Chapter 21: NCCI Edits and Bundling: What Pays Together, What Doesn't, and Why Unbundling Is a Federal Case

"The edit file is not the payer being difficult. It is the payer having written down, in advance, which combinations it does not believe." — constructed

Overview

Chapter 5 introduced Account 31-2245 as the audit cautionary file. Chapter 17 §17.9 coded it and Case Study 1 followed the money — \$25,720.80 across 42 claims, eleven of them probably defensible and unprovable.

This chapter explains the file that would have caught it before the first claim went out.

The National Correct Coding Initiative is a set of edits, published by CMS, free, updated quarterly, that says which code combinations are and are not payable together and — crucially — which of those can be overridden with a modifier and which cannot.

Three facts about it are worth having before you read further.

It is public. You can look up any code pair, right now, without permission and without a subscription, and know before you bill what will happen. Account 31-2245 is eighteen months of nobody doing that.

It is not arbitrary. Every edit rests on a stated policy, and the policies are published in a manual that runs to several hundred pages and that almost nobody in the field has opened. §21.5 is about reading it before you override, and it is the most useful habit in this chapter.

And it is enforceable in a way most coding rules are not. Deliberate, systematic unbundling is a False Claims Act theory — Chapter 5 §5.3's territory — with the full apparatus of treble damages, per-claim penalties, and whistleblower suits behind it. §21.10 is not a scare section. It is the reason this chapter is starred.

In this chapter, you will learn to:

• Say why an edit file exists
• Read a procedure-to-procedure edit
• State exactly what indicators 0, 1, and 9 permit
• Tell the three MUE indicators apart, and know which cannot be appealed
• Find the Policy Manual chapter governing your claim
• Apply the standards of medical and surgical practice
• Decide whether an edit may be overridden
• Explain how unbundling becomes a federal case
• Tell an NCCI edit from a payer's own

21.1 Why an edit file exists

Start with the problem the file solves.

CPT contains codes that overlap. Not by accident — because procedures genuinely contain other procedures. A surgical approach includes an incision. An endoscopic repair includes getting the scope there. A comprehensive study includes its components.

Nothing in CPT itself prevents reporting the parts and the whole. Chapter 13 §13.5's parentheticals catch some of it. The "(separate procedure)" designation catches more. But a code set describing services cannot, by itself, enumerate every combination that would double-count.

Somebody has to write down which combinations are not payable together. That is the edit file.

Three things follow.

The edits encode policy, not preference. Each one rests on a stated rationale — the standards of medical and surgical practice, an anatomic relationship, a CPT instruction, or a coding convention. §21.6 covers the largest of these.

They are published in advance. This is genuinely unusual. A payer telling you, before you bill, exactly which combinations it will not pay, and why, is a form of fairness that the billing world does not extend in most other areas.

And they are checkable by you. The files are free and downloadable. A practice that is surprised by an NCCI denial was not surprised — it just had not looked.

⚠️ Where Claims Die

The most expensive NCCI failure is not a denial. It is a payment.

An edit that fires produces a denial, which enters a work queue, which someone works. Annoying, visible, and self-correcting.

An edit that is overridden by an automatic modifier produces a payment — and the claim is now an overpayment carrying an assertion nobody evaluated. Chapter 17's Account 31-2245 is exactly this: the macro made the denials stop, and the denials stopping is what made it invisible.

Which produces a genuinely counterintuitive operational rule: a rising NCCI denial rate may be a department working correctly, and a denial rate that falls to nothing after a configuration change is a finding, not a success.

21.2 Procedure-to-procedure edits: Column One and Column Two

The core of the system. Every procedure-to-procedure edit is a pair of codes with a direction.

   COLUMN ONE  /  COLUMN TWO

     Column One  ....  the code that IS payable
     Column Two  ....  the code that is NOT payable
                       when reported with Column One
                       for the same patient, same date,
                       same provider

Two things about the direction, and both matter.

The direction is fixed. An edit pairing A (Column One) with B (Column Two) does not also mean B blocks A. If you report both, the payer pays A and denies B, regardless of which you listed first, which is why Chapter 18 §18.8's line-sequencing rule and this rule are different rules solving different problems.

And Column One is generally the more comprehensive or higher-valued service — the one that contains the other. Generally, not always: some edits exist for reasons other than containment, and §21.7 covers those.

The file itself

CMS publishes separate PTP edit files for practitioner services and for outpatient hospital services, and the edits differ between them — the same code pair can be edited in one setting and not the other. Use the file for your setting.

Each row carries: the Column One code, the Column Two code, the effective date, the deletion date if the edit has been retired, and — the field everything in §21.3 turns on — the modifier indicator.

