By Jamie Noorlander, Director of Coding Applied Research, and Kate Rattanni, Sr. Manager, Coding & Quality
Every October 1, CMS refreshes the ICD-10-CM code set, and every year that refresh quietly reshapes how health systems document and get paid for care. This year's update (FY2027) brings 190 new codes, 30 deletions, and 4 revisions — a net addition of 160 codes to a set that will total nearly 74,900 active codes. That's actually a smaller cycle than we've seen in recent years — but smaller doesn't mean less consequential.
If there's one theme running through this year's update, it's specificity. Codes are catching up to advancements in medical diagnostic capabilities — genetic testing, gene sequencing, and precision medicine are letting clinicians identify conditions to a degree the code set hasn't always kept pace with. In other words, the code set is doing exactly what it should: evolving alongside the care it describes.
A few clinical areas see meaningfully more specificity this year:
Most of the 30 deletions follow a familiar pattern: a broader parent code retires in favor of the more specific children introduced this cycle. That's exactly where the real-world risk sits — a claim still coded to a now-deleted code after October 1, or one that defaults to an unspecified code when a more specific ICD-10 option is already available, both of which tend to surface as denials weeks later rather than on day one. Encounters that span the cutoff (say, a procedure on 9/30 with discharge on 10/2) add another layer of complexity that coders have to catch manually.
More specific codes only help if the clinical note supports them. If a physician writes "plantar fasciitis" without laterality, no coder — human or AI — can assign the specific code; they fall back to unspecified, and the point of the update is lost. Every ICD update has the potential to increase documentation burden upstream to providers, and most health systems don't hand out one-on-one training for an ICD cycle the way they might for an Evaluation & Management CPT change. The advantage of AI-powered coding here isn't just speed — it's being able to catch documentation that hasn't caught up with the new specificity ICD-10 now requires, while still coding correctly and precisely whenever the documentation does support the newly available codes.
Code set updates shouldn't be an annual fire drill for health systems or for the technology that serves them. Real readiness starts before CMS publishes the final file and follows a few principles.
Impact before inventory. The question isn't "what changed?”, it's "what changes for whom?" Mapping the additions, deletions, and revisions against the ICD chapters they touch reveals which service lines will actually feel this update, and historical claims volume shows where the impact concentrates, so preparation follows real-world effect rather than the order of the change log.
Catch failures before they become denials. Every October creates the same two risks: claims still coded to a now-retired code, and encounters that span the cutoff. Both surface as denials weeks later, unless the system is built to catch them on day one.
Speed with judgment. AI gets an organization to readiness fast; a human in the loop is what ensures fast never comes at the cost of quality.
This is how we approach every update cycle — which is why, for our customers, October 1 arrives and coding just keeps working. Looking further out, specificity isn't slowing down, and it shouldn't. ICD-11 moves toward base codes assembled with extensions in software rather than one flat list humans have to hold in their heads. That's the direction we've already built for: specificity lives in the tooling, so documentation can finally keep pace with the care it describes.
