Medicare Gives Clinical AI a Billing Identity Before It Has a Price Theory: The CY 2027 OPPS SaMS Proposal, the New O1 Status Indicator, the New Technology Bridge, and the Lab-Analysis Migration
🎧 Podcast episode for paid subscribers only. Also available on Spotify.
To listen to paid episodes in Apple or Spotify, link your Substack subscription via the show settings on those platforms (instructions inside the Substack app under Subscriptions → Podcast).
Abstract
The CY 2027 OPPS proposed rule, CMS-1850-P, does not create a universal payment category for clinical AI and does not price algorithmic value from first principles. Comments are due August 31, 2026.
What it does: proposes an outpatient taxonomy for certain algorithmic clinical services, a new status indicator (O1) that makes many of them separately payable, and a temporary New Technology APC band to hold rates roughly steady while CMS keeps arguing with itself about cost, evidence, packaging, and overuse.
The vocabulary tell: CMS renames Software as a Service (a cloud business model) to Software as a Medical Service, or SaMS. Cute, and telling.
Not all of it gets paid separately. Across source-defined groups totaling 36 HCPCS codes designated as SaMS in Table 61, a subset would carry O1 while others retain their existing status indicators (such as Q1, E1, N, or M) and packaging rules. Do not read this as “all the AI codes now get paid.”
The rates are stability, not valuation. AI-QCT stays at 950.50 dollars even though a handful of early claims implied placement in the lowest New Technology APC payment band. Optellum keeps its assignment even though claims implied a ~90 percent cut. This is legibility, not a value model.
The sleeper is the lab move: CMS proposes to treat repeated algorithmic analysis of stored test data as an “other diagnostic test,” pulling it off the Clinical Laboratory Fee Schedule.
This did not start in 2026. HeartFlow got separate OPPS payment back in 2018. MedPAC has been warning about bundle erosion and overuse the whole time.
Table of Contents
CMS gives clinical AI a billing identity before it has a price theory
What SaMS actually is, and the quiet burial of SaaS
Why OPPS was never built to pay for software
Three cohorts, not one category, and O1 is a flag not a seal
The New Technology bridge, or rate stability wearing a pricing costume
The claims-data paradox: AI-QCT and Optellum
The lab migration is the real structural move
This started in 2018: HeartFlow and the lineage nobody remembers
MedPAC’s counterweight and the value question CMS ducked
What O1 does not certify
CMS gives clinical AI a billing identity before it has a price theory
The temptation with this proposal is to write the headline “Medicare starts paying for AI,” collect the clicks, and move on. Resist it, because it is wrong in both directions at once. Medicare has paid for algorithmic clinical software since 2018, so this is not a debut. And the CY 2027 OPPS rule does not actually price the value of AI, so this is not a valuation either. What it is, and this is the useful framing, is CMS handing a set of algorithmic clinical services a billing identity before it has figured out what any of them are worth. Legibility first, valuation later, maybe never.
That gap between “we can now see it in the claims” and “we know what to pay for it” is the entire story, and it is a genuinely hard problem rather than bureaucratic dithering. The stuff CMS is trying to pay for does not behave like the stuff OPPS was built to pay for. There is no gurney, no vial, no observable material resource to cost out. There is a subscription, a per-click license, and a proprietary model whose actual cost the vendor has every incentive not to disclose. So CMS is doing the only thing it can do at this stage, which is create a place to put these services on a claim, flag some of them as separately payable, and buy itself time with a temporary rate band while it keeps asking the questions it has been asking, unresolved, for years. Everything below is a variation on that theme.
