Article Summary: Kathleen Haslam, Director of Outpatient, Professional, and Regulatory Auditing at Penstock, examines how medical coding changes can cause payment integrity edits to become outdated and lead to improper payments. CPT, HCPCS Level II, NCCI, MUE and ICD-10 updates can change code definitions, combinations, modifiers and billing requirements while existing payment integrity logic continues operating under earlier rules. Haslam explains why automated code-set updates alone cannot account for the clinical and operational judgment required to determine how coding changes affect edit logic, fee schedules, benefit configurations and contract language. She also explains how chart review identifies payment errors that claims data may not reveal by comparing valid, clean claims with the underlying clinical documentation. Using the 2021 office visit coding revisions as an example, the article demonstrates how a claim can appear correct while the medical record does not support the billed service under updated coding requirements. The article concludes that health plans should maintain continuous payment integrity edit review aligned with medical coding update cycles, supported by clear ownership, human oversight and access to medical records.

‍

Every January, when I was a director of compliance at ahealth system, the same thing happened around the third week.

Someone would come to me holding a stack of denials. Acode we used constantly had changed on Jan. 1, retired, redefined or foldedinto another code, and nobody had updated it in our system. We'd been billingthe old one for three weeks, and by then a hundred claims were already out thedoor.

One code nobody updated doesn't produce one bad claim. Itproduces hundreds, and they all come back.

Each one becomes work. Someone pulls the record. Someonedecides whether the new code accurately describes that service, which is adocumentation question rather than a find-and-replace. Someone corrects theclaim and resubmits it. You pay for the same claim twice in staff time, and ifenough weeks pass, you're past the filing deadline and the money is gone.

I do payment integrity work now, on the other side of theclaim, and I see the same failure in reverse. A plan pays something itshouldn't because an edit was written against a code set that has sincechanged. Take a procedure once billed alongside the imaging that guided it. Thecode was revised to cover both, but the edit still allows the pair, so theimaging gets paid twice.

Nobody in either story was careless. Both organizationswere busy with the things that announce themselves: patient care, quality,member service. Code maintenance never announces itself. It waits.

‍

How Medical Coding Changes Create Improper Payments

Medical coding changes because medicine changes.Procedures get redefined, services once reported separately are combined, codesare created and retired.

The 2026 CPT set alone carried 418 changes, according to theAMA:288 new codes, 84 deletions and 46 revisions. And that's one code set on oneschedule. HCPCS Level II updates quarterly. So do the National Correct CodingInitiative's procedure-to-procedure edits and Medically Unlikely Edits (MUEs). ICD-10changes in October and April. Corrections can arrive in between.

The updates tell you what changed. They don't tell youwhat it means for your payments.

‍

Why Automation Alone Can't Keep PaymentIntegrity Edits Current

The obvious response is to automate it: subscribe to thefeeds, ingest the files, let the system keep itself current. That's genuinelygotten easier, and it should have. Software catches a retired code faster andmore reliably than people can: it's a known value, and a system can flag andreject it.

The harder case is a valid code that no longer means whatit once did. The rules for a code pair, a modifier or a unit count haveshifted. The claim still looks plausible.

An updated code set is data. An edit is a judgment. Whena revision folds one service into another, a human decides whether the editneeds rewriting rather than remapping. When a code retires, someone has to knoweverywhere it lives, in edit logic, fee schedules, benefit configuration andcontract language, and confirm each one changed.

These failures are almost never technical. The updategets split across teams, everyone assumes someone else's system caught it, andno one's accountable for confirming it was done. Automation without an ownerisn't safer. It's a faster way to be wrong in more places.

‍

Why Chart Review Catches What Claims Data Misses

Edits check a claim against rules. Chart review checks aclaim against what happened.

That distinction matters most in exactly the situation acode change creates. The codes on the claim are valid. Nothing in the datacontradicts anything else in the data. The only thing that reveals the problemis the documentation, because the question is no longer whether the codeexists. It's whether the code still describes what was actually done.

Back in 2021, the office visit codes were revised to basecoding on total time or medical decision making. The claims looked fine on thesurface, but when you reviewed the office notes, the documentation frequentlydidn't meet the new requirements.

No edit could have caught that. The codes were valid, theclaim was clean, and the only place the problem existed was in the note.

That takes someone who can read a note against thecurrent definition of a code and judge whether the documentation supports whatwas billed.

‍

Why Payment Integrity Edits Need Continuous Review

Edits need a review cycle matched to the code calendar,not the budget calendar. And the review has to be able to reach the record.Anything that stops at the claim will keep confirming the claim is internallyconsistent, which was never the question.

The codes will change again in October, and again inJanuary. Every edit in production was written against an earlier version of therules, and none of them will announce when they've gone out of date. That workhas to belong to someone, and it has to happen before the claims go out ratherthan after the denials come back.

‍