2025 Terminology Updates are Coming-If you work with health data, it matters.

Ah, September.? Kids are heading back to school, the NFL regular season is under way…and it’s time to get ready for new versions of many of the standardized coding systems used in health care every day!? For some of these (like ICD-10-CM), the new 2025 versions go into effect on the first of October.

While those who perform or bill for health care services are generally highly aware of these annual changes, they may be less top-of-mind for those who build content or logic that uses these codes for secondary purposes: clinical decision support (CDS), population management, research, etc.? However, they are important to proactively address.

The key building blocks for secondary use of standardized codes are “value sets”, which are simply lists of all the codes from a particular vocabulary that represent a single clinical concept.? For example, a value set of ICD-10-CM codes for the disease myasthenia gravis in adults would have two entries (G70.00 and G70.01).? A value set of ICD-10-CM codes for fracture of the femur might have thousands of entries, but the principle is the same.? Value sets are referenced in logical statements that are executed by a software application for whatever the secondary use case is.

However, a value set cannot just be built once and forgotten about.? Standardized coding systems all undergo periodic updates (usually on a regular schedule). With each update, new codes are added in order to fill in gaps in the prior version, or to keep up with advances in medical knowledge and the emergence of new conditions, treatments, and tests.? If you don’t update your value sets with any newly-added codes that correspond to the concept that the value set represents, your value set will fall out of date.? When those new codes start being recorded on patient records, your logic will fail to pick them up, and accuracy will be compromised: CDS rules will alert users when they shouldn’t (or fail to alert users when they should), quality measures will overestimate (or underestimate) the adherence of care to established quality measures, research results will be skewed…you get the idea.

The 2025 edition of ICD-10-CM has 252 new billable codes.? A quick perusal of these shows that, as is typical, there are clusters of multiple codes that follow a consistent pattern.? For instance, there are 61 new codes for hematologic malignancies that specify whether the malignancy is in remission or not.? In the 2024 version, codes existed to describe these same hematologic malignancies, but without specifying whether or not remission had been achieved.? There are also one-off new codes, such as R41.85, “Anosognosia” (which describes the state of a patient having a neurologic deficit but being unaware of it).

In addition to new codes, one thing to watch out for are changes to the description for an existing code that modifies its meaning.? This is frowned upon by informaticists (particularly terminologists), and with good reason.? It violates one of the core principles of terminology best practices as articulated in Dr. James Cimino’s classic 1998 “Desiderata” paper , known as “concept permanence”, which means that a given code, once published, should always have the same meaning.? If its meaning changes over time, the resulting uncertainty requires more care in interpretation (at best) or, like overlooked new codes, can result in automated logic yielding results at variance with clinical reality.

This isn’t terribly common with the most popular standardized health vocabularies.? Most changes to the description accompany a code are mere stylistic changes (or sometimes, correct a typographical error).? As an example, in 2025, the description for ICD-10-CM A13.81, “Rieger’s anomaly” was changed to “Rieger anomaly” (for some reason, there has been a trend in recent years to avoid apostrophes in eponymic diagnostic terms)-Worthy of note only to point out its insignificance.

On the other hand, some changes in ICD-10-CM descriptions between 2024 and 2025 do result in a significant change in meaning.? I found five in my review of the 2025 update, as shown in the table below:

When a standardized vocabulary is updated, in addition to adding new codes and changing the descriptions for existing codes, the update may remove some previously-released codes. The verbiage varies among vocabularies, but these are generally termed "retired", "inactivated", "deprecated", or something similar. This means that the vocabulary publisher no longer considers the codes to be appropriate for use going forward. However, it doesn't mean you should remove them from your value sets. In fact, since they may persist in patient records for the indefinite future, it is important to retain them in your queries and logic (unless they're somehow seriously flawed, like they represent an ambiguous or nonsensical concept, which is very rare).

I hope this article has been useful—Please leave comments and/or questions.? And if any use of standardized vocabularies, i.e. value-set content or logic, falls under your scope of responsibility, make sure to establish schedule of regular maintenance to address vocabulary updates.? For reference, here is the update schedule for some of the most commonly-used vocabularies:

• ICD-10-CM, as noted above, is updated yearly, going into effect on October 1 (link ).
• CPT-4 “category I” codes (the main portion of CPT-4) are updated yearly.? The updated code set is published every fall, and takes effect the following January 1 (link ).
• HCPCS level II is updated quarterly, in January, April, July, and October (link ).
• SNOMED CT US Edition is updated twice yearly, in March and September (link ).
• RxNorm publishes an incremental update on a weekly basis, with full releases monthly (link ).
• LOINC is updated twice yearly, in February and August (link ), though prior to August 2021, updates occurred in June and December.

Logos Informatics Consulting is a clinical informatics consultancy whose mission is to drive client success through optimal management of clinical data. If we can be of assistance, please reach out to use at [email protected].