GHI III Panel Response

This appendix reproduces the organizational-capacity material submitted by SoundBridge in response to panel item SOUN-04, which requested "a team slide: who is here, what they have done, and the gaps you know you are missing." The request followed the read-out at §7, in which the team dimension scored 1.56 — the panel's most unanimous finding, and accordingly the single largest available score mover within the revision window. Because the request contained three distinct demands, the company answered in two slides: one enumerating present personnel, one enumerating absent capability. The source document is SoundBridge_Team_and_Gaps_Row8_08302026.pptx, dated 08302026, comprising two slides. The material below reproduces that deck and the commentary accompanying its submission.

A.1 Composition of the team

SoundBridge's operating group consists of six people. The staffing shape is stated explicitly rather than left to inference: one full-time founder and five fractional or advisory operators, each named and each under engagement. Clinical and reimbursement responsibilities are deliberately vested in two different individuals, and neither of those individuals is the founder — a separation the company presents as a compliance feature rather than an artifact of scheduling.

The slide records each engagement in fuller form. Chris Boardman, Founder, President & CEO, is engaged full-time; he is an Oscar-nominated, six-time Emmy-winning composer with forty years in how music moves people, and is the origin of the protocol. He leads Year-1 deployment. Dr. Herman Williams, Chief Medical Officer, is engaged fractionally; he is a physician executive drawn from hospital management and healthcare consulting, and owns the clinical function and clinical governance. Nio Queiro, Chief Operating Officer, is engaged fractionally; she brings 33 years in revenue cycle management across hospitals and health systems, is CEO of The Queiro Group, and owns reimbursement. James Lakes, Chief Growth Officer, is engaged fractionally; he spent 16 years at Microsoft, then VMware and Salesforce, latterly in healthcare sales, and owns partner contracting. Kimberly Sena Moore, Music Therapy Consultant, is engaged in an advisory capacity; she is a board-certified music therapist, Associate Professor at Florida Gulf Coast University, and has published in the Journal of Music Therapy. Alfie Payne, Chief Technology Officer, is engaged fractionally; his remit covers software architecture, compliance and data security, and he built patient onboarding, record retrieval and music curation.

The first slide opens with the admission rather than closing with it: none of the six has ever run a federally qualified health center. The company's stated reasoning is that a health center investor will raise the point regardless, so the slide raises it first. That gap, together with five others, is carried onto the second slide.

A.2 Relevant experience

The aggregate experience represented by the six is stated as the answer to the second element of the panel's question. It comprises forty years of composing for film and television; thirty-three years of hospital revenue cycle; sixteen years at Microsoft, followed by VMware and Salesforce; a physician executive from hospital management; and a board-certified music therapist who teaches and publishes.

A.3 Declared capability gaps

Six gaps are declared. Each is given a name, a remedy, a date and, where applicable, a price. The company's framing is that a gap without an owner and a date is not a disclosure but a hedge.

A.3.1 Summary form

A.3.2 Expanded form

The second slide states the same six gaps against present coverage, making explicit what — or who — carries each function today. The two tabulations differ in column structure and in content, and both are reproduced.

A.4 What the round does not fund

The company states the boundary of the round explicitly. The round does not buy a sales team, a clinical delivery organization, or an in-house research function. SoundBridge employs no clinical staff at all: the health center employs the facilitator, the supervision and the provider of record. The consequence is that the clinical delivery organization is not a SoundBridge cost line and is not proposed to become one within this round.

A.5 Derivation of the gap analysis

The six gaps are derived from the risks present in the company's own operating model rather than from a generic organizational chart — that is, each is the personnel counterpart of an identified model risk rather than a vacancy inferred from a template hierarchy. Two features of the derivation warrant separate statement.

• The separation of clinical and reimbursement functions across two individuals, neither of whom is the founder, is presented as a structural compliance property of the current organization rather than as a future commitment.
• The finance and governance row is offered as evidence rather than as apology: the founder identified and corrected two structural errors in his own pro forma, now at version 78, before any external party raised them. The company presents this as the basis for the controller and independent board seat proposed at the seed, on the reasoning that a model whose author found its own errors is a model whose remaining error rate is worth reducing by independent hand.

B.1 Status of the engagement

The panel's request under SOUN-05 was stated in a single sentence: "Obtain an independent fair-market-value opinion on the fee-splitting structure." The request arose from the read-out at §4.2, where the service fee then in place took roughly 90% of Medicare professional-service reimbursement. Four advisors raised the point. Two of them named the legal exposure directly.

The item bearing on this row is not a source document in the ordinary sense. It is a brief prepared for the valuer — Row09_FMV_Brief_for_Evaluator, fifth edition, 31 August 2026 — built on Pro Forma v78 at the settled $261.00 license, supplemented by material held in other rows of the package. The brief itself is finished and restated on the $261.00 license as of 31 August 2026. The opinion is not. It is priced at approximately $25,000 and is not yet funded.

The position may be stated in three parts. The fee to which the four advisors objected no longer exists. A full brief for the valuer now exists. The opinion itself has not been bought.

B.2 The instrument being valued has changed

The arrangement the panel examined took 49.5% of whatever the partner collected. That percentage, and its dependence on the partner's collections, is what the four advisors flagged.

It has been replaced by a flat license of $261.00 per episode, composed of $65 a month for the three protocol months and $22 a month for the three follow-up months. The license is fixed in advance. It is the same for every payer. It is payable whether or not the practice bills or collects anything.

The valuer is therefore being asked to opine on a materially narrower arrangement than the one the panel saw. The change is not presentational: it removes the collection-contingent character of the fee that generated the original objection, and it substitutes a term that can be read off the agreement without reference to what any payer ultimately pays.

B.3 The coding error disclosed in the brief

The brief opens with an error in SoundBridge's own model rather than with an argument in the company's favor.

The model claimed a code for monitoring the musculoskeletal system. Nothing in the protocol performs musculoskeletal monitoring. The code was selected because it carried a published price, and the code that correctly describes the service does not.

The claim is withdrawn. Withdrawing it removes $119.96 from what the partner collects. Because the license is fixed while the collection base shrinks, the withdrawal raises SoundBridge's share of the pot from 31.2% to 35.8% — that is, the correction moves the headline ratio in the direction least favorable to the company.

The disclosure is made by the company against itself. The company's stated position is that a smaller number it can defend is worth more than a larger one a reviewer can take apart.

B.4 Structure and evidentiary classification of the brief

The brief assembles the material a valuer would require and sorts it into three classes: published federal data, published federal rules, and SoundBridge's own work. Work originating with the company is marked as unverified throughout, and four specific items carry an explicit flag instructing the reader not to rely on them without independent checking. The classification is structural rather than advisory — it is applied to every input, so that the boundary between externally verifiable material and company-generated material is visible on the face of the document.

B.5 Two methodological contributions of the current edition

Two features distinguish the fifth edition from its predecessors.

B.5.1 The partner's costs shown two ways

The full cost of a facilitator is $115,062 a year. Of that total, $38,160 is building and back-office overhead the health center already pays. That $38,160 is a real cost if the partner hires a new person. It is not a real cost if the partner moves an existing employee onto the program.

Presenting a single fully loaded figure therefore conceals a decision that belongs to the partner rather than to the arrangement. The brief accordingly shows the partner's retained economics under three costing treatments.

The same transaction reads as unsignable under the first treatment and comfortably good under the third. Neither treatment is wrong. Which one applies is a fact about that partner's hiring decision, not a fact about the license fee.

B.5.2 Payer mix shown as a range

Payer mix is presented as a range rather than as a point estimate. The 93.8% adjustment carried in the model is an inherited assumption that has not been checked by anyone, and it multiplies every downstream figure. Expressing it as a range rather than a number makes the dependence explicit rather than absorbing it silently into the result.

B.6 Where the fee sits against published comparable bands

The following position is stated by SoundBridge against its own interest.

The full-service band describes vendors who supply clinical staff. SoundBridge supplies none. The license therefore sits inside a band that does not describe the arrangement it prices — which is the panel's original objection, arrived at here through the company's own correction rather than through a reviewer's challenge.

The earlier claim that the fee sat "just above the intermediate band" is formally withdrawn.

B.7 The eight questions put to the valuer

The brief puts eight questions to the valuer. The fifth is the one with the widest consequences: whether the opinion should be a ratio at all. The collection base moves with three variables SoundBridge does not control — the coding outcome, the partner's payer mix, and the partner's staffing choice. A ratio opinion is stale the moment a second partner in a different state signs. An opinion on the structure would last longer.

1. Which benchmark governs.
2. What the non-software parts are worth.
3. Whether per-patient is the right unit.
4. Whether FQHC economics or physician-practice benchmarks apply.
5. Whether the opinion should be a ratio at all.
6. Whether an opinion should instead be rendered on the structure, which would last longer than a ratio opinion that goes stale on the signing of a second partner in a different state.
7. Validation of the fifty-state analysis.
8. How to treat the implementation fee.

B.8 The pricing gap, resolved

Earlier editions of the brief disclosed an unreconciled gap: the published rate card totaled $261.00 an episode while the model carried $269.29.

That gap is resolved. The license is $261.00 — the rate card figure, computed from four terms stated in the agreement: $65, $22, three protocol months, three follow-up months.

The figure of $269.29 was a manual override pinned during the v77 rebuild, so that correcting the collection base could not move the top line. It was never a price, and the workbook shows why: $269.29 implies an active rate of $67.76, or $67.06 / $22.70 on a proportional split. No rate card produces those figures.

The valuer is to be asked to opine on $261.00. The change reduces five-year cumulative EBITDA from $17,439,360 to $16,588,688, and SoundBridge's share of partner collections from 36.9% to 35.8%.

B.9 Material held in other sections

No standalone source document exists for this row. The material bearing on it is held in a brief prepared for the valuer together with cross-references into four other rows of the package.

B.10 Current pricing position

The following statement of the current pricing position is provided for the valuer's reference.

B.11 Conditions for closure

Four steps were identified as outstanding. One has since been completed.

• Fund and commission the valuation engagement (~$25K), with the selection mechanism and reference date settled per Row 10 §1.
• Obtain counsel's short written view on the threshold Anti-Kickback question — whether SoundBridge is in a position to refer or generate federal health care program business at all — before the full opinion (Row 10 §8).
• Secure at least one FQHC CFO reference on the record, whether through the five named calls in Row 11 §10 or through the advisor's three committed FQHC introductions, which are currently gated at SOUN-21 pending resolution of this very question.
• Settle the $261 / $269.29 question — completed 31 August 2026. The license is $261.00.

This appendix reproduces, in neutral form, the instructions issued by SoundBridge to its regulatory counsel, Kevin Page. The underlying source is the second edition of those notes, dated 30 August 2026, which supersedes an earlier note of 28 August. The second edition is built on Pro Forma v78. Its status at the time of writing is that counsel is briefed but deliberately on hold; a scoped engagement letter is to follow once the panel-response phase closes. No opinion had been delivered when the note was written.

The instructions respond to a request recorded in the evaluation panel's item SOUN-06, which asked the company to "obtain qualified healthcare regulatory advice, including Anti-Kickback Statute clearance." The read-out at §4.2 records that this was offered as GHI's strong recommendation, made in its capacity as a prospective partner rather than as counsel to the company. The company's position, stated in a single line, is that counsel is engaged and fully briefed, and that he has been told to wait, on purpose, until the remainder of the panel responses are finished.

The material below is organized as the source note is organized: a substantive treatment of the engagement, the scope, the corrections, the governing federal material, the pricing decision and the requested order of work; followed by the supporting detail, which records what changed between editions, the recitals and covenants sought, a date correction, and the disposition of the superseded first edition.

C.1 The deliberate hold on the engagement

The first line of the note states the position directly: counsel is on hold. The reasoning given is procedural rather than substantive. A scoped engagement letter should name exactly what is needed and in what order. Sending questions piecemeal to a compliance lawyer is expensive and produces worse answers. The hold is therefore presented not as a delay in obtaining advice but as a decision about the sequence in which advice is commissioned, taken so that the eventual instruction is complete and ordered when it issues.

The consequence for the record is that, as of the second edition, the company holds no regulatory opinion. What it holds is a briefed adviser and a defined queue of questions. The note is explicit that the queue will be converted into a scoped engagement letter once the panel-response phase closes, and that the engagement letter, not the note, will be the operative instrument.

C.2 The scope of work to be requested

Three items are identified as the work counsel will eventually be asked to perform:

• The partner agreement, with the fair market value opinion written in as a condition precedent — the fee clause does not take effect until the opinion arrives.
• The Anti-Kickback and Stark opinion.
• A view on corporate practice of medicine, which decides whether the staffed model is ever usable.

Counsel is expressly not asked to value anything. The note gives two reasons, both of which are about the reception of the eventual opinion rather than about counsel's competence. A compliance lawyer cannot certify his own client's price; and the panel that asked for the opinion would discount it if he tried. The valuation function is thus deliberately kept outside the regulatory engagement, and the fair market value opinion is instead positioned as an external input on which the fee clause is made conditional.

The conditions-precedent construction carries a commercial purpose as well as a compliance one. Because the fee clause does not take effect until the opinion arrives, the partner agreement can be signed before the valuation is complete without the parties committing to a price that the valuation might not support.

C.3 Corrections made against the company's own interest

Two corrections between the first and second editions run against SoundBridge's own position. The note records both, and records that the company made them of its own motion.

The first concerns the share of partner collections. The 28 August note told counsel that the figure was 31.2%. At that level, a low valuation was unlikely, and a one-way price ratchet was correspondingly cheap for the company to concede. After the coding correction, the figure is 35.8% — inside the full-service band. The direction of the correction is upward, not downward. The consequence is that a low opinion is now more likely than counsel had been told, and the ratchet that had looked inexpensive is more expensive. The note instructs counsel to price that in.

The second concerns a product feature that does not exist. The 28 August note asked counsel to draft a covenant requiring the partner to watch a live dashboard. No such real-time alert was ever built, and it does not appear in the specification. The second edition withdraws the request in terms:

Please do not draft that.

What is wanted in its place is a covenant putting acute risk with the practice. The substitution is treated in the note as a straightforward matter of aligning the contractual drafting with the product as it exists and with the allocation of clinical responsibility the company intends.

C.4 The federal provision that supports the structure

The note draws counsel's attention to CMS FAQ A60, which addresses the employment of a health center's behavioral health care manager. The FAQ states that such a care manager may be employed by or contracted to the health center —

not to another company.

If that language governs, its effect is categorical: no vendor can supply its own care managers to a health center and take a share of the resulting revenue. It rules out the staffed model, and it makes the structure SoundBridge has adopted the only lawful shape available.

The note nonetheless asks counsel to check the provision rather than rely on it. Three features are given as the reasons for caution. It is dated December 2019. It references retired codes. It is not repeated in current CMS guidance. The note adds a fourth consideration of a different kind: the provision happens to favor the company, which is precisely why it should be verified rather than leaned on. The instruction is therefore to test a favorable authority before it is used, on the view that an unverified provision that supports the company's structure is a liability in the hands of a reviewer rather than an asset.

C.5 The pricing question, now settled

Earlier notes had asked counsel to resolve a discrepancy between two figures: a published rate card totaling $261.00 an episode, set against a model carrying $269.29. That question is no longer open. The company has decided, and the decision is recorded so that counsel does not spend time on it.

The license is $261.00. It is computed from four terms that belong in the agreement itself: $65 per enrolled patient per month for three protocol months, plus $22 per month for three follow-up months. The drafting instruction attached to this decision is that the agreement should state those four terms and never a per-episode total, so that the arithmetic stays visible on the face of the contract.

C.6 The requested sequence of work

The note specifies the order in which the answers are wanted:

1. The threshold question in writing first: is SoundBridge even in a position to refer or generate federal healthcare business? The founder's view is that it is not: the company takes no referrals, makes none, and bills nothing. If that is right, most of the analysis is shorter than people expect.
2. The conditions-precedent clause, because it unblocks a partner signature.
3. The escalation covenant that replaces the withdrawn dashboard one.
4. Corporate practice — yes or no, no memo needed.
5. The full AKS/Stark opinion.
6. The BAA and the SOAP-note attestation language.

The sequence is ordered by dependency rather than by importance. The threshold question comes first because a negative answer shortens everything that follows; the conditions-precedent clause comes second because it unblocks a signature; and the full opinion comes late, after the questions that determine its scope have been settled.

C.7 A safe harbor that does not apply

One dead end is recorded in the note expressly to save counsel the time of exploring it. The FQHC safe harbor at 42 CFR 1001.952(w) runs the wrong way. It protects things given to a health center. The SoundBridge fee runs from one. On that reading it does not apply, and the note closes the point rather than leaving it open for analysis.

C.8 Recitals and covenants requested

The note asks for eight items to be carried in the partner agreement. Each is wanted as a durable contractual fact rather than as a description of current practice, on the stated reasoning that the valuer, the panel and any later reviewer will read the contract before they read the deck.

• SoundBridge supplies no clinical personnel and no clinical labor, and employs none of the three Collaborative Care team members
• The health center employs all three roles and retains all clinical decision-making
• The health center owns the escalation protocol, designates the supervising clinician, and holds the judgment and the liability
• The health center performs all billing and revenue cycle; SoundBridge submits no claim to any federal healthcare program
• The fee is fixed in advance, does not vary with reimbursement rates, payer mix, volume of federal healthcare business or billing success, and is payable whether or not the practice bills or collects
• Neither party is required to refer patients to the other
• The health center determines coverage and billability; SoundBridge makes no warranty of reimbursement
• The health center designates a facilitator who is available to furnish services face-to-face in person, and who is employed by or under contract to the health center and not to any other company

C.9 Changes from the first edition

The complete second-edition note is filed alongside the summary document in the Row 10 folder, as Row10 — Notes for Kevin Page (v2, 30 August 2026). Its eight sections cover: why counsel is receiving it and the hold; what will be asked for when the scope arrives; conditions-precedent mechanics; recitals and covenants; the federal conditions on employment and on capability; coverage risk and the RTM correction; AKS and Stark inputs; corporate practice; data, documentation and the revised alert covenant; and the order of work.

