SolvingHealth · the model

The attested outcome is the asset

Derived, not asserted. Three things happened between 5 July and 10 September 2026 that together decide what an orthopaedic practice is selling for the next decade. Here is the derivation, the arithmetic, the constraints, and what would prove it wrong.


1. The law

Start with the only premise that does any work. A selection layer — a search engine, a directory, a referral network, an answering system — can only sort on attributes it can read. Everything else about you is prose, and prose is not comparable.

The first attribute you make machine-readable becomes the attribute you get selected on.

This is not a claim about artificial intelligence. It is a claim about ranking functions, and it has been true of every marketplace ever built. What is new is that the selection layer for surgical care now exists, and it has exactly one populated column.

2. Exhibit A — the selection layer arrived

On 1 September 2026, two companies announced a partnership routing musculoskeletal patients end to end. By their own figure, roughly 15% of members progress past conservative care; those members are guided to participating orthopaedic providers with a bundled price up front, capable of combining surgeon, facility and anaesthesia into a single number.

The mechanism is named in the vendor's own product list: navigation infrastructure, powered by a named agent, used to route patients to transparently priced care. A named agent, in production, selecting which orthopaedic surgeon a patient reaches.

What it sorts on
Price. Not because anyone decided price matters most — because a decade of transparency rules populated the price field and nobody populated an outcomes field. The agent is not choosing price over quality. It is choosing the only thing in the record.

This is not a criticism of either company. Transparent bundled pricing is a real good and the 15% figure is an honest one. It is an observation about routing layers in general.

And the network is opt-in, which reads as protection until you apply the difference between a list and an answer. A list has many winners. An answer has one, or maybe three. Absence from the machine-readable set does not make you second-best; it leaves you out of the answer.

3. Exhibit B — CMS is populating the other column

The ACCESS Model (Advancing Chronic Care with Effective, Scalable Solutions) is a CMS Innovation Center model under §1115A of the Social Security Act. It began 5 July 2026 and runs ten years. Voluntary, Original Medicare, four clinical tracks — and one of them is musculoskeletal: chronic musculoskeletal pain.

It replaces activity payment with Outcome-Aligned Payments: a recurring payment for managing a qualifying condition, with full payment tied to achieving measurable outcome targets. For the MSK track those targets are validated patient-reported outcome measures of pain, mood and function.

The sentence that changes the board
CMS states the model will “promote transparency by publishing risk-adjusted health outcomes … so patients and referring clinicians can make informed choices,” and will maintain a public directory of participants, the conditions they treat, and their risk-adjusted outcomes. ⚠ The date is specific: CMS says publication begins Winter 2028, in the ACCESS Directory and on Data.CMS.gov. That is a dated commitment, not something already visible — and the gap between now and then is exactly the window in which a practice can build the record it will be published on.

The payer is populating the outcomes column itself — publicly, risk-adjusted, for musculoskeletal care. The column the routing agent could not read is being filled in by the largest payer in the country.
The door most clinicians will actually use
You do not have to become an ACCESS organisation to be paid inside it. CMS states that primary care and referring clinicians can refer patients to participating organisations, receive electronic updates on their patients' progress, and bill a new co-management payment for documented review of patient updates and associated coordination activities — medication adjustments, problem-list updates.

That is a materially lower barrier than participation, and it is the entry point we expect most practices to take first.

And note the entry requirement. Participants must be Medicare Part B-enrolled organisations and must designate a physician Clinical Director responsible for clinical oversight and compliance. A named accountable clinician stops being a differentiator and becomes a condition of admission.

Source: cms.gov/priorities/innovation/innovation-models/access, read 10 September 2026 (page last modified 12 August 2026). Rolling application start dates are listed; the next one printed on that page falls after its own last-modified date, so ⚠ re-verify current dates against CMS before acting. Not legal or billing advice.

4. Exhibit C — the governance pattern is already specified

If agents are going to act inside clinical care, somebody has to say how they are held. The most serious public answer is ARPA-H's ADVOCATE programme — $62.7M over four years, building toward the first FDA-authorised clinical agentic system. ⛔ Nothing is FDA-authorised; authorisation is the programme's goal, with a first-of-its-kind package due within 24 months of award.

The architecture is the useful part:

Nowhere in that architecture did anyone buy more human reading. Faced with making an autonomous clinical agent trustworthy at national scale, with the regulator in the room, nobody proposed a larger review committee. They bought a machine that checks the machine, and put the named human at escalation and accountability — not at accuracy.

That distinction is load-bearing and we state it as a limit on our own product: a human in the loop buys accountability, not accuracy. The evidence does not support “reviewed, therefore more accurate.”

5. The constraints — where you may actually stand

Two payment mechanics decide which populations are reachable. Both are commonly modelled wrongly.

