idea-018 · desirability deck
Every slide states a condition that would have to be true, then reports where it stands using the candidate file's own claim and its own confidence tag. A deck never upgrades a tag, invents a number, or recommends anything, and there is no ask slide.
The chair: the user. The adult carrying a long-bone fracture that stopped healing, the orthopaedic trauma surgeon at the visit where the serial radiographs are read, and the DMEPOS supplier's intake staff who have to put the device in a box and defend the paperwork.
How to read this: every slide is a condition, not a conclusion. A bracketed claim reference points into knowledge-base/candidates/idea-018.md — the number is the claim's position in its ## Claims list and the tag is copied from it, never adjusted here. A no-claim marker means nothing in the file speaks to the condition at all. See docs/deck-spec.md.
Would have to be true: A skeletally mature adult would have to sit in an orthopaedic clinic with two sets of radiographs taken at least 90 days apart showing that healing at a long-bone fracture has ceased for three or more months, and there would have to be enough such adults per year for a product line to exist.
Where it stands: The moment is real and is defined in Medicare's own words — NCD 150.2 holds that nonunion of a long bone "is considered to exist only when serial radiographs have confirmed that fracture healing has ceased for 3 or more months", with a minimum of two sets of radiographs, multiple views, 90 days apart [claim 9: verified], and the candidate's intended_use is written to restate that criterion. Skull, vertebral and tumour-related nonunions are outside it [claim 10: verified]. How many such adults exist annually is the file's weakest leg: the population estimate is the Generator's own unsourced 5-10%-of-~6-million figure, recorded as permanently unconfirmed absent a human-attached licensed source [claim 17: unconfirmed], and the Medicare E0747 utilisation count that would floor it could not be looked up in either direction [claim 15: unconfirmed].
What would settle it: The CMS "Medicare Durable Medical Equipment, Devices & Supplies — by Geography and Service" public-use release (data.cms.gov) for E0747 allowed services, which the file names as its own first check and which no connector here can read yet (docs/backlog.md 1.13). The epidemiology leg needs a human to attach a licensed orthopaedic-trauma source; the market connector has already answered unverifiable on it.
If it's false: If the eligible population is an order of magnitude smaller than the Generator's guess, Factor 3 (TAM) falls toward its 2/5 kill floor and the narrowing that makes the regulatory case also removes the market — every slide in the viability deck downstream of population becomes moot.
Would have to be true: The patient would have to put a coil on the limb and keep the field on for the prescribed hours every day across a multi-month course — a daily act they perform zero times today — and the low-profile garment form factor would have to be the reason they keep doing it in month three.
Where it stands: Stated in the file as an assumption and nothing more: the idea depends on patients wearing the device for the prescribed hours and on a garment-integrated coil materially improving adherence over existing form factors, with no adherence data retrieved for any existing bone growth stimulator [claim 29: unconfirmed]. The adoption connector returned unverifiable on it — a patient-behaviour claim needs primary research. The nearest real adherence figure in the file sits in that same claim's verifier note and is recorded there as context rather than evidence: the TRUST trial reported moderate compliance, with 73% of patients administering at least half of recommended treatments, for a different device in a different indication [claim 29: unconfirmed].
What would settle it: Instrumented wear-time data from a first cohort of real nonunion patients — which this device's own onboard log is designed to produce — or published adherence data for the incumbent devices, retrieved by a literature-detail sweep rather than assumed. No connector in this repo answers patient behaviour.
If it's false: The differentiating feature stops differentiating. If wear time is unchanged by form factor, the candidate is commodity PEMF hardware against four entrenched brands, and the Moat factor (already 1/5 on claim 21) has nothing left in it but the adherence log.
Would have to be true: Three separate people would each have to say yes before a single device reaches a patient: the orthopaedic surgeon would have to reach for a bone growth stimulator at that visit at all rather than book revision fixation; having reached for one, they would have to write for a new entrant instead of Orthofix, Zimmer Biomet or Bioventus; and a DMEPOS supplier would have to add it to a catalogue built around those same incumbent brands.
Where it stands: The surgeon's first choice — stimulator versus revision surgery — is nowhere in the candidate file; this deck put it on record [claim 31: unverified]. The second choice is in the file as an explicit assumption: the idea depends on orthopaedic trauma surgeons prescribing a new entrant over incumbents with decades of familiarity and established DMEPOS distribution [claim 28: unconfirmed], returned unverifiable by the adoption connector. The supplier leg was likewise unrecorded and is now a claim [claim 32: unverified]. Those incumbents are named only on the Generator's background knowledge [claim 19: unconfirmed] — the 510(k) lookups that came back refuted were answering a different question, since these devices are PMA-marketed and this repo has no PMA or registration-and-listing connector.
What would settle it: Structured interviews with orthopaedic trauma surgeons who manage nonunion, and with two or three DMEPOS suppliers who already dispense E0747 — a paid primary-research line item, as claim 28's verifier pass says in terms. The partial proxy is the CMS DMEPOS utilisation release showing how often incumbents are actually dispensed, which is blocked on the same missing file as claim 15.
