FY2027 ICD-10-CM: what's official, and why the counts disagree

Every year the ICD-10-CM code set turns over on 1 October, and every year the same thing happens: a dozen articles appear announcing how many codes were added, and they do not agree with each other.

This year I went and checked what the official sources actually say — and the answer turns out to explain the disagreement. This post covers the dates that are genuinely official, the files CDC and CMS publish, why the counts differ, and the part that actually costs money if you miss it.

#The dates that are official

These are the parts nobody disagrees about, and both come straight from the source.

The CDC's ICD-10-CM Files page states that "FY27 ICD-10-CM codes should be used for healthcare services provided from October 1, 2026, through September 30, 2027".

The CMS ICD-10 page puts it in claims terms: "Use these files for discharges occurring from October 1, 2026 – September 30, 2027 and for patient encounters occurring from October 1, 2026 – September 30, 2027."

Worth being precise about what that means, because it trips people up: the boundary is the date of service, not the date you key the claim. A service on 30 September 2026 is coded from FY2026 even if it is billed in November. There is no transition period and no overlap.

#What CDC and CMS actually publish

CMS publishes the FY2027 release as a set of files:

• 2027 Addendum
• 2027 Code Descriptions in Tabular Order
• 2027 Code Tables, Tabular and Index
• 2027 Conversion Table
• 2027 POA Exempt Codes
• FY 2027 ICD-10-CM Coding Guidelines

The CDC's NCHS directory carries the same release. Its FY2027 folder lists, verbatim:

Note the dates. The code files landed in June; the conversion table followed in August. If you pulled the release in June and have not been back, you do not have the conversion table — which, as below, is the one worth having.

#Why every article gives a different number

Here is the thing I did not expect to find, and it is the useful part of this post.

Neither CDC nor CMS publishes a headline count of new codes. I checked both pages specifically for one. The CDC files page states the effective dates and links the directory; it gives no figures. The CMS page lists the files and the date range; it gives no figures either.

So every number you have read — and I have seen figures ranging from around 190 to around 238 new codes for FY2027 — is somebody's own count, derived from the addendum. Those counts are not wrong so much as differently scoped. Reasonable people counting the same file disagree on:

• Billable codes versus all codes. A new category that exists only as a header, with the billable detail one level down, is either one new code or several depending on what you are counting.
• Restructuring. When a code is expanded into more specific children, the parent may be deleted and the children added. Counted one way that is several additions and a deletion; counted another it is one revision.
• Revisions to descriptions versus changes to the codes themselves.

None of that matters much if you are reading an article for orientation. It matters a great deal if you are quoting a number in training material or an internal memo, because the figure you cite will not match the one your colleague read somewhere else — and neither of you will be able to point at an official source that settles it.

#The deletions are the part that bites

Additions announce themselves. You reach for a code, it is not there, you look it up and find the new one. Annoying, fixable, and it happens while you are paying attention.

Deletions do the opposite. A code that went inactive on 1 October sits quietly in the places nobody re-reads:

• Charge templates and order sets, which were built once and have worked ever since
• Superbills and cheat sheets, including the laminated one taped inside a cupboard
• Saved favourites in whatever tool you look codes up in
• Standing orders and recurring visit templates

None of these fail loudly. They produce a clean-looking claim with an invalid code on it, which comes back as a denial some weeks later — by which time it has happened to a batch, not to one claim.

This is what the conversion table is for. ICD-10-CM-CONVERSION-TABLE-FY2027.xlsx maps codes that are inactive as of 1 October 2026 to their replacements. It is the single most useful file in the release for anyone who is not rebuilding a code set from scratch, and it is the one that arrived late — in August, two months after the rest.

#What to check before 1 October

A short list, in the order that catches the most for the least effort.

1. Run your top codes against the conversion table. Not the whole code set — the fifty or so codes that make up the bulk of your volume. If none of them moved, you are largely fine. If one did, you have found the problem before it found you.
2. Open the templates. Charge templates, order sets, superbills, the favourites list in your lookup tool. These are where deleted codes survive.
3. Read the FY2027 guidelines, not just the code list. The guidelines PDF is a separate file for a reason: instructions change too, and a code that still exists can acquire a new Excludes1 note that changes what you may report alongside it.
4. Check what your software actually loaded. Vendors update on their own schedule. "We support FY2027" and "the build on this machine has FY2027" are different statements, and the second is the one that matters on 1 October.
5. Agree the cutover rule with billing. Date of service, not date of entry. Write it down before the week it matters.

#A note on the app I build

I write this as the person who makes Codra, so treat it as disclosed interest rather than a neutral recommendation.

Codra is an offline ICD-10-CM reference — the whole FY2027 code set ships inside the app, along with the alphabetic index and the instructional notes. It has a screen for exactly the problem above: it lists what was added, deleted and revised for the fiscal year, and reconciles your own favourites and lists against the new set, so a deleted code in your favourites surfaces in the app rather than in a denial. Deleted codes are kept and marked rather than vanishing.

In the spirit of this post, I should say what it shows and where the figure comes from: Codra's build derives its change lists from the official addendum, and for FY2027 it reports 238 new, 21 deleted and 19 revised. That is our count from the official file under our counting rule — not an official figure, because as established above, there is not one. If you need a number you can defend, derive it yourself and state the rule.

#Frequently asked questions