "This was not a duplicate, and here is why" is. Everything on this page is downstream of that distinction.

There is also a route that is not an appeal at all. CMS lists a duplicate denial the provider believes was wrongly identified as a duplicate among its examples of clerical error, and clerical errors are handled as reopenings rather than appeals. For a claim you can simply adjust or correct, CMS says that is the more efficient path and no reopening request is needed.

A remark code sometimes tells you all of this on the remittance itself. N111 reads "no appeal right except duplicate claim/service issue", which is the same rule written into three words.

Read the letters in front of the 18

This code is one of the few that names its own group code. The text attached to code 18 says to use it only with OA, except where a state workers' compensation regulation requires CO. That is a rule from the code list itself.

OA-18: Other adjustment. What the code's own text asks for. The amount goes to neither side, because a duplicate does not decide who pays for anything.
CO-18: Contractual obligation. Named in the code's own exception for state workers' compensation rules, and approved much more broadly than that by the industry combination table. Your side of the ledger, so the money answer does not change.
PR-18: Outside every rule. Nothing about a duplicate belongs on a patient statement. If you have a PR-18, call the payer rather than posting it.
PI-18: Outside both rules, and sent anyway by some dental and Medicaid payers. Still not the patient's.

Almost everyone searches for this code as CO-18, and there is a reason the standard's OA rule does not settle it. The industry table of approved combinations lists seven entries for code 18, and every one of them approves CO as well as OA, while reprinting the OA-only note in the next column. Two authorities, one file, opposite answers.

Payers land all over that gap. One Medicare contractor publishes this denial as CO-18. One state Medicaid program's crosswalk uses OA, CO and PI on different internal codes. One large dental carrier uses PI on one row and no group code at all on the other.

The useful conclusion is small but real. Do not treat an unexpected letter here as a payer error worth chasing. Every group code except PR is defensible on this code, and none of them changes what you do next.

What to do next

Find the original claim

Search the patient and the date of service in the payer's portal, not in your own system. Your system may not have the original at all if it came from somewhere else.
If you cannot find one: the payer may have matched against a claim from another office, or the match may be wrong. That is the one version of this code worth disputing, and you will need the claim number they matched to.
Find out what happened to it

Three answers, three different jobs. This is the branch that decides your whole afternoon.
It paid: you are done. Post the payment against the original and close the second claim. Nothing was lost except the cycle. It is still processing: also done. Wait. Resending again only adds another duplicate. It denied: the duplicate is a distraction. Work the first claim's denial code, which is the real problem, and ignore the 18.
Ask whether the service really happened twice

If the answer is yes, the claim needs to show that. The payer is not arguing with you. Its test simply could not tell two services apart.
Say which side, or how many. CMS names this use directly, pointing at the modifiers for the right and left side of the body as the way to show a line is not a duplicate of the one above it. Modifiers that mark a repeat of the same procedure on the same day do the same job. Pick the one that describes what was actually done. Sometimes there is no modifier that fits. CMS says as much: not every procedure code has one available. Then the claim goes to a person at the payer to review, and your documentation is what decides it. Check the units first. Sometimes two of something belongs on one line as two units rather than on two lines. Which one that payer wants is in its own billing guide.
Use the corrected claim route, not a new claim

If you need to change something on a claim the payer already has, there is a specific way to send it that tells the payer this replaces the earlier one. Sending a plain claim instead is the single most common way to create a duplicate denial.
Reference the original claim number. Every payer's corrected claim process needs it, and a corrected claim without it is just another copy. Medicare's manual is blunt about the consequence: a claim that duplicates one previously denied is treated as a brand new claim unless something on it shows it is a resubmission carrying extra information.
Stop the resends at the source

If duplicates keep arriving, the cause is usually a habit rather than a claim. Look at how long your staff waits before resending, and whether two people work the same accounts.
Check your clearinghouse reports too. A batch sent twice shows up there before it shows up as a wall of denials three weeks later.

18 against 97

These two get mixed up constantly, because both mean "we are not paying you again for this". They are different events, and the difference tells you what to do.

Code	What the payer is actually saying
18	The same line arrived twice. Nothing was judged. Go and look at the first one. The fix is usually nothing.
97	A different service on the claim already includes the payment for this one. Two different procedures, one payment. The fix is a modifier or a write-off. See the CO-97 page.

Code

What the payer is actually saying

The same line arrived twice. Nothing was judged. Go and look at the first one. The fix is usually nothing.

A different service on the claim already includes the payment for this one. Two different procedures, one payment. The fix is a modifier or a write-off. See the CO-97 page.

The quick test: with an 18, the other line is the same procedure code. With a 97, it is a different procedure code.

They also need different modifiers, and this is where people lose weeks

The two problems have two separate sets of modifiers, and using one set on the other problem does nothing. CMS says so directly in the correct coding manual: the repeat procedure modifiers do not get past a bundling edit.

Modifier	What it is for
76	The same procedure repeated by the same clinician. A duplicate answer.
77	The same procedure repeated by a different clinician. A duplicate answer.
91	A clinical laboratory test genuinely repeated. Not for rerunning a test to confirm a result, and not for problems with the specimen or the equipment. CMS says misusing it points at abuse.
RT and LT	Right and left side. CMS publishes an article on these specifically to avoid duplicate denials. Put them on separate lines: both on one line is rejected as incorrect coding.
50	Both sides. CMS wants one claim line with one unit. Two lines, or two units with side modifiers, is denied as incorrect coding.
59 and X series	A bundling answer, not a duplicate answer. These belong on a 97. They will not clear a duplicate edit.
78 and 79	Neither. These are for the global surgery period. Plenty of articles list them as a duplicate fix, and no Medicare source does.