🔢 Code It

Reading one row.

text Column One Column Two Effective Deleted Modifier 29827 29822 (date) ---- 1

What it says, in words: "29822 is not separately payable when reported with 29827 for the same patient on the same date by the same provider. This edit is active. A modifier MAY override it if the criteria are met."

What it does not say: that a modifier is appropriate here. That indicator is a permission, not an authorization — §21.3 — and Account 31-2245 is what happens when a practice reads it as the second thing.

And notice what is not in the row. There is no explanation. The reason is in the Policy Manual, in the chapter covering the musculoskeletal system, and §21.5 is about going and reading it.

Actually looking one up

This takes about two minutes and almost nobody in a billing office has done it, so it is worth walking through concretely rather than describing.

   1. Go to CMS's NCCI page. It is public; there is no login.

   2. Choose the right file:
        · PRACTITIONER PTP edits   → professional claims
        · OUTPATIENT HOSPITAL PTP  → facility claims
      They differ. Using the wrong one gives a wrong answer
      confidently.

   3. Download the current quarter. The files are large and
      they are ordinary spreadsheets.

   4. Search for your Column One code. The rows under it are
      every code that is not payable with it.

   5. Read the MODIFIER INDICATOR column.

   6. Check the DELETION DATE. A row with a deletion date in
      the past is history, not a live edit.

Two things about doing this that are worth knowing before you try.

The files are big. Hundreds of thousands of rows. This is a spreadsheet-filtering task, not a reading task, and anyone who can filter a column can do it.

And the edit exists at the CODE PAIR level, not the claim level. You are asking "can these two codes ever be reported together" — not "is my claim right." The file answers the first question. Steps 2 through 4 of §21.8 answer the second.

A practice that downloads the file once and searches its own top twenty code pairs against it has done, in an afternoon, more edit research than most practices do in a decade.

21.3 The modifier indicator: 0, 1, and 9

One character, and it decides everything.

Indicator 0 is absolute

This is the one billing offices waste the most time on.

An indicator of 0 means the services are considered to be so integral to one another that no circumstance separates them. A claim reporting both with modifier 59 will deny, and the appeal will also deny, and the second denial is not a failure of the appeal — there is nothing to appeal to.

This is §20.10's four-way triage in a specific case: an indicator-0 denial is "bundled." Remove the charge. Nothing to appeal.

Indicator 1 is a permission, not an authorization

An indicator of 1 says "a modifier may override this edit." It does not say "this edit should be overridden."

Read that twice, because the entire compliance content of this chapter is in the gap between those two sentences.

The indicator tells you the system will accept a modifier. Whether the modifier is appropriate is a clinical and documentary question, answered by reading the operative note — Chapter 14 §14.5's discipline, one more time:

Before appending modifier 59, find the sentence in the operative note.

Account 31-2245's edit carried an indicator of 1. The macro read that as permission to append. Eleven of forty-two claims were probably justified and the other thirty-one were not, and the indicator was identical on all forty-two.

⚖️ Compliance Check

A modifier indicator of 1, applied at a rate approaching 100%, is a self-report.

Chapter 17's Case Study 1 made this point about modifier frequency. Here is the mechanism underneath it.

Indicator 1 exists because a minority of cases genuinely separate. That is what "may override" means — the edit is right most of the time and wrong sometimes, and the modifier is for the sometimes.

A practice overriding an indicator-1 edit on essentially every occurrence has asserted that the minority case is its normal case. That claim may be true — some practices genuinely do unusual work — and it is a claim, visible in claims data, that the practice has made without knowing it made it.

The measurement is one query: for each Column One / Column Two pair you override, what percentage of the time do you override it? Chapter 17's Case Study 1 gave the threshold: above 90% is a finding.

21.4 Medically unlikely edits and the MAI

PTP edits are about combinations. MUEs are about quantity.

A medically unlikely edit is the maximum units of service a provider would report for a single code, for a single beneficiary, on a single date of service, under most circumstances.

Chapter 20 §20.3 said too-many-units is what MUEs are built to catch. This is that.

MUEs are based on anatomic considerations (you have two kidneys), code descriptors (a code reading "bilateral" cannot be reported twice), CMS policy, nature of the service, and claims data analysis.

The MUE Adjudication Indicator

And here is the field almost nobody knows exists, which decides whether an MUE denial is worth working at all.

MAI 2 IS THE ONE PEOPLE WASTE MONTHS ON.

An MAI of 2 means the units reported are impossible or forbidden as a matter of policy — you cannot remove three kidneys, and a code whose descriptor says "bilateral" cannot describe four sides.

These denials are not appealable in any meaningful sense. The correct response is to find out why your claim reported those units, because the answer is almost always a units error of the kind Chapter 20 §20.3 describes, or a duplicate.

Three practical rules.

Look up the MAI before working an MUE denial. One field decides whether you are writing an appeal or fixing a charge.

