Which ICD-10-CM Codes Have Been Retired? A Twelve-Year Lifespan Census

Annual change lists answer one year at a time: what arrived, what left, what was reworded. They cannot answer the question a coder actually asks when looking at an older record, which is whether a code was valid then, and if not, when it stopped being valid. Answering that needs every order file laid side by side.

ICD-10-CM took effect in the United States on October 1, 2015, which is fiscal year 2016, so the twelve order files from FY2016 to FY2027 cover the code set's entire working life here. Read together they show 99,085 distinct codes, of which 75,909 have been valid for submission in at least one year.

A code leaves that reportable population in one of two ways, and the difference decides what a coder should do about it:

• Deleted. The code is gone from the order file. 619 once-reportable codes have left this way. A claim carrying one is rejected as an invalid code, which is at least an unambiguous error message.
• Converted to a header. The code is still in the file, still has its title, and still looks entirely normal, but it has acquired children and is no longer valid for submission. 411 once-reportable codes are in this state. This is the expensive case: nothing about the code looks wrong, and the denial says the code is not billable rather than not real.

One more result falls out of the comparison, and it is worth knowing because it removes a whole class of worry: no code has ever been withdrawn and later restored. Across twelve years there is not a single gap in any code's run. Once a code leaves, it stays gone.

Each row counts the codes that were valid for submission in that fiscal year and not in the next one, so the retirement takes effect on October 1 of the following year.

Retirement is not spread evenly. Two years account for more than half of it: the FY2017 update, the first after the partial code freeze that preceded adoption, and the FY2023 update, which reorganised large parts of the external-cause chapter. A typical year since then retires about thirty codes.

The chapters are lopsided too. Of the 1,030 retirements, 602 fall in the injury, poisoning and external-cause chapters, where CMS has repeatedly replaced broad categories with more specific families. The remainder are scattered thinly: 56 in the endocrine chapter, 38 in digestive, 37 across eye and ear, and single or low double digits elsewhere.

Converted to a header.D69.1, qualitative platelet defects, was reportable from adoption in FY2016 through FY2026 — eleven straight years. In FY2027 it gained D69.11 and D69.19 and stopped being valid for submission. The code still sits in the file with the same title. A coder reporting it from October 1, 2026 is not using a deleted code; they are using a category header, and the claim comes back unpaid.

Deleted outright. The H40.11X- family for primary open-angle glaucoma left the order file entirely when the category was restructured. There is no header left behind and no title to recognise; the code simply is not in the file any more.

Reportable for a single year.318 codes were valid in exactly one fiscal year before retiring. D49.5, neoplasm of unspecified behavior of other genitourinary organ, is one: valid in FY2016 only. These are the hardest to catch, because a coder who learned the code set in any later year has never seen them, and an audit of old records will.

The same code, four different titles. Retirement is not the only way a code changes under a coder. 1,235 codes have had their long title revised at least once without the code itself changing, and Z68.43 has carried four distinct titles. Quoting a description from memory is therefore no safer than quoting a code from memory.

That last row is the one that surprises people. For all the annual churn, 91.9 percent of the codes reportable today have been reportable since the day the United States adopted ICD-10-CM.

1. Establish the fiscal year of the date of service, not the date of billing. A fiscal year runs from October 1 to September 30, so a March 2024 encounter is FY2024.
2. Look the code up in that year order file. If it is absent, the code did not exist; if present with the valid flag, it was reportable; if present without it, the code was a header that year and needed a more specific child.
3. When a code has retired, check which way it went. If it became a header, the replacement is one of its children and the documentation usually already supports the choice. If it was deleted, the replacement may sit elsewhere in the tabular and the record may need a query.
4. For claims spanning October 1, split by encounter date rather than applying one year files to the whole batch.

The release history pages list every code in each annual bucket, and Claim Check flags codes that are not valid for submission in the current year.

• Treating "deleted" and "no longer reportable" as the same set. Four in ten retirements never left the file at all; they became headers.
• Checking the current year file for an older date of service. A code valid in FY2019 can be absent, reportable or a header today, and only the FY2019 file settles what it was then.
• Assuming a code that disappeared might come back. None has, in twelve years.
• Trusting a remembered description. Over twelve hundred codes have been reworded without changing.
• Reading a converted header as a coding error by the clinician. The documentation is usually fine; the code set simply became more specific, and the child code is already supported.

The FY2016 to FY2025 files are the archived icd10cm_order_YYYY.txt releases on the CDC/NCHS FTP site; FY2026 and FY2027 are the CMS "code descriptions in tabular order" releases. Every figure on this page is a mechanical comparison of those twelve files, produced by the same script that builds the open-data pack and re-run on 2026-09-14. No figure is taken from a secondary source. The per-code table behind the page, and the shorter list of retired codes alone, are published as CSV and JSON under CC0 so that any row here can be checked against the CMS file it came from.

One limit worth stating: the order files record what CMS published for each fiscal year, so a mid-year addition appears in the file for the year it was published. The FY2024 file in the CDC archive includes the April 1, 2024 additions for that reason.