Modifier

What it is for

The same procedure repeated by the same clinician. A duplicate answer.

The same procedure repeated by a different clinician. A duplicate answer.

A clinical laboratory test genuinely repeated. Not for rerunning a test to confirm a result, and not for problems with the specimen or the equipment. CMS says misusing it points at abuse.

RT and LT

Right and left side. CMS publishes an article on these specifically to avoid duplicate denials. Put them on separate lines: both on one line is rejected as incorrect coding.

Both sides. CMS wants one claim line with one unit. Two lines, or two units with side modifiers, is denied as incorrect coding.

59 and X series

A bundling answer, not a duplicate answer. These belong on a 97. They will not clear a duplicate edit.

78 and 79

Neither. These are for the global surgery period. Plenty of articles list them as a duplicate fix, and no Medicare source does.

Other codes that look like a duplicate

Code	How it differs from 18
B13	Says the payment may have been made in an earlier payment. Softer than 18, and it points at a different remittance rather than a different claim. It shares remark codes with 18, so the remark will not tell the two apart.
B20	The service was partly or fully carried out by another provider. The closest real neighbor to 18, and the answer when two offices billed for the same work.
M86	A remark, not a reason code. The plan already paid for the same or a similar procedure inside a set time window. That is a frequency rule, not a duplicate, and days or months separate the two services.
M80	A remark saying the service is not covered when done in the same session as something already processed. It normally rides on code 97, not on 18.
N522	A remark for a duplicate of a claim being handled as a crossover between payers. Worth knowing, because it means the second claim was never yours to send.
151	You billed more of something than the payer will pay for. It counted, and the count was too high. A duplicate was never the question.
45	Priced below your charge. Money was paid and reduced, rather than stopped. See the CO-45 page.
16	The claim could not be processed as sent. Sometimes a claim you resent after a 16 comes back as an 18, because the payer kept the first one. See the CO-16 page.

Code

How it differs from 18

B13

Says the payment may have been made in an earlier payment. Softer than 18, and it points at a different remittance rather than a different claim. It shares remark codes with 18, so the remark will not tell the two apart.

B20

The service was partly or fully carried out by another provider. The closest real neighbor to 18, and the answer when two offices billed for the same work.

M86

A remark, not a reason code. The plan already paid for the same or a similar procedure inside a set time window. That is a frequency rule, not a duplicate, and days or months separate the two services.

M80

A remark saying the service is not covered when done in the same session as something already processed. It normally rides on code 97, not on 18.

N522

A remark for a duplicate of a claim being handled as a crossover between payers. Worth knowing, because it means the second claim was never yours to send.

151

You billed more of something than the payer will pay for. It counted, and the count was too high. A duplicate was never the question.

Priced below your charge. Money was paid and reduced, rather than stopped. See the CO-45 page.

The claim could not be processed as sent. Sometimes a claim you resent after a 16 comes back as an 18, because the payer kept the first one. See the CO-16 page.

When not to appeal

The original claim paid. There is nothing to win. Post it and close the second claim.
The original is still processing. An appeal on a claim nobody has decided yet goes nowhere.
You genuinely sent the claim twice. That is not a dispute, it is a housekeeping note.
The first claim denied for another reason. Appeal that reason, on that claim. The 18 is noise.
The service really happened twice and the claim did not say so. A corrected claim with the right coding is the route, and it is faster.
It is an institutional or DMEPOS claim rejected as an exact duplicate. Medicare's manual says there are no appeal rights on those at all.

Two versions are worth pushing on. If the payer matched your claim to something that is not yours, or to a service on a different date, that is a wrong match rather than a duplicate: get the claim number they matched to and start there. And if the denial came from a suspect duplicate review rather than an exact match, a person made a judgment, which means your record can change it.

Does 18 show up in dental?

Barely, and that is the finding that matters. Dental plans deny duplicates constantly. They almost never call them 18.

One large dental carrier publishes a document mapping its own processing policies to the standard codes, running to roughly 870 policy rows. Two of them use code 18. One carries remark N111 and no group code at all. The other carries N522, about a duplicate of a crossover claim, with group code PI.

Where the dental duplicates actually go

Into other codes entirely, and both of them mean something else.

Code used instead	Which policies land there
119	Six policies whose wording is plainly about a duplicate are filed under the benefit maximum code. See the CO-119 page.
96	Policies named "duplicate claim received", "duplicate exam found" and similar go under the not-covered code, mostly with remark M86 or N130. See the CO-96 page.

Code used instead

Which policies land there

119

Six policies whose wording is plainly about a duplicate are filed under the benefit maximum code. See the CO-119 page.

Policies named "duplicate claim received", "duplicate exam found" and similar go under the not-covered code, mostly with remark M86 or N130. See the CO-96 page.

So do not look for an 18 to find your duplicates. On a dental remittance the code will usually be 96 or 119, and the only thing telling you it was a duplicate is the sentence the payer printed next to it. Reading that sentence is the whole skill on dental remittances.

A state Medicaid crosswalk shows the same split

One state program maps its plain "exact duplicate of a previously paid claim" message to code 18 as you would expect. Its dental specific edits do not: multiple exams go to one code, fragmented visits to 97, and repeated cleanings to 119.

That is the pattern across both documents. A duplicate at the claim level gets code 18. A duplicate at the level of dental treatment gets coded as a benefit rule, because to the plan that is what it is.

Related codes

Back to all denial codes

RemitBench reads dental EOB PDFs and turns them into posting-ready lines, in your browser, with no upload. It does not read medical remittances today.

Open the app

Opens your mail app with the message written. Nothing is sent from this page.