DowncodedMedicare Reposted the Roster It Called Final
CMS replaced the CY2027 Ambulatory Specialty Model participant list on October 3 and said so nowhere. The roster did not move: same 5,533 clinicians, same 232 markets, same $1.3 billion of CY2024 Medicare billing behind the plus-or-minus 9 percent the model puts on them. Exactly one column changed, and it was the one that costs nobody a dollar.
Last November, in the middle of the CY2026 physician fee schedule, CMS wrote a rule that drafts clinicians into a payment model they are not allowed to decline (90 FR 49266, published November 5, 2025).
The list of who got drafted went up on September 3. 5,556 rows, 5,533 clinicians, 232 markets, a cardiology cohort and a low back pain cohort, and a file name that said Master_Final.
A month later it went up again.
The file changed, the roster didn't
The participant dataset on data.cms.gov now serves a different CSV. Its name is 202609017_Revised_Master_Final_Participant_List_Public_CY27.csv, which manages to carry both "Revised" and "Final," plus a nine-digit date stamp that is not a date. The file is stamped October 3 and sits in the October folder. The catalog flag that would tell a monitoring script something moved did not flip until October 5.
Nothing announces it.
The model's own page was last touched October 2, roughly eleven hours before the new file landed, and it still reads "The list of final ASM participants for the 2027 performance year is now available." No note, no correction, no reason to re-pull.
So we re-pulled both, and diffed all 36 columns.
The roster is identical. Same 5,556 rows, same 5,533 NPIs, same two cohorts split 3,341 and 2,215, same 232 CBSA codes, same 2,085 clinicians flagged small practice.
Run the CY2024 Medicare non-drug payments for the list and you get $1,318,088,916, to the dollar the same number the September file produced.
Nobody was added and nobody got out.
One column moved
Of 36 columns, exactly one differs: MIDDLE_NAME.
In the September file, every one of the 5,556 rows carried a middle name. 1,581 of them carried a middle name that was byte-identical to the clinician's first name.
The data dictionary, which CMS has not revised, says the field holds "the middle name of the ASM participant."
Check those NPIs against NPPES and the diagnosis is immediate. The eight we pulled all return middle_name: null.
Their own enrollment record says they have no middle name, and the roster had copied the first name in on top of the blank.
The repost clears exactly those rows, 1,578 of them, plus two more that got a real middle name supplied, and it touched zero single-letter initials.
1,580 of 1,581 fixed (chart below), and one row, somehow, still has it.
What a month of revising did not reach
Four things in that file contradict themselves, and all four survived the repost untouched:
- 23 NPIs appear on two rows each. None of the 23 pairs are identical rows, and every pair names two different organizations, which is the tell: the file is keyed to a clinician-organization pair while the model adjusts a clinician.
- 7 of those NPIs carry both a
Yesand aNoin the small-practice column. - 4 carry two different CBSA assignments, in two cases across a state line.
- 4 carry two different specialty assignments, all four inside the low back pain cohort.
The small-practice column is the one that bites. The dictionary defines it as whether the NPI is part of a practice of "15 or fewer clinicians within the NPI's associated Taxpayer Identification Number," singular, and seven clinicians here have two TINs and two answers.
The file does not say which one governs.
Those seven billed $2,124,285 in CY2024 Medicare non-drug payments. At the model's 9 percent, that is $191,186 of adjustment hanging on a field with two values, and the largest of the seven is carrying $74,707 of it alone.
Final Thoughts
Payment adjustments in this model run from -9% to +9% for the first two performance years and widen to 12 by the last, and the clock starts January 1, 2027. A roster that decides that is worth getting right, and the correction CMS shipped is the one correction that moves no money at all.
The practical version is shorter. If you matched your clinicians against the September file on full name, 28.5% of the rows were carrying a middle name that was not a middle name, and your match rate told you something false about which of your physicians are in this thing. That is fixed now. Pull the file again, match on NPI, and if a clinician comes back twice, note that CMS has not decided either.
Thanks for reading.