MAI 1 denials are frequently a legitimate reporting problem — the units were real and needed to be split across lines with modifiers.

And MAI 3 denials are the only ones where the medical record is the answer, which makes them the only ones worth building an appeal packet for.

The MUE values you cannot see

Most MUE values are published, in the same free downloadable form as the PTP edits, with the MAI alongside.

Some are not. CMS has withheld certain MUE values from publication, on the reasoning that publishing a maximum invites billing up to it.

Two consequences, and they pull in opposite directions.

You cannot always check in advance. For an unpublished MUE, the first indication that a value exists is a denial. That is genuinely frustrating and there is no workaround.

And it does not change what you should do. Report the units actually performed. An MUE is a detection threshold, not a target — and the reason some values are unpublished is precisely that somebody, somewhere, treated a published one as a target.

The correct posture toward an unpublished MUE is the correct posture toward a published one: it should never be the constraint that decides your units. The dose administered decides the units (Chapter 20 §20.3). The MUE tells you when somebody will look.

🧮 Run the Numbers

A claim reports 8 units of a code whose MUE is 3.

The MAI decides everything:

```text MAI 1 → the MUE applies PER LINE. 3 units on line 1, and the remaining 5 on additional lines with appropriate modifiers IF clinically justified and documented. ► A reporting problem with a reporting solution.

MAI 2 → 3 is an absolute maximum. 8 is not possible. ► DO NOT APPEAL. Find out why your system said 8.

MAI 3 → 3 is a clinical benchmark. 8 may be legitimate and unusual. ► Appeal, with documentation. This is the case where the record can change the answer. ```

Same denial, same units, three completely different correct responses — and a billing office that does not look up the MAI will treat all three the same way and be wrong twice.

21.5 The NCCI Policy Manual, and reading it before you override

The edit files say what. The Policy Manual says why.

The National Correct Coding Initiative Policy Manual for Medicare Services is published by CMS, free, updated annually, and organized into a general correct coding policies chapter followed by chapters corresponding to the CPT sections.

It is several hundred pages, and you are not going to read it straight through. Nobody does.

What you do is read the chapter covering the claim in front of you — and it is genuinely readable, written in reasonably plain language, and organized so you can find the relevant discussion in a couple of minutes.

What is in it

The general chapter covers the principles that run through everything: the standards of medical and surgical practice, medically unlikely edits, modifier use, the treatment of anesthesia, and the "separate procedure" designation.

The section chapters cover the specific policies for their code ranges — and this is where you find sentences like "debridement is not separately reportable when performed in the same anatomic region as a repair," which is the sentence Account 31-2245 needed.

The habit

Before you override an edit, read the Policy Manual section that governs it.

Three reasons this is the highest-value habit in the chapter:

It tells you what the edit is protecting, which tells you what would genuinely separate the services. You cannot document your way past an edit if you do not know what it is about.

It frequently tells you the answer directly. A great many "can I bill these together" questions are answered in a paragraph, and the paragraph is more authoritative than any consultant.

And it is what a reviewer will read. An appeal that quotes the Policy Manual and explains why this case falls outside it is arguing on the reviewer's own ground. An appeal that asserts the services were distinct is arguing on nothing.

21.6 Standards of medical and surgical practice

The largest single principle behind the edits, and it is worth stating in the manual's own shape.

Services integral to the performance of a procedure are included in that procedure and are not separately reportable.

"Integral" means the service is a necessary or usual component — something that must happen, or ordinarily happens, for the primary procedure to be performed.

Examples of what is integral to essentially any procedure:

   · cleansing, shaving, prepping the skin
   · draping and positioning the patient
   · inserting intravenous access for medication
   · sedation, where not separately reportable
   · local, topical, or regional anesthesia by the operating physician
   · surgical approach and exposure
   · surgical cultures
   · wound irrigation
   · surgical closure and dressings
   · application of splints or immobilizers at the operative site

Every item on that list has been billed separately by somebody, and the list exists because they did.

Two extensions worth carrying:

A more extensive procedure includes a less extensive one at the same site. If a lesser procedure is attempted and a greater one performed, report the greater. Chapter 17 §17.1 said the same about the surgical package.

And a complication that does not require a return to the operating room is generally not separately reportable during the postoperative period — Chapter 17 §17.2's global period, from the edit side.

Four more principles from the general chapter

Worth knowing because each one settles a category of question:

Sequential procedures. When a physician attempts a procedure by one approach, abandons it, and completes it by another, only the completed procedure is reported. Chapter 18 §18.5's laparoscopic-to-open conversion is the specific case; this is the general rule it comes from.

