Feasibility and sizing

Kicker

Three of this Inquiry's load-bearing recommendations were never tested, and all three change shape when you test them. Pharmacy routing is two products, not one: the fill-outcome half is buildable now with the vendor MEDvidi already runs — DoseSpot's published API carries a per-clinic HasRxFill flag, a Surescripts medication-history feed with LastFillDate, an IsDrugAvailable / CURRENTLY_UNAVAILABLE transparency endpoint, and MEDvidi's own portal already models FullFill / PartialFill / NoFill — while the store-level stock half does not exist as an obtainable feed for any prescriber in the United States and will not, because retail inventory lives in the pharmacy's own ERP. The DEA cliff is smaller and harder than the study says: smaller because the cohort that has to be converted before 1 January 2027 is roughly 13,600 patients, which is about five clinician-days each across a 117-person panel; harder because the statute has no referral cure — 21 U.S.C. §829(e) requires that the prescribing practitioner personally conduct the in-person evaluation, so CVS MinuteClinic, urgent care and a PCP referral letter are all legally worthless here, and the 2023 "qualifying telemedicine referral" MEDvidi once shipped was a proposed rule that was never finalised. The permanent cure attaches to a practitioner–patient pair, which means every 1099 clinician who quits re-orphans their entire cured panel — a liability nobody in this study has priced. And the payer floor is now a number, not a shrug: Medicare's published national non-facility allowed amounts for 2024 are $171.35 for a psychiatric diagnostic evaluation (90792) and $119.71 for a moderate-complexity established visit (99214), and nurse practitioners are paid 85% of that by regulation — so MEDvidi's blended $169.51 cash visit becomes about $107.68 net if it takes insurance with an NP-heavy panel, meaning volume must rise 1.53× just to stand still.

Source legend

---

1. What we found

1.1 "Pharmacy fill data" is four different objects, and only three of them exist

Two Findings — A1 §6 and A6 §6 — nominate real-time pharmacy routing as the highest-leverage unclaimed build in the company, described as "which pharmacy near this patient both stocks this drug and accepts telehealth prescriptions" [A6]. That sentence bundles four separate data objects with four separate answers [REASON]:

The study's recommendation is written as if it needed #4. It does not. #1, #2 and #3 together produce a ranked pharmacy list with observed success rates, which is the product; #4 would only make it deterministic. Conflating them is why nobody checked — the whole idea looks impossible if you start at the hardest quarter of it.

1.2 DoseSpot's own published API surface, read from the source

DoseSpot ships one Swagger 2.0 specification per subscription plan at my.dosespot.com/webapi/v2/swagger/docs/<Plan>, publicly and without authentication [DS-API]. The seven plans differ materially:

Endpoint count by plan (2026-08-19)                                      [DS-API]
  Full + EPCS   155      Hybrid + EPCS  128      Jumpstart      113
  Full          141      Hybrid         117      Jumpstart+EPCS 124      ReadOnly  53

Which plan MEDvidi holds is inferable rather than stated. POST /api/patients/{patientId}/ prescriptions/sendEpcs appears in Full + EPCS only; GET /api/patients/{patientId}/narx (the PDMP NarxCare report) appears only in the three EPCS plans [DS-API]. A1 found the EPCS prescription states EpcsSigned, EpcsError and ReadyToSign in MEDvidi's clinician bundle, a Prescription Issue Resolution module with PDMP in the loop, and a hand-built "Doctor Cabinet" UI rather than an embedded vendor iframe [A1]. A hand-built UI is DoseSpot's Full Integration model, which its own site describes as "access to over 250 DoseSpot API calls" and "create screens and workflows from scratch" [DS-WEB]. Taken together, MEDvidi is on Full + EPCS — the top plan, and the only one that carries every endpoint discussed below [REASON].

That matters because the following are already inside MEDvidi's contract, not a purchase decision:

[DS-API]

1.3 RxFill exists, is named, and MEDvidi can already read whether it is switched on

The single most decisive artefact in this Area is a boolean. DoseSpot's Clinician model contains a nested object ClinicianClinicErxSummary with exactly these fields [DS-API]:

ClinicianClinicErxSummary                                                [DS-API]
  ClinicId  HasNewRx  HasRefills  HasRxChange  HasCancel  HasRxFill  HasEpcs  HasEpa

HasRxFill is a per-clinician, per-clinic capability flag. Its presence means RxFill is a supported DoseSpot transaction that is enabled or disabled per clinic — and that MEDvidi's own account can be queried today to find out whether it is on. That is a ten-minute internal API call, not a research project.

RxFill is the NCPDP SCRIPT dispense notification: the pharmacy telling the prescriber what happened to the script. Surescripts names it explicitly on its own e-prescribing product page, read 2026-08-19 [SS]:

"In-workflow tools help pharmacists and prescribers collaborate efficiently using NCPDP-defined transactions, including NewRx, RxFill, RxChange, RxTransfer, RxRenewal, NewRxRequest and CancelRx."

and, in the EHR-vendor workflow description:

"The patient picks up the prescription, and the prescriber is notified. The RxFill transaction provides certainty that the patient has picked up their medication."

