Sometimes Interoperability Is Just a Modifier: CMS’s CY 2027 Imaging Site-Neutral Cut, the Hospital Machine-Readable File, 835 Allowed-Amount Percentiles, and the Contract-Semantics Price RFI
🎧 Part I Podcast free on Spotify.
🎧 Part II 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
CMS calls interoperability a “necessary foundation,” but the July package builds no single omnibus interop rule. It builds narrower pieces: standardized claims and status data, ACO rosters and digital measures, FHIR exchange, a targeted imaging site-neutral cut that runs on modifiers and codes, and a versioned hospital price file it is trying to make both valid and comparable.
Two OPPS components, two different legal statuses. The noncontrast-imaging payment change is a CY 2027 proposal, and CMS says it would apply beginning January 1, 2027 if finalized. The hospital price-transparency section is an RFI sitting on top of already-live machine-readable-file rules and finalized CY 2026 schema changes.
Site neutrality, precisely: pay the physician-office-equivalent rate (a 40 percent relativity adjuster) for imaging-without-contrast in APCs 5521 through 5524 and certain composite APCs, but only at grandfathered off-campus departments billing modifier PO. Rural sole-community-hospital departments and on-campus departments are exempt. This is the third narrow service-family step, not universal parity.
CMS’s case, in numbers: 70 codes carry more than 95 percent of the off-campus volume, excepted-department rates run about 2.5 times office rates, and CMS pegs CY 2027 savings at 260 million dollars and roughly 8.5 billion over 2027 through 2036.
Why it’s an interop story: the cut can only exist because the claim identifies the service, the APC family, the setting, and the PO or PN modifier. Payment reform rides on transaction semantics.
Transparency has moved from a posting chore to a governed public dataset: schema, validator, NPIs, 835-derived percentiles, an attestation, and now a monthly enforcement file. The next frontier is contract meaning, not file shape.
Table of Contents
Interoperability is not one API, sometimes it is a modifier
The site-neutral proposal, precisely bounded
CMS’s utilization case, with the numbers
Why a payment cut belongs in an interoperability story
Price transparency graduates from a posting chore to a public dataset
The CY 2026 quality layer that is already live
The 2027 RFI goes after contract meaning, not file shape
The consumer layer is still a mess
Enforcement becomes data too
Computable is not comparable
Interoperability is not one API, sometimes it is a modifier
CMS keeps calling interoperability a necessary foundation, and the phrase does a lot of quiet work, because it sets up an expectation the July package never fulfills. Nobody shipped an omnibus interoperability rule. What actually exists is a scatter of narrower mechanisms that each depend on shared, stable data: standardized claims and status transactions, ACO assignment rosters and digital measures, FHIR-enabled exchange, a targeted imaging payment cut that runs entirely on service and location identifiers, and a hospital price file CMS is trying to make both machine-valid and actually comparable across hospitals. Interoperability here is not a product. It is a precondition that shows up in different costumes.
The most useful reframing of this installment is that interoperability is sometimes not an API at all. Sometimes it is the humble billing modifier that lets a payment rule exist in the first place. The two big pieces here, an imaging site-neutral proposal and a hospital price-transparency RFI, look unrelated, one about paying less for CT scans and one about parsing contract terms out of a JSON file. What ties them together is that both are entirely dependent on identifiers meaning the same thing everywhere, and both are attempts to turn messy real-world variation into something a computer can act on. They just serve different users and sit at different points on the road from “the file parses” to “the numbers are comparable.”
The site-neutral proposal, precisely bounded
The imaging proposal has to be stated with its boundaries intact, because the fastest way to be wrong about it is to call it site neutrality without the qualifiers. CMS proposes to pay the physician-fee-schedule-equivalent rate, implemented as a relativity adjuster equal to 40 percent of the OPPS amount, for imaging-without-contrast services assigned to APCs 5521 through 5524 and certain specified multiple-imaging composite APCs, but only when those services are furnished in an excepted off-campus provider-based department billing modifier PO. Nonexcepted off-campus departments, the ones billing modifier PN, already get the office-equivalent rate, so this proposal is about closing the gap for the grandfathered, excepted departments that had been protected. And rural sole-community-hospital departments would be exempt from the change entirely.