Anesthesia by the operating physician. Local, regional, or moderate sedation provided by the physician performing the procedure is generally not separately reportable, subject to the specific rules for moderate sedation (Chapter 18 §18.11). This is why Account 10-4471's lidocaine has been off the claim for three chapters running.

"Medically unlikely" as a coding concept, not just an edit. Some code combinations are edited not because one includes the other but because reporting both describes something implausible — which is §21.7's mutually exclusive logic, and which the general chapter states as a principle before the edits implement it.

And the "separate procedure" designation. Chapter 13 §13.8 defined it and Chapter 17 §17.9 applied it. The Policy Manual explains its relationship to the edits directly: a code so designated is not reported when it is a component of a larger procedure at the same session, and a great many PTP edits simply implement that designation. When you see an edit involving a "(separate procedure)" code, the designation is the reason and you have already read it.

Notice how much of this chapter is a restatement of things Part III already told you. That is not redundancy — the edits are Part III's rules with an enforcement mechanism attached, and a coder who understood the surgical package, the add-on rules, and the "separate procedure" designation already knows most of what the edit file says.

🎓 Exam Watch

The standards-of-practice list is a rich source of exam items, and they all look the same: a scenario describing a procedure plus something on the list, asking how many codes.

The answer is almost always one.

Three that recur:

• Local anesthesia by the operating physician — included. (Account 10-4471's lidocaine, Chapter 17 §17.2.)
• Surgical approach — included. An exploratory step that becomes the approach to a definitive procedure is not separately reported.
• Wound irrigation and closure — included, unless the closure rises to a separately reportable repair. (Chapter 17 §17.5: simple is included; intermediate and complex are not.)

21.7 Mutually exclusive and comprehensive/component logic

Two distinct rationales sit behind PTP edits, and knowing which one you are looking at changes what argument is even available.

Comprehensive / component

One code contains the other. The comprehensive code describes a service that includes the component code's service.

This is the majority of edits and it is the one §21.6's principle produces. The argument that sometimes works is that the component service was performed at a different session, site, lesion, or encounter — genuinely separate rather than part of the comprehensive service. That is exactly what modifier 59 and the X{EPSU} modifiers assert (Chapter 14 §14.5), and it is why they exist.

Mutually exclusive

The two procedures could not reasonably be performed together, or represent two different ways of doing the same thing.

An open and a closed approach to the same fracture. Two different methods of the same repair. An initial and a subsequent service of a type that admits only one.

The argument that sometimes works is narrower, because the objection is not "this is included" but "these cannot both have happened as reported." A modifier asserting distinctness does not answer that. What answers it is a different anatomic site or a different session — and if neither is true, one of the codes is wrong.

The practical difference: for a comprehensive/component edit, ask "was this separate?" For a mutually exclusive edit, ask "did both of these actually happen?" They are different questions and billing offices ask the first one for both.

(CMS consolidated its separately published mutually exclusive edit file into the PTP file some years ago. The logic did not go away; it is now carried within the PTP edits, and the Policy Manual still discusses both rationales.)

🔍 Check Your Understanding

Three denials. Same modifier indicator of 1. Which of them can a modifier fix?

A. A lesion excision and a simple repair at the same site, same session.

B. A lesion excision on the left forearm and an intermediate repair of a separate laceration on the right forearm, same session.

C. Closed treatment and open treatment of the same fracture, same session.

---

Answers:

A — no, and not because of documentation. Simple repair is included in excision (Chapter 17 §17.5). This is comprehensive/component, the component genuinely was part of the comprehensive service, and no circumstance separates them because there was no separate circumstance. Remove the line.

B — yes. Comprehensive/component, and the separating circumstance is real: a different lesion, a different site, and an intermediate rather than simple repair. Override with XS, and the note already names both forearms.

C — no, and for a completely different reason. This is mutually exclusive. The objection is not "one includes the other"; it is "these cannot both have happened." A distinctness modifier answers a question nobody asked. One of the two codes is wrong, and the fix is to determine which treatment was actually performed — Chapter 17 §17.7, and "closed treatment" means the physician did not open it.

The lesson: the modifier indicator was 1 in all three. It told you the system would accept a modifier. It told you nothing about whether one was appropriate, and in two of the three cases the right answer was to change the claim rather than to modify it.

21.8 Overriding an edit: what documentation must already exist

The chapter's operational core, and it is short because the rule is short.

   BEFORE YOU OVERRIDE AN EDIT

   1. Is the modifier indicator 1?
        0 → STOP. Nothing overrides this. Remove the charge.
        9 → the edit does not apply.
        1 → continue.

   2. What does the POLICY MANUAL say this edit protects?
        Read the section. Two minutes.