The receiving side is already modelled inside MEDvidi. A1 found the app's fill-outcome enum — FullFill, PartialFill, NoFill, Active, Discontinued, Completed, CancelRequested, CancelPending, Cancelled, CancelDenied, Changed — surfaced to patients as "Medication filled / partially filled / not filled" [A1]. Those are DoseSpot's MedicationStatus values verbatim, and they appear in the spec in two places: as the payload of POST /patients/{id}/prescriptions/{rxId}/medicationStatus (a write), and as a field on PrescriptionLog (a read) [DS-API]. So the schema supports both a human marking the outcome and a feed populating it. Which one is actually populating MEDvidi's records is the question that decides whether §1.8's verdict is "already have it" or "one contract away" — and it is answerable from the inside in an afternoon [REASON].

1.4 Surescripts sells the productised version of exactly this

If RxFill is not enabled, the outcome data is still purchasable as a finished product. Surescripts First-Fill Abandonment, read on its own product page 2026-08-19 [SS]:

"Surescripts First-Fill Abandonment proactively monitors new prescriptions sent across the Surescripts network and flags when patients don't pick them up. Care teams get the data they need, delivered at the frequency they select…"

Its published specification, verbatim from the FAQ [SS]: the payload is "information on the patient, provider, medication, dose, route of administration, date written and days unfilled" plus trend data "across clinics, prescribers and medications"; the cadence is "daily or weekly … of your choosing"; the thresholds and abandonment definition are customer-configurable; the look-back is 12 months with continuous monitoring; and it is sold to EHR vendors, health systems and healthcare analytics vendors. Surescripts calls it "the only data product on the market that proactively identifies when new prescriptions are abandoned."

Surescripts' own framing numbers on the same page: "27% of new prescriptions are not filled due to payer rejections and abandonment by patients" and "98 million new prescriptions were abandoned by patients at pharmacies in 2023" [SS].

Separately, Medication History for Ambulatory returns "12 months of detailed, accurate medication history from pharmacy benefit managers (PBMs) and pharmacies", explicitly including "pickup date" and "cash-pay prescriptions", and its FAQ says the data is "an accurate list of medications that have been purchased by the patient… from pharmacy fill data as well as pharmacy benefit plan claims" [SS]. Surescripts reports 3.79 billion medication histories delivered in 2025 [SS]. DoseSpot already resells this as GET /patients/{id}/medications/history, described in DoseSpot's own spec as "a patient's medication history according to PBM insurance claims submitted to the Surescripts ePrescribing network… A patient must give consent to retrieve this data" [DS-API].

Note the asymmetry in what Surescripts publishes: 2.64 billion e-prescriptions filled in 2025 and 32.9 million RxChange transactions in 2025 — but no RxFill volume anywhere [SS]. A network that publishes one transaction's volume and not another's is telling you which one is universal and which one is not.

1.5 Store-level stock: not obtainable, and the reason is structural

No pharmacy chain publishes an inventory API to prescribers. The systems that hold store-level stock are the pharmacy's own supply-chain platforms — Tecsys, Oracle NetSuite, Microsoft Dynamics 365 Supply Chain and the chains' internal dispensing systems — all of which sit on the pharmacy side of the network boundary and none of which expose a prescriber-facing endpoint [REASON], supported by a 2026 survey of the pharmacy inventory-software market that lists exactly these as the category and describes them as intra-organisation visibility tools. Nothing in the NCPDP SCRIPT transaction set carries inventory: the transactions are NewRx, RxFill, RxChange, RxTransfer, RxRenewal, NewRxRequest and CancelRx [SS], and none of them is a stock query.

The one availability signal that does exist inside the prescribing workflow is DoseSpot's transparency endpoint, and its shape tells you its scope [DS-API]:

GET /api/patients/{patientId}/transparency/alternativePharmacies?nDC=<ndc>
  -> PharmacyAlternative[]
       IsDrugAvailable      : boolean
       NotDispensableReason : "CURRENTLY_UNAVAILABLE" | "DOES_NOT_DISPENSE"
       Pharmacy             : { NCPDPID, StoreName, Lat, Lng, ServiceLevel, … }
       PharmacyOffers       : [{ OfferType, Price, Quantity, EstimatedDelivery }]
         OfferType ∈ { "Cash", "InsuranceEstimate", "Prime", "RxPass" }

Prime and RxPass are Amazon Pharmacy programme names, and EstimatedDelivery is a mail-order field. So this is a digital/mail pharmacy marketplace feed, almost certainly not a retail chain-store stock feed [REASON]. It is still the closest thing to a live availability signal in any e-prescribing API, and DOES_NOT_DISPENSE is a genuinely useful per-pharmacy per-drug flag. Its coverage for Schedule II stimulants at retail is unverified from outside and is question 4.1 below.

1.6 The supply signal that is free, public and machine-readable — and the numbers are bad

openFDA publishes the FDA drug-shortage database as a JSON API at package-NDC granularity, with a per-record availability field. Queried 2026-08-19 [FDA]:

Share of presentations not fully available, 2026-08-19                   [REASON] on [FDA]
  Mixed amphetamine salts  (19 + 12) / 73  = 42.5%   ; hard-unavailable 19/73  = 26.0%
  Lisdexamfetamine         (2 + 34) / 108  = 33.3%   ; hard-unavailable  2/108 =  1.9%
  Methylphenidate          (4 + 8)  / 38   = 31.6%   ; hard-unavailable  4/38  = 10.5%

Two things follow. First, the Adderall-generic shortage is in its forty-seventh month — posted 2022-10-12, still Current, with records reverified as recently as 2026-08-17 [FDA]. This is not a 2023 news story; it is the operating environment MEDvidi is in right now, and it is the reason its own FAQ has a page called /faqs/pharmacy-issues/ [A1].