The differences between the two editions are set out below.

C.10 A date correction

Section 3 of the note states that G0511 and G0512 "both stopped being reportable on 1 January 2026." That is wrong for G0511. The verified position is as follows.

G0511 was replaced by individual care management codes required from 1 January 2025, with a CMS transition under which it remained billable through 30 September 2025. G0512 was replaced from 1 January 2026, when health centers became required to report 99492, 99493, 99494 and G2214 individually.

That wording is to be used in place of the section as drafted. The Document Control page in the parent folder carries the same correction.

C.11 Superseded material retained

The first edition is retained in the same folder for the record, under the title SUPERSEDED 28Aug — Row10_Notes_for_Regulatory_Counsel.docx. It was retitled so that it is not mistaken for the current instruction. It carries the withdrawn real-time-alert design, the superseded 32.2% share figure, and the "Chris Brown" byline.

Note on provenance and version

This appendix reproduces, in third-person register, the supplement to the FMV brief circulated internally as Row11_The_FQHC_Ledger.docx. It was prepared in response to the evaluation panel's request recorded as SOUN-07, which asked the company to "Show the math from the FQHC's side of the table, and find one FQHC CFO who has said yes to the split." The read-out at §4.2 gave the reason the request mattered: if the partner's economics do not work, neither do the company's.

The document answers the first half of that request and does not answer the second. The arithmetic was completed and it altered SoundBridge's own model; no FQHC finance chief has endorsed the arrangement on the public record, and the reason offered is that no such person appears on the public record at all.

The figures throughout are stated on Pro Forma v76. The workbook of record is now v78, and three quantities have since moved. Partner collection with three add-on units falls from $836.76 to $729.82; the SoundBridge license as a percentage of collections rises from 32.2% to 35.8%; and what the partner keeps against a fully-loaded staff cost moves from +$88.14 (10.5%) to −$10.53 (−1.4%). The collection fell because of the RTM coding correction recorded in Row 12; the license fell from $269.29 to $261.00 when the fee was settled on the published rate card on 31 August 2026. The base case stated in this appendix therefore no longer holds: on a fully-loaded facilitator the partner is now slightly under water at the current fee. The v76 figures are retained here without adjustment because this document is the diagnosis on which the subsequent corrections were built.

SOUNDBRIDGE HEALTH · SUPPLEMENT TO THE FMV BRIEF

The FQHC's Ledger

The document sets out what the partner actually banks, what it actually costs them, what a 30-minute session with no documentation burden can actually deliver, and what a health center CFO would say if one had ever been asked on the record.

Prepared 28 August 2026 · Second pass, corrected · Built on Pro Forma v76 sheets 'Inputs Global' §12–§17, 'Inputs FQHC' rows 30–38, 'FMV Comparables' · Not legal or valuation advice

Status. Every correction recommended in Section 11 of the source document is implemented in Pro Forma v77, 28 August 2026. The v76 references below are retained deliberately, because the document is the diagnosis and v77 is the remedy — the two are intended to be read together. The document addresses GHI SSM project plan Row 11; the second half of that row, a named FQHC CFO on record, remains open and is addressed in Section D.10.

D.1 Summary

The partner-economics check in v76 is arithmetically correct and rests on two inputs that are both wrong, in opposite directions. It charges the facilitator at $18 a session, which is far too low. It assumes six sessions a day, which is also far too low. Correcting both produces a working deal — at a partner margin inside the benchmark band, at the fee as priced, for reasons the pro forma does not currently state.

The central case. A facilitator delivering 30-minute sessions with AI-generated documentation can be scheduled at roughly 14.5 sessions a day. At an 85% show rate that realizes 12.3. With the three 99494 add-on units that the protocol's own session minutes support, the health center collects $836.76 an episode and keeps $88.14 — 10.5% net after its own staff — and SoundBridge's share of collections is 32.2% at an unchanged price. That is mid-band on Steve Page's 5–15% and at the top of the arguable 20–30% vendor band.

The independent check. AIMS Center's caseload guidance puts a full-time behavioral health care manager at 130 hours of direct care a month, which is 14.2 SoundBridge sessions a day and a concurrent caseload of 78 — inside AIMS's published 60–80 band. Two derivations, one from the clock and one from the canonical CoCM staffing authority, land on the same number. Pro Forma v76's six sessions a day implies a concurrent caseload of 33, roughly half AIMS's floor for an FQHC deployment.

The correction to the first pass. Every figure in the earlier version of the document used 30-minute sessions, and the break-even thresholds are unchanged. What was wrong was the verdict attached to them: the first pass took v76's six-a-day as the realistic case and dismissed the higher figures as unachievable. That treatment was inverted. Six a day is the assumption that cannot survive, because the platform exists precisely to remove the documentation load that would otherwise cap the day at six.

D.2 The ledger as v76 states it

The ledger is stated per episode. An episode comprises six enrollment months (three protocol, three follow-up) and three payer-recognized billable months.

The arithmetic reconciles exactly. $754.24 gross × 0.938404 × 0.95 = $672.39; the license spread over six enrollment months is $44.88. The 10.0% sits mid-band against Steve Page's observed 5–15%.

The ledger arrives at that result, however, on two numbers that a health center CFO will not accept.

D.3 Input 1: Facilitator cost per session at $18.00

The figure is a blue input at 'Inputs Global' row 121, extracted in v75 from a hardcoded $180.

The two cells are mutually inconsistent inside a single workbook. The honest replacement is the evaluator brief's own stack: facilitator $66,339, UDS-derived overhead $38,160, licensed supervisor share $10,562, giving $115,062 per FTE-year all-in, with the psychiatric consultant carried separately at $54.99 an episode and the 15%-of-collections line dropped so that the UDS multiplier is not counted twice.

D.4 Input 2: Six sessions a day, and why it is the weaker number

Sessions are 30 minutes. The facilitator writes nothing. The session is recorded, transcribed, formatted to a SOAP note and timestamped by the platform; the caseload provider reviews the transcript and the provider of record attests. There is consequently no charting block at the end of the day and no note-writing between patients. What remains between sessions is a room change and a click.

The day budget, computed at 220 working days, is as follows.

Six a day is 3.0 contact hours out of eight. That is the productivity of a clinician who documents, not of a facilitator whose documentation is machine-generated. The figure is flagged in the workbook as a management assumption and should be replaced.

The AIMS check

The AIMS check is the derivation the document recommends putting before the panel. AIMS Center's Caseload Size Guidance for Behavioral Health Care Managers (updated 23 October 2025) recommends 60–80 patients for a full-time behavioral health care manager — verbatim, 80 for "common adult mental health conditions… doing some or all work virtually," and 60 for "a Medicaid/uninsured population in a FQHC/RHC setting." Both of its worked examples total 130 hours of direct care per month.

Two independent routes arrive at 13–14. The canonical CoCM staffing authority and a stopwatch agree with one another, and the pro forma is the outlier.

What actually limits throughput

The binding constraint is not documentation but no-shows. A 1:1 scheduled telehealth slot that the patient misses is a dead half-hour unless it is backfilled, and behavioral health no-show rates in safety-net settings are high. That is the variable the pilot has to measure, and it is the reason the central case below is stated on realized sessions rather than scheduled ones.

D.5 The ledger restated

The restatement assumes ten sessions per episode, 220 working days, $115,062 all-in per facilitator FTE, and the fee held at $269.29. The table reports net per episode and the percentage of collections retained.

The central case is 14.5 scheduled at an 85% show rate. At three add-on units the partner retains 10.5% and SoundBridge takes 32.2% of collections. Both figures are inside their respective benchmarks, and neither required a price change to reach.

D.6 The add-on units as a function of the timestamp log

The add-on units are not an assumption; they are the timestamp log. The protocol of record is four 30-minute sessions in month 1 and three in each of months 2 and 3. Set against the CoCM time thresholds, the position is as follows.

The midpoint reading is now better supported than the documents of record allow. NC Medicaid's December 30, 2025 coverage update states of 99494, verbatim, "At least 16 additional minutes must be provided to use this code." The APA's published time table corroborates it arithmetically — the one-unit band opens at 86 minutes, which is 70 + 16, not 70 + 30. Meadows publishes the same 16–30 minute range.

Three consequences follow.

First, v76 books none of this. 'Inputs Global' section 16 holds the 99494 attach rate at 0.00. The $672.39 collection contains zero add-on units. Every figure in the "+3 units" and "+4 units" columns above is unbooked collection sitting on the partner's side of the table.

Second, the $672–$892 "defensible range" is exactly the zero-to-four-add-on-unit range. Four units at $61.46, taken through the same payer-mix and clean-claim haircuts, gives $891.55. That reconciliation was not previously stated and merits statement: the collection uncertainty the FMV brief flags is not vagueness about rates, but one specific, answerable coding question.

Third, the audit log is the evidence. Every session is timestamped by the platform. The add-on unit is not claimed on an estimate of care-manager minutes; it is claimed on a machine-generated record of them. That is the same architectural argument §5 of the evaluator brief already makes about supervision, applied here to billing, and it is stronger in this application because the standard failure mode across CoCM programs is precisely that the time is spent and never logged.

Two corrections to cocm-countable-time.md §4 are recorded. That document states that four add-on units "is exactly the assumption Pro Forma v76 already carries" — v76 carries zero. It also states "one third of the MUE ceiling of twelve," which is right at the episode level (four units a month across three months) but reads as a monthly claim; the MUE is 4 units per date of service, and CoCM is reported on one date of service per month, so month 1 at two units is half the monthly ceiling, not one sixth. Under an MAI 2 or 3 edit, exceeding the cap denies every unit on the line, not only the excess — a point worth establishing before anyone models a fifth unit.

D.7 Break-even, and why the fee is second-order

The table reports realized sessions per working day at which the episode funds the staff.

The final column shows the effect of giving the license away entirely: break-even moves by 6.5, 3.5 and 2.9 sessions a day respectively. Claiming three add-on units moves it by 4.8. The countable-time question is therefore worth about as much to the health center as the entire fee, and the productivity question is worth more than both.

The finding is not that the fee is defensible but that the fee is not the main variable, and the arithmetic above is what establishes it.

Fee headroom

The headroom analysis runs in the opposite direction to the panel's concern: at realistic throughput the fee as priced is conservative rather than aggressive.

The $269.29 as priced sits below the 10%-retention ceiling from about 12.3 sessions a day upward. The document draws no case from this for raising the fee — the vendor-share diagnostic is what the panel reacted to, and the company's position is that it should come down, not up — but the model should stop implying that the partner is being squeezed.

D.8 Four corrections to the denominator, all FQHC-specific

The $672.39 is built from PFS national non-facility rates, and for an FQHC that methodology is CMS's own: FQHC care-coordination lines are paid at the national non-facility amount with no GPCI or GAF, separately from and in addition to the PPS encounter. A national-rate model is therefore more accurate for a health center than for a private practice. Four adjustments nonetheless apply, three of them downward.

Coinsurance. Medicare pays 80% of the lesser of charge or rate; the patient owes 20%, and a sliding-fee panel frequently does not pay it. Medicare is 19.0% of the model's billable panel, so full leakage is 3.8% of collections, about $26 an episode. Commercial cost-sharing runs the same way and is not quantified here.

Medicaid rates. Medicaid is the real exposure and it is 57% of the billable panel. Colorado's FY26 integrated care schedule prices 99492 at $119.40 against Medicare's $160.32 — 26% lower. The 93.8% payer-mix adjustment is an inherited v69.1 ratio; one Colorado-like state would put it nearer 88%. This is the assumption most worth replacing from a remittance file.

PPS absorption. Some states absorb CoCM into PPS entirely. In California, health centers receive the same PPS payment regardless of billing code, so marginal CoCM collection may be near zero. Texas and North Carolina pay CoCM outside the encounter rate; New York pays a monthly case payment. There is no national rule, and each Wave-1 state requires checking before it opens.

Opportunity cost. Opportunity cost is invisible in the model and runs against the company's position. A co-located behavioral health visit bills a full FQHC PPS encounter — $207.72 CY2026 Medicare base before GAF, and typically more under Medicaid PPS — while a whole month of CoCM bills $144.96 to $160.32 with no wrap. West Health's Carlo and Wardlow put the point directly, stating that this

"provides a powerful financial incentive to maintain co-location models in FQHCs and to avoid CoCM implementation."

The answer offered is not that CoCM pays better per encounter. It is that co-location is capacity-bound by a licensed clinician the center cannot hire, and CoCM is not.

One date correction is recorded. G0512 was retired effective January 1, 2026, not 2025 — that date belongs to G0511. cocm-countable-time.md §6 has this right; the continuation brief does not.

D.9 The one federal rule that helps

CMS's Care Management Services in RHCs and FQHCs FAQ, Q60, states verbatim:

"The behavioral health care manager furnishes both face-to-face and non-face-to-face services. This person works under the general supervision of the RHC or FQHC practitioner and may be employed by or working under contract to the RHC or FQHC, not to another company."

If that language governs, a vendor cannot supply its own employed care managers to a health center and take a share. It forecloses the full-service template in FQHCs and validates the license-only structure SoundBridge has already chosen.

Two cautions attach. The FAQ is dated December 2019 and instructs billing G0511 and G0512, both now retired; CMS's current FQHC booklet (MLN006397, March 2026) does not repeat the language, and no replacement FAQ was located. NACHC's March 2026 CoCM tips say only that the care manager "may either be employees of the health center or working under contract," omitting the qualifier. The position is therefore a live conflict for counsel rather than a settled point — but it is the rare compliance finding that runs in the company's favor.

A related point is already noted in cocm-countable-time.md §6.4: the care manager must be available to furnish face-to-face services in person even though doing so is not required. That requirement is satisfied under the license template with a center-employed facilitator, and remains a live question under the full-stack template.

D.10 The CFO question

No FQHC CFO has said yes to this split on the public record, and none has said no. The category is empty.

Roughly thirty-five searches across vendor case studies, PCA and NACHC materials, HRSA and CMMI documentation, AIMS implementation work, the Commonwealth Fund, Milbank, the trade press and the peer-reviewed implementation literature returned no named FQHC finance executive endorsing — or rejecting — a behavioral health vendor arrangement in which the vendor takes a material share of the health center's collection. What the search does establish is why.

CoCM enablement vendors went to health systems, not health centers. Every enumerable Concert Health partnership is a system — CommonSpirit, Froedtert, Mercy, AdventHealth, WellSpan. Iris Telehealth's FQHC offering is priced on a different instrument entirely, a flat hourly rate per scheduled hour of provider time. The closest documented analogue in a government source is MindHealthy supplying virtual care managers, psychiatric consultants and registry technology to One Health in Charlotte — and One Health is a primary care group, not an FQHC, with no payment terms disclosed.

The nearest comparable cuts both ways. The Community Clinic Association of Los Angeles County published a CoCM financial sustainability toolkit for California FQHCs, built by Curt Degenfelder (Curt Degenfelder Consulting) and Sarah Arnquist (SJA Health Solutions), whose break-even illustration puts a half-time LCSW at roughly $124,900 of annual revenue against $120,000 of cost. That is the same shape as Section D.7 — thin, workable, and with little margin for a vendor share. The comparable is unfavourable to the company's position and is reported for that reason.

How the absence should be read. It is a fact about the market, not a signal about the fee, and it is the same fact the workbook already records: no CoCM enablement vendor publishes pricing, which is why the comparables are drawn from adjacent RPM/RTM enablement. The recommended treatment is to state plainly that there is no public precedent, that this is why an independent opinion is being commissioned rather than a benchmark cited, and that the first partner agreement will be the first data point.

Five calls are proposed as capable of producing a real reference, listed in order of expected yield:

• Curt Degenfelder, Curt Degenfelder Consulting — built the CoCM financial model for California health centers and speaks at NACHC. The most likely person in the country to know whether any health center CFO has signed a share-of-collections BHI deal.
• Sarah Arnquist, SJA Health Solutions — co-author; worked the Los Angeles health centers directly.
• The CFOs of the five LAMMHA CoCM health centers — AltaMed, Eisner Health, St. John's Community Health, TCC Family Health, Clinica Msr. Oscar A. Romero. The FQHCs most likely to have already run this arithmetic.
• AIMS Center billing and financing office hours, University of Washington — they field exactly this question and will know of unpublished arrangements.
• James D. Sinkoff, Deputy CEO and CFO, Sun River Health — a large-FQHC CFO who already speaks publicly on health center payment economics.

the advisor's three committed FQHC introductions are the other route, and a CFO reference obtained that way is worth more to the valuer than anything in the public record.

D.11 What this changes

Figures drawn from Pro Forma v76 are modeled. CY2026 rates are CMS Physician Fee Schedule national non-facility amounts. Throughput figures assume 30-minute sessions, ten sessions per episode, 220 working days, and the evaluator brief's $115,062 all-in facilitator FTE cost; the psychiatric consultant is carried as a separate per-episode line and the 15%-of-collections overhead line is dropped to avoid double-counting the UDS multiplier. Show rates and facilitator productivity are modeled, not observed, and are the two quantities the pilot measures directly. The document is a commercial and factual supplement, and is not legal advice, not a valuation, and not a substitute for either.

