Saturday, June 27, 2026

AMA CPT: Coding Policy EARLY RELEASES: Cat I, Cat III, PLA, APP S AI, ADMIN MAAA

 AMA has released a website that displays "early release" updates in policy in five areas:

  • Category I Codes
  • Category III Codes
  • PLA Codes
  • Appendix S: The "AI Taxonomy"
  • Administrative MAAA codes
Find them all here:


Early Release
The Category I, Category III, PLA, and administrative MAAA code sets all include some form of early-release codes. Category III, PLA, and administrative MAAA codes are early release by definition: Category III codes are released twice a year, and PLA codes are released quarterly. Category I codes are rarely released early, with vaccines being one exception.
_____

PLA (Incl. new rules & oddities)
The PLA code section includes policy text found just ahead of the PLA listing at the end of the Pathology chapter, p. 701 in my 2026 book. For example, it says that PLA codes include, but are not limited to, ADLTs and CDLTs "as defined under the Protecting Access to Medicare Act of 2014 [PAMA]." However, PAMA defines ADLTs but nowhere defines CDLTs.

AMA then states that "these assays" — which could refer grammatically to PLAs, ADLTs, or CDLTs — may include, but are not limited to, MAAA, GSP, Infectious/Microbiology, or Chemistry. Descriptors follow any coding conventions described elsewhere, such as those for MAAA or GSP.

This policy section also notes that tests with physician work are not eligible for PLA codes and that "Algorithm-only analyses of existing test results, without accompanying biomarker analysis, do not meet eligibility criteria for a PLA code." This has been used to exclude AI tests based on H&E slides from PLA. PLA tests must be "sole course," which means that tests with, for example, software interpretation performed elsewhere, are not PLA codes.
_____

Appendix S, AI

Appendix S, on AI definitions, was extensively revised over multiple sequential CPT meetings, and I believe the final May 1 meeting text is presented here, with a dateline of June 8. While the chapter is called, "Taxonomy for artificial intelligence in medical services" the text immediately states, "the term “AI” is not defined in this taxonomy." This is one of the weirder statements I’ve seen.

Apparently, if two people vehemently disagree over whether a service is AI or not, under each person’s own unwritten definition, this chapter can nonetheless classify the type of AI that it is, but not whether it is.

Concurrent Inconsistent Versions Not Flagged

Also note, if you seek Appendix S from the "early update" page, you'll get what I think is 2027 text here:

But if you come into AMA from Google for Appendix S, you'll get the 2025 and earlier version instead:

###