Second, the actionable unit is not the store, it is the NDC. A prescriber who knows that a specific strength from a specific manufacturer is Unavailable nationally can write a different strength, a different manufacturer's product, or a different molecule before the script goes out. That is a decision-support feature that costs one nightly job against a free federal API, needs no vendor contract, and nobody in this category ships it [REASON].

1.7 The variable that actually decides whether a stimulant script gets filled is not stock

In May 2022, CVS Health — the largest US pharmacy chain — stopped filling controlled-substance prescriptions issued through Cerebral and Done Health. Walmart did the same, and individual CVS and Walgreens locations plus some grocery-chain pharmacies blocked those prescribers [PRESS-RX]. The block was per prescribing organisation, not per drug and not per store.

That reframes the whole product. The routing question a MEDvidi patient actually faces is:

1. Will this chain fill a script from this practice? — a corporate policy, binary, changeable overnight, and knowable only from MEDvidi's own observed outcomes [REASON].
2. Will this individual pharmacist accept a telehealth-issued CII? — MEDvidi's own FAQ already concedes this "varies from state to state" [A1].
3. Is the molecule/strength in supply? — free from openFDA [FDA].
4. Does this store have it on the shelf today? — unobtainable.

Only (4) is impossible, and it is the least important of the four. (1) and (2) are the ones that produce "day 13, three different pharmacies, paid $195 for nothing" [A1], and they are learnable only from MEDvidi's own history — which is the moat argument, restated correctly: the asset is not a data feed anybody can buy, it is a longitudinal per-chain, per-store, per-state, per-drug success table that only a practice with 180,000 encounters a year can accumulate [QUANT] [REASON].

1.8 Verdict on the pharmacy build

Buildable now, in three tiers, with no vendor contract required for the first two.

The workaround path gaps.md asked about — deriving fill success from MEDvidi's own tickets and patient-reported outcomes — is not a workaround. It is tier 0, and it is the best tier, because it is the only one that captures per-chain policy toward MEDvidi specifically. What it takes: the support tickets are already typed (PHARMACY_ISSUE document type, an IVR deep-link on option 6, a clinician-side Prescription Issue Resolution module [A1]), so the missing work is (a) a structured outcome field on the ticket — chain, NCPDP ID, refusal reason, resolution — replacing free text; (b) a patient-side "did you get it?" prompt fired N days after eRxSent, which the portal's existing notification machinery already supports; and (c) a nightly job that folds both into a ranked table keyed on NCPDPID (which pharmacies/search already returns) [DS-API] [A1] [REASON]. That is one PM, one engineer, one analyst, and roughly a quarter — not a platform programme.

The single question that would settle the rest: read Clinician.ClinicErxSummaries[].HasRxFill for MEDvidi's clinics in the DoseSpot API. If it is true, MEDvidi is already receiving dispense notifications and has been discarding them. If false, the ask to DoseSpot is one line.

---

1.9 What 1 January 2027 actually requires — read from the statute, not the FAQ

MEDvidi's own patient-facing page says the flexibilities are extended through 2026-12-31 [A4]. Verified independently and it is correct, but the instrument matters. 90 FR 61301, published 2025-12-31, effective 2026-01-01, expiring "at the end of the day December 31, 2026" [REG-FR4]. As of 2026-08-19 there is no fifth extension and no final Special Registration rule: a Federal Register query for all DEA documents published since 2026-01-01 mentioning "telemedicine" returns exactly one hit, an unrelated MAT dispensing rule at 91 FR 34754; the current public-inspection list contains 87 documents and zero from DEA [REG-FRQ]. 134 days remain; 19.1 weeks [REASON].

The permanent cure, in DEA's own words in that rule [REG-FR4]:

"Once a practitioner has conducted at least one in-person medical evaluation of a particular patient, the specific requirements of the Ryan Haight Act related to remote prescribing of controlled substances no longer apply to that specific practitioner-patient relationship. This permits the practitioner to remotely prescribe controlled substances to that patient indefinitely, regardless of how much time has passed since the initial in-person medical evaluation or whether that evaluation was for a separate medical concern…"

Now the part the study got wrong. The statute reads [REG-RH]:

§829(e)(2)(A) "The term 'valid prescription' means a prescription that is issued for a legitimate medical purpose in the usual course of professional practice by— (i) a practitioner who has conducted at least 1 in-person medical evaluation of the patient; or (ii) a covering practitioner."

§829(e)(2)(B)(i) "The term 'in-person medical evaluation' means a medical evaluation that is conducted with the patient in the physical presence of the practitioner, without regard to whether portions of the evaluation are conducted by other health professionals."

"The practitioner" is the prescriber. A referral does not cure §829(e). A visit to CVS MinuteClinic, an urgent care, a PCP, or a mobile-nurse vendor does nothing whatsoever for a MEDvidi clinician's authority to prescribe a stimulant to that patient. The "qualifying telemedicine referral" MEDvidi published in May 2023 [A4] came from the March 2023 NPRMs, which drew 38,369 comments and were never finalised [A4] [REG-FR4]. Rebuilding that flow as designed would rebuild a compliance artefact for a rule that does not exist.