SoundBridge Health · Response to the GHI III advisory panel

Row 12 — Remote Therapeutic Monitoring: the coding position, the federal basis for it, the product specification that supports it, and the formal review now in progress.

Prepared by Chris Boardman · Second edition, 30 August 2026 · Federal citations retrieved 29 August 2026 · Payment amounts are CY2026 Physician Fee Schedule national non-facility rates, post-correction · Financial figures are drawn from Pro Forma v78, 29 August 2026 · Reimbursement and coding position owned by Nio Queiro, Chief Operating Officer · Not legal advice and not a coding audit.

What changed in this edition. Nothing in the coding position changed. The second edition records the officer of record for that position — see Section E.9 — and adds her signature block. The first edition, dated 29 August 2026, carried the founder's name only.

The short answer. The panel asked for independent confirmation of the company's RTM coding. Working the question produced a correction to SoundBridge's own materials, a clear answer to the threshold issue underneath it, and a narrower revenue line than the company previously carried. SoundBridge reports 98975 and 98980/98981. It does not report 98978. CPT 98977, which appeared in earlier materials, is withdrawn — it was wrong. The device requirement that governs all of it is met, and Section E.2 sets out why in CMS's own words. Formal review by the AMA and by the partner's Medicare contractor is in progress.

E.1 The correction

Pro Forma v77 and one briefing document carried CPT 98977 as what the company internally called a "musculoskeletal proxy." The CY2026 descriptor reads "device(s) supply for data access or data transmissions to support monitoring of musculoskeletal system, 16–30 days in a 30-day period." Nothing in the SoundBridge protocol monitors a musculoskeletal system. The code was selected because the behavioral-health device code is unpriced and 98977 is priced. That is a pricing rationale, not a coding rationale, and the label "proxy" concealed the problem rather than qualifying it.

98977 is withdrawn from the model and from every document. It is stated first because a reviewer checking descriptors would find it in about a minute, and the company's judgment is that the error is better handed over than found.

The correction is made, not merely identified. Pro Forma v78 is built and verified: 98977 is out, 98975 and 98980 are corrected to actual CY2026 rates, no device-supply code is claimed, and the workbook's "v78 Changes" tab records every cell that moved and why. Nothing in this document is a plan to fix something later.

E.2 The threshold question is the device, and it is answered

Before any question of which code applies, RTM requires that the supplied article be a medical device as FDA defines it. That is where the analysis actually lives, and the answer is specific and favorable.

CMS wrote a definitional requirement, not a clearance requirement. From the CY2022 Physician Fee Schedule final rule, verbatim:

"We note that, for both sets of codes, the device used must meet the FDA definition of a device as described in section 201(h) of the Federal Food, Drug and Cosmetic Act."

The requirement is restated in the CY2024 final rule as "at least one medical device as defined in section 201(h) of the FFDCA." The phrases "FDA-cleared" and "cleared by the FDA" do not appear in the CY2022 rule at all. Where clearance language surfaces in these rules it appears in commenters' submissions, and CMS did not adopt it.

Three considerations turn that observation from an argument into a position.

First, CMS imposed an express clearance requirement elsewhere, for behavioral health, in the same rulebook. For the digital mental health treatment codes G0552–G0554, CMS requires a device "cleared under section 510(k) … or granted de novo authorization by FDA and classified under 21 CFR 882.5801." CMS knew precisely how to condition payment on a marketing authorization. It did so for those codes and not for RTM. Reading a clearance requirement into RTM anyway is not conservatism; it is the application of a different rule than the one CMS wrote.

Second, Class II classification does not imply that a 510(k) is required. The largest company billing RTM today, Sword Health, operates on a 510(k)-exempt listing — product code ISD, 21 CFR 890.5360, Class II, exempt from premarket submission. Hinge Health's clearances cover Enso, a pain-relief hardware device, and not its monitoring platform. Clearance is the exception in RTM, not the norm.

Third, FDA has published the position the platform occupies. From FDA's guidance on device software functions:

"Even though these software functions may meet the definition of medical device, the FDA intends to exercise enforcement discretion for these software functions, because they pose lower risk to the public."

FDA's list of those functions includes, verbatim, software that helps

"patients with diagnosed psychiatric conditions … maintain their behavioral coping skills by providing a 'Skill of the Day' behavioral technique or audio messages that the user can access when experiencing increased anxiety,"

together with symptom questionnaires and the tracking and trending of self-entered data.

That is the position: an article meeting the section 201(h) definition, which is what the codes require, sitting in the low-risk category for which FDA does not enforce premarket requirements.

A limit is stated plainly. Enforcement discretion is a published policy, not a statutory safe harbor, and FDA's own wording is that these functions "may meet" the device definition. The position is sound, and it is not immune to a change in FDA policy.

E.3 The product specification that supports the position

The position set out above holds only while the product stays in the low-risk category. The platform is in development, and the specification below is the design of record.

One item changed, and it changed on paper rather than in code. An earlier design contemplated a risk alert raised on the treating practitioner's dashboard in real time. It was described in the company's notes to counsel and in the briefing note; it was never built. It has been taken out of the specification.

The reason is that FDA's Clinical Decision Support guidance is explicit on the point:

"FDA does not consider software functions intended for a critical, time-sensitive task or decision to meet Criterion 4, because an HCP is unlikely to have sufficient time to independently review the basis of the recommendations,"

and its worked example of a device function is software that analyzes patient information to detect a serious condition and

"generate an alarm or an alert to notify an HCP."

Urgency alone is disqualifying, whether or not the software recommends a treatment — and the SoundBridge design never did.

The real-time alert originated in answer to a provider's question about how the platform would work, rather than as a designed clinical feature. Nothing in the pro forma depends on it.

Treatment-intensity signals remain in the specification. They surface in the registry the psychiatric consultant already reviews on its normal cadence, with the inputs and the rule logic disclosed in plain language so that the clinician can independently review the basis. They carry no urgency and recommend no treatment. Fidelity scoring, transcription, SOAP-note generation and the audit log are unchanged.

Documents to be corrected are identified here. The notes to regulatory counsel of 28 August and the third-edition briefing note describe the earlier design. Both are being revised to this specification. The discrepancy is noted in this document rather than resolved by quiet reissue.

E.4 Escalation is the employing practice's responsibility

SoundBridge does not operate an escalation policy, does not furnish crisis response, and does not make or participate in any clinical decision. The health center adopts the escalation protocol as its own, names the supervising clinician reachable during session hours, and holds the clinical judgment and the liability. SoundBridge supplies the certification curriculum in which the facilitator is trained to recognize risk indicators and to follow the practice's protocol, and supplies the documentation architecture that records what happened.

This allocation is not new and was not devised for the present question. The health center has employed all three Collaborative Care roles since the commercial model was settled, for a reason unrelated to RTM: clinical judgment belongs to licensed clinicians who are accountable to their own employer, their own supervisor and their own board. Escalation is clinical judgment. It follows from who holds the license, not from anything the company wanted to achieve downstream.

Escalation is not a billable feature. It generates no revenue for SoundBridge, appears in no code the company reports, and nothing in the license fee depends on it. This is stated explicitly because the alternative reading — that a safety function was allocated for a commercial or regulatory purpose — would be a fair thing for a reviewer to worry about, and is not what happened.

Two consequences follow from the allocation rather than motivating it, and the first is the one that matters.

• It is the safer design. The facilitator is on the call with the patient. A dashboard alert requires that somebody happens to be watching a dashboard, and it degrades exactly when the clinic is busiest. Trained in-session recognition responds in seconds, with a warm handoff to the health center's own on-call clinician, documented in the record, and 988 as a backstop rather than as the protocol. It is also squarely consistent with the company's workforce thesis — SUMMIT and COBRA both show supervised non-specialists delivering structured protocols effectively where training is competency-gated and supervision continuous.
• It keeps acute risk out of software, and therefore out of the device question in Section E.2. That is a consequence of the allocation, not a reason for it. Had the allocation run the other way, the answer would have been to build the escalation properly and accept the regulatory consequence, rather than to move a safety function to suit a classification.

E.5 What the company reports, and what it does not

The management codes work for the following reason. 98975, 98980 and 98981 name no body system and no therapy modality. They are condition-agnostic on their face, and the parent descriptor's own subject matter is "therapy adherence, therapy response, digital therapeutic intervention" — which is what the platform captures.

98978 is not reported for the following reason. It is the only behavioral-health device-supply code CPT provides, and its descriptor names cognitive behavioral therapy. The SoundBridge protocol is documented throughout — in the care plan, the session note and the clinical evidence map — as a manualized psychoeducational skills-training curriculum that is specifically not CBT and not psychotherapy. The NCCI Policy Manual, Chapter I, states that "a provider should not report a CPT code out of the context for which it was intended" and that the provider "must have performed all of the services noted in the descriptor." A claim line reading "monitoring of cognitive behavioral therapy" behind a chart that expressly disclaims CBT is a contradiction inside one record, and it needs no local coverage determination to be actionable.

The company considered and rejected changing the clinical characterization to fit the code. The characterization is correct as it stands, it is what the evidence supports, and it is not for sale to a coding convenience.

One further point is recorded here: no authority anywhere defines "cognitive behavioral therapy" for coding purposes. Not the AMA, not CMS, not any Medicare Administrative Contractor, and not any payer policy the company could locate. The term operates as a discriminator in exactly one place in the entire code set — this descriptor. That absence is why the question is open rather than settled, and it is precisely what SoundBridge has put to the AMA.

E.6 What the correction does to the model

The device-supply leg is where recurring monthly RTM revenue sits in the musculoskeletal and respiratory markets. Removing it leaves set-up and treatment management. RTM falls from $304.00 an episode to $184.04, and the partner's collection falls with it in all three states of the open Collaborative Care add-on question.

The consequence is disclosed rather than left to discovery. A lower collection denominator raises SoundBridge's share of what the partner collects, and at three add-on units that share now sits inside the full-service comparable band rather than at the ceiling of the intermediate one — which is the panel's original objection, arrived at by the company's own correction. On a fully-loaded facilitator cost the three-unit case, previously the base case at 10.5% partner retention, is now marginally negative.

That is a Row 9 question rather than a Row 12 one, and it is answered there. Pro Forma v78 carries both the fully-loaded and the incremental facilitator ledgers, and presents the payer-mix assumption as a stated range rather than a point. What belongs in this appendix is only this: the coding correction was made against the company's own commercial interest, the model has been rebuilt to carry it, and none of it was found by a reviewer.

Two dependencies the model must now carry are identified.

• 98980 requires practitioner time, not facilitator time. The descriptor reads "physician or other qualified health care professional time" — deliberately, in contrast to the RPM codes, which say "clinical staff/physician/other qualified health care professional." Twenty minutes of practitioner time per month plus one interactive communication with the patient is a real burden on the partner, and the health center ledger does not currently carry it. CMS's health center guidance permits auxiliary personnel to furnish care coordination under general supervision, which is in tension with the CPT descriptor. The issue is unresolved, and it is named.
• Minutes cannot be counted twice. RTM treatment-management time may not also be counted toward Collaborative Care. Because 98980 requires practitioner minutes while the company's Collaborative Care case rests on the facilitator's 120/90/90 contact minutes, the two lines are not obviously competing — but the arithmetic has not been done, and it will be done before either figure goes to a valuer.

One item runs in the company's favor. The CY2027 proposed rule would deny payment where remote monitoring is "delivered by contractors" rather than by clinical staff employed by the practice. The SoundBridge structure — the health center employing all three roles — is on the right side of that proposal by construction, and was arrived at for the reasons in Section E.4 rather than in anticipation of it.

E.7 What the company is not claiming

• That 98977 was ever correct. It was not.
• That the protocol is cognitive behavioral therapy.
• That FDA has reviewed or authorized anything. It has not, and none is required for the codes the company reports.
• That enforcement discretion is a safe harbor. It is a published policy that FDA can revise.
• That the platform is built. It is in development, and Section E.3 is the specification rather than a description of shipped software.
• That the partner covenant in Section E.4 is drafted. It is with counsel.
• That the partner economics are settled. They are not, and the RTM correction made them harder rather than easier. The three-state range and the two facilitator ledgers in Pro Forma v78 are the current answer, and that answer belongs to Row 9.
• That any RTM revenue is bankable before a pilot generates adjudicated claims.

E.8 The formal review in progress

The foregoing is the company's coding opinion. It is therefore the input to an independent review rather than the output of one, and it is not presented as anything else. Formal review is under way on both of the questions it turns on.

A written AMA response on the scope of 98978 would be the only authority on that question in existence, which is why it is listed first. The MAC inquiry is free, names the company's exact fact pattern, and is evidentially meaningful even though it binds nobody — and it goes in the partner's name, because the health center is the biller and SoundBridge is not.

No adviser is being asked to confirm a conclusion the company reached first. Where a review comes back against the company, this document will be reissued rather than defended.

E.9 Officer of record

The reimbursement and coding position stated in this document is owned by Nio Queiro, Chief Operating Officer. She leads revenue cycle for SoundBridge Health, brings 33 years of revenue cycle management across hospitals and health systems, and identified the reimbursement pathway on which the company's model is built. She has reviewed the code selection in Section E.5 and the reasoning behind it: the reporting of 98975, 98980 and 98981; the withdrawal of 98977; and the decision not to report 98978. The position is hers as an officer of the company, and she will answer for it.

What that signature is, and what it is not, is stated directly. It records that an officer with three decades in revenue cycle owns this position and stands behind it. It is not independent confirmation of it. No officer's review can be, however expert, and this document does not offer it as one — the independent routes are in Section E.8 and they remain in progress. The distinction matters enough to state twice: an officer owning a position is a governance fact; an outside authority confirming it is an evidentiary one. This document supplies the first and is seeking the second.

The founder prepared the analysis and the model. The reimbursement and coding position it states is the Chief Operating Officer's, and it was signed on that basis.

This document was prepared for the GHI III advisory panel in response to project plan Row 12. Code descriptors are drawn from a payer reproduction of the CY2026 CPT set and have not been verified against the AMA codebook, which is paywalled; the CPT 2026 Professional Edition is on order so that the RTM guideline paragraph can be read directly. Payment amounts are computed from the CMS Physician Fee Schedule indicator file as modified 1 July 2026 and were recalculated independently against that file. The CY2022 and CY2024 preamble passages quoted in Section E.2 were retrieved from the Federal Register's full-text index. This document states the company's understanding and its open questions; it is not a legal conclusion, not a coding audit, and not a valuation.

The document was executed by signature. Nio Queiro, Chief Operating Officer, SoundBridge Health, signed as officer of record for the reimbursement and coding position stated above; Chris Boardman, Founder, President and Chief Executive Officer, SoundBridge Health, signed as well.

E.10 Sources

CMS — CY2022 PFS final rule, 86 FR 64996, §II.E.4.(37) · CY2024 PFS final rule, 88 FR 78818 · CY2025 PFS final rule fact sheet (DMHT clearance condition) · CY2026 PFS final rule fact sheet (CMS-1832-F) · CY2026 correcting amendment, 12 March 2026 · CY2027 PFS proposed rule fact sheet · MLN901705, Telehealth & Remote Monitoring, December 2025 · MLN006397, Federally Qualified Health Center, March 2026 · PFS indicator file, 2026 · NCCI Policy Manual, Chapter I · 42 CFR 405.2464

FDA — Examples of Software Functions for Which FDA Will Exercise Enforcement Discretion · Clinical Decision Support Software, January 2026 · General Wellness: Policy for Low Risk Devices, January 2026 · 21 CFR 882.5801 · Product code ISD, 21 CFR 890.5360, Class II 510(k)-exempt · GUDID, SWORD Digital Therapist

Other — Anthem CG-MED-91, CY2026 RTM descriptors · APTA RTM practice advisory · AMA CPT Network · AMA Summary of Panel Actions, September 2024

First edition. Prepared August 30, 2026, at pre-seed stage, in response to a reviewer question. No protocol-specific efficacy data yet exists.

This appendix addresses the question of whether the brain entrains to the tempo of music: what the published evidence supports, what it does not support, and what the SoundBridge protocol depends on instead.

F.1 Position statement

The temporal structure of music modulates and synchronizes neural activity, and the motor system tracks tempo. That relationship is well established, and it is the relationship the protocol uses to pace breath and movement.

Whether a specified tempo produces a corresponding autonomic shift is an open question in the field — tested directly by three groups and demonstrated by none of them. The SoundBridge protocol does not depend on it.

The physiological active ingredient is paced breathing at 5.5 to 6 breaths per minute, for which the evidence is strong. Music is what paces that breathing, and what makes a patient willing to perform it ten times. Adherence is the company's central empirical claim, and completion is what the pilot measures first.

The material that follows is the evidence for those three statements, in that order. The question posed in the title of this appendix contains two claims that are routinely given one name; section F.3 separates them, and the sections after it take each in turn.

F.2 Summary of the evidence

"Entrainment" names two different claims. The first is auditory–motor: cortical responses appear at the beat frequency, and the motor system tracks tempo. The second is autonomic: heart rate, sympathetic tone or endogenous brain rhythms drift toward the tempo. The first claim is established. The second is not.

The auditory–motor claim is well supported. EEG frequency-tagging shows responses at the beat frequency and at metrical subharmonics when a listener merely imagines a meter. Manipulating tempo across 50 to 200 bpm changes both the strength and the direction of motor–auditory connectivity. The entire clinical literature on rhythmic auditory stimulation rests on this coupling.

The autonomic claim has been directly tested and has failed three times. Microneurography found no change in sympathetic nerve activity across five tempo conditions. A closed-loop study driving a pulse above each listener's live heart rate found no entrainment and no synchronization. A 2026 experiment yoking music tempo to each listener's own baseline heart rate found little evidence of entrainment — only a small general arousal rise, identical across conditions.