Remote therapeutic monitoring — four rules

RuleConsequence
Monitoring in a global period may be billed only by practitioners not receiving the global paymentThe operating surgeon is locked out of their own post-operative patient for the global period
One practitioner per patient per 30 daysThe monthly slot is exclusive and may already be held by day 91
RTM requires no established patient relationship (physiologic monitoring does)Reach extends past the existing panel
Services may be delivered under general supervisionThe only line that lets one clinician's accountability cover a population rather than a schedule

And the arithmetic that settles it. The management line is twenty minutes per patient per calendar month, captured in roughly 53% of monitored months:

300 patients × 0.53 × 20 min = 53 hours per month

Impossible for a surgeon; ordinary for delegated staff. The binding constraint is hours, not dollars — which is why the valuable engineering is in preparing, routing, evidencing and attesting that interaction, never in the monitoring device.

⛔ And the two rails are mutually exclusive

CMS states that ACCESS participants and their affiliated entities may not submit Medicare fee-for-service claims for other services furnished to their ACCESS-aligned beneficiaries during an active care period — only ACCESS codes may be billed for aligned beneficiaries.

So ACCESS is not additive to a monitoring model. It is an alternative, chosen per patient. Any plan that stacks both on the same beneficiary is wrong before it starts.

5b. One more force, and a caution about how it is being repeated

The Ambulatory Specialty Model is the sharpest of the lot for this argument, because it is the one that scores a person rather than a hospital. CMS describes it plainly: “ASM is a mandatory model with five performance years that run from January 1, 2027, through December 31, 2031,” testing performance-based payment adjustments for specialists treating heart failure or low back pain.

Orthopaedic surgery is named in the low-back-pain cohort. Eligibility is historic: at least twenty attributed episodes from the relevant episode-based cost measure, in the calendar year two years before the performance year. A practice in a selected geography does not choose to participate.

⚠ And a claim to handle carefully
The CY2027 Physician Fee Schedule proposes replacing the G2211 visit-complexity add-on with a modifier — MOD1, raising the associated E/M payment by about 16% as a percentage rather than a flat dollar amount, with a second tier, MOD2, at roughly 32% and available only to ACO participants.

⛔ That is a proposed rule, not a final one. We have seen it repeated in the present tense — “G2211 becomes MOD1” — and it does not yet. It is the rule the comment period exists for. If it finalises as proposed it widens the gap for proceduralists, who have comparatively little longitudinal evaluation-and-management volume to append a percentage to. Until then it is a forecast about a draft.

One widely circulated professional-society summary states MOD1 at 32%; CMS's own fact sheet and multiple independent summaries say 16% for MOD1 and 32% for the ACO-only MOD2. Where secondary sources disagree, read the rule.

Sources: cms.gov/priorities/innovation/innovation-models/asm and the ASM participant FAQ; CY2027 PFS proposed rule fact sheet and the Federal Register text. Read 10 September 2026. ⚠ Payment-adjustment magnitudes and participant counts circulate widely in secondary summaries — verify against the rule before quoting a percentage.

6. What all of it converges on

Four independent forces, pushing from different directions, all reward exactly one artifact.

ForceWhat it demands
The routing layerAn outcome attribute a machine can read, or you are absent from the answer
ACCESSRisk-adjusted outcomes published from Winter 2028; MSK scored on PROMIS PF, PROMIS PI and PGIC; a named physician Medical/Clinical Director required at entry. Referring clinicians may bill a Co-Management Payment (G0677 for MSK, $30 per service, up to 3× per 12-month care period) without joining the model.
ASM — mandatory, five performance years 1 Jan 2027 – 31 Dec 2031Scores an individual named clinician. The low-back-pain cohort explicitly includes orthopaedic surgery, neurosurgery, anaesthesiology, pain management, interventional pain management and PM&R, for clinicians with ≥20 attributed episodes in the calendar year two years prior. There is no opt-out.
Episode models — TEAM (live, 30-day, five surgical episodes) and CJR-X (90-day joint episodes including PT, from 2028)A named clinician or institution scored across a period, on the perimeter of the operation. ⚠ TEAM began in 2026 but CMS states “Track 1 will have no downside risk and lower levels of reward for the first year, or up to three years for safety net hospitals” — so exposure arrives in the second performance year, not the first. ⛔ TEAM reaches knee, hip, ankle and spinal fusion — not shoulder, foot or hand.
Agent governanceAn inspectable, per-claim rationale with a human accountable at escalation
An outcome record — collected on a schedule, risk-adjustable, signed by a named clinician, hash-anchored so the signature is checkable by anyone, portable enough to outlive the vendor that collected it, and machine-readable so a routing layer can select on it.

That is the asset. Not the model, not the app, not the device. Four different pressures converge on one artifact, which is what distinguishes a strategy from a list of initiatives.