The seven statutory exceptions at §802(54) are, verbatim in structure [REG-RH]:

There is also the covering practitioner doctrine at §829(e)(2)(C): a practitioner may prescribe without an in-person evaluation if acting "at the request of" a practitioner who has conducted an in-person evaluation or an evaluation through the practice of telemedicine within the previous 24 months, and who "is temporarily unavailable" [REG-RH]. A CEO will reach for this — "can one doctor do the in-person and the rest cover?" The honest answer is no as a routing model: the doctrine is conditioned on the primary's temporary unavailability, and using it as the standing assignment mechanism for a 117-clinician pool would be a sham. Whether the "or telemedicine within 24 months" clause preserves anything for visits conducted before 2027-01-01 is a real question for counsel and is parked at 4.4.

1.10 Sizing the cohort

The inputs, and what each is worth:

INPUTS                                                                    tag
  2026 run-rate visits            180,000/yr   medvidi.ai "15,000+ monthly
                                               encounters", adjudicated    [QUANT]
  Visits per patient-year         3.43         120,000 / 35,000, company's
                                               own figures                 [A3]
  ADHD return rate "RR1"          41% -> 53%   MEDvidi's own roadmap       [A5]
  Served states                   35           AZ CA CO CT FL GA ID IL IN KS
                                               KY ME MD MA MI MS MO MT NE NV
                                               NH NM NY NC ND OH OR PA TN TX
                                               VT VA WA WI WY               [A4]
  Clinician panel                 50 or 117    contested; both used         [A3][A5]

STEP 1 — annual treated patients                                       [REASON]
  180,000 / 3.43 = 52,478  -> 52,500

STEP 2 — share on a controlled substance, and share on Schedule II
  Not published anywhere. Modelled three ways, central case argued below.
    low      50% controlled x 70% of those CII = 35.0%  -> 18,400 CII patients/yr
    central  65% controlled x 75% of those CII = 48.8%  -> 25,600 CII patients/yr
    high     80% controlled x 80% of those CII = 64.0%  -> 33,600 CII patients/yr

  Why 65/75 is the central case: 28 of the 32 programmatic condition-state pages
  are ADHD [A2]; the drug-name post inventory is stimulant-led [A2]; insomnia
  (zolpidem, eszopiclone) and anxiety (benzodiazepines) add Schedule IV;
  weight-loss GLP-1s, ESA letters, SSRIs and non-stimulant ADHD drugs are not
  controlled at all. This is an ESTIMATE and the largest single uncertainty
  in this section.

STEP 3 — point-in-time active Schedule II base
  Not every annually-treated patient is active on 1 Jan 2027. MEDvidi's own
  published ADHD return rate moved 41% -> 53% [A5]; use 53% as the share who
  form a continuing relationship.
    25,600 x 0.53 = 13,559  ->  ~13,600 active Schedule II patients
    range at 41% / 65% return: 10,500 – 16,600

~13,600 patients need an in-person evaluation by their own MEDvidi prescriber before 2027-01-01, or their next stimulant script is unlawful.

STEP 4 — rate                                                          [REASON]
  13,559 / 19.1 weeks = 710 patients per week, nationally
  13,559 / 117 clinicians = 116 patients per clinician
  13,559 /  50 clinicians = 271 patients per clinician
  At 24 in-person exams per clinician-day (8 h at 20 min):
     116 / 24 =  4.8 clinician-days per clinician   (117-panel)
     271 / 24 = 11.3 clinician-days per clinician   (50-panel)

That is the number to say out loud in the interview: five to eleven days of in-person clinic per clinician permanently immunises the entire book. Not a programme. A fortnight.

1.11 Where those patients are

Distributing the active CII base across the 35 served states in proportion to 2024 state population [CENSUS] — a proxy, since MEDvidi does not publish state mix, and one that probably understates California and New York because those are the states with the oldest and best-ranked programmatic pages [A2] [REASON]:

Concentration                                                          [REASON] on [CENSUS]
  Top 5 states  = 44.1% of served population -> 5,980 patients
  Top 10 states = 63.8% of served population -> 8,651 patients
  Bottom 15 states (NM NE ID NH ME MT ND VT WY + 6) < 5% combined

So a top-five-state programme covers 44% of the exposure for 249 clinician-days if every invited patient attends — thirteen clinician-days a week across five states, or 112 days at the 45% attendance rate assumed in §1.13 [REASON]. Wyoming, Vermont and North Dakota together are about 200 patients and should be written off rather than served.

1.12 Is the referral supply there? Wrong question — here is the right one

Because a third party's in-person visit cures nothing (§1.9), the retail-clinic supply question gaps.md posed does not decide anything. For completeness, that channel has been contracting anyway: Walmart shuttered all 51 Walmart Health centers, Walgreens closed 160 VillageMD locations and CVS closed MinuteClinic sites including 25 in Los Angeles, all announced in 2024, leaving MinuteClinic (~1,100 sites) as effectively the only national retail clinic network [PRESS-RX]. Even at full strength it would not have helped.

The three things MEDvidi can actually buy:

Option A — MEDvidi's own clinicians hold in-person clinic days. The only option that cures §829(e) directly. Requires: rented clinical space in a metro, the clinician physically present, the clinician licensed and DEA-registered in that state (which they already are, since they treat patients there [A5]), and enough patient density to fill a day.

Option B — originating-site partnership under §802(54)(A). The patient physically attends a DEA-registered hospital or clinic and the MEDvidi clinician appears by video from anywhere. This is the classic telehealth hub-and-spoke and it is permanent — no in-person evaluation by the prescriber is required at all, for any visit, as long as the patient is at a registered site. It is also operationally heavy: it converts every future visit into a scheduled trip, which destroys the product. Worth building for the subset that will never attend a one-off exam, not for the base.

