This inquiry set out to name the binding constraints behind American mental-health access and score the leading fixes against them. Three kill conditions fire and become the filing's load-bearing shape: there is still no national series for outpatient wait or new-patient acceptance by payer (KC1); the circulating "95% of beds gone since 1955" figure is single-root advocacy arithmetic on state-hospital beds, not a whole-system census (KC2); and commercial parity evidence largely cannot see the ERISA self-funded majority (KC3). Against that backdrop, usable Medicaid supply — not licensed headcount — is the outpatient wedge: psychiatrists' Medicaid acceptance was 43.1% in the national NAMCS frame, and a four-city secret-shopper study found only 17.8% of directory-listed Medicaid clinicians offered a new-patient appointment. Mental disorders are 12.7% of 2023 disabled-worker awards, not roughly one-third — that prior is withdrawn; worker stock (~28.6%) belongs to GBMT-12. A ten-architecture scorecard, red-teamed and then structurally blinded, ranks #5 crisis continuum first under the Medicaid-SMI weighting, #2 CCBHC first under commercial-parent and state-commissioner weightings, and #10 do-nothing last under every weighting. Asylum-scale bed rebuild never leads (H7 Supported). Status: twice checked — one blind re-score is not "final" (M10).
Three kill conditions fire — and become the filing
KC1 fires. No federal — and no ≥80%-population non-federal — series publishes outpatient mental-health appointment wait or new-patient acceptance by payer. Circulating "48-day average wait" figures are vendor or association surveys. CMS has written Medicare Advantage network wait standards and moved toward Medicaid managed-care secret-shopper rules; standards are not a measured national median series. Under the protocol, that absence is the whitepaper's headline: the country argues about a shortage it does not measure at the point of use. Every O1 score below is therefore a proxy — clinic demo waits, Lifeline volume, NQTL corrections, secret-shopper snapshots — and those proxies are not one ladder.
Phase 0 · ws02 · GAO-22-104597 · HRSA workforce briefs citing National Council · CMS MA / Medicaid managed-care rulemakingsKC2 fires. The circulating "~95% of psychiatric beds eliminated since 1955" / "558k → ~35k" figure traces to Treatment Advocacy Center / Torrey-lineage arithmetic on state (and historically state/county) hospital beds. It is not a whole-system bed census once general-hospital psychiatric units, private psychiatric hospitals, and VA beds are in frame. This filing drops absolute "beds since 1955" claims and scores contemporary Construct A/B inventories plus IMD financing geography only.
Phase 0 · ws03 · TAC staffed-bed surveys · NRI state-hospital reportingKC3 fires. Most commercially insured lives sit in ERISA self-funded plans outside the state insurance exams that generate much of the public "parity compliance" evidence. Federal EBSA can reach those plans and has forced CAA comparative-analysis corrections covering more than 7.6 million participants — but that is investigative reach, not a census. Commercial O4 is scored band-only / unknown for the self-funded majority.
ws06 · DOL/HHS CAA reviews · state exam geography · May 2025 nonenforcement of new 2024 MHPAEA provisionsLicensed ≠ available — Medicaid is the wedge
HPSA designations and license tallies count clinicians who may never open a Medicaid panel. Two independent methods put usable Medicaid access below 50%: Bishop et al. (JAMA Psychiatry 2014) found psychiatrists' Medicaid acceptance at 43.1% versus 73.0% for other physicians in the national NAMCS frame; Brahmbhatt et al. (JAMA 2024) secret-shopped directory-listed Medicaid clinicians across four large cities and found only 17.8% reachable, accepting, and offering a new-patient appointment. Oregon claims work shows phantom directory participation for mental-health specialties. Commercial acceptance is materially higher in the NAMCS frame (private noncapitated 55.3% for psychiatrists vs 43.1% Medicaid) — the wedge is the payer, not an absolute absence of licensees.
ws02 · Bishop et al., JAMA Psychiatry 2014 · Brahmbhatt et al., JAMA 2024 · Oregon claims reconstructionsRates and paperwork bind before "stigma" (H5 Supported): Allegheny MCO rate-change elasticity is modest (~0.16) and mostly expands quantity for existing patients; CCBHC PPS packages move access metrics where fee bumps alone do not. Do not promise that a national rate floor opens new-patient panels at H2 scale — architecture #1 scores directionally correct and elasticity-limited.
ws05 · Allegheny rate study · Mathematica/ASPE CCBHC demonstration · Decker-style nulls on stigma-only storiesAwards are 12.7% — the one-third prior is false
Protocol H8 wagered that psychiatric awards were a care-system overflow valve and near-plurality of working-age awards. The SSA awards table kills the plurality leg: among 2023 disabled-worker awards, mental disorders were 12.7%; musculoskeletal was 34.0%. Absolute stock on the rolls is still large — worker mental-disorder stock ≈ 28.6% is GBMT-12's object, not this filing's. The one credible local-access paper located this pass rejects a simple scarcity→awards overflow story (scarce care → more awards); treatment capacity can enable applications by creating diagnostic pathways. Do not score architectures on false awards plurality.