If it's false: If surgeons route these patients to revision fixation, the addressable population is not the nonunion incidence but the much smaller non-operative subset, and Slide 1's denominator is wrong. If suppliers will not stock it, there is no route to the patient at all — in the DMEPOS channel the supplier, not the manufacturer, is the seller of record.
Would have to be true: The patient would have to accept three costs: waiting out the documented three-month cessation window and a second radiograph set before they are eligible at all; wearing the coil daily for the course; and paying whatever their share of the device comes to.
Where it stands: The waiting cost is not a design choice and is verified — the coverage criterion itself requires at least two radiograph sets separated by a minimum of 90 days with healing ceased for three or more months [claim 9: verified], and the candidate's intended use deliberately restates it. The money cost cannot be stated: no claim in the file carries a price. The DMEPOS purchase allowance is an unsourced "low thousands" guess [claim 16: unconfirmed], and whether the device is even paid in its own right rather than absorbed into another payment is unchecked in either direction [claim 23: unconfirmed]. That the patient accepts the resulting cost share is an assumption this deck put on record [claim 33: unverified].
What would settle it: The CMS DMEPOS fee schedule public-use file for E0747 — noting the file's own caveat that amounts vary by state and by purchase-versus-rental modifier, so one number may not exist — plus the beneficiary cost-share rules for the DME benefit. Then a survey of patients at that price point.
If it's false: If out-of-pocket cost is the binding constraint, the adherence assumption never gets tested, because the device is never dispensed. It also shifts the whole desirability case onto whatever the supplier can bill, making Slide 2 of the viability deck load-bearing here as well.
Would have to be true: The prescribing surgeon would have to be able to point to at least one controlled study of PEMF in established nonunion — not fresh fracture — before putting a patient on a multi-month course with an unfamiliar brand.
Where it stands: The file records this as a named hole rather than a claim of evidence: the published PEMF-in-established-nonunion literature was not retrieved in either direction [claim 25: unconfirmed], though the verifier's sweep found the field is not empty — 80 matches including a 2024 systematic review update — with none read or appraised, so none of it carries a tag. What is verified points the other way and a sceptical surgeon will raise it: the TRUST trial, 501 patients across 43 North American trauma centres, found LIPUS after tibial fracture fixation "does not accelerate radiographic healing and fails to improve functional recovery" [claim 24: verified]. The file is explicit that this is a different modality in a different indication — the fresh-fracture indication this candidate gives up — and not evidence about PEMF in nonunion. The proposition the prescriber would actually want is stated on the feasibility deck [claim 40: unverified].
What would settle it: A literature-detail sweep over the leads in claim 25's verifier note (PMID 39387850, 42270187, 41244766 among them), appraised for whether they address established nonunion as distinct from fresh fracture. A few hours, free, and the cheapest unclosed item in the file.
If it's false: If the nonunion literature turns out to be as negative as the fresh-fracture literature, Factor 6 (mechanism and clinical risk) drops rather than holds at 2/5, and the narrowing loses its clinical rationale as well as its commercial one — the indication would be narrow and unsupported.
Would have to be true: In a first handful of real users, a surgeon who has just read a nonunion radiograph would have to choose the device unprompted, and the device's own log would have to show field-on minutes tracking the prescribed schedule through month three rather than decaying after week two.
Where it stands: Both halves are unbacked. The prescribing choice rests on [claim 31: unverified] and [claim 28: unconfirmed]; the wear-time half on [claim 29: unconfirmed]. No connector reaches either, and both verifier passes said so in the same words: this needs primary research, and a model's own estimate of stakeholder behaviour would be a fabricated claim.
What would settle it: Twenty structured prescriber interviews at nonunion-managing trauma centres, and a wear-time observation in the first dispensed cohort. Neither is a lookup; both are Stage 7 customer-discovery work items, and this slide is their specification.
If it's false: Nothing downstream survives it. Every other desirability slide assumes the device gets prescribed and worn; if the first cohort shows neither, the remaining conditions are untestable rather than false.
Every condition above with nothing verified behind it — the [no claim] markers, plus the unverified claims this deck itself put on record. Read this slide first.
adoption returned unverifiable; primary research only.
literature-detail sweep over claim 25's own leads.
Six of the eight rest on what a person would do, and no connector in this repo answers that. Note the shape as well as the length: the rubric scores no factor for whether a user wants the thing, so on this candidate the one verified line on the whole deck (claim 9) describes a coverage criterion, not a preference.
If only one thing from this chair could be checked: Whether an orthopaedic trauma surgeon, at the visit where the serial radiographs confirm an established nonunion, reaches for a bone growth stimulator at all — and then for this one [claim 31: unverified], [claim 28: unconfirmed].
This one rather than the others because it sits upstream of all of them. Adherence (Slide 2), cost tolerance (Slide 4) and evidence appetite (Slide 5) are only testable on patients who have been prescribed the device, and every one of those patients arrives through this single decision by someone who has prescribed Orthofix or Zimmer Biomet for years. It is also the condition the file is quietest about: claim 28 records the brand-switch half as an assumption, and until this deck the prior question — stimulator or revision surgery — was not on record anywhere.
Naming it is not a recommendation, a gate, or a kill. It is the answer to "check what first."