CMS's July NCCI update adds 2,199 practitioner edit pairs. No modifier will bypass 424 of them.

The Q3 2026 NCCI edit files took effect on July 1. If your July denials are landing heavier than June's, this is the most likely reason, and the fix is not the one most billing teams reach for first.

CMS posted the files on June 1, 2026, with a July 1, 2026 effective date. The National Correct Coding Initiative updates its procedure-to-procedure (PTP) and Medically Unlikely Edit (MUE) tables every quarter, and CMS states plainly that it "does not provide a look-up service or a clean claims tool." Nobody sends you a summary. The change files are posted, they go live on the first of the quarter, and your clearinghouse rejections are how most practices find out what moved.

So we counted them.

What is actually in the July file

CMS publishes the full PTP tables behind an American Medical Association CPT license click-through, but it also publishes a free, ungated quarterly additions, deletions and revisions change set. That change set is the diff, and it is the primary document for this story. We downloaded it and tallied the rows.

Two things jump out. The PTP side moved by thousands of pairs while the MUE side moved by dozens. And the practitioner and outpatient hospital tables moved almost in lockstep, differing by 22 additions and 4 deletions, which is what you would expect when most new edits are clinically driven rather than setting-specific.

The number that matters: 424 pairs no modifier will touch

A PTP edit is a pair. CMS's own booklet on using the NCCI tools puts it plainly: when both codes are reported for the same patient on the same date of service, "the Column 1 code is eligible for payment and the Column 2 code is denied," unless both codes are clinically appropriate and you use an appropriate NCCI-associated modifier. Whether you can rescue the Column 2 code therefore comes down to one field in the file, and CMS prints the legend right in the header of its own change file:

• 0 = not allowed
• 1 = allowed
• 9 = not applicable

That is the modifier indicator. It decides whether an NCCI modifier can bypass the edit at all.

Of the 2,199 new practitioner pairs, our count of the change file finds:

The outpatient hospital file splits almost identically: 417 of its 2,177 new pairs carry indicator 0.

Those 424 pairs are the practical story of this quarter. On an indicator-1 edit, a correctly documented, appropriately applied NCCI-associated modifier can unbundle the pair, and the appeal is a documentation argument you can actually win. On an indicator-0 edit there is no modifier to append. Appending one anyway does not create an exception. It creates a record of putting modifiers on edits that forbid them. That pattern is exactly what a payer audit is built to find.

This is the reflex that costs small practices money in July. A denial lands, the biller recognizes the edit, reaches for the modifier that worked last quarter, and resubmits. On 424 new pairs that resubmission is a guaranteed second denial, and the staff time spent on it is gone.

The deletions are quiet money

The 708 deleted practitioner pairs get almost no attention and they are worth reading. A deleted edit means CMS is no longer bundling that combination, so a pairing your team stopped billing together, or has been writing off, may be payable again as of July 1.

Most practices never revisit this. The edit went in at some point, someone built a rule in the practice management system or wrote a note in a cheat sheet, and the rule outlived the edit. Deletions are the one part of the quarterly file that can put money back on the table rather than take it off, and they are the part nobody reads.

Only two pairs changed their modifier indicator

CMS publishes a separate, very short file for pairs whose modifier indicator changed. This quarter it contains two pairs, identical across the practitioner and outpatient hospital files: 0027U with 81270, and 93295 with 93294. CMS lists both at modifier indicator 1 in the Q3 file.

Worth being precise about what that does and does not tell you. The free change file gives the new indicator, not the prior one, so it establishes what the indicator is now and not which direction it moved. If a pair in that list is one your practice bills, read the current value off the Q3 file rather than inferring the change from a previous quarter's cheat sheet.

The absence of indicator churn is itself the finding. Coders sometimes brace each quarter for a wave of bypassable edits going non-bypassable. Q3 2026 was two pairs. The pressure this quarter is volume: 2,199 brand-new edits landing at once.

The MUE side barely moved

MUEs cap the units of a code that CMS will pay for one patient on one date of service. The July change is small: 74 additions and 7 deletions for practitioner services, the same 74 and 7 for outpatient hospital, and for DME suppliers 3 deletions with no additions at all.

A short MUE list is good news for a small practice, because MUE denials are the ones your front-end scrubber is least likely to catch. It does not mean zero exposure. Seventy-four newly capped codes is still seventy-four codes that will start rejecting at a unit count nobody on your team has memorized.

Why a small practice feels this before a large one

A hospital system runs the quarterly files into a claim-scrubbing engine, and the edits are loaded before July 1. A three-provider practice frequently does not. It finds out through denials, one at a time, four to six weeks later, when the remittance advice arrives and the July cash flow is already short.

That lag is the entire competitive disadvantage, and it is not a technology problem. The change files are free, they are small, and they are posted a month ahead of the effective date. What is missing is a standing task on someone's calendar for the first business day of every quarter.

What we counted, and what we did not

In the interest of showing our work: every count above is our own tally of the row counts in CMS's published Q3 2026 change files, which are linked in the sources below. Anyone can download them and reproduce the numbers.

We deliberately did not reproduce the code pairs themselves. The PTP tables are CPT data, and CPT codes and descriptions are copyright the American Medical Association. The aggregate counts are facts about a public CMS file; the tables are licensed content, and the license lives with CMS and the AMA, not with us. Pull them from the source.

We also did not characterize what the individual new edits mean clinically. That would require reading the NCCI Policy Manual, which CMS revises annually and which is the document that explains the rationale behind an edit. If a specific new edit is hitting your claims, the manual is where that rationale lives, and it is the document to open. Confirm a code's official description in the Code Book before you build a billing rule around it, and use the encoder when you are checking the codes on a specific encounter.

For the documentation side of an indicator-1 appeal, the same principle applies here as in any other payer review: the record either supports the separate service or it does not. CMS's booklet is explicit that the supporting documentation has to be in the patient's medical record. Build the packet against the criterion by name and keep the evidence together in an evidence worksheet, so the appeal arrives as one packet rather than a scavenger hunt. When the note is ambiguous about whether a distinct procedural service happened, that is a query, not a modifier. Our provider query templates cover the phrasing that keeps a query compliant.