   3. Does the DOCUMENTATION — already written, already signed —
      establish the specific circumstance that separates them?
        Different session? Different site? Different lesion?
        Different incision? Different practitioner?
        NO → STOP. You do not have the modifier.
        YES → continue.

   4. Is there a MORE SPECIFIC modifier than 59?
        XE separate encounter · XS separate structure
        XP separate practitioner · XU unusual non-overlapping
        Use it.

Step 3 is the whole thing, and it has a tense.

"Already written, already signed."

Chapter 17's Case Study 1 is the reason for that tense. Eleven claims were probably justified and were repaid because the operative notes said "debridement performed" and nothing that could be written eighteen months later would make them say more.

THE DEFENSE IS CONTEMPORANEOUS DOCUMENTATION, AND IT CANNOT BE BUILT RETROACTIVELY.

Two corollaries most people miss.

A query is not a fix. Asking a surgeon eighteen months later whether the debridement was in a different region is asking them to remember. Chapter 38 §38.3's rules govern queries and this is outside them.

And an override that is correct still needs the note to say so — because the payment is not the end of the transaction. A correctly overridden edit that pays is still subject to review, and at review the only thing that exists is the record.

📋 Read the Chart

Source: operative report, arthroscopic shoulder procedure Encounter: outpatient surgery What it says:

"…the rotator cuff tear was identified at the supraspinatus insertion and repaired with two suture anchors as described above.

Attention was then turned to the subacromial space, anatomically distinct from the repair site. Extensive bursal debridement was performed for symptomatic bursitis, addressed as a separate problem from the cuff pathology."

What it means: this note supports the override and Account 31-2245's did not. The difference is one clause.

• It names a different anatomic region — the subacromial space, not the supraspinatus insertion
• It says so explicitly — "anatomically distinct from the repair site" — rather than leaving a reader to infer it
• It states a separate indication — symptomatic bursitis, not cleanup of the repair

What to do about it: override the edit — with XS, separate structure, not 59. Chapter 14 §14.5: use the specific modifier where one applies. XS says why; 59 says only "somehow distinct."

What NOT to do: treat this note as a template. It is a description of what happened. A surgeon who dictates this paragraph in a case where the debridement was at the repair site has not documented a distinct service; they have documented a false one, and the difference between a helpful template and a compliance disaster is exactly that.

Where it appears: in the eleven cases out of forty-two. This is the note the practice needed and did not have, and it takes four seconds longer to dictate than the one they had.

21.9 The shoulder claim, edit by edit

Account 31-2245, run through this chapter's machinery. Chapter 17 §17.9 coded it; this is what the edit file would have said.

The claim:

Line 2 — 29826

29826 is an add-on code, and add-on codes carry a ZZZ global indicator (Chapter 17 §17.2). It is reported with its primary procedure by design.

There is no edit to run. An add-on code and its primary are not a bundling question; they are a required pairing. Chapter 13 §13.7's rules govern, and the line is correct.

Line 3 — 29822 with 29827

Run the pair. Column One 29827. Column Two 29822. Modifier indicator 1.

Step 1: indicator is 1. A modifier may override. Continue.

Step 2 — read the Policy Manual. The musculoskeletal chapter addresses arthroscopic debridement directly: debridement in the same anatomic region as a repair is not separately reportable. Now you know what the edit protects — it is protecting against counting the cleanup of the repair site as a second procedure.

Step 3 — the documentation.

The operative note documents debridement of the same structures that were repaired.

STOP.

Not "there is no sentence supporting distinctness." Worse: there is a sentence establishing the opposite. The note affirmatively describes the thing the edit exists to prevent being billed twice.

Step 4 is never reached.

What the correct claim was

Two lines. And for the eleven claims where the debridement was genuinely in a different anatomic region, three lines — with XS rather than 59, because separate structure is what was true, and Chapter 14 §14.5 says to use the specific modifier where one applies.

🗂️ The Encounter — running the edit for 99214 + 20610

Account 10-4471's line 1 is 99214-25 and line 2 is 20610-RT. Run it.

The pair exists. An E/M service reported with a minor procedure on the same date is a procedure-to-procedure edit — Column One 20610, Column Two 99214 — and the modifier indicator is 1.

Step 1: indicator 1. A modifier may override. Continue.

Step 2 — the Policy Manual. The general chapter addresses this directly: an E/M service on the same date as a procedure is included unless it is significant and separately identifiable above and beyond the usual pre- and post-procedure work. That is modifier 25's test verbatim (Chapter 14 §14.4), and it is not a coincidence — the modifier and the edit are two halves of one policy.

Step 3 — the documentation. Figure 4.2 supplies four elements, three of which have nothing to do with the knee: three chronic conditions individually assessed with plans, prescription drug management, two tests ordered with stated reasons. The documentation establishes the circumstance. Continue.

Step 4 — a more specific modifier? No. The X{EPSU} modifiers apply to non-E/M services; modifier 25 is the specific modifier for this circumstance, and it is already on the line.

The override is correct. ✓

And the payer denied it anyway on day 17 — CO-97 with RARC N19.