Option C — hybrid in-state clinicians. Recruit or convert clinicians who hold both a telehealth panel and physical practice space, so that the in-person evaluation happens in their own office and the pair is cured permanently. This is the cheapest per patient and the slowest to stand up.

What an in-person clinical encounter is worth, as a market anchor: Medicare's published national non-facility allowed amount is $105.33 for 99203 and $160.28 for 99204 — that is what the American health system pays for a 20-to-45-minute in-person new-patient visit, inclusive of beneficiary cost-sharing [CMS-PFS]. Cash retail runs above it. So $85–$215 is the defensible band for buying an in-person encounter, and MEDvidi's own contractor rate card is $30–$60 per visit [A3].

1.13 Cost per preserved patient

PROGRAMME MODEL — top-5-state cohort                                    [REASON]
  Cohort invited                              5,980   (§1.11)
  Attendance rate assumed                       45%   ESTIMATE. These patients
                                                      chose MEDvidi to avoid a
                                                      waiting room. Range 30–60%.
  Patients attending                          2,691
  Throughput                             24 / day     8 h at 20 min
  Clinician-days required                       112

  Loaded clinician-day cost:
    clinician time     24 visits x $50            = $1,200   [A3] rate card top
    space              exam room, per diem        =   $250   ESTIMATE
    travel/lodging when out of home metro         =   $300   ESTIMATE
                                            low $1,200  central $1,750  high $2,200

  Clinic cost         112 x $1,750                     = $  196,204
  Build + operate the flow (eng, ops, comms, 1 qtr)    = $  300,000   ESTIMATE
  TOTAL                                                 = $  496,204

  Cost per preserved patient  496,204 / 2,691           = $   184.41
     at $1,200/day: $161.49    at $2,200/day: $203.16

WHAT A PRESERVED PATIENT IS WORTH                                       [REASON] on [A3]
  Forward-year revenue, follow-ups only:  2.43 x $159 = $386.37
  Full patient-year (A3's figure):        195 + 2.43 x 159 = $581.37
  Payback on $184.41 at $386.37/yr = 5.7 months
  Payback on $184.41 at $581.37/yr = 3.8 months

WHAT DOING NOTHING COSTS                                                [REASON]
  Current ARR (company's own, 2026-05-05)                    $30,000,000  [QUANT]
  Central estimate: Schedule II patients = 48.8% of base        (§1.10)
  Revenue attributable to the Schedule II line             ~$14,600,000/yr
  This is not a one-time loss: after 2027-01-01 no new telemedicine-only
  Schedule II patient can be started at all, so the line does not decay,
  it stops.

A $0.5M programme that protects a $14.6M line is not a close call. The honest caveats: the 48.8% share is the study's weakest number, the 45% attendance rate is a guess, and both should be replaced with internal data on day one.

1.14 The consequence nobody has priced: the cure attaches to a contractor

Because the exemption is per practitioner–patient pair [REG-FR4], an in-person evaluation is an asset that belongs to the clinician, not the clinic. MEDvidi's clinicians are 1099 contractors [A3]. Three consequences, none of which appears in any pass-1 Finding:

1. Routing dies. A1 found a booking flow that offers "another provider" when the preferred one is unavailable [A1]. After the cure, reassignment to an uncured clinician makes the next Schedule II script unlawful. Every cured patient must be pinned to one named prescriber, and the product has to enforce it — a hard constraint on the scheduler, the waitlist and the AI receptionist.
2. Churn re-orphans panels. At an assumed 30% annual contractor churn — plausible for 1099 telehealth, unverified — 13,559 × 0.30 ≈ 4,070 patients per year lose their cure and need re-examination in person, forever [REASON].
3. The state-licensure matrix gets a third dimension. Today it is clinician × state. After the cure it is clinician × state × cured-patient set, and adding a state no longer just needs a licence, it needs a clinician willing to travel there.

This is the argument that turns compliance work into a strategy conversation: the cure is a moat only if MEDvidi employs the moat. A company with 13,600 cured pairs held by contractors has a depreciating asset; a company that converts its top prescribers to employment, or builds Option C's hybrid in-state model, owns it.

---

1.15 The payer floor is public, and here it is

A3 correctly refused to guess a commercial rate [A3]. Medicare does not require guessing. The table below is CMS's own published national average allowed amount — Medicare payment plus beneficiary coinsurance and deductible, i.e. total revenue to the practice — for calendar-year 2024 claims, non-facility place of service [CMS-PFS]:

Derived combinations [REASON] on [CMS-PFS]:

  99213 + 90833 = $151.00       99214 + 90833 = $185.34

Two coding notes a CEO will test. First, MEDvidi's 30-minute initial maps to 99204 ($160.28) or, for a psychiatric practice, 90792 ($171.35). Second, MEDvidi's 15-minute follow-up does not reach 99214 on time (which needs 30–39 minutes) but plausibly does on medical decision-making, because prescription drug management is a moderate-risk element and ADHD is a chronic condition — so the realistic follow-up code is 99214 ($119.71) with 99213 ($85.37) as the conservative floor [REASON]. Every number below is shown both ways.