7. The stack, as components of that one artifact

LayerIts job in the artifact
chanioContext. The corpus stays on the machine. The best clinical corpus is precisely the one that cannot be uploaded — so local-first is a working requirement, not a preference, and the method survives the model of the month.
SurgeonValueProduction. The surgeon-side line that generates the record as a by-product of work already being done — capture, coding audit, prior auth, registry, PROMs, monitoring.
ClinicalSwipeSupply of signatures. Named licensed physicians who review and sign. ACCESS turns the named Clinical Director into an entry requirement, which turns signature supply into a market.
HarnessHealthThe gate. Every clinically meaningful output is held as DRAFT until a named physician attests it. No code path sends an unsigned draft to a patient, payer, or record.
HashCareProof. A SHA-256 fingerprint anyone can verify. This is what makes the record portable — the attestation remains checkable after the vendor that produced it is gone.
JointCoach · MSKvalueThe MSK instance. The patient-side collection and the episode economics for the one track ACCESS names.
Kinematic JointThe clinical grounding. What the measurements mean, and why a patient-specific reference beat a universal one in three subspecialties independently.

Read down that column and the estate is not a portfolio. It is one production line for one artifact, and each site is a station on it.

7b. Managed episodes and the question of whose record it is

Capable companies already sell episode management into exactly these models — coordination and analytics across TEAM, CJR-X and ASM, patient-reported outcome collection taken on at risk, weekly cost tracking written back into the hospital's EHR, virtual therapy networks, digital pre- and post-operative programmes. That is real work, competently done, and worth paying for. A practice facing mandatory downside risk with no infrastructure should look hard at it.

We build a different thing, and the difference is worth stating precisely rather than competitively.

If a vendor collects the outcomes and reconciles the episode, the outcome record is theirs — in their schema, in your EHR instance, under a contract. That is fine while the relationship lasts.

But the forces in section 6 all end at the same place: a public directory, risk-adjusted, with your NPI on it. When outcomes are published against your name, you want to have signed them yourself, and to hold a copy that stays verifiable after any vendor relationship ends.

That is the entire reason the attestation is hash-anchored and the record is portable. Not because managed services are wrong — because an attestation you cannot verify without the vendor is not portable, and a record you did not sign is not yours.

Both can be true at once. Many practices should buy episode management and keep their own attested, portable copy of what was signed in their name.

7c. The way this fails in practice, which is not technical

Everything above is a claim about payment mechanics and machine-readability. None of it survives contact with an organisation that has no owner for the pathway.

The clearest statement of this we have seen came as a comment on someone else's post, from a health-system chief strategy officer, about a large hospital innovation programme:

“The number nobody publishes is how many of those [companies] are in routine clinical use today. Graduating a cohort and deploying are different outcomes. A years-long track only closes that gap if a service line owns the pathway and its budget, rather than the innovation office holding both.”

That is the same missing denominator that shows up everywhere in this field. Programmes publish the numerator — companies enrolled, pilots run, models deployed — and not the denominator, which is how many reached routine use and what happened to the rest. An absence of published failures is not evidence of none.

Applied here, honestly: an attested outcome record is worthless if no service line owns producing it, and no amount of correct reasoning about ACCESS, ASM or routing layers changes that. The binding constraint is almost never the technology, and we would rather say so than sell around it. Ask any vendor, including us, what the denominator is.

8. What would prove this wrong

An argument that cannot fail is not an argument. Five things would break it:

  1. The routing layer stays marginal. If AI-mediated referral never reaches meaningful volume, the selection pressure never arrives. This is a forecast, not a measurement — we say “expected,” never “is,” and nobody has clean numbers, us included.
  2. CMS does not publish the directory, or publishes it so coarsely that nothing is comparable. The published-outcomes promise is the load-bearing half of Exhibit B.
  3. Risk adjustment fails in MSK badly enough that published outcomes are ignored by referrers as unfair.
  4. Attestation gets automated away — if a regulator accepts machine-only sign-off at scale, the scarcity of the named signature disappears.
  5. Portability turns out not to matter because one incumbent registry becomes the de facto record and nobody needs proof that survives it.

Disclosure. SolvingHealth builds physician-attestation and remote-monitoring infrastructure for orthopaedic practices. This argument favours the category we work in. Every primary source above is public; check them rather than us.

Limits we hold ourselves to. Nothing we build is risk-adjusted today, and nothing we publish ranks any provider. We do not claim a human reviewer makes AI output more accurate. We do not claim FDA clearance or authorisation for anything, and ⛔ nothing in ADVOCATE is FDA-authorised — that is the programme's goal. Payment figures vary by locality and year; none of this is billing or legal advice, and a compliance review is not optional.

Open questions we have not resolved. Whether software-conducted interaction counts toward the twenty-minute management requirement (we doubt it, and it sets the ceiling on the whole monitoring model). The code-level split of the monthly monitoring rate, which is why we publish no blended per-patient figure. Current ACCESS application dates, which must be re-verified against CMS.

Sources read 9–10 September 2026: cms.gov (ACCESS, TEAM, episode models) · arpa-h.gov (ADVOCATE) · telehealth.hhs.gov (remote monitoring billing) · the routing partnership announcement on the vendors' own newsroom. Where a claim did not verify, it is not on this page.

← SolvingHealth