The legal scaffolding matters because it is what makes a non-budget-neutral, service-specific cut defensible. CMS invokes the statutory authority to adopt a method for controlling unnecessary growth in the volume of covered outpatient services, and it points to the D.C. Circuit’s decision in American Hospital Association v. Azar, which upheld exactly this kind of service-specific, non-budget-neutral payment reduction. That citation is not decoration. It is CMS signaling that it has been to court on this theory before and won, which is why it feels comfortable targeting a single service family rather than adjusting rates across the board.
The boundaries, stated plainly, are the whole point. This is not universal OPPS-to-office parity. It applies to a defined imaging family, only at grandfathered off-campus hospital departments, with a rural carve-out, and it does not touch on-campus departments or the broad run of outpatient services. Anyone who describes this as CMS adopting general site neutrality, or who forgets the rural exemption, or who sweeps on-campus departments into it, has overstated a deliberately narrow policy into something it is not. Precision on scope is the difference between describing the proposal and misrepresenting it.
CMS’s utilization case, with the numbers
CMS builds its argument on a concentration story, and the numbers are worth carrying because they are what a comment letter will either accept or attack. CMS says roughly 70 HCPCS codes account for more than 95 percent of imaging-without-contrast volume in off-campus departments, even though the relevant APCs contain 337 codes in total, and the same proposal covers APCs 5521 through 5524 plus imaging composite APCs 8004, 8005, and 8007. That concentration matters, because it means a narrowly targeted policy hitting a short list of codes captures almost all the volume. CMS adds that more than 90 percent of those top 70 codes overlap between excepted and nonexcepted departments, which is the empirical backbone of the “same service, different rate” claim: the departments are doing substantially the same imaging, just under different payment treatment.
The growth figures are the motive. From 2016 through 2025, CMS says volume for the top codes in excepted departments rose more than 38 percent and spending rose 33 percent, with the increment amounting to about 126 million dollars in 2025 alone, and CMS estimates the proposal would reduce Medicare spending by about $260 million in the first year. More striking, utilization per fee-for-service beneficiary climbed more than 67 percent over a period when fee-for-service enrollment actually fell by roughly 17 percent. Volume up sharply while the covered population shrank is exactly the pattern the “control unnecessary growth” authority was written for, and CMS is clearly building that record on purpose.
Then the payment gap, which is what makes it worth doing. CMS calculates that excepted-department OPPS rates average about 2.5 times the office rate for this family, and that beneficiary cost sharing runs more than twice the office amount, because a beneficiary generally pays 20 percent of the higher rate. So the patient is penalized for where the scan happens, not for anything clinical. CMS estimates CY 2027 savings of 260 million dollars, split as about 190 million to Medicare and about 70 million to beneficiaries, and projects 8.5 billion in lower net Part B spending across 2027 through 2036. To its credit, CMS says it considered the obvious confounders, coding changes, guideline shifts, case mix, and pandemic rebound, and it leans on MedPAC’s finding of substantial severity overlap and no statistically significant cost effect for aligned low-complexity services, while noting that genuinely needed extra observation, medication, or diagnostic services remain separately billable. The honest editorial formulation is that CMS is testing whether comparable service identity plus observed setting patterns justify a narrower rate. It is not claiming every patient or every facility is identical, and it would be a straw man to say it is.
Why a payment cut belongs in an interoperability story
At first glance an imaging payment cut has nothing to do with interoperability, and that reaction is exactly why it belongs here. The cut can only operate because the claim carries a specific stack of identifiers that all have to be present and correct: the HCPCS code for the service, the APC family it maps to, the provider setting, the off-campus status, and the PO or PN modifier that says which kind of department billed it. Strip out any one of those and the policy cannot find its target. The rate reduction is not applied by a human reading a chart. It is applied by logic keying off standardized data fields.