Where physiology does move with music, the pathway runs through respiration. Three independent groups converge on this. Autonomic change appears only when fast tempo is paired with fast breathing; the condition producing the greatest autonomic–cerebral coherence is also the condition producing the most regular breathing. The music is not acting on the heart. It is acting on the breath, and the breath is acting on the heart.

That is an argument for the protocol's design, not against it. The protocol pairs music with paced breathing and rhythmic movement. The mechanism the evidence supports is precisely the mechanism the curriculum uses. What the evidence does not support is a claim that a specified tempo produces a corresponding physiological shift, and SoundBridge does not make that claim.

Tempo does reliably move arousal — as a level effect, not a frequency effect. Faster music raises rated arousal, ventilation, respiratory rate and heart rate in a graded way. It does not cause any of those rhythms to converge on the beat. That distinction constitutes the whole of the answer to the question.

The music layer's role is pacing and engagement, and its incremental physiological contribution is untested. No trial has compared music-paced breathing against paced breathing alone. The company states that as an open question and has designed the pilot to begin answering it.

F.3 The question, split

Most statements of the form "the brain entrains to the music" slide between these two without noticing. The first is a claim about neural tracking of a stimulus. The second is a claim about frequency convergence of an endogenous rhythm. They are different physical claims about different systems, and they carry different evidentiary standing.

F.4 What is established

F.4.1 Cortical tracking of beat and meter

Nozaradan et al. (2011) used EEG frequency-tagging with a 2.4 Hz beat and recorded a sustained steady-state response at the beat frequency. When listeners imagined a binary meter, an additional response appeared at 1.2 Hz; imagining a ternary meter produced responses at 0.8 Hz and 1.6 Hz. The meter was never acoustically present — it was mental imagery — which is evidence that the response is partly endogenous rather than purely stimulus-driven. N = 8.

F.4.2 Tempo modulation of motor–auditory connectivity

Nicolaou et al. (2017) (N = 21) presented piano excerpts at 50, 100, 150 and 200 bpm with phase-randomized noise controls of identical acoustic content. Tempo changed both the strength and the directionality of long-range motor–auditory connections, most prominently in the alpha (8–12 Hz) and beta (12.5–18 Hz) bands. The same group showed movement-related motor activity tracking tempo variance in listeners who remained physically still.

This is the coupling on which the applied literature rests. Rhythmic auditory stimulation in gait rehabilitation — the best-established therapeutic use of musical rhythm, and the pathway described by Thaut et al. (2015) — depends on exactly this auditory–motor mechanism.

F.4.3 Bearing on the protocol

What this body of evidence supports in the SoundBridge protocol is the rhythmic-movement component, and the use of tempo to pace movement and breath. It supports nothing about heart rate or sympathetic tone, and section F.5 explains why the two must be kept apart.

F.4.4 A caveat stated unprompted

Even the supported claim carries a live methodological dispute. Whether frequency-tagged responses reflect entrainment of an endogenous oscillator or a train of evoked potentials is actively contested. Damsma et al. (2025), a modeling study, found that tempo-dependent enhancement at the beat frequency is reproduced equally well by an oscillator model and by a purely evoked model, with the evidence leaning slightly toward the evoked account. The measurement does not distinguish the mechanisms.

The practical consequence is that a statement of the form "music at 140 bpm produces brain rhythms at 2.33 Hz" does not appear in SoundBridge material. The defensible statement is that the temporal structure of music modulates and synchronizes neural activity, and that the motor system tracks tempo.

F.5 Findings not supported by the evidence

The proposition that physiology drifts toward the musical tempo has been tested head-on, three times, by three groups, using three different methods. All three tests are null.

F.5.1 Direct sympathetic recording

Bretherton et al. (2019) used microneurography — a tungsten microelectrode recording muscle sympathetic nerve activity directly from the peroneal nerve — across five tempo conditions from 60 to 180 bpm, alongside heart-rate variability, blood pressure and respiration. Sympathetic nerve activity did not change across tempo conditions. Parasympathetic indices did rise, but at both extremes — 60 and 180 bpm — with baseline low-frequency power predicting who responded. That is an autonomic effect. It is parasympathetic and non-monotonic, and it is not a drive toward tempo.

A precision point about that sample is recorded here because it cuts against the company's own reading: the study analyzed 52 participants overall, but only 24 underwent microneurography, and single sympathetic units were confirmed in 10. The sympathetic null therefore rests on a smaller confirmed base than the headline sample suggests, and it should be weighed accordingly.

F.5.2 Closed-loop pulse

Mütze et al. (2020) drove a rhythmic pulse at 25%, 40% and 55% above each participant's live measured heart rate, and separately at their heart rate. They found neither entrainment nor synchronization, with large individual differences. Their conclusion:

there is "no direct and simple correlation between the musical tempo and the heart rate contrary to what is often suggested in everyday psychology."

F.5.3 Tempo yoked to individual baseline

Juslin et al. (2026) manipulated ambient music tempo by ±3% and ±30% relative to each listener's own measured baseline heart rate, in a 2×2×2 design with 36 participants, with and without finger-tapping. Cardiac activity neither aligned with the target tempo nor shifted toward it. What appeared instead was a similar small increase in heart rate and arousal across all conditions — a general arousal effect of musical rhythm, not tempo-matching.

The provenance of that last study is material. Rhythmic entrainment is one of eight mechanisms in Juslin's own BRECVEMA framework, the standard taxonomy of how music induces emotion. It entered that framework grounded largely on a mid-century clinical observation. Juslin's own experimental work has manipulated four of the eight mechanisms and left entrainment out. When his group finally tested it directly, the paper's title asked whether it is "a plausible hypothesis in need of evidence." The researcher who named the mechanism is the one who published the null.

F.5.4 The 2025–26 literature

Nothing published in 2025–26 reverses this position. A 2026 review reports the conventional pattern — slow tempo toward parasympathetic dominance, fast tempo toward increased heart rate — but describes modulation of autonomic tone, not beat-matching. A 2025 study found neural responses peaking at target rates and harmonics, and framed this explicitly as frequency-specific entrainment "rather than frequency-shifting." A 2025 biofeedback study synchronizes music to the heartbeat — the opposite direction of causation.

F.5.5 What each source can and cannot carry

F.6 The respiratory pathway

This finding reframes the question, and it carries more weight than any other in this appendix.

Bernardi et al. (2006) (N = 24, randomized crossover) presented six musical styles spanning 55 to 150 bpm. Faster tempo produced graded increases in ventilation, respiratory rate, heart rate and blood pressure; slower tempo produced the reverse. The relationship tracked tempo, not stated musical preference or training.

That same study refutes frequency-matching within its own data. The ratio of musical tempo to breathing rate ranged from 4.1 to 8.7 across conditions. If entrainment were occurring, that ratio would sit at a stable small integer. It does not. This is a proportional arousal response, not frequency locking — and Bernardi et al. (2006) is the study most often cited in support of entrainment.

Its headline finding is worth stating alongside: inserted two-minute pauses produced the greatest relaxation of any condition, below baseline on blood pressure, heart rate and ventilation. Silence outperformed music.

Watanabe et al. (2015) isolated the pathway. Crossing metronome-paced respiration (15 vs 20 breaths/min) with acoustic tempo (60 vs 80 bpm), autonomic change appeared only in the fast-tempo-plus-fast-breathing condition — interaction F(1,15) = 9.79, p < .01, d = 0.83. Fast tempo at slow breathing produced no autonomic effect at all.

Mollakazemi et al. (2019) (N = 14) found that the condition producing the greatest autonomic–cerebral coherence was the slow song — which also produced the narrowest respiratory frequency bandwidth, that is, the most periodic breathing.

Three independent groups, three methods, one conclusion: the music is not acting on the heart. It is acting on the breath, and the breath is acting on the heart.

That is simultaneously an argument for pairing music with breath control — which is what the protocol does — and an argument that the breathing is the active ingredient. SoundBridge holds both readings.

F.7 Tempo as a level effect

Tempo is real. It operates as a level effect, not a frequency effect.

The film-scoring tradition has known for a century that tempo drives audience response: intimate scenes are scored slow, chase scenes fast, largely irrespective of genre. That craft knowledge is not folklore, and the empirical literature corroborates it.

Husain et al. (2002) (N = 36) crossed tempo (60 vs 165 bpm) with mode and found a clean double dissociation: tempo affected arousal but not mood, F(1,32) = 91.82, p < .001; mode affected mood but not arousal. Fernández-Sotos et al. (N = 63) found η² = .401 for happiness and .311 for tension across 90/120/150 bpm.

Tempo is also the most cross-culturally portable structural cue in music — which matters directly for a health center population. In Balkwill and Thompson (1999), Western listeners with no exposure to Hindustani classical music judged joy from tempo at r = .75 and sadness at r = −.92. In Fritz et al. (2009), a study of Mafa listeners in northern Cameroon with no prior exposure to Western music, both groups classified faster pieces as happy and slower pieces as fearful. Tempo crosses cultural and musical-literacy boundaries better than mode or harmony do. That is an argument for specifying tempo bands in a multilingual, multi-ethnic patient panel, and it is well supported.

The precise scope of this evidence should be noted. Nearly all of it measures perceived or felt arousal level — a graded response to how fast the music is. None of it measures frequency convergence. Three further precision points follow, each of which a reviewer may raise:

• Perceived and felt emotion diverge about 40% of the time. In one direct test, the positive relationship — the listener feels what the music expresses — held in only 61% of cases. An audience recognizing "this is a chase" is fully compatible with their pulse doing nothing.
• Tempo is confounded in practice with loudness, note density, register, dissonance and percussiveness. Where these are separated, note density has produced larger effects than tempo, and percussiveness has driven skin conductance where tempo did not.
• No specific bpm value is privileged. Preferred and spontaneous tempo clusters near 120 bpm, and the tempo-to-arousal slope is moderated by each listener's own spontaneous motor tempo. The mapping is relative to the person, not absolute.

The bearing on the question is this. The scoring tradition demonstrates that tempo communicates and induces arousal — a level effect, and one this protocol uses deliberately. The entrainment hypothesis asserts something structurally different: that physiology converges on the tempo — a frequency effect. A century of practice is strong evidence for the first and silent on the second.

F.8 The function of the music layer

If the music layer's incremental physiological contribution is unproven, the honest question is what it is doing in the protocol. There are two answers, and only the first is a physiological claim.

First: the music paces the breath and the movement. This is the auditory–motor coupling established in section F.4, and it is the mechanism on which the curriculum is built. It is also — and this is the point most easily missed — the only version of entrainment with any evidence of driving behavioral persistence. In the single study where music improved adherence to a physical activity program, the benefit came entirely from tempo-to-pace synchronization; the arm receiving pleasant personalized music without synchronization performed worse than no music at all. The entrainment that survives scientific scrutiny and the entrainment that has adherence evidence behind it are the same mechanism, and that mechanism is not "music is enjoyable." It is that a beat makes a repeated movement easier to sustain.

Second: the music is why a patient attends ten sessions. Music engages reward circuitry in a way few low-cost, self-administered stimuli do — established causally by Ferreri et al. (2019), a double-blind crossover in which levodopa and a dopamine antagonist moved musical pleasure and willingness-to-pay in opposite directions. The limit of that finding is stated plainly: no published study establishes that the reward value of music causes sustained engagement with a task or intervention. The step from reward to adherence is a hypothesis the company holds, not a finding it can cite.

What can be cited for retention is different and stronger. Windle et al. (2020), across 29 randomized trials and 5,294 patients, found that matching patients to a treatment they preferred cut dropout from 33.5% to 18.8% — RR 0.62. That is an argument about autonomy and choice, not about content pleasantness. And in a population where a documented 44% of eligible patients never engage with Collaborative Care at all, and where the recorded barriers are stigma, privacy, perceived non-need and competing demands, a session organized around music is a session a patient can attend without it looking like a mental health appointment. That flanks a barrier the literature actually records.

F.9 Claims not made

The following claims are set out here so that a reviewer does not have to locate them independently.

• SoundBridge does not claim that a specified tempo produces an autonomic shift toward that tempo. It is unproven by anybody; it is an open question in the field, not a gap peculiar to SoundBridge.
• SoundBridge does not claim that music produces particular brainwave states associated with relaxation.
• SoundBridge does not claim that adding music to paced breathing improves on paced breathing alone. No trial has tested it in that direction.
• SoundBridge does not claim that patient-chosen music is physiologically superior to assigned music. The largest body of evidence in the field says it is not; personal relevance is justified as an engagement and adherence choice.
• SoundBridge does not cite the Cochrane music-therapy review as evidence for this protocol. Its effects are for therapist-delivered music therapy and do not transfer to a facilitator-delivered listening-and-breathing curriculum.

F.10 The evidentiary weight of null findings

A null is not automatically stronger than the claim it displaces, and the honest reading of this literature has limits worth naming.

The samples are small. Bretherton et al. (2019) confirmed single sympathetic units in 10 participants; Juslin et al. (2026) ran 36; Mollakazemi et al. (2019) ran 14; Nozaradan et al. (2011) ran 8. These are laboratory studies with the sample sizes that invasive or high-density recording imposes. None of them is a population-scale finding, and a well-powered future study could revise any of them.

Absence of a demonstrated effect is not proof of no effect. What the literature establishes is that three groups looked directly for autonomic entrainment with three different methods and did not find it. That is meaningfully different from — and weaker than — a demonstration that it cannot occur.

The strongest claim available in either direction is modest. SoundBridge does not assert that music cannot shift autonomic state toward a tempo. The company asserts that no one has shown it, that the investigators who looked hardest did not find it, and that a protocol should not be built on a premise in that condition.

That is why the matter is an endpoint rather than a wound. The cortical-to-autonomic bridge at a specified tempo is unproven by anybody. It is an open question in the field, not a gap peculiar to SoundBridge, and the pilot treats it as a named hypothesis to test rather than a premise to assume.

F.11 Pilot endpoints

The cortical-to-autonomic bridge at a specified tempo is unproven by anyone. Stating it as a named hypothesis the pilot is designed to test is a stronger position with a scientific reviewer than asserting it as a premise — because an assertion is checkable in ten minutes and a hypothesis is not.

The pilot as scoped is a single-arm feasibility and outcomes study. It can establish engagement, adherence and pre-post symptom change. It cannot separate protocol effect from regression to the mean, expectancy, or the Collaborative Care wrapper itself. The controlled comparison that would settle the music question — music-paced breathing against paced breathing alone — is the right use of a subsequent round, and SoundBridge would report a null.

F.12 Position restated

The position is unchanged from section F.1, and unqualified by anything in between.

The temporal structure of music modulates and synchronizes neural activity, and the motor system tracks tempo. That is well established, and it is what the protocol uses to pace breath and movement.

Whether a specified tempo produces a corresponding autonomic shift is an open question in the field — tested directly by three groups and demonstrated by none of them. The SoundBridge protocol does not depend on it.

The physiological active ingredient is paced breathing at 5.5 to 6 breaths per minute, for which the evidence is strong. Music is what paces it, and what makes a patient willing to do it ten times. Adherence is the company's central empirical claim, and completion is what the pilot measures first.

F.13 References

Balkwill LL, Thompson WF. A cross-cultural investigation of the perception of emotion in music: psychophysical and cultural cues. Music Perception. 1999;17(1):43–64. doi.org/10.2307/40285811

Bernardi L, Porta C, Sleight P. Cardiovascular, cerebrovascular, and respiratory changes induced by different types of music in musicians and non-musicians: the importance of silence. Heart. 2006;92(4):445–452. doi.org/10.1136/hrt.2005.064600

Bretherton B, Deuchars J, Windsor WL. The effects of controlled tempo manipulations on cardiovascular autonomic function. Music & Science. 2019;2. doi.org/10.1177/2059204319858281

Damsma A, de Roo M, Doelling K, Bazin PL, Bouwer FL. Tempo-dependent selective enhancement of neural responses at the beat frequency can be mimicked by both an oscillator and an evoked model. Cerebral Cortex. 2025;35(9):bhaf258. doi.org/10.1093/cercor/bhaf258 — a correction notice was subsequently published, Cerebral Cortex 2026;36(3):bhag039.

Ferreri L, Mas-Herrero E, Zatorre RJ, et al. Dopamine modulates the reward experiences elicited by music. PNAS. 2019;116(9):3793–3798. doi.org/10.1073/pnas.1811878116 — N = 27, double-blind within-subject crossover.

Fritz T, Jentschke S, Gosselin N, et al. Universal recognition of three basic emotions in music. Current Biology. 2009;19(7):573–576. doi.org/10.1016/j.cub.2009.02.058

Husain G, Thompson WF, Schellenberg EG. Effects of musical tempo and mode on arousal, mood, and spatial abilities. Music Perception. 2002;20(2):151–171. doi.org/10.1525/mp.2002.20.2.151

Juslin PN, Harmat L, Barradas G, Omstedt G, Redtzer V. Rhythmic entrainment of heart rate as a mechanism for musical emotion induction: a plausible hypothesis in need of evidence? Psychology of Music. 2026;54(1):42–62. doi.org/10.1177/03057356241302809 — experimental study, N = 36, tempo yoked to individual baseline heart rate.