That is the most important sentence in this chapter.

A correct override is not a guaranteed payment. The edit permitted the modifier, the documentation supported it, the modifier was right, and the payer applied its own edit policy and denied the line.

§21.11 is about why that is possible. And Chapter 30's appeal succeeds — on day 59, forty-two days later — because the documentation that justified the override in the first place is the same documentation that wins the appeal. You do not build it for the appeal. You build it for the claim.

The two overrides, side by side

This book now contains one override that fails and one that succeeds, and they differ at exactly one step.

Everything is the same until step 3. Same indicator, same permission, same structure. The difference is whether a sentence exists in a note that was signed before anybody thought about the edit.

And notice the last row of step 1 through 4 that differs before the outcome does: on one claim a macro appended the modifier and on the other a person read a note. That is the whole chapter.

21.10 Unbundling as a False Claims Act theory

This section is why the chapter is starred.

Chapter 5 §5.3 established the False Claims Act's definition of "knowingly": actual knowledge, deliberate ignorance, or reckless disregard for the truth or falsity of the information. No specific intent to defraud is required.

Systematic unbundling fits that definition with unusual comfort, and it is worth being precise about why.

The edits are public. A practice cannot claim it did not know a combination was not payable when the file stating so is free and downloadable. "We didn't know" is available as an argument only to somebody who did not look — and not looking, at volume, is what deliberate ignorance means.

The pattern is quantifiable. Unbundling produces a signature: a code pairing, a modifier, and a rate. It requires no clinical judgment to count and no chart to detect.

The claims are numerous. The False Claims Act's penalties attach per claim, and unbundling is by its nature repetitive — it is a configuration, applied to every occurrence. Forty-two claims is a small number in this context.

And whistleblower suits are the common origin. The Act's qui tam provisions permit a private party to bring an action on the government's behalf and to share in the recovery. A great many unbundling cases begin with a coder, a biller, or a former employee who noticed a macro.

⚖️ Compliance Check

What separates an error from a case is not the error. It is the pattern and the response.

A single unbundled claim is a coding error. It happens; it gets corrected.

The same unbundled combination on every claim for eighteen months is a configuration — and a configuration is a decision, made once, applied at volume, by an organization rather than by a person. That is the shape enforcement looks for.

And what turns a discovered configuration into something worse is the response. Chapter 5 §5.1's sixty-day rule attaches on identification. An organization that finds a macro like Account 31-2245's, understands it, and does not repay is no longer arguing about whether it knew. It knows.

The genuinely reassuring version of this section: the organizations that get into trouble are almost never the ones that made an error. They are the ones that had no mechanism for finding errors, and then had no mechanism for responding to one. Both mechanisms are cheap. Chapter 37 builds them.

What a coder should actually do with this

This section exists to be useful, not to frighten, and the useful content is short.

You are not the person who decides whether something is a False Claims Act problem. That is counsel's question and it should never be yours.

What you are is frequently the first person who can see it — because you are the one looking at the claims. Three things follow.

Report a pattern, not an instance. "This claim looks wrong" goes in a work queue. "This code pair carries modifier 59 on ninety-four percent of our claims" goes to compliance, and it is a materially different conversation.

Report it in writing, promptly, and keep a copy. Not defensively — because a dated record of when something was raised is protective for the organization as much as for you. Chapter 5 §5.1's sixty-day clock starts on identification, and identification has a date.

And do not investigate it yourself. Pulling thirty charts to see whether you are right is well-intentioned and it is the wrong move: it delays the report, it may complicate an investigation, and it is not your job. Raise it and let the people whose job it is do the rest.

The qui tam origin story people tell is a coder who found something and blew a whistle. The far more common story — and the one every organization would prefer — is a coder who found something, reported it internally, and it got fixed. Chapter 37 §37.8 is about making the second story the likely one, and the honest version is that it depends almost entirely on whether the first report was received well.

21.11 Payer edits that are not NCCI

The last thing you need, and the source of a great deal of confusion.

NCCI edits are Medicare's. Other payers have their own.

Three categories.

Payers that adopt NCCI. Many commercial payers and most state Medicaid programs use NCCI edits, in whole or in part, sometimes with a lag and sometimes with modifications. Medicaid has its own NCCI edit files, which are related to but not identical with the Medicare files.

Payers with proprietary edits. Large commercial payers license or build clinical editing software with its own rule sets — some derived from NCCI, some from specialty society guidance, some from the payer's own claims analysis. These edits are frequently not published, or are published only in summary.