This is what data-defined payment looks like in practice. A modifier, two characters on a claim line, is what lets CMS carve one site-and-service combination out of the vast undifferentiated stream of outpatient claims and pay it differently. That is interoperability in the least glamorous, most operational sense: not an app, not a FHIR endpoint, but the boring agreement that PO means one thing and PN means another, consistently, across every hospital in the country. It does not create clinical-data interoperability, and nobody should claim it does. What it demonstrates is the dependency running underneath the entire series, which is that payment reform is impossible without consistent transaction semantics. You cannot pay differently for a thing you cannot reliably identify, and identification is a data problem before it is a payment problem.
Price transparency graduates from a posting chore to a public dataset
The second half of this installment is where interoperability climbs one rung, from claims identity to price-data identity, and the hospital price-transparency regime is the case study. The obligation itself is old news by now. Since January 1, 2021, hospitals have had to publish a comprehensive machine-readable file plus consumer-friendly shoppable-service information. The machine-readable file has to cover gross charges, discounted cash prices, payer-specific negotiated charges, and de-identified minimum and maximum negotiated charges across hospital items and services. And since July 1, 2024, hospitals have been required to produce that file in one of CMS’s standardized layouts, the wide CSV, the tall CSV, or JSON, following an official data dictionary, with the standardized schema and validator formally taking effect January 1, 2025. That standardization is the moment the file stopped being a free-form disclosure and became a schema-governed artifact.
The reason this matters more than a compliance checkbox is who CMS expects to use the file. CMS explicitly envisions employers, researchers, innovators, price-tool developers, EHR vendors, consumer apps, and aggregators all reusing the machine-readable file. That intended-audience list reframes the whole exercise. The file is not a static page a shopper glances at once. It is a piece of public data infrastructure, meant to be ingested, joined, and built on by third parties at scale. Once you accept that framing, everything CMS does next, the schema tightening, the validator, the identifiers, follows logically, because infrastructure has to be consistent in ways a one-off disclosure never did.
The CY 2026 quality layer that is already live
The part people miss is that the transparency regime already got materially stricter this year, and this piece is live rather than proposed. Beginning January 1, 2026, with enforcement starting April 1, hospitals that report percentage-based or algorithm-based rates, rather than a clean dollar figure, must derive and encode statistics from their remittance data: the 10th percentile, the median, and the 90th percentile of allowed amounts, plus the count of underlying allowed amounts, computed from 835 remittance advice or equivalent data over a 12- to 15-month lookback, and CMS’s final schema implements those fields in version 3.0. In the same move, CMS removed the old “estimated allowed amount” concept. So a hospital can no longer wave its hands with an estimate when the negotiated charge is a formula; it has to go into its actual remittance history and compute a distribution.
The accountability wrapper tightened too. Hospitals must identify the relevant active Type 2 NPIs, the organizational provider identifiers, and include an attestation from a CEO, president, or responsible senior official swearing the file is true, accurate, complete, and contains all the information needed to derive a dollar amount when the negotiated charge is not directly knowable. That is a named executive putting their signature on the data, which changes the compliance calculus considerably.
CMS’s official data dictionary version 3.0 is where the regulation literally becomes machine validation, and the specifics show how tightly the screws turned. The schema replaced the prior affirmation object with an exact attestation, added an attester name field and a list-valued Type 2 NPI field, removed the estimated-amount field, and added median amount, 10th percentile, 90th percentile, and count fields. For a percentage or algorithm charge, the remittance count is now required, not optional. And in a nice privacy touch, counts from one through ten are reported as the literal category “1 through 10” rather than the exact small number, to reduce the risk of reidentifying a specific contract from a thin cell. That last detail is the tell that CMS is thinking about this file as a real dataset with real reidentification risk, not as a compliance formality.
The 2027 RFI goes after contract meaning, not file shape
Having largely won the battle to make the file parse, CMS is now opening the harder war, which is making the contents mean the same thing across hospitals, and that is what the 2027 RFI is about. CMS says its compliance reviews keep finding that important contract context shows up inconsistently and that free text is a nightmare to parse. So it asks whether things like outlier payments, stop-loss clauses, rate tiers, carve-outs, and other adjustments should get their own dedicated structured fields, and if so, whether those fields belong at the item-or-service level, the service-category level, or the contract level. Those are not trivial questions; they are the difference between a price file you can actually compare and one that is technically valid and practically useless.