Mollakazemi MJ, Biswal D, Elayi SC, et al. Synchronization of autonomic and cerebral rhythms during listening to music. Physiological Research. 2019;68(6):1005–1019. doi.org/10.33549/physiolres.934163

Mütze H, Kopiez R, Wolf A. The effect of a rhythmic pulse on the heart rate: little evidence for rhythmical "entrainment" and "synchronization". Musicae Scientiae. 2020;24(3):377–400. doi.org/10.1177/1029864918817805

Nicolaou N, Malik A, Daly I, et al. Directed motor-auditory EEG connectivity is modulated by music tempo. Frontiers in Human Neuroscience. 2017;11:502. doi.org/10.3389/fnhum.2017.00502

Nozaradan S, Peretz I, Missal M, Mouraux A. Tagging the neuronal entrainment to beat and meter. Journal of Neuroscience. 2011;31(28):10234–10240. doi.org/10.1523/JNEUROSCI.0411-11.2011

Thaut MH, McIntosh GC, Hoemberg V. Neurobiological foundations of neurologic music therapy: rhythmic entrainment and the motor system. Frontiers in Psychology. 2015;5:1185. doi.org/10.3389/fpsyg.2014.01185

Watanabe K, Ooishi Y, Kashino M. Sympathetic tone induced by high acoustic tempo requires fast respiration. PLOS ONE. 2015;10(8):e0135589. doi.org/10.1371/journal.pone.0135589

Windle E, Tee H, Sabitova A, et al. Association of patient treatment preference with dropout and clinical outcomes in adult psychosocial mental health interventions. JAMA Psychiatry. 2020;77(3):294–302. doi.org/10.1001/jamapsychiatry.2019.3750 — 29 RCTs, N = 5,294.

Scope and limitations

This appendix is a review of published literature prepared to accompany a pre-seed investment discussion. It is not a clinical guideline, not a systematic review, and not a claim of efficacy for the SoundBridge protocol, for which no trial data yet exists. Effect sizes are reported with confidence intervals where the source provides them, and where the literature is weak — small samples, conference proceedings, absent sham controls, unblindable interventions — that limitation is stated alongside the citation rather than left for the reader to discover.

Appetite for music-based care in the health center population: what the evidence establishes, what it does not, and what the pilot will settle.

Prepared August 28, 2026; revised for distribution August 30, 2026. Citations verified against primary sources.

G.0 Scope and purpose of this appendix

This appendix responds to a panel comment directed at Section 4.3 of the main submission. The comment under response reads:

"Answer the demand-side question: appetite for music-based care in an older, lower-income Medicare and FQHC population. Section 4.3 — deserves an answer even if clinical evidence arrives tomorrow."

The appendix answers that comment in full. It also corrects one factual premise the comment contains regarding the composition of the health center patient population, because that premise bears directly on market size, payer mix and pricing, and because every subsequent question would otherwise inherit the error. The correction is set out in Section G.1. The answer occupies Sections G.2 through G.6.

One framing observation precedes both. The comment treats demand as a binary. The evidence does not settle a binary; it settles direction. The open question is not whether this population has an appetite for music as a health modality. It is what share of that existing appetite converts into enrollment in a specific clinical product at a specific price. That is a legitimate diligence question, it is considerably narrower than the comment implies, and a pilot answers it more efficiently than additional literature can. Section G.5 states precisely what remains open and what result would establish that SoundBridge is wrong.

Every figure below was retrieved from the primary source or the publisher's abstract. Where a figure could not be confirmed against the primary source it was removed rather than softened, and the exclusion is recorded in Section G.7.

G.1 Correction of record: the health center population is neither older nor predominantly Medicare

The HRSA Uniform Data System for 2024 reports 32,387,774 patients across 1,359 awardees. The age distribution of that panel is as follows.

Mean patient age is roughly 35. The payer mix is Medicaid/CHIP 49%, private 22%, uninsured 18%, Medicare 7%, and dual-eligible 4%.

The question as framed therefore addresses 12% of the patient panel and 7% of the payer mix. Medicare matters, and it is the fastest-growing segment, having added 227,601 patients between 2023 and 2024, but it is the tail of this market rather than its body. SoundBridge bills through Collaborative Care and Remote Therapeutic Monitoring, both of which are payer-agnostic. The addressable population is 19 million working-age adults first, and 3.8 million Medicare and dual-eligible patients second.

The income half of the comment's premise is accurate. Ninety percent of health center patients are at or below 200% of the federal poverty level, and 67% are at or below the poverty line. Sections G.2 and G.3 establish that the demand evidence is strongest, not weakest, in precisely that stratum.

G.2 The demand ladder: five independently sourced steps

The argument advanced here does not rest on a single study. It is a chain that begins with near-universal behavior and narrows progressively toward the specific target population, each rung standing closer to that population than the one before it.

Rung 1 — Near-universal music engagement among older adults

The University of Michigan National Poll on Healthy Aging, The Sound of Music (2024), surveying adults aged 50–80, reports:

• 56% listen daily, and a further 29% listen at least several times a week (85% combined)
• 98% report at least one health-related benefit, comprising 75% stress relief, 65% mood, and 61% memory
• 41% say music is very important to them, and more than half of Black and Hispanic respondents say so

The health center panel is 40% Hispanic and 17% Black. The cultural salience of this modality is accordingly highest in the population for which SoundBridge is building.

The limit of this rung, stated plainly: listening to music is not the same as consenting to clinical music care. What Rung 1 establishes is the behavioral base from which any adoption would have to be drawn, and that base is very large.

Rung 2 — Existing use of and interest in non-pharmacological care, with coverage as the gate

The University of Michigan National Poll on Healthy Aging, Use of and Interest in Integrative Medicine Strategies, July 26, 2022 (n = 2,277, adults 50–80), reports:

• 66% have used at least one integrative medicine strategy to prevent or treat a health concern
• A further 21% have not used one but are interested in trying
• Among those who said such services were not covered or were unsure, 84% said they would be likely to try them if insurance covered them (34% likely, 50% somewhat likely), and 66% did not know whether they were covered

This is the rung of greatest consequence for a reimbursed model. It indicates that low observed utilization of integrative care in this population should not be read as low appetite: awareness and coverage suppress it. SoundBridge enters through an existing covered benefit, which removes the barrier the poll identifies.

Two caveats are recorded here because they are material. First, this is a poll report rather than a peer-reviewed study. Second, cost was not the leading barrier: among those neither using integrative strategies nor interested in them, "no interest or no need" ranked first at 33%, ahead of cost at 27% and unawareness at 26%. That finding is carried forward into Section G.4 rather than set aside.

Rung 3 — Demand for complementary care measured inside an actual health center

Ho DV, Nguyen J, Liu MA, Nguyen AL, Kilgore DB. Use of and Interests in Complementary and Alternative Medicine by Hispanic Patients of a Community Health Center. J Am Board Fam Med. 2015;28(2):175–183.

The study surveyed 150 patients at an FQHC serving a medically underserved, predominantly Hispanic community in Orange County, California. The majority were insured through Medicaid or the county Medical Services Initiative, and more than 90% were below 200% of the federal poverty level. Findings:

• 63% had used at least one complementary or alternative modality
• 72% (95% CI 65–79%) wanted additional such resources offered at the health center
• Among the services patients wanted offered: massage 30%, mind-body relaxation techniques 16%, yoga 13%, acupuncture 11%
• Patients were generally receptive to discussing these approaches with their clinicians

Ho et al. (2015) is not a music study. It is direct evidence against the premise that a safety-net population is inherently uninterested in non-pharmacological care, and it is evidence measured inside an actual health center rather than in an affluent sample.

Rung 4 — Music specifically, already in use, in a low-income behavioral health population

Prasad K, Prasad A, Dyer NL, Bauer BA, Soderlind JN, Fischer KM, Croghan IT, Kaufman CC, Rosmarin DH, Wahner-Roedler DL. Use of Complementary and Integrative Medicine Among Low-Income Persons With Mental Health Disorders. Mayo Clin Proc Innov Qual Outcomes. 2025;9(1):100585.

The study surveyed 102 low-income patients at outpatient evaluations in a community behavioral health center; depression was present in 87% and anxiety in 85%. Findings:

• 72% had used at least one complementary or integrative modality
• Music was the third most common modality, used by 36%, behind prayer (41%) and spirituality (37%)
• Perceived benefit: mood 49%, stress 49%, sadness 43%
• More than two-thirds had never discussed this use with their physician

Prasad et al. (2025) is the strongest demand-side citation available, because it stacks four attributes of the target market simultaneously: low income, mental health diagnosis, safety-net setting, and music specifically, already in use for the outcomes the protocol targets. The two-thirds who never disclosed the practice to a clinician constitute the demand-side statement of the thesis itself: the behavior exists, and no delivery mechanism captures it.

A scope note on this source is warranted. The published abstract characterizes the population only as low-income. It does not report Medicaid or Medicare eligibility, and SoundBridge does not describe it as such in any of its materials.

Rung 5 — Revealed preference at program scale, with payers already contracting

These figures should be read against the relevant benchmarks: adult psychotherapy dropout of 19.7% (669 studies, 83,834 clients); depression-app dropout of 26.2% raw and 47.8% adjusted; and real-world Collaborative Care at Montefiore's seven practices, three of them FQHC-designated, where 44% of eligible patients never engaged at all.

G.3 The income premise and its bearing on the thesis

The barrier operating in this population is not indifference to treatment. It is a mismatch between patient preference and the treatments on offer:

• Minority patients are 1.8 to 2.6 times more likely to prefer counseling over medication (OR 2.6 Black, 2.5 Asian-Pacific Islander, 1.8 Hispanic — Givens 2007)
• 29% of adults who skipped needed mental health care said they did not think it would help (KFF 2023)
• 21% of Black adults could not find a provider with a shared background, versus 10% of White adults (KFF 2023)

This is a preference problem. Preference is what a non-stigmatizing, culturally salient, music-based protocol addresses. The low-income half of the comment's premise therefore operates in favor of the thesis rather than against it.

G.4 Claims not made, and the counter-signals

SoundBridge does not claim that the evidence proves strong demand for music-based care among older, low-income Medicare and health center patients. That statement would be stronger than the evidence supports. Six findings cut against the thesis. All six are set out below, each accompanied by SoundBridge's reading, together with the delivery data underlying the largest of them.

G.5 The precise statement of what remains open

The following sequence is established by the evidence assembled in Section G.2: music engagement → perceived health benefit → complementary-care utilization → music specifically used as a health modality by low-income behavioral health patients → stated demand for such services inside a health center → willingness to try when covered.

The following is not established: what percentage of referred health center patients will consent, initiate, and complete a ten-session protocol at the stated price.

That is a conversion question rather than an appetite question. It cannot be answered by additional literature, and it does not need to be, because it is answered by the pilot's first three endpoints, which exist for this purpose.

Criteria that would falsify the thesis: a completion rate at or below the psychotherapy benchmark would undercut the adherence thesis on which the model rests. Such a result would be reported.

G.6 On the final clause of the comment

The clause "deserves an answer even if clinical evidence arrives tomorrow" is correct, and SoundBridge's position is stronger than that clause allows.

In safety-net behavioral health, engagement is the binding constraint rather than efficacy. Collaborative Care has 79 randomized trials behind it, and the national health center depression remission rate at 12 months is nevertheless 13.8%. Montefiore demonstrates the mechanism: 44% of eligible patients never engaged, and a quarter of those who did never returned. Efficacy that patients do not consume produces 13.8%.

Demand is therefore not a question standing beside the clinical question. It is the primary claim. Music functions as the adherence mechanism rather than as an asserted physiological active ingredient, and completion is the central empirical claim the pilot is built to test.

G.7 Appendix: citations considered and excluded

Three meta-analyses of music interventions in older adults were considered for this section and left out. All three are real and were verified. They are recorded here so that the record of what was considered is complete.

A structural point applies to the three together. They are not independent lines of evidence. Wang (2023) and He (2026) both searched CNKI, Wanfang, VIP and CBM and draw on a substantially shared pool of Chinese-language trials that Dhippayom (2022), which searched only international databases, structurally could not include; eleven of He's fourteen trials are Chinese-language. Presenting all three as converging would overstate the evidence.

The substantive reason for exclusion is separate. All three are efficacy evidence, and this is the demand section. Placing clinical meta-analyses here would concede the frame that demand must be argued from clinical benefit. It must not, which is the point of the comment's own final clause.

G.8 Sources

Population and system — HRSA UDS 2024 National Report, Tables 3A, 5, 6A, 6B · KFF, Community Health Center Patients, Financing, and Services · NACHC, Health Center Workforce policy paper · Commonwealth Fund, 2024 FQHC Survey

Demand — University of Michigan NPHA, The Sound of Music (2024) · University of Michigan NPHA, Use of and Interest in Integrative Medicine Strategies (July 26, 2022), n = 2,277, doi:10.7302/4879, poll report, not peer reviewed · Ho DV et al., J Am Board Fam Med. 2015;28(2):175–183, doi.org/10.3122/jabfm.2015.02.140210, PMID 25748757 · Prasad K et al., Mayo Clin Proc Innov Qual Outcomes. 2025;9(1):100585, doi.org/10.1016/j.mayocpiqo.2024.11.002, PMID 39758507, open access · Johnson et al., Community of Voices, J Gerontol B (2020) · Erkkilä et al., BJPsych (2011) · Mass Cultural Council, social prescribing outcomes · Art Pharmacy longitudinal outcomes, Frontiers in Public Health (2026)

Benchmarks and counter-signals — Swift & Greenberg, Premature Discontinuation in Adult Psychotherapy (2012) · Torous et al., dropout in depression app trials (2019) · Blackmore MA et al., Psychiatr Serv. 2022;73(8):842–848, doi.org/10.1176/appi.ps.202000924 · Hemmy L et al., J Alzheimers Dis. 2026;111(2):890–896, doi.org/10.1177/13872877261437271 · McCreedy EM et al., METRIcAL, JAMDA (2022) · Conard M et al., METRIcAL secondary analysis (2025), PMC12138331 · Six SP et al., JMIR Ment Health (2021) · Givens et al. (2007) · KFF Survey of Racism, Discrimination and Health (2023) · Carr et al., SYNCHRONY feasibility study (2023)

G.9 Verification policy

Every figure in this document was retrieved from the primary source or the publisher's abstract and verified against it. Where a figure could not be confirmed, it was removed rather than softened, and the exclusion is stated in Section G.7. This document responds to a comment on Section 4.3. It is not a claim of efficacy for the SoundBridge protocol, for which no trial data yet exists.

This appendix reproduces, in third-person form, the company's response to the panel's Row 15 question (SOUN-11), prepared 29 August 2026 and built on Pro Forma v78. It is neither legal advice nor a valuation, and it is not a substitute for either. It states the company's position and its open questions.

The panel's question was put as follows:

State whether SoundBridge supplies, brokers, or assumes the care manager and psychiatric consultant staffing.

The question matters, as the panel's read-out at section 4.4 records, because the three verbs imply three different companies, and one of them changes the revenue split. The subject of the question is the behavioral health care manager seat and the psychiatric consultant seat within the Collaborative Care model: whether SoundBridge supplies, brokers, or assumes either of them.

The answer, in one sentence, is none of the three. SoundBridge does not supply, broker, or assume either role. The health center employs the behavioral health care manager, employs the psychiatric consultant, and employs the treating practitioner of record. SoundBridge occupies none of the three Collaborative Care seats and employs none of the people in them.

H.1 The three verbs, answered separately

The row offers three options because they are three different things, and denying one is not denying the others. Each is answered below with the fact that would falsify the denial. Two formulations of the same answer are preserved here, since they differ in wording: the condensed statement given in the summary treatment, and the fuller statement given in the source document. In each, the third column states the fact whose existence would defeat the denial in the second, so that the denial is not merely asserted but made checkable.

On the third verb there is a second reading, and it deserves a direct answer as well. SoundBridge does model the partner's staffing. The facilitator is carried at $115,062 per FTE-year all-in, and the psychiatric consultant at $54.99 per episode, in the partner-side ledger. Modeling a cost the partner bears is not assuming it. The distinction is visible in the fee: the license is payable in full whether the partner staffs the program well, badly, or not at all.

H.2 The two clinical seats

The two seats named in the question are stated individually below, together with the treating and billing practitioner seat, so that each can be checked against the company's answer separately rather than as a single undifferentiated claim.

The psychiatric consultant seat is the cleaner of the two and should be read first. SoundBridge supplies nothing into it whatsoever. Treatment-intensity signals surface in the registry the consultant already reviews on its normal cadence; that is a display, not a service, and it replaces no part of the consultant's function.

H.3 The point at which the position is most testable

Facilitator certification is the only thing SoundBridge supplies that touches the care-manager seat, and a careful reader will press on it. The company's position is set out below in a form that permits it to be checked:

• The health center selects, hires, schedules, supervises, pays, and may dismiss the facilitator. SoundBridge does none of these and has no consultation right over any of them.
• Certification attests that a person the partner has already employed can deliver the protocol to script. It is a competency credential in a vendor's own method, not a clinical license and not a condition of employment the vendor controls.
• SoundBridge does not require that any named individual be assigned, and does not require that the person be assigned to the program at all.

The edge case is flagged here by the company itself, rather than left to be discovered by a reader who would then have found it unannounced. If SoundBridge holds an unqualified right to withdraw a facilitator's certification, that right is the closest thing in the arrangement to influence over who occupies the seat — not because it terminates employment, which it cannot, but because it can render an employee unable to deliver the program. The remedy is drafting, and it should be drafted before the FMV valuer sees the agreement: de-certification limited to specified, evidenced fidelity failures, with notice, a remediation period, and re-certification on cure. This is on the list for counsel and is not yet in the draft.

Everything else about the seat belongs to the health center. Selection, hiring, scheduling, supervision, payment and dismissal are the center's, and SoundBridge has no say in any of them.