And payer-specific medical policies that function as edits — a policy stating that a given combination requires records, or is not payable, or is payable only with specific documentation.

What this means operationally

An NCCI edit tells you what Medicare will do. It does not tell you what a commercial payer will do.

Account 10-4471 is the illustration: the modifier 25 override was correct under NCCI and the commercial payer denied it anyway, applying its own policy. Neither party was wrong — they were applying different rule sets, and the practice won on appeal because the documentation was good enough to satisfy the stricter one.

Which edit denials are worth appealing

A six-way triage, and it is Chapter 20 §20.10's four-reason framework applied to edits specifically.

   NCCI, indicator 0 ........... DO NOT APPEAL. Remove the charge.
   NCCI, indicator 1,
     documentation supports ..... APPEAL, quoting the Policy Manual
                                  section and the documented circumstance
   NCCI, indicator 1,
     documentation does not ..... DO NOT APPEAL. Correct the claim.
   MUE, MAI 2 .................. DO NOT APPEAL. Find the units error.
   MUE, MAI 1 or 3 ............. workable — split the lines, or appeal
                                  with documentation
   PROPRIETARY edit ............ GET THE POLICY FIRST, then decide

Three of the six say do not appeal, and a billing office that internalizes that will redirect a substantial amount of effort from letter-writing to charge correction — which is faster, cheaper, and correct.

Three habits:

Check NCCI first, because it is free and it catches the majority.

Then check the payer's published policy for your highest-volume combinations. It is finite work — most practices have a dozen combinations that matter.

And treat a denial from a payer that adopts NCCI differently from one that does not. The first can be researched against a public file. The second requires the payer's own policy, and the phone call that gets it is a better use of an hour than an appeal written blind.

📞 On the Phone

"This line is denying as incidental and I can't find an NCCI edit for it."

You probably cannot, because there probably is not one. This is a proprietary edit, and the call has one goal: find out what rule you are up against.

Ask, in this order:

"Is this an NCCI edit or a plan-specific edit?" A direct question with a direct answer, and it saves the rest of the call if the answer is NCCI.

"Which policy or edit rule is being applied?" Payers generally have a name or number for it, and a named rule can be looked up, requested, and quoted back.

"Is the policy published, and where?" Many are, in provider portals nobody reads. Get the link and keep it.

"Is the edit modifier-eligible, and which modifier?" Some proprietary edits accept overrides and some do not, exactly like NCCI's indicator.

What works: "I've checked the NCCI PTP file and I don't find this pair. Can you tell me whether this is a plan-specific edit, and if so which policy it's under and whether it's published?"

What does not work: "NCCI allows this." NCCI is not their rule set, and asserting it invites a conversation about a file that does not govern.

And write down what you learn, against the code pair. This is the second time in two chapters — Chapter 20's anesthesia rounding rule was the first — that the point of a call is to capture a payer-specific rule that will govern every future claim. The office that keeps that list stops making these calls.

Summary

The National Correct Coding Initiative is CMS's published, free, quarterly-updated statement of which code combinations are not payable together. A practice surprised by an NCCI denial had not looked.

Procedure-to-procedure edits pair Column One (payable) with Column Two (not payable with it). The direction is fixed. Separate files exist for practitioner and outpatient hospital settings.

THE MODIFIER INDICATOR

0 — no modifier will override, ever. Remove the charge; nothing to appeal. 1 — a modifier MAY override. A PERMISSION, NOT AN AUTHORIZATION. 9 — the edit does not apply.

An indicator-1 edit overridden at a rate approaching 100% is a self-report — it asserts that the minority case is your normal case.

Medically unlikely edits cap units. The MUE Adjudication Indicator decides whether a denial is workable: MAI 1 per line, splittable with modifiers · MAI 2 ABSOLUTE — do not appeal, find out why your system said that · MAI 3 a clinical benchmark, appealable with documentation.

The NCCI Policy Manual says WHY. Free, annual, a general chapter plus section chapters. Read the section that governs your claim before you override — it tells you what the edit protects, frequently answers the question outright, and is what a reviewer will read.

Most MUE values are published; some are deliberately not — because a published maximum invites billing up to it. An MUE is a detection threshold, not a target, and the dose or the anatomy decides the units.

Standards of medical and surgical practice: services integral to a procedure are included in it. Prepping, draping, positioning, access, local anesthesia by the operating physician, approach and exposure, irrigation, closure. A more extensive procedure includes a less extensive one at the same site. Sequential procedures — attempted one way, completed another — report only the completed one. And a great many edits simply implement the "(separate procedure)" designation.

Much of this chapter restates Part III with an enforcement mechanism attached. A coder who understood the surgical package, the add-on rules, and the "(separate procedure)" designation already knows most of what the edit file says.

Comprehensive/component edits ask "was this separate?" Mutually exclusive edits ask "did both of these actually happen?" — and a distinctness modifier does not answer the second.