CMS also goes after the identifier problem directly, and the example it gives is the perfect illustration of why semantic interoperability is hard. It asks about standardizing payer, plan, product, network, and employer identifiers, because the same insurer shows up in the wild as “Blue Cross,” “BlueCross,” “BC,” and “BCBS,” and those variants make cross-hospital matching a mess. If you cannot reliably tell that two hospitals are describing the same payer, you cannot compare what that payer pays them, and the whole point of the dataset evaporates.
This is the clean conceptual line worth drawing for a technical audience: the shift from syntactic interoperability, meaning can the file pass the schema, to semantic interoperability, meaning does the same field mean the same thing across hospitals and contracts. CMS has mostly solved the first and is now attacking the second. The crucial status caveat is that the RFI is a set of questions, and CMS explicitly anticipates future guidance and notice-and-comment rulemaking before any of it becomes required. The RFI does not amend the underlying transparency regulation. So these proposed structured fields are things CMS is thinking about, not fields hospitals must populate in 2027, and presenting them as finalized requirements would be a status error.
The consumer layer is still a mess
Underneath the machine-readable-file story sits the consumer-facing layer, and CMS more or less admits it is still a mess. The current rules require hospitals to post 300 shoppable services, including 70 that CMS specifies, or every shoppable service if the hospital offers fewer than 300. Hospitals can satisfy this with a standardized file or by qualifying through an internet price-estimator tool. In practice, CMS reports inconsistent formats, confusion about whether prices include facility charges, professional charges, or both, ambiguity around ancillary and bundled services, and cases where a cash price appears in an estimator but not in the machine-readable file, which means the two artifacts contradict each other.
So the 2027 RFI asks a batch of consumer-layer questions too: whether to update the list of 70 specified services, whether to require a common shoppable-service file format, whether to end the deemed-compliance shortcut for estimator tools, whether to require the estimator’s underlying data as a separate file, whether to standardize consumer-facing fields, and whether to make ancillary services, implants, and bundles explicit rather than buried. Every one of those is a real usability problem for anyone trying to actually shop for care. And every one of them is, again, a question rather than a 2027 requirement. The consumer layer is being actively rethought, but it has not been resolved, and the honest read is that the shopper experience remains the weakest, least standardized part of the whole regime.
Enforcement becomes data too
The final move completes the pattern that runs through the entire oversight rail: enforcement itself becomes a dataset. In July 2026, CMS launched a monthly public-use dataset of enforcement activities and outcomes, carrying the hospital and location, address, the CMS-assigned transparency identifier, the action or outcome, and the date. Since 2024, CMS has had authority to disclose compliance assessments and actions well beyond the formal civil monetary penalties it used to be limited to publicizing. So the government is now publishing, on a recurring cadence, a structured record of who complied, who did not, and what happened.
Line that up with everything else in this installment and you get a recurring architecture that is worth naming explicitly, because it recurs across CMS’s whole approach: a schema, a validator that checks against it, organizational identifiers that make entities matchable, an executive attestation that puts accountability on the data, and an enforcement record that turns the whole thing into oversight. That is the same skeleton, dressed for price transparency, that shows up in nursing-home surveys and drug-price reconciliation. CMS builds the format, checks conformance, identifies the actors, extracts a sworn statement, and publishes the consequences.
One caution CMS itself issues is worth honoring: a closed enforcement case does not establish present compliance. A hospital that resolved an old case is not thereby certified as compliant today, and reading the enforcement dataset as a clean bill of health would misuse it. And the deeper caveat, the one that should temper any excitement about price transparency as a shopping tool, is that even at its best the machine-readable file describes contracted or historical allowed amounts, not a guaranteed individualized out-of-pocket quote for a specific patient. The 2027 RFI is trying to make those historical amounts comparable across hospitals without pretending they are personalized price quotes, and that distinction has to survive intact. Computable is not the same as comparable, and comparable is not the same as an actual quote. CMS has largely standardized the file’s shape. Standardizing what the numbers mean, and being honest about what they can and cannot promise a patient, is the work that remains
.