H.4 Durability: contractual rather than descriptive

The position is contractual. The absence of clinical labor is drafted as a recital and a covenant in the partner agreement, alongside the partner's covenant that acute-risk escalation and all crisis response are its own. A valuer is being asked to value a contract, not a snapshot of how the company happens to operate this quarter.

It is also, in this setting, close to compulsory. CMS's health center care-management guidance asks the question directly and answers that the behavioral health care manager may be employed by or under contract to the RHC or FQHC, not to another company. If that provision governs, no vendor can supply or broker a care manager into a health center and be paid out of what the center collects for that person's work. It makes the arrangement SoundBridge already sells the only shape this can lawfully take in the FQHC channel.

Two caveats belong with that reading and are carried in every document that cites it. The FAQ is dated December 2019 and instructs billing through G0511 and G0512, which the source document describes as both retired on 1 January 2026 — a statement corrected at H.8 below. NACHC's March 2026 guidance omits the qualifier, and CMS's current health center booklet does not repeat it. The provision is therefore treated as a live question for counsel, not as a settled rule. The company's answer to Row 15 does not depend on how it resolves — the arrangement is structured this way regardless — but the strength of the answer does.

H.5 The staffed template, disclosed

A second template exists in company materials, under which SoundBridge would employ clinical personnel. It is not the base case, it is not what is being sold, it is not what the FMV brief values, and it is not modeled in Pro Forma v78. Counsel has been asked to sequence the corporate-practice question behind the FAQ discussed above, because if that provision governs, the staffed template is foreclosed in health centers by federal payment rule before state doctrine is reached. The founder's working conclusion is that the staffed template is not a real product in health centers.

It is disclosed here because a panel member who finds it later, unmentioned, will read it as concealment. It is mentioned in the FMV brief for the same reason.

H.6 Why supplying labor is unnecessary as well as impermissible

The preceding section addresses whether SoundBridge may supply clinical labor into a health center. This one addresses whether it would need to, and the answer is that it would not.

Serving the estimated ten-state gap requires roughly 188 facilitator FTEs at the lower presenting-rate estimate and 470 at the higher, against 907 unlicensed behavioral health FTEs already employed by the 192 health centers in those states.

The constraint is not headcount. The workforce is already on the partners' payrolls. What it lacks is a structured intervention it is permitted and equipped to deliver, and a supervision record that makes delivering it defensible. That is what the license supplies — and, per the partner ledger, redirecting existing staff rather than hiring net-new is the basis on which the partner's economics work at the current price.

H.7 Consequences for the valuation

Three consequences follow from the position set out above, and the valuer should carry each forward. The first of them is stated against the company's own interest and is not softened in this appendix any more than it is in the FMV brief.

• The vendor-share comparison is not like-for-like. The license sits inside the full-service comparable band at three and four add-on units while supplying no clinical labor at all. That is stated against the company in the FMV brief and is not softened here.
• The full-service band may describe an arrangement that cannot lawfully exist in this setting. If the FAQ governs, the comparables that include supplied care managers are not available comparables for an FQHC.
• SoundBridge bears no staffing risk: no wage inflation, no vacancy, no turnover, no supervision liability. A valuer should price the arrangement accordingly — in both directions.

H.8 A date correction

Section 4 of the source document states that G0511 and G0512 "both retired on 1 January 2026." That statement is wrong as to G0511. The verified position is that G0511 was replaced by individual codes required from 1 January 2025, with a CMS transition through 30 September 2025, and that G0512 was replaced from 1 January 2026. See the Document Control page.

Sources and disclaimer

Sources: CMS, Care Management Services in Rural Health Clinics and Federally Qualified Health Centers FAQs, December 2019 (Q60/A60); CMS Behavioral Health Integration FAQs, December 5, 2023; NACHC, Reimbursement Tips: Psychiatric Collaborative Care Model, March 2026; SoundBridge Health Two-Template GTM Strategy Matrix (50 States), July 23, 2026 (company work product, independent validation sought); SoundBridge Health Pro Forma v78, 29 August 2026. Capacity and workforce figures are as stated in the evaluator brief, fourth edition. This appendix states the company's position and its open questions. It is not legal advice, not a valuation, and not a substitute for either.

Source document: SBH_Product_Story_Row16_v2.docx, prepared 29 August 2026, second edition 30 August 2026. Clinical claims are limited to what evidence-map.md (third edition) supports; population figures are drawn from HRSA UDS 2024. The document is neither legal nor valuation advice. Its status at the time of writing is answered: the slide copy exists, but it has not been built into the deck.

The panel's request, recorded as SOUN-12, was stated as follows:

Add the missing product story: a slide on why music and mental health, and the after state for the patient.

The read-out at §4.5 records why the request mattered: an advisor who knows the product personally still left the deck unconvinced of the need.

The governing proposition of the entire account is that music is the adherence mechanism and slow-paced breathing is the active ingredient. The evidence map states this explicitly — Links 5, 6 and 8 do not support music as the physiological agent — and any version of the material that implies otherwise is checkable and wrong in about ten minutes. The honest version is also the stronger presentation, because engagement is the binding constraint in safety-net behavioral health and music is an engagement claim rather than a therapy claim.

I.1 What the account has to supply

The existing package explains the care gap, the workforce answer, the coding position and the economics. It does not explain why a patient would undertake the protocol, or what that patient holds at the end. A reader can follow the entire package and still not know what the product is for.

Two failure modes are available, and the current materials risk both.

• Claiming that music is the therapy. The Cochrane music-therapy effects (SMD −0.98, clinician-rated) belong to qualified music therapists and do not transfer to a facilitator-delivered listening-and-breathing curriculum. Citing them invites a clinician to dismantle the section.
• Answering a product question with a reimbursement question. The package already does this elsewhere: everything about the patient is at present expressed as a code, a minute, or a denominator.

The temptation, in other words, is to claim that music is the therapy. That claim is checkable and wrong, and a clinician would take the section apart in ten minutes. The honest version is stronger in any case. In safety-net mental health the problem is not that treatment fails; the problem is that patients never arrive, or arrive once. Collaborative Care has 79 randomized trials behind it and the national remission rate is still 13.83%. At Montefiore, 44% of eligible patients never engaged. Treatment nobody consumes produces 13.83%. Engagement is therefore the constraint, and music is an engagement claim, not a therapy claim.

I.2 The account

The material occupies two slides rather than one. The first carries the argument; the second is the only patient-level slide in the deck and should not be compressed to make room for the first. The material reproduced in I.2.1 and I.2.2 was drafted as presentation copy and is presented here in that form. The line that governs both slides is that music is why the patient stays, and breathing is what treats them.

I.2.1 The argument

The constraint is engagement, not efficacy. Collaborative Care has 79 randomized trials behind it. The national health center depression remission rate at twelve months is 13.83%. At Montefiore, 44% of eligible patients never engaged, and a quarter of those who did never came back. Efficacy that patients do not consume produces 13.83%.

Music is the thing this population will actually accept.

These figures stand against 29% of adults who skipped needed mental health care because they did not think it would help, and a 1.8–2.6× stated preference for counseling over medication among Black, Hispanic and Asian patients. Cultural salience is highest exactly where the care gap is.

Where the approach has been tried, people finish. A community choir trial enrolled 61% and retained 92% at six months. An individual music-therapy RCT recorded 18 of 20 sessions attended, with 12% dropout against 24% in the control arm. The relevant benchmarks are psychotherapy dropout at 19.7% and depression-app dropout at 26–48%.

What actually treats the patient is the breath. Ten sessions teach paced breathing at approximately 6 breaths per minute, which engages baroreflex resonance; gain persists at rest between sessions. Pooled across 26 RCTs, the effect sizes are depression g = −0.40, anxiety −0.32, and stress −0.35. Music paces the breath and carries the patient back for session two.

A footnote is to be carried on the slide itself. The Montefiore 44% is a referral-conversion figure — the share of eligible patients who never engaged at all. The 85% in the financial model is an attendance figure among patients already enrolled. The two measure different things and are not in conflict, but a program with Montefiore's engagement profile would not sustain the modeled throughput. Both are first-order pilot measurements, and the delivery-feasibility endpoint is where they are settled.

I.2.2 The patient trajectory from day zero to month six

This is the only patient-level slide in the deck, and the one to retain if only one survives.

A condensed rendering of the same comparison appears in the summary treatment of Row 16, with materially different wording, and is reproduced here for completeness.

SoundBridge does not claim to raise the remission rate; no trial data exists. The company claims to capture it. The pre-assessment also creates index events, which adds patients to the partner's denominator. Whether capture nets positive depends on the worklist being used.

The evidence that would show the account to be wrong is stated on the slide. A ten-session completion rate at or below the 19.7% psychotherapy benchmark would undercut the adherence thesis on which the whole model rests. A remission rate at or near 13.83% would indicate that the protocol adds nothing to usual care. Either result should change the plan, and the company would report it.

If the deck can carry only one slide, the instruction is to keep the patient-trajectory slide and fold the argument into its headline: cut the demand table to the two >50% figures, drop the completion benchmarks to the notes, and keep the falsifiers, which are the cheapest credibility in the section.

I.3 The argument beneath it

The argument starts from the failure, not from the modality. The safety-net behavioral health problem is not that the treatment does not work. It is that the patient never arrives, or arrives once. Seventy-nine trials of Collaborative Care and a 13.83% national remission rate constitute the whole argument in one sentence, and the pairing reframes the product from a novel therapy — which invites a demand for efficacy data the company does not have — to a delivery mechanism for a therapy that already works, which is a claim about conversion and is what the pilot measures.

Three reasons account for disengagement in this population, and music answers each of them.

The honest mechanistic statement belongs on the slide rather than in a footnote. Music tempo moves respiration and autonomic arousal, and the effect is breath-mediated: autonomic change appears only when tempo and breathing move together. That is simultaneously the argument for pairing music with breath control and the argument that the breathing is doing the work. Stating it plainly costs nothing and secures the reviewer's trust for the rest of the deck.

I.4 The after state in the patient's sequence

The after state, stated in one sentence: at the end of three months the patient has something they can do themselves, in a moment when they need it, without an appointment — and the clinic has proof it happened.

Deterioration is handled explicitly. Each session is evaluated against a rule set authored by a licensed psychiatrist, and treatment-intensity signals surface in the registry the psychiatric consultant already reviews. Acute risk is handled in session by the facilitator under the health center's own escalation protocol, with a warm handoff. That provision is not a software feature and is not billed. It should be stated before it is asked about.

I.5 Claims the account does not make

Four lines are disclaimed, all of which a reviewer can check.

• Not that music therapy works. The Cochrane effects are for qualified music therapists, short-term, with no quality-of-life effect. It is a different intervention.
• Not that music plus paced breathing beats paced breathing alone. The best test is null (Kow 2018). This is Link 5, and it is open.
• Not that patient-chosen music is physiologically superior to assigned music. This is not supported.
• Not that the protocol raises remission. It improves capture — and because the pre-assessment creates index events, it adds patients to the partner's Measure 21a denominator. Whether it nets positive depends on the worklist being used. This is the same claim the patient-trajectory slide carries, stated the same way; the two must not drift apart.

The falsification criteria appear on the patient-trajectory slide rather than behind it. Naming the result that would kill the thesis is what secures credibility for everything else in the section, and it costs nothing the company was not going to measure in any case.

I.6 Anticipated objections

Does music engagement decline with income?

It does. Adults under $60K, those aged 65–80, those in fair or poor health, and socially isolated adults all listen less daily. That is an argument for a prescribed, scheduled, facilitator-led protocol rather than a self-serve application. The people least likely to reach for music on their own are the reason a delivery mechanism has to exist.

Has the approach stalled in nursing homes?

It has. In the Music & Memory pragmatic trial only 40% of residents had any staff-facilitated music use in a given week, falling to 15% during the pandemic. That is a delivery failure — overloaded staff, no protocol, no documentation path, no reimbursement. Each of those four is what this product supplies.

Is the arts-prescription evidence thin?

Some of it is. SYNCHRONY enrolled 12.7% of those screened and averaged 10.5 of 42 sessions. SoundBridge does not cite it, and a 42-session commitment is part of the reason: ten sessions is a materially different ask. What the company does cite is Massachusetts at an 83% attendance rate, with two statewide insurers paying.

Why not simply teach breathing?

That is the right question, and it is Link 5, open. The answer is that a breathing program is a compliance product, and the completion literature for compliance products is poor. Music is the reason the patient returns for session two, and completion is the company's central empirical claim.

Does an 85% show rate conflict with a cited 44% who never engage?

The two measure different populations — conversion at referral versus attendance once enrolled — and the model's sensitivity is to realized sessions per day, which falls over below 10.2. The evidence map carries this as a named open item rather than a resolved one, and the pilot's first three endpoints settle it.

I.7 Provenance and one reconciled figure

Every figure above is carried from an existing document of record. Nothing here is new research.

One open item remains, now narrowed. The 13.83% figure is HRSA UDS Table 6B line 21a, national reporting. fmv-brief-evaluator.md states it as the 2024 national average, while evidence-map.md §6 records that the reporting year is not stated in the extract the company holds. HRSA's CY 2024 state performance-indicator publication leaves the national benchmark column blank for this measure, so that sheet will not settle the question; the figure has to be read off the UDS national data table. The instruction is to pull it from there, cite the year on the slide, and align the ten-state table, which is captioned 2025 while every other health center figure in the package is 2024.

The Montefiore reconciliation is now stated rather than deferred: on the argument slide, in §6, and in the provenance table above. It had previously been carried as an open item four inches from the figure it appeared to contradict. The model assumes an 85% show rate while the company's own citation says 44% never engage; those measure different things — 44% is conversion at referral, 85% is attendance once enrolled — and both are first-order pilot measurements.

Second edition, 30 August 2026. Draft. Not legal advice, not a clinical claim, and not a substitute for the pilot.

This appendix addresses panel item SOUN-13, which asked SoundBridge to support the $53M five-year EBITDA figure visibly, or to scale it to a number that could be defended. The reason the panel attached weight to the request is recorded at §4.6 of the read-out: an unsupported terminal number retroactively discounts the rigorous work presented earlier in the deck.

The appendix is organized in two vintages, and the two disagree by design. Sections J.1 through J.7 present the current treatment, computed on Pro Forma v78 at the settled license of $261.00, dated 31 August 2026. Section J.8 reproduces the source document as it was written on 28 August 2026 on Pro Forma v77 at a $269.29 license, together with the v70 Assumption Audit on which it drew. The v78 figures are current. The v77 figures are superseded and are retained for the record, because the reasoning they carry — in particular the account of the two structural defects and the four cost plugs — remains the evidentiary basis for the current number and is not reproduced elsewhere. Where the two vintages report different values for the same measure, the v78 value governs.

The workbook of record is SoundBridge_Health_Pro_Forma_v78_09012026.xlsx: the build with Inputs Global B189 set to 2. The 29 August build it replaces has been moved to an OLD subfolder. The status of the row is Answered. The deck still carries $53M and has not yet been updated.

J.1 The single change that accounts for the fall

The terminal number fell because SoundBridge removed the fee to which the panel objected. The defensible figure is $16,588,688 of five-year cumulative EBITDA, computed on Pro Forma v78 at the settled license of $261.00.

The superseded model charged the partner a service fee that came to 49.5% of what the partner collected — $575.77 on an episode collection of $1,163.17. A fee of that construction is precisely the kind that raises the Anti-Kickback question. The company replaced it with a flat license of $261.00: $65 a month for three protocol months plus $22 a month for three follow-up months, set in advance, and identical for every payer.

The plan runs 96,997 episode starts over five years. At the old fee those same episodes bring in approximately $30.5M more revenue, and because a license carries almost no incremental cost, nearly all of that additional revenue would have fallen to EBITDA. That one change closes 84% of the gap between $16.6M and $53M.

This is a reconstruction rather than a restatement: it applies the old fee rate to the present volume plan. It is directional, and it is checkable.

J.2 Restatement on Pro Forma v78 at the settled license

The restatement was performed twice, and the two passes had different outcomes.

Item 2 of the reconciliation list asked for five-year EBITDA to be restated on v78. That was done, and the figure did not move. The Row 12 RTM correction lowered what the partner collects but did not touch what SoundBridge charges, so the company's revenue line was unaffected by it.

The figure then moved for a different reason. On 31 August the license was settled on the published rate card at $261.00, rather than the $269.29 manual override the model had been carrying. That is a genuine price change and it does flow through to the terminal number.

Five-year cumulative EBITDA on v78 is $16,588,688. The decision cost $850,672, which is the price of holding one number that survives a valuer.

The year-by-year sequence is −$1.14M, −$0.73M, $0.76M, $5.00M, $12.69M. The model loses money in Years 1 and 2 and crosses into profit in Year 3. The lowest point the bank balance ever reaches is $44,127, in month 3.

J.3 Four figures that are readily conflated

Four figures in the current model sit within $5M of one another and mean entirely different things. They are easily confused and should be labeled on every occurrence.

J.4 Sensitivity to the valuation outcome

The fair-market-value opinion could land below $261.00. The table below is computed on v78 and replaces the corresponding table at §5 of the source document, which was built on stale cached values and overstated every outcome.

The table should be read without softening. A 15% retention would take five-year EBITDA to $4.27M. That is a real haircut and not a rounding error. The earlier version of this table reported $13.6M at the same point. It was wrong.

J.5 Sensitivity to signing cadence

What actually protects the model is signing partners faster rather than pricing the license higher. This sensitivity is also computed on v78.

At 1.5× plan — 34.5 go-lives instead of 23 — the model returns to plan economics even at a materially lower license fee.

The caution attaching to that result is stated plainly in the source and is preserved here: it is a sales answer to a pricing problem, and it holds only if the pipeline supports 34.5 go-lives. Row 18 is where that has to be shown, and it is not yet shown.