ws08 · SSA Annual Statistical Report on the DI Program (Chart 10 / awards tabulation) · GBMT-12 Phase 0 stock · Swenson pathway evidenceCommunity and crisis beat asylum rebuild — with the diversion caveat
Equal-effort steelman in §7: jurisdictions that invested in community capacity show measured outpatient and crisis-adjacent gains without proportional asylum-scale bed restoration. CCBHC demonstration metrics — adult time-to-eval improving in DY1→DY2 (9.0→5.4 days), high open-access reporting — are the outpatient package (scored at O1=4 after red team refused a 5). Middle-tier continuum evidence (Michigan mobile crisis; behavioral health crisis centers) shows arrest/ED-adjacent gains without large bed rebuilds. 988 itself grew answered/routed volume dramatically (Vibrant/GAO/KFF; ~2.5× January answered contacts 2022→2024; ~19.1M routed Jul 2022–Sep 2025) without a multi-state causal ED/arrest diversion series — H4 Supported on volume, thin on diversion.
ws04 · ws05 · ws07 · Vibrant KPI extracts · GAO-26-108114 · Mathematica/ASPE CCBHC · Swartz AOT stack (selected-band only)IMD exclusion remains the default Medicaid purchase constraint for adult stays in facilities over 16 beds; §1115 SMI/SED waivers are a patchwork, not national repeal (H3 Supported on statute/CMS path). McBain HCRIS finds no significant higher Construct B bed rates in waiver states so far — financing geography ≠ proven bed rebuild. Architecture #4 (bed rebuild / state-hospital reinvestment) never leads after pass-2: contemporary need documentation is not instrument effect.
The scorecard — twice checked, ranks still provisional
Ten architectures scored on five anchored objectives (realized access, acute/crisis capacity, payer-taking workforce, financial protection/parity, state-capacity load), then ranked under three explicit weight vectors. Red team cut CCBHC's O1 from 5→4 and O2 from 4→3, and parity's O1 from 4→3. A structurally blinded re-score (batch msgbatch_013vUDQ6mt6JV8mz2H95zfsx) differed on 10 of 50 cells; 7 corrected, 3 kept. Leadership held.
| # | Architecture | O1 | O2 | O3 | O4 | O5 |
|---|---|---|---|---|---|---|
| 1 | Medicaid behavioral rate floor + admin simplification | 3 | 3 | 4 | 3 | 4 |
| 2 | CCBHC expansion as default safety-net model | 4 | 3 | 4 | 3 | 4 |
| 3 | IMD repeal or broad MH IMD waiver | 3 | 4 | 3 | 3 | 4 |
| 4 | Bed rebuild / state hospital reinvestment | 3 | 3 | 3 | 3 | 4 |
| 5 | 988 + mobile crisis + stabilization continuum | 4 | 4 | 3 | 3 | 4 |
| 6 | Parity enforcement with ERISA teeth | 3 | 3 | 3 | 4 | 3 |
| 7 | Assisted outpatient treatment expansion | 3 | 4 | 3 | 3 | 4 |
| 8 | Primary-care BH / Collaborative Care | 3 | 3 | 3 | 3 | 3 |
| 9 | Workforce liberalization | 3 | 3 | 3 | 3 | 3 |
| 10 | Do-nothing comparator | 3 | 2 | 2 | 2 | 4 |
| Weighting | 1st | 2nd | Last |
|---|---|---|---|
| W1 — Medicaid-SMI enrollee | #5 Crisis continuum (3.65) | #2 CCBHC (3.55) | #10 (2.40) |
| W2 — Commercially insured parent | #2 CCBHC (3.60) | #5 Crisis (3.50) | #10 (2.55) |
| W3 — State BH commissioner | #2 CCBHC (3.70) | #1 Rate floor · #5 Crisis (tie 3.60) | #10 (2.70) |
Pass-1's all-three CCBHC headline did not survive red team. Pass-2 / post-blind: #5 leads W1; #2 leads W2 and W3; #10 last everywhere. #2's W2 lead is still an O1-proxy from Medicaid safety-net clinic metrics — not commercial secret-shopper proof. #6's O4=4 is examined-band CAA corrections only — it does not close KC3. #3 IMD and #7 AOT are numerically identical after reconcile for different reasons; #8 and #9 are all-neutral ignorance-density rows, not mild endorsements. Measurement infrastructure is a named missing scored row (deviation #19).
ws09-scorecard.md · ws09-red-team-log.md · ws09-rescore-log.mdSequencing — measure first, then the instruments that already run
Publish order for the desk, and a rough policy order implied by the board:
| Order | Move | Why it sits here |
|---|---|---|
| 1 | National by-payer realized-access series (measurement architecture) | KC1 is the headline; VA access standards are the domestic precedent the civilian system lacks. Missing scored row — owed next or explicit exclusion. |
| 2 | Complete 988 + mobile crisis + stabilization continuum | W1 leader; volume already national; diversion still unproven — fund the middle tiers the evaluations actually speak to. |
| 3 | CCBHC expansion as default safety-net model | Leads W2/W3; measured wait-to-eval / open-access package under Medicaid-dominant populations. |
| 4 | Rate floor + admin simplification; examined-band parity with federal teeth | Directionally correct, elasticity-limited (#1); #6 closes some examined-plan gaps without resolving ERISA-majority O4. |
| Later | IMD waiver geography; targeted acute/forensic capacity; AOT | Financing and selected-band tools — not the lead "fix mental health" story under O1-binding weights. |
| Do not lead with | Asylum-scale bed rebuild; awards-plurality rhetoric; headcount-only workforce liberalization | H7 demotes #4; H8 fails as written; #8/#9 are evidence-deficit, not endorsement. |
The honesty box