Overriding: indicator 1 → read the Policy Manual → find the circumstance in documentation that is already written and already signed → use the most specific modifier. The defense is contemporaneous documentation and it cannot be built retroactively.

Account 31-2245's line 3 fails at step 3 — not for lack of a supporting sentence but because the note affirmatively documents debridement of the repaired structures. The correct claim was two lines, and three for the eleven genuinely distinct cases, with XS rather than 59.

Unbundling is a False Claims Act theory because the edits are public (so not looking is deliberate ignorance), the pattern is quantifiable from claims data, penalties attach per claim, and qui tam suits commonly originate with an employee who noticed a macro. What separates an error from a case is the pattern and the response.

And what a coder does about it is narrow and real: report a pattern rather than an instance, in writing, promptly, keeping a copy — and do not investigate it yourself. The decision is not yours; the observation frequently is.

NCCI is Medicare's. Other payers adopt it partially, or run proprietary edits that are not published, or apply medical policies that function as edits. A correct NCCI override is not a guaranteed payment — Account 10-4471's line 1 was overridden correctly and denied anyway on day 17.

And three of the six edit-denial categories say DO NOT APPEAL: indicator 0 · indicator 1 without supporting documentation · MAI 2. A billing office that internalizes that redirects real effort from letter-writing to charge correction.

And the Encounter's override is correct at every step, which is why Chapter 30's appeal wins on day 59 using the same documentation that justified the modifier in the first place. You do not build it for the appeal. You build it for the claim.

Key Terms

National Correct Coding Initiative (NCCI) — CMS's published set of edits stating which code combinations and unit quantities are payable. Free and updated quarterly. (Ch.21)

Procedure-to-procedure (PTP) edit — a directional pairing of two codes in which one is payable and the other is not when both are reported for the same patient, date, and provider. (Ch.21)

Column One — in a PTP edit, the code that is payable. (Ch.21)

Column Two — in a PTP edit, the code that is not separately payable when reported with Column One. (Ch.21)

Modifier indicator — the field stating whether a modifier may override a PTP edit: 0 never, 1 possibly, 9 not applicable. (Ch.21)

Medically unlikely edit (MUE) — the maximum units of service a provider would report for a code, for one beneficiary, on one date. (Ch.21)

MUE adjudication indicator (MAI) — the field stating how an MUE is applied and whether it can be exceeded: 1 a claim line edit, 2 an absolute policy-based date-of-service edit, 3 a clinical benchmark appealable with documentation. (Ch.21)

Comprehensive/component — the edit rationale under which one code's service includes the other's. (Ch.21)

Mutually exclusive — the edit rationale under which two procedures could not reasonably both have been performed, or represent two ways of doing the same thing. (Ch.21)

Standards of medical and surgical practice — the principle that services integral to a procedure are included in it and not separately reportable. (Ch.21)

Edit override — the use of a modifier to permit separate payment of a Column Two code, permissible only where the indicator allows it and the pre-existing documentation establishes the separating circumstance. (Ch.21)

NCCI Policy Manual — CMS's free annual publication stating the policies behind the edits, organized as a general chapter plus chapters by CPT section. (Ch.21)

Proprietary edit — a payer's own clinical editing rule, frequently unpublished, that is not an NCCI edit. (Ch.21)

Sequential procedure — a procedure attempted by one approach and completed by another; only the completed procedure is reported. (Ch.21)

Qui tam — the False Claims Act provision permitting a private party to bring an action on the government's behalf and share in the recovery. (Ch.21)

Spaced Review

1. Why does an edit file exist? Name the problem it solves that CPT cannot solve for itself.

2. State exactly what modifier indicators 0, 1, and 9 permit. Which one is a permission rather than an authorization, and what is the difference?

3. A claim reports 8 units of a code whose MUE is 3. Give the correct response under each of the three MAI values.

4. (Chapter 20) Which of the four reasons a valid code does not pay is an indicator-0 denial? What is the remedy?

5. Name six services that are integral to essentially any procedure and therefore not separately reportable.

6. Distinguish comprehensive/component from mutually exclusive. What question does each invite, and why does a distinctness modifier answer only one of them?

7. State the four steps before overriding an edit. Which step has a tense, and why?

8. (Chapter 17) Account 31-2245's line 3 fails at step 3. State precisely why — and explain why "there was no supporting sentence" is not the right answer.

9. Name the four features of unbundling that make it a comfortable False Claims Act theory.

10. Account 10-4471's modifier 25 override was correct at every step and the payer denied it anyway. Explain how both of those are true.

11. Describe, in six steps, how you would look up whether two codes are edited against each other. Which file do you choose, and what happens if you choose the wrong one?

12. Some MUE values are unpublished. State the reason, and say why it does not change what you report.

13. (Chapter 13) A code is marked "(separate procedure)." What is the relationship between that designation and the edit file?

14. You notice a code pair carrying modifier 59 on ninety-four percent of your organization's claims. State exactly what you do, in three parts — and state one thing you should not do.