J.6 The origin of the $53M figure

The closest figure anywhere in the current model is $52.3M — Year 5 revenue at three times plan cadence. That is a different measure, in a different year, on a different assumption. What $53M originally was matters less than the following: the current number is built line by line from a workbook in which every inherited assumption is marked kept, killed or restated with a reason. A reader can check it in about ten minutes. $53M never could be checked in that way.

J.7 Consequences for the company's presentation materials

• The number in the deck must be changed to $16.6M. The deck and the workbook currently contradict each other, and a visible contradiction is worse than an unproven claim.
• The figure must be labeled precisely — "five-year cumulative EBITDA," never simply "EBITDA," and never placed next to Year 5 revenue without both labels.
• The reason for the change should be stated in one line: a fee that took 49.5% of what the partner collects was removed; it cost about $30M of five-year revenue, and the company did it before being asked twice.

A founder who identified two structural defects in his own model and corrected both presents a different proposition from one whose terminal number was never supported. Team was the panel's lowest-scoring dimension, at 1.56.

The full detail follows. Section 5 of the source document below carries the superseded sensitivity table; the corrected v78 figures appear at J.4 above.

J.8 The source document as written on Pro Forma v77 (superseded figures retained)

The material in this section reproduces Row17_The_Terminal_Number.docx, prepared 28 August 2026, built on Pro Forma v77 and the v70 Assumption Audit, and not offered as legal or valuation advice. Its subject was why the five-year EBITDA figure fell from $53M to $17.4M, what the then-current number was precisely, and why nothing in its plausible range was existential. Every figure in this section is superseded by the v78 figures above and is retained for the record.

J.8.1 The answer in one sentence, as originally stated

The number fell because SoundBridge removed the fee the panel objected to. Charging 49.5% of what the partner collects produced $53M. Charging a flat $269.29 license produces $17.4M. One change accounts for roughly five-sixths of the difference, and it is the change Section 4.2 of the read-out asked for.

The read-out was correct that an unsupported terminal number retroactively discounts the rigorous work in front of it. The source document supported the number not by defending $53M, which is not defensible, but by showing what replaced it and why.

J.8.2 The reconciliation

Pro Forma v69.1 charged an MSO service fee derived cost-plus and landing at 49.5% of the CoCM/RTM collection. On the v69.1 episode collection of $1,163.17 that is $575.77 per episode. Pro Forma v77 charges $269.29, flat, set in advance, uniform across payers.

The plan then carried 96,997 episode starts over five years. At the old fee those same episodes would have produced $29.7M more revenue — and because the license line carries almost no incremental cost, nearly all of it would have fallen to EBITDA.

That single change closes 84% of the distance between $17.4M and $53M. The remainder sits in the second structural defect described below and in the volume build.

This is a reconstruction rather than a restatement. It applies the v69.1 fee rate to the v77 volume plan. The v69.1 workbook is not in front of the reader and its own volume build differed. The point it establishes is directional and checkable: the dominant driver of the fall is the fee, and the fee fell because the company changed it deliberately.

J.8.3 What else changed, and where it is recorded

v70 was a rebuild of v69.1 rather than a re-forecast. Every inherited assumption was marked kept, killed or restated on the Assumption Audit sheet, which carries forward unchanged into v77. Nine entries bear on the terminal number.

The two structural defects, in the workbook's own words:

"REVENUE WAS STACKED, NOT COMPOUNDING. On v69.1 'Volume Drivers', every partner ramped linearly to a fixed maturity and then stayed flat for the rest of the model. No account ever grew after month 18. Total revenue therefore rose only because new partner rows were added underneath. That is a stacked services forecast: growth is a function of sales headcount, not of the installed base."

"THE FEE WAS A PERCENTAGE OF REIMBURSEMENT. v69.1 charged an MSO service fee derived cost-plus and landing at 49.5% of the CoCM/RTM collection. However it is derived, a fee that moves with what the partner collects from federal healthcare programs invites the Anti-Kickback and Stark question."

The Assumption Audit marks the first KILLED — "THE CENTRAL FIX" — and the second KILLED.

Four cost plugs were also identified, and they are the part of the audit that warrants attention.

These four ran in the opposite direction from the fee: removing them raised EBITDA. The terminal number fell despite them, which the source document states plainly — the rebuild was not an exercise in marking the forecast down. It corrected errors in both directions, and the fee dominated.

J.8.4 What $17.4M is, precisely

Three figures in the model as then built sat within $5M of one another and meant entirely different things. They are easily confused and should be labeled on every occurrence.

The plan reaches 23 active partners and 54,228 episode starts in Year 5, on 96,997 starts across the five years. Years 1 and 2 are loss-making (−$1.13M, −$0.70M); the model crosses into profit in Year 3.

One figure in the package as then circulated required relabeling for the same reason. The panel note prints "Year 5 revenue $17.7M" and "five-year EBITDA $17.4M" four paragraphs apart, and a reader will conflate them.

J.8.5 Where $53M might have come from, and why it does not matter much

The figure nearest to $53M anywhere in the model as then built is $52.3M — Year 5 revenue at three times plan signing cadence, taken from the cadence sensitivity. That is a different measure, in a different year, at a signing rate three times the plan.

Whether the deck's $53M was that, or a v69.1 EBITDA at the old fee, or something else again, is a question for the deck's own build. The useful statement is not what $53M was. It is that the current number is built line by line from a workbook in which every inherited assumption is marked kept, killed or restated with a reason. That is a claim a reader can check in about ten minutes, and $53M never was such a claim.

J.8.6 Why nothing in the plausible range is existential

The fair-market-value opinion could land below the price then carried. Five-year EBITDA across that range was presented as follows. This table was built on stale cached values and overstated every outcome; it is superseded by the v78 sensitivity at J.4 and is retained here only as the source document recorded it.

The offset was already modeled. A signing cadence of 1.5× plan — 34.5 go-lives rather than 23 — returns 99% of plan Year 5 revenue at a fee at the top of the software-only comparable band. The valuation outcome changes how many partners the plan needs. It does not change whether the model works.

The caution recorded with that result is preserved: the offset is a sales answer to a pricing problem, and it holds only if the pipeline supports 34.5 go-lives. Project plan Row 18 is where that is to be demonstrated, and it is not yet demonstrated.

J.8.7 What this requires of the deck, as originally stated

The number must be changed. The deck says $53M; the workbook says $17.4M. Leaving both in circulation is the single most checkable inconsistency in the package, and it is worse than an unsupported number because it is a visible contradiction rather than an unproven claim.

The figure must be labeled precisely — "five-year cumulative EBITDA," not "EBITDA," and never adjacent to Year 5 revenue without both labels.

The reason for the change should be stated in one line: a fee that took 49.5% of what the partner collects was removed; that cost roughly $30M of five-year revenue, and the company did it before being asked twice. A founder who caught two structural defects in his own model and fixed both is a materially different proposition from one whose terminal number was never supported — and Team was the panel's lowest-scoring dimension, at 1.56.

J.8.8 Provenance of the source document

Figures in Section J.8 are drawn from Pro Forma v77, 28 August 2026, and from the Assumption Audit sheet carried forward from v70. The v69.1 reconciliation at J.8.2 applies the v69.1 fee rate to the v77 volume plan and is a reconstruction, identified as such. The document is not legal advice, not a valuation, and not a substitute for either.

J.8.9 Also in the folder

SoundBridge_Health_Pro_Forma_v78_08292026.xlsx — the financial model as of 29 August 2026, carrying the RTM coding correction described in Row 12, both the fully-loaded and incremental facilitator ledgers, the payer-mix range, and the "v78 Changes" tab recording every cell that moved from v77 and why. The cached values in that build are stale by design until Excel recalculates on open. It is the build superseded by SoundBridge_Health_Pro_Forma_v78_09012026.xlsx, the workbook of record identified at the head of this appendix, and it has been moved to an OLD subfolder.

Forty-eight named organizations, 191 named people, and the two signatures the plan requires first.

Prepared 31 August 2026. Verified against Pro Forma v78, 29 August 2026. Supersedes the Row 18 scaffold of 28 August. Not legal or valuation advice.

K.0 Summary of the appendix

Eight of nine advisors scored go-to-market identically because the package named no accounts. The pipeline now names 48 organizations across the ten target states, with an entry-point person and title for each, 191 contacts in total, and a drafted first-touch message per account. Every one of those accounts stands at stage Identified. None has been contacted. That is the current state of the pipeline, and it is reported without qualification. The plan does not require twenty signatures to be credible at present; it requires two, in months 1 and 4, with a bench behind them. This appendix names which two, states why those two were selected, and specifies what has to move for each.

K.1 The scope of the Row 18 request

The read-out criticism was specific: no named pipeline. It was not a request for market size — the ten-state gap analysis answers that question. It was not a request for a target-state rationale — the fifty-state matrix answers that question. It asked for organizations, people, and a stage for each.

The failure mode in the earlier package was the substitution of a market answer for a pipeline answer. Everyone who might buy was expressed as a population, and how many was expressed as a revenue sensitivity. This appendix does not proceed in that manner. It names organizations, and it is explicit about the stage each has reached.

K.2 The target-state decision

The Row 18 scaffold left one strategic question open, and that question had to be closed before any table could be populated.

The decision is that the pipeline is built from the ten low-resistance states only — CO, IN, IA, MN, NH, OK, OR, PA, UT, WY. The fifty-state matrix, the ten-state gap analysis and Pro Forma v78 all run on those ten states, and the pipeline now runs on them as well. There is one strategy in the package rather than two.

The cost of that decision is stated plainly. The named relationships already on file were built for the earlier seven-state list — MO, KY, WV, TN, VA, AR, OH — which has been jettisoned. Those contacts (Community Care of West Virginia, HCC Network, MCHC, New River Health, Williamson Health, Cabin Creek) were warm, and a warm name is of greater value than a cold list. They are retired from the pipeline. SoundBridge is trading real warmth for strategic consistency, on the view that re-opening seven states at this point would undercut the one analysis the panel has not questioned, and that a pipeline whose accounts sit outside the stated strategy reads either as inattention or as two forecasts running concurrently.

The benefit of the decision is that the conference calendar and the state primary care association roster have both been rebuilt against the ten current states and are named in Section K.7. The fall cluster previously on file sat in jettisoned states; it no longer does.

Central Ozarks Medical Center is deferred rather than retired, and the pro forma label requires amendment. Pro Forma v78's 'Cohorts FQHC' tab names partner 1 as "COMC — Central Ozarks Medical Center (pilot)", a Missouri health center carried forward from v69.1 and never revised when the target states changed. COMC remains a live relationship. It is not far enough advanced to hold a month-1 slot, and re-engagement is gated on three conditions that do not yet exist: the closing of the panel response phase, the construction of the software compliance code, and the buyer-specific sales enablement set described in Section K.9. Once those conditions are cleared, COMC re-enters the pipeline, which places it in the Wave 2 window from month 14, at which point the Missouri scope tier has to be answered before the account converts.

What requires change is the label, not the relationship. The cohort names drive nothing in the model — signing months are taken from the cadence table at 'Inputs Global' rows 137–156, and the names are text — so no number in the model is affected. However, a panel member opening Pro Forma v78 finds a Missouri pilot in the month-1 slot while this document names a Colorado health center, which is precisely the two-strategies problem the decision above was made to end. The recommended edit is to rename 'Cohorts FQHC' B5 and B6 to the two Wave 1 accounts identified in Section K.5, and to carry COMC into a Wave 2 row. It is a text change to three cells.

K.3 The plan to which the pipeline reconciles

Pro Forma v78 carries cadence 2 — quarterly staged, with Wave 2 opening after the durability read. This is not a target. It is the schedule on which every downstream number in the model is built, and a pipeline that does not reconcile to it constitutes a second forecast.

The first three signing months carry the whole of the near-term commercial obligation. Partners 1 and 2 sign in months 1 and 4. Partner 3 does not sign until month 14. The ten-month gap between them is the durability read. The pipeline that has to be evidenced at present is therefore not twenty accounts. It is two signatures and a credible bench behind them, which is a considerably easier proposition to demonstrate than the row implies.

The figure of 23 does not denote 23 health centers, and the distinction defines the scope of this row. The 23 comprises the twenty FQHC partners above plus three enterprise or payer contracts signing at months 30, 42 and 54, carried on their own cadence on Pro Forma v78's 'Cohorts VBC' tab. Row 18 owns the twenty. The remaining three represent a different buyer, a different sale and a different pipeline, and nothing in this appendix should be read as evidencing them.

These figures have been verified against Pro Forma v78 directly. The cadence switch at 'Inputs Global' B131 is set to 2; the twenty signing months listed above are column C of the cadence table at rows 137–156 and match to the month. Activation lag is 'Inputs FQHC' B19. The v78 Changes tab records the rebuild as one subject — the RTM line — together with the two FQHC ledgers and the payer-mix range; the partner schedule was not touched in either the v77 or the v78 rebuild. Note that v78's cached values are stale by design until Excel recalculates on open; every figure above is a stored input rather than a computed cell.

K.4 The three qualifying questions applied to every named account

A name alone does not constitute a pipeline. Each of the following questions has an arithmetic behind it, and each is answered per account in Sections K.5 and K.6.

K.4.1 Whether the organization can staff the deployment

The unit of the answer is the facilitator FTE, and it derives from throughput rather than from headcount.

The model's reference organization is 48 panel-holding providers at 18.6 facilitator FTEs at full adoption — approximately 5,040 episodes a year, or 105 episodes per panel-holding provider per year. Across the Year 5 base of 23 partners, the plan requires roughly 200 facilitator FTEs, about 22% of the 907 unlicensed behavioral health FTEs currently employed across all 192 health center awardees in the ten states.

Two consequences follow, and both govern how the accounts below should be read. First, 18.6 × 23 is 428, not 200, so the plan already assumes that partners average well below mature adoption. Phasing is built into the model rather than introduced as a concession for the purposes of this appendix. Second, no account in this pipeline is asked to staff full adoption at signature. A one-to-two FTE phase-one deployment yields 271 to 542 episodes a year, which for a 50,000-patient health center is on the order of 5 to 11% of the patients already receiving a behavioral health service at that organization. That represents a redirection of existing staff rather than a hiring program, and it is the answer given per account below. The sizing applies state-level behavioral health service-use rates from UDS 2024 to each organization's patient count; it is an order-of-magnitude check, and the per-account figure should be replaced with the organization's own UDS Table 6A line before a signature is executed.

K.4.2 Whether the state's scope of practice supports a non-licensed facilitator

All ten target states are tier 1 under the license template — the group supported on the basis as it stands, with named statutory bases: CO 12-245-703 · IN 25-23.6-4.5-2 · IA title act · MN 245I.04 · NH 330-A:34 / 329-B:28 · OK 59 O.S. 1353 · OR 675.825(4) · PA independent-practice limit · UT 58-60 Part 6 · WY 33-38-103. Of 51 jurisdictions classified 16 / 26 / 9, the ten sit entirely within the first group.

This is the one qualifying question the pipeline answers close to uniformly, and that uniformity is a direct consequence of the target-state decision recorded in Section K.2. Independent validation of the tiering has been requested as part of, or alongside, the FMV engagement.

Minnesota is the outlier within that group and should be read as such. The other nine bases are scope provisions: they describe what an unlicensed person may do. Minnesota 245I.04 is a credentialing chapter, in that it names the categories of person rather than the boundaries of the act. For adult clients the applicable category is mental health rehabilitation worker, which requires a high school diploma plus one of six named qualifiers — an associate degree, two years of postsecondary education, RN status, three years of lived experience, three years as a caregiver, or 2,000 hours of health or human services work — together with the training mandated at 245I.05 subd. 3(c) and treatment supervision under 245I.06 by a mental health professional or certified rehabilitation specialist. The lighter mental health behavioral aide category, which requires only a high school diploma, is limited to child clients and does not reach the SoundBridge population.

The tier classification holds. A named, usable statutory pathway exists, which is more than several tier-2 states can offer, and the qualifiers are ones that a health center's existing behavioral health staff will already meet. What differs is the shape of the answer: in Minnesota the facilitator is credentialed into a defined category rather than operating inside a scope boundary, and that credentialing step is a genuine implementation task carrying a genuine timeline. It is named as the blocker on the Minnesota account in Section K.5, and it is the one state among the ten in which the scope answer costs time rather than merely requiring confirmation.

K.4.3 What the state's Medicaid program actually pays for CoCM

There is no national rule, and this is the weakest column in the pipeline. Colorado is the only one of the ten states for which SoundBridge holds a priced answer: the FY26 schedule prices 99492 at $119.40 against Medicare's $160.32, 26% lower, on a panel that is roughly 57% Medicaid. The other nine states are unconfirmed and are marked as such below. The evidence from other states demonstrates that the answers diverge sharply — Texas and North Carolina pay CoCM outside the encounter rate, New York pays a monthly case payment, and California appears to absorb it into the PPS encounter entirely, which would render the marginal collection near zero. An account in a PPS-absorbing state is not a qualified account, and SoundBridge will not assert nine answers it has not read.

The resolution path, per state and before a signature in that state, is the state Medicaid behavioral health fee schedule, confirmed against the state primary care association named in Section K.7. This step is inexpensive and is scheduled in Section K.9.

One stale code reference is being corrected at source. Two Utah account records describe the opportunity as "CoCM/G0511." G0511 was retired 1 January 2025 and G0512 on 1 January 2026. The account logic is unaffected; the code reference is stale and is being corrected at source.