1.16 The 85% haircut, and the second in-person cliff

42 CFR 414.56(c), current text: nurse-practitioner and clinical-nurse-specialist allowed amounts "may not exceed 85 percent of the physician fee schedule amount for the service" [CFR]. MEDvidi's panel is MD, DO, NP, PMHNP and PA [A3], with the company's own historical claim that "50% of healthcare providers are MD" [A3]. So roughly half of every insured visit would be paid at 85%.

Medicare telehealth status, as of August 2026: the Consolidated Appropriations Act, 2026 extended the pandemic-era flexibilities through 2027-12-31; behavioural health telehealth in the beneficiary's home is permanent and not subject to the geographic restrictions; audio-only is permitted for behavioural health where the patient cannot or will not use video [KFF-TH].

And here is the second cliff. Medicare's own statutory in-person requirement for tele-mental-health delivered in the home — an in-person visit no more than six months before the initial telehealth appointment, and annually thereafter — is currently delayed and takes effect in January 2028 [KFF-TH]. That is the same structural obligation as the DEA's, from a different regulator, arriving twelve months later. A MEDvidi that solves the DEA cliff by standing up in-person capability has already solved the Medicare one; a MEDvidi that solves it by exiting Schedule II has not, and would hit it again in 2028 if it went after Medicare volume. This is the strongest single reason to build the in-person capability as a permanent asset rather than a 2026 fire drill. [KFF-TH] is a secondary source and the exact effective date should be confirmed against 42 U.S.C. §1395m(m)(7) before it is quoted to anyone.

1.17 Commercial and Medicaid, relative to that floor

Commercial. Milliman's parity work found that in 2017, across 37 million commercially covered lives, primary-care office visits were reimbursed 23.8% higher than behavioural-health office visits relative to the Medicare fee schedule, and that average behavioural-health office-visit reimbursement remained below Medicare allowed amounts across 2013–2017 [MILL]. That is seven-to-nine-year-old data read through secondary summaries, and it is the weakest evidence in this Finding. Directionally it says the same thing every telepsychiatry operator says out loud: in behavioural health, commercial does not pay a premium to Medicare the way it does in medical/surgical. Treat Medicare allowed as roughly the centre of the commercial distribution for these codes, not the floor [REASON].

Medicaid. The Medicaid-to-Medicare fee index for 2024 [KFF-MCD]:

Medicaid 99214, applying the index                                      [REASON] on [CMS-PFS][KFF-MCD]
  National  119.71 x 0.75 = $89.78    ; NP at 85% = $76.32
  CA        119.71 x 0.67 = $80.21    ; NP at 85% = $68.17
  TX        119.71 x 0.63 = $75.42    ; NP at 85% = $64.10
  Medicaid 90792 national 171.35 x 0.75 = $128.51

A Medicaid follow-up delivered by a PMHNP in Texas is worth about $64 against MEDvidi's $159 cash. Medicaid is not a growth channel for this business at this cost structure; it is a mission decision.

1.18 What taking insurance would do to the model

BLENDED CASH PRICE TODAY                                                [REASON] on [A3]
  Initial share = 1 / 3.43 = 29.2% ; follow-up share = 70.8%
  Blended = 0.292 x 195 + 0.708 x 159 = $169.51
  (A5's $166.20 is the same number computed from a slightly different mix)

BLENDED INSURED EQUIVALENT, same visit mix                              [REASON] on [CMS-PFS]
  90792 + 99214, physician rate   = $134.77  =  79.5% of cash
  90792 + 99213, physician rate   = $110.44  =  65.2% of cash
  90792 + 99214, NP at 85%        = $114.55  =  67.6% of cash
  90792 + 99213, NP at 85%        = $ 93.87  =  55.4% of cash
  99205 + 99214, physician rate   = $147.10  =  86.8% of cash   (best realistic case)

PER-VISIT CONTRIBUTION, cash versus insured                             [REASON] on [A3][CMS-PFS][CFR]
  Cash follow-up   : 159.00 - 35 clinician - 4.91 card (2.9% + $0.30) = $119.09
  Insured follow-up: 99214 x 0.85 = 101.75 - 6% RCM = 95.65 - 35       = $ 60.65
  Cash initial     : 195.00 - 45 clinician - 5.96 card                 = $144.04
  Insured initial  : 90792 x 0.85 = 145.65 - 6% RCM = 136.91 - 45      = $ 91.91
  RCM at 6% of collections is an ESTIMATE at the mid of the common
  outsourced-billing range; it excludes denial write-offs, AR carrying
  cost, credentialing and the coinsurance that has to be collected from
  the patient anyway.

THE BREAK-EVEN                                                          [REASON]
  Cash, net of card processing            = $164.28 per visit
  Insured NP-blended, net of 6% RCM       = $107.68 per visit
  Volume multiple to hold revenue flat    = 164.28 / 107.68 = 1.53x

MEDvidi would need 53% more visits to earn the same revenue. Against that: credentialing in 35 states across a dozen payers (12–18 months, per the company's own history of having had payer relationships in six states and abandoning them in 2023 [A3]), claims and denials infrastructure, prior authorisation on stimulants — which is exactly the class of drug payers prior-authorise most aggressively — accounts receivable replacing negative working capital, and the loss of the $100 no-show fee, which no payer contract permits it to keep [A3].

And in favour: the addressable market stops being "people who will pay $195 out of pocket". That is not a small thing, and it is why a Director of Revenue Cycle Management was hired in July 2026 [A7]. But the arithmetic says a payer strategy is a volume bet at a 35% lower unit price, not a margin improvement, and the honest framing in the room is: which patients do we want that we cannot get today, and is a 1.53× volume lift from them plausible inside two years?

Note the interaction with §1.16 that makes this decision worse than it looks: Medicare's own in-person requirement for home-based tele-mental-health starts in January 2028 [KFF-TH]. A payer-forward MEDvidi inherits two in-person obligations, one from DEA and one from CMS, on a one-year lag.

---

2. Capability / object table

---

3. Reconciliation notes

R1 — A1 and A6 both call pharmacy routing the top build; both are half right. A1 §6 says "pharmacy-level fill-success data already exists in the fill-status model, so the platform can rank pharmacies by observed telehealth-eRx acceptance per state per drug" [A1]. That is correct and this Finding confirms the mechanism down to the DoseSpot endpoints. A6 §6 adds "which pharmacy near this patient both stocks this drug and accepts telehealth prescriptions… because it requires operational data nobody has bothered to accumulate" [A6]. The "stocks" half is wrong — it is not un-accumulated, it is un-obtainable. Say the recommendation as A1 said it, never as A6 said it.

R2 — A4's DEA mitigation depends on a rule that was never finalised. A4 §6 proposes rebuilding the 2023 referral flow as a permanent capability [A4]. The 2023 flow implemented the March 2023 NPRMs' "qualifying telemedicine referral", which drew 38,369 comments and never became law [A4] [REG-FR4]. Under the statute as it stands, the only cure is an in-person evaluation by the prescribing practitioner [REG-RH]. A4's underlying instinct — build it as an asset, not a fire drill — survives intact; the mechanism has to change from "collect referrals" to "run in-person clinic days, and pin the pair".

R3 — A4's "referral graph as an asset" becomes "the cure is held by a contractor". Same insight, sharper form, and it produces a concrete organisational recommendation (convert top prescribers to employment, or recruit hybrid in-state clinicians) that no Finding reached.

R4 — A3 was right to refuse a commercial rate and wrong to park the question. Medicare's published allowed amounts and 42 CFR 414.56 give a hard floor, and Milliman gives a directional ceiling that sits close to that floor for behavioural health. The question A3 parked as "answerable only by a contracted rate" is answerable to within about ±20% from public sources [A3] [CMS-PFS] [CFR] [MILL].

R5 — Panel size 50 vs 117 changes the cliff programme by 2.3×, not the verdict. At 117 clinicians the cure is 4.8 days each; at 50 it is 11.3 days each [REASON]. Both are feasible inside 19 weeks. The contested number [QUANT] C2 does not need resolving to make this decision, which is worth saying in the room.

R6 — the 180,000-visit run-rate, not 120,000. All sizing here uses the adjudicated 2026 run-rate [QUANT]. Using 2025's 120,000 would understate the exposed cohort by a third.

R7 — DoseSpot Connect is adjacent, not the same product. The incumbent vendor already sells a patient-facing SMS surface to "compare prices, find savings, and choose the optimal pharmacy", and claims "50% of first fills are ready for pickup within three hours" [DS-WEB] — vendor marketing on an unstated population, not a MEDvidi baseline. It does mean tier 0 has a buy option, not only a build option.

---

4. Open Questions / Parked

1. Is HasRxFill true for MEDvidi's clinics? Settled by: one authenticated read of Clinician.ClinicErxSummaries[] in MEDvidi's own DoseSpot account. This is the single highest-value question in this Finding.
2. Is MedicationStatus on MEDvidi's prescriptions populated by a feed or by staff? If staff, the fill-outcome dataset is a manual artefact of unknown coverage. Settled by: distribution of MedicationStatus by User field on PrescriptionLog over 90 days.
3. Does DoseSpot's transparency/alternativePharmacies return anything useful for Schedule II stimulants at retail, or only for mail/digital pharmacies? The Prime/RxPass offer types suggest the latter [DS-API]. Settled by: one call for a common MAS NDC against a real patient ZIP, or one question to the DoseSpot account manager.
4. Does §829(e)(2)(C)'s "or an evaluation of the patient through the practice of telemedicine, within the previous 24 months" preserve a covering-practitioner pathway for pairs formed during the flexibility window? A genuine question of statutory reading with material consequences. Settled by: outside counsel, before November.
5. Is a fifth extension or a final Special Registration rule at OMB right now? Not established. reginfo.gov refused connections from this environment on 2026-08-19 and the check was not completed. Settled by: the RIN 1117-AB40 entry on reginfo.gov's pending-review list, checked weekly from now to December. DEA sent the fourth extension to OMB 51 days before publication [A4], so an OMB appearance in early November is the signal to watch.
6. What share of MEDvidi patients hold a Schedule II prescription? The 48.8% central estimate drives every number in §1.10–1.13 and has no source. Settled by: one query on the prescription table grouped by Schedule.
7. What share of patients will attend an in-person visit? The 45% assumption drives cost per preserved patient. Settled by: a 500-patient test in one state in September.
8. What is MEDvidi's actual state mix? §1.11 uses population weights. Settled by: visits by state, one query.
9. What is the contractor churn rate on the clinician panel? Drives the recurring re-cure burden in §1.14. Settled by: HR.
10. Would MEDvidi's payer mix even be commercial? A cash-pay ADHD practice's patients skew employed and insured, but the ones who would arrive because of insurance may skew Medicaid, where the rate is $64–$90 [KFF-MCD]. Settled by: a payer-mix survey on the existing intake, which already asks about insurance for the pharmacy benefit [A3].
11. Exact effective date and statutory cite for the CMS tele-mental-health in-person requirement. [KFF-TH] is secondary. Settled by: 42 U.S.C. §1395m(m)(7) and the CY2026 PFS final rule.
12. Milliman's precise commercial-to-Medicare ratio for behavioural health office visits. Read through summaries, not the original report. Settled by: the original PDF, or a current replacement analysis.