K.5 Wave 1: the two signatures, named

Four criteria were applied, in order, to the 48 accounts:

• A deliverable contact at present, comprising a valid MX record plus a documented address format. Eleven of the 48 clear this criterion; see Section K.8.
• An entry point who owns integrated behavioral health or the primary-care referral workflow, rather than a cold CEO. The sale is a program conversion, not a software purchase.
• Existing embedded behavioral health with no evident formal CoCM billing. This condition makes the staffing answer a redirection rather than a hire.
• A resolvable Medicaid CoCM answer before signature.

The two selected accounts are deliberately different in shape. One is a large multi-site system phased at a single site, which tests whether the model survives contact with an organization that has its own behavioral health leadership. One matches the model's reference organization, so that the durability read also tests the economics as modeled. A pipeline of two identical accounts would read months 4 through 14 twice and learn a single thing.

Partner 1, signing month 1: Peak Vista Community Health Centers, Colorado Springs, Colorado

The rationale for placing this account first is as follows. Colorado is the only one of the ten states in which the Medicaid CoCM answer is priced rather than assumed, so the collection side of the durability read is measurable from month 1, and the durability read is the gate on the entire Wave 2 schedule. Peak Vista is additionally the largest health center in the ten states by patient count, operates embedded behavioral health with no formal CoCM registry billing, and is running at roughly a −3% margin with behavioral-health-receptive leadership, which renders net-new reimbursable revenue a straightforward internal case rather than a budget contest. The scope question in Colorado is settled at 12-245-703 and costs no time.

The cost of placing this account first is also stated. Eight hundred and fifty staff across 20 centers does not constitute the shortest sales cycle available, and the C-suite is mid-transition. That risk is accepted deliberately: a fast signature into an unpriced Medicaid state would produce a durability read that could not be interpreted, which is the more expensive failure of the two.

Partner 2, signing month 4: United Family Medicine, Saint Paul, Minnesota

The rationale for the second position, and for the suitability of that slot, is as follows. United Family Medicine is the closest account in the entire pipeline to the model's reference organization, so it measures the economics the pro forma actually asserts rather than a deployment artifact. It is a teaching clinic, comprising faculty plus family-medicine residents, which produces unusually high primary-care density per patient together with a durable referral engine, and a manualized, measurement-based protocol is a natural fit for a residency. Its decision chain is the shortest of any candidate account: CMO to CEO and CFO, all named.

The three months between the two signatures are what this account requires. The Minnesota credentialing step is a genuine implementation task carrying a genuine timeline, and the month-4 slot affords a runway that a month-1 slot would not. The organization's existing behavioral health staff will already meet the 245I.04 qualifiers; what consumes time is the mandated training and the supervision structure. Both should be confirmed before signature rather than after it.

K.6 Wave 1 bench: five accounts behind the two

All five bench accounts have deliverable contacts at present and an entry point who owns either behavioral health or the referral workflow. All stand at stage Identified.

One deliberate outlier is held separately from the bench. The Iowa Clinic (250+ providers across 40+ specialties, CMO Kevin Cunningham, MD — deliverable) is not an FQHC. It is a physician-owned multi-specialty group with a commercial-dominant payer mix and the largest behavioral health gap in the set. It sits outside the FQHC PPS question entirely, which makes it the one account in the pipeline that tests the economics with no PPS-absorption risk and the strongest RTM payment odds. It is carried as a control on the whole PPS thesis rather than as a wave slot. The CMS FQHC care-management constraint at Q60/A60 does not reach it; the license-only structure is unchanged in either case.

K.7 Wave 2 and the full pipeline: 48 accounts across ten states

Wave 2 opens after the durability read, with a first signature in month 14. It draws from the balance of the CRM. The full distribution, reported with contact deliverability, is the accurate map of where the pipeline can act at present and where it cannot.

The segment mix across the 48 accounts is 45 FQHCs, 2 independent physician groups, and 1 hospital-affiliated organization. The priority mix as scored is 20 PURSUE, 22 STRONG, 6 NURTURE. Aggregate scale across the 48 is roughly 1.74 million patients and 4,400 clinical providers, noting that "providers" in this context counts medical, dental and behavioral health staff together and is not the panel-holding count on which the model runs. See Section K.8.

K.8 Channel logic

Four routes are available, presented in descending order of conversion and ascending order of scale.

1. Advisor and GHI introductions. An advisor has committed three FQHC introductions. These are the highest-conversion names available to SoundBridge and they are deliberately gated: project plan Row 25 holds them until the revenue-split question resolves at Rows 9 and 11. They are absent from the tables above for that reason, and not because they do not exist. They are additive to the 48 rather than drawn from it, and on the conversion arithmetic set out in Section K.9 they matter more than their number suggests.

2. Primary care associations. The state primary care association is the aggregation point for health centers and the fastest route from a single conversation to a room of members. A PCA relationship is worth more than any single account and is the mechanism behind the State Cascade upside case in Pro Forma v78. All ten are named and current.

3. Conferences, on an attend rather than exhibit basis. The calendar has been rebuilt against the ten target states, and the timing now constitutes a live constraint rather than a planning note.

Three events are recommended, for the following reasons. CHAMPS/NWRPCA reaches ten accounts including one Wave 1 signature, and Colorado's own PCA runs no standalone fall event, so this is the only fall route to Colorado leadership. PACHC is chosen over the Iowa conference on the same dates because Pennsylvania is the largest state block in the pipeline and has zero deliverable email contacts; it is the one state in which a conference substitutes for a channel that does not currently function. MNACHC on 29 September is closest in time and sits in the state holding the month-1 signature, and attending before that signature is of greater value than attending after it. CFHA is the one national event worth weighing against these three, because eight of the strongest accounts are entered through a behavioral health owner and CFHA is where those individuals convene.

4. Direct outreach. Ninety-six two-email sequences are drafted in the relationship-first voice, opening with a listening ask rather than a pitch. They are gated on Section K.9.

K.9 Unresolved conditions and their resolution paths

Four gaps remain. None is concealed, and each carries a named next step.

Contact deliverability is the binding operational constraint. Only 11 of the 48 accounts have a valid MX record together with a documented address format. Fourteen are flagged as deliverability risks and twenty-three require format verification before any send. Pennsylvania, Oregon, New Hampshire and Wyoming — 17 accounts in total — have none. Nothing in the pipeline is blocked on strategy; it is blocked on address verification, which represents a week of work rather than a quarter. Both Wave 1 accounts and all five bench accounts fall within the deliverable eleven, which is why they occupy the positions they do.

The buyer-facing enablement set does not yet exist, and it gates more than its description suggests. Every account above is entered through one individual, and no health center signs on one individual's assessment. The set has to match the chain the CRM actually maps, which is not the generic enterprise chain. Across 191 named contacts at 48 health centers, the roles that recur are CMO or medical director (49), CEO or executive director (47), CFO (38), and behavioral health owner (36). Two observations follow, and both correct the framing adopted in earlier drafts of this row.

The behavioral health owner is a buyer in their own right, and is the entry point for the month-1 account. Sherri Sharp at Peak Vista is a VP of Behavioral Health; Mimi Gardner at HealthLinc is a Chief Behavioral Health Officer. Their question concerns neither clinical validity nor margin. It is whether the arrangement renders the program they already operate financially self-sustaining without displacing their licensed staff. No enablement piece currently answers that question, and it is the first question the flagship account will ask.

There is no compliance officer to whom correspondence can be addressed. Zero of the 191 named contacts hold a compliance, integrity or privacy title, and at a health center of 30 to 150 providers, compliance is normally carried by the CFO, the COO, a quality and risk director, or outside counsel rather than by a standing C-suite role. A standalone compliance one-pager therefore addresses a reader who is largely not present. The compliance content belongs inside the CFO piece, which is where the question is actually raised.

The compliance content is furthermore an argument rather than a defense. The license-only structure was selected precisely so that neither party carries fee-splitting exposure: SoundBridge supplies software at a flat per-patient fee fixed in advance and uniform across payer types, payable whether or not the practice bills or collects, and the health center employs the facilitator, the supervision and the provider of record. Stark is close to inapplicable on those facts, since there is no physician financial relationship and no designated-health-service referral, and the live question is Anti-Kickback and fair market value, which is Row 9 and is being answered by an independent opinion rather than asserted. The CMS FQHC care-management position at Q60/A60 operates in the same direction: if it governs, requiring the behavioral health care manager to be employed by or under contract to the health center and not to another company, it forecloses the full-service vendor template outright and renders the SoundBridge structure the only lawful one. It is the strongest structural argument available, and earlier drafts of this row wrongly listed it as an objection to be overcome. It should be presented with its caveats intact — dated December 2019, referencing retired codes, not repeated in MLN006397 of March 2026, and omitted from NACHC's March 2026 guidance — as a live question held in SoundBridge's favor rather than as a settled rule.

One further actor, specific to health centers, is absent from the CRM entirely. These organizations are 501(c)(3)s with HRSA-mandated patient-majority boards, and a new reimbursed service line typically reaches the board. The CRM names no board contacts and records no board step, which constitutes a gap in the pipeline rather than in this appendix; the CEO piece should nonetheless be written so that it survives being forwarded to a board that is majority patients rather than administrators.

The enablement set therefore comprises four pieces, and not the four named previously: behavioral health owner, CMO, CFO with the compliance content folded in, and CEO written to survive the board. The CFO piece depends on the two FQHC ledgers now built in v78; the compliance content within it depends on the Row 9 and Row 12 outputs. This is the difference between a first touch that receives a reply and a second meeting that converts, and it is also the gate on re-engaging Central Ozarks as described in Section K.2.

Panel-holding provider counts are the one field the CRM cannot supply. The provider figures above count medical, dental and behavioral health staff together. The model runs on panel-holding providers, and the two differ materially: a 300-provider health center does not have 300 panels. Every staffing answer above is therefore stated as a phased FTE count and an episode volume, rather than as a scaled multiple of the 48-provider reference. The source that closes this gap is HRSA UDS Table 5 per awardee.

Medicaid CoCM treatment is unread in nine of the ten states. Colorado is priced. The remainder are marked unconfirmed throughout, and no account in an unconfirmed state should reach signature before its fee schedule has been read. The source that closes this gap is the state Medicaid behavioral health fee schedules, confirmed against the PCA contacts named in Section K.8.

The conversion arithmetic does not work on 48 accounts, and this is stated here before a reader computes it. Twenty signatures against a 48-account cold list implies a 42% eventual conversion rate, which is not a credible figure. The resolution is that the CRM is not the universe. There are 192 health center awardees across the ten states; the CRM holds 48 of them, or 25%. Twenty signatures against 192 addressable organizations is 10.4%, which is an ordinary enterprise conversion rate. The pipeline as it stands is therefore the qualified top of a substantially larger addressable set, and the plan depends on three elements that the tables above understate: widening beyond the CRM into the remaining 144 awardees, the PCA channel, and the three gated advisor introductions.

K.10 The next four weeks

K.11 Data provenance

Account data is drawn from the SoundBridge GTM CRM for the ten low-scope states, current to August 2026, with organization scale taken from HRSA Uniform Data System 2024 reporting and IRS Form 990 filings where noted. Financial and throughput figures are taken from Pro Forma v78, 29 August 2026. Contact names are researched from public leadership pages; email addresses are inferred from documented domain formats and are marked by deliverability confidence rather than verified by send. Scope-of-practice tiering is taken from the fifty-state matrix and awaits independent validation. Medicaid CoCM treatment is confirmed for Colorado only. This appendix is not legal, valuation or reimbursement advice.

Source document: Row 19 — The Reimbursement Architecture (Summary and Supporting Detail), 1 September 2026. The folder and the document were created on the same day, after the compendium was first compiled. The material is built on Pro Forma v78 at the settled $261.00 license. Status: answered.

The panel's request, recorded as SOUN-15, was as follows:

"Lead with the reimbursement architecture: CY2026 national rates, the 60% fee-sensitivity stress test, and the tranched milestone-gated raise."

This is the only item in the review set in which the panel asked the company to lead with something it had praised rather than to correct something it had criticized. Five separate advisors credited the reimbursement architecture without coordinating with one another, and the read-out, at §3, described it as the most sophisticated reimbursement engineering in the cohort. The package as it stood at the time of the read-out answered every criticism the panel had made and documented none of the praise.

L.1 The hazard the row addresses

The three elements the panel credited were credited as they stood in a deck that has since been comprehensively superseded. Restating the praise without restating the substance would amount to quoting a review of a different product.

Two of the three credited elements no longer exist in the form the panel saw. The fee against which the 60% stress test was run has been removed. The rate build has been rebuilt downward. Repeating the compliment as originally written would place a favorable quotation next to figures that no longer support it, which is the single easiest discrepancy for a reviewer to identify.

The row is therefore organized element by element: what was credited, what the element is now, and what changed. In two of the three elements the current position is materially different from the position that was credited.

Before the row was prepared, this item had neither a folder nor a document; both now exist. The organizing principle of the document is that the substance, rather than the compliment, is what carries forward. Stated in a single line: the three things the panel praised were praised in a deck that has since been comprehensively superseded, so restating the praise without restating the substance would be quoting a review of a different product.

L.2 Element 1: the CY2026 rate build

The build is constructed code by code at a $33.40 conversion factor rather than assumed.

CoCM at $634.62 plus RTM at $184.04 gives a gross of $818.66. At a 93.8% payer mix and a 95% clean-claim rate, the partner collects $729.82.

Why the current build is stronger than the version that was praised

The RTM line was rebuilt downward. CPT 98977 had been carried at $40 a month. It is the musculoskeletal device-supply code, and nothing in the protocol monitors a musculoskeletal system. It is withdrawn with no replacement claimed, and the gross collection fell from $938.62 to $818.66. That is $119.96 an episode removed from the company's own case, identified by SoundBridge, before any reviewer raised it.

The rate build is stronger than the version that was praised, and it is stronger because it is smaller. The withdrawal of 98977 is reported here for the same reason it is reported at Rows 9 and 12: a smaller figure the company can defend is worth more than a larger figure a reviewer can take apart.

Two supporting points

Two features of the build bear directly on the objections a reviewer would otherwise raise first.

• These lines pay at the national rate with no geographic adjustment, separately from and on top of the PPS encounter. The usual objection to national-rate modeling therefore does not apply to them.
• The three add-on units rest on a timestamp log rather than on an assumption. Contact of 120, 90 and 90 minutes clears the 70, 60 and 60-minute thresholds by full thirty-minute blocks, without recourse to the midpoint rule.

These are the two places at which a reviewer would otherwise press first. Both are stated as verifiable properties of the build rather than as arguments about it: the first is a feature of how the codes are paid, and the second is a feature of the record the company holds.

L.3 Element 2: the fee-sensitivity stress test

What the panel credited was a service fee set at 90% of partner collections and stressed down to 60%. On a $1,240 episode, that is $1,116 falling to $744 — a cut of 33%.

That structure no longer exists, because the same panel objected to it at §4.2. A percentage cannot be stress-tested as a percentage once it has ceased to be a percentage. The original stress test measured the wrong thing — not because it was badly constructed, but because its subject was removed.

Rebuilt in the units a valuer actually works in, namely what the partner retains, the test now runs deeper.

Partner retention is the unit a valuer works in, and rebuilding the test on that basis produces a deeper cut than the one that was praised. That the replacement test is harder than the original is the point of this element, and it is stated rather than implied.

A second sensitivity exists that the original did not contain. Signing cadence moves Year 5 EBITDA from $12.69M at plan to $40.12M at three times plan. Cadence moves the answer by multiples; the fee moves it by tens of percent. That is the correct form of the argument that nothing in the model is existential. It is also a sales answer to a pricing problem, and for that reason it depends on Row 18. The element as a whole therefore concedes that the test the panel credited has been retired, and offers in its place a test constructed on the quantity that determines the valuation, computed to a five-year EBITDA figure at the floor that the earlier form of the test did not produce at all.

L.4 Element 3: the tranched, milestone-gated raise

This is the element that survived unchanged, and the gate on it is now a named condition rather than a rhetorical one.

The raise comprises three post-money SAFEs: a $200K bridge at $8M, Tranche 1 of $1.1M at $10M, and Tranche 2 of $1.4M at $14M with a 20% discount. The total is $2.7M. The founder retains 61%; SAFE overhang is 23%.

What makes the gate real rather than rhetorical is the condition attached to Tranche 2. It is conditioned on the month-9 milestone and on the independent fair-market-value opinion at Row 9 — an external opinion the company does not control, addressing the exact question four advisors raised. It is not a date against which the company marks its own homework.

The tranche gate is strengthened by Row 9 rather than weakened by it: an open FMV opinion functions as a named external condition on Tranche 2 rather than as an unanswered question. Row 9 remains the one genuinely open row in this set, and this element depends on it by design.

The cash trough is $44,127 in month 3, and the plan is never cash-negative on $2.7M. Exit at 8×, 12× and 15× Year-5 EBITDA, after a 30% haircut, gives $71.1M, $106.6M and $133.3M.

L.5 Claims the row does not make

The following limitations are stated in the summary of the source document and are not softened in its supporting detail.

• A lower valuation reduces all of it.
• The coding position is stated and officer-signed but not independently confirmed. Four review routes are running and none has reported.
• Medicaid is priced in Colorado only and unread in nine states, some of which may absorb CoCM into the PPS encounter entirely.
• The 93.8% payer-mix factor is inherited from a superseded workbook and should be replaced from remittance data rather than defended. That factor multiplies every figure in Element 1 and has never been checked against remittance data.

L.6 Items outstanding

• The deck that GHI III reviewed still requires manual placement in the Row 19 folder; the connector cannot upload binaries.
• The Deck GHI III Reviewed — Supersession Notice, which explains why that deck no longer describes the company, is written and filed in the parent folder.