---

5. What this does NOT cover

• Pharmacy-side economics. Why a pharmacy refuses a telehealth CII — DEA registrant responsibility, wholesaler thresholds, PBM audit exposure — is a real and separate mechanism. Only the prescriber-visible surface is worked here.
• Vendor pricing. Neither Surescripts nor DoseSpot publishes rates. Feasibility is established; procurement is not.
• Clinical judgement. Whether non-stimulant ADHD pharmacotherapy (atomoxetine, viloxazine, guanfacine, bupropion — none of them scheduled) is an acceptable substitute for a stimulant patient facing the cliff is a clinical question, not this Finding's. It is named as an option and not evaluated.
• State-law overlays on the in-person requirement. A4 established that Arkansas, South Carolina, Alabama and West Virginia impose their own in-person rules [A4]; those are all unserved states. Whether any served state adds an obligation on top of §829(e) was not checked.
• The Special Registration proposal's arithmetic (worked thoroughly in A4) and credentialing mechanics and timelines for a payer strategy, payer by payer and state by state. Only the rate arithmetic is done here.
• Facility-rate telehealth billing. POS 02 versus POS 10, modifiers 93 and 95, and the originating-site facility fee Q3014 all materially change insured revenue and are not modelled; §1.18 uses non-facility rates throughout, which is the favourable case for a home-based telehealth practice and should be verified before it is quoted.
• Anything behind authentication. No account, no login, no form submission, no host enumeration. Every artefact here is a public unauthenticated GET of a document.

---

6. What this means for a VP of Product

The favourite recommendation survives, but only if you say it in the right half. Walking in and saying "we should build real-time pharmacy routing" invites the CEO to ask where the stock data comes from, and there is no good answer — retail inventory lives in the chain's own ERP and no prescriber in America can see it. Walking in and saying "the reason patients pay $195 and then fail to get treated is that we do not know which chains will fill our scripts in which states, we already collect that in support tickets as free text, and a ranked fill-success table plus the free FDA shortage feed plus turning on RxFill would let us steer the patient before the script goes out" is a different conversation — it is specific, it is cheap, it uses two systems the company already pays for, and it converts the single largest complaint category into a defensible asset. The distinction between those two sentences is the whole value of having checked.

The 2 January problem is smaller than the study implies and structurally worse. Smaller: roughly 13,600 active Schedule II patients need an in-person evaluation, which is five to eleven clinician-days per clinician, and a top-five-state programme covering 44% of the exposure costs about half a million dollars — 112 clinician-days at a 45% attendance rate, 249 if everyone comes — against a $14.6M revenue line, a three-to-six-month payback. Structurally worse: there is no referral cure. §829(e) says the prescribing practitioner must have conducted the in-person evaluation. The flow MEDvidi shipped and deleted in 2023 implemented a proposed rule that never became law, so rebuilding it as designed would rebuild a compliance artefact for a regulation that does not exist. And because the exemption attaches to a practitioner–patient pair held by a 1099 contractor, every clinician who quits re-orphans their whole cured panel — at 30% churn that is roughly 4,000 patients a year, forever. That is the sentence that turns this from a compliance project into an org-design argument: if the cure is the asset, the company has to own the people who hold it. Nobody in the room will have said that.

The cut list this produces is short and specific. Do not build the referral-collection flow. Do not buy a pharmacy inventory feed — there isn't one. Do not stand up an originating-site network before testing whether patients will attend a single one-off exam, because Option B converts every future visit into a trip and destroys the product for the sake of the 30% who won't come once. Do not open new states before December: each new state is a new cured-pair matrix to maintain, and A5's "stop adding states" recommendation [A5] now has a regulatory reason on top of its clinical one. And do not run the in-person programme nationally — Wyoming, Vermont and North Dakota together are about 200 patients and are worth writing off rather than serving.

On insurance, be the person who already did the arithmetic. The Director of Revenue Cycle Management hired in July 2026 [A7] means the CEO is thinking about this, and the candidate who arrives with "Medicare pays $171.35 for a 90792 and $119.71 for a 99214, nurse practitioners are capped at 85% of that by 42 CFR 414.56, behavioural-health commercial rates historically sit at or below Medicare, so our blended $169.51 becomes about $107.68 net and we would need 53% more volume to stand still" is not arguing against a payer strategy — they are pricing it. The right position is that insurance is a volume bet at a 35% lower unit price, not a margin improvement, and it should be run as a bounded experiment in one or two states against a named segment MEDvidi cannot reach today, not as a company-wide conversion. And the thing to add that nobody else will: CMS has its own in-person requirement for home-based tele-mental-health starting January 2028. A payer strategy inherits a second in-person obligation twelve months after the DEA one. Which means the in-person capability is not a 2026 fire drill at all — it is the prerequisite for both the regulated present and the insured future, and that is the single strongest reason to fund it now rather than to hope for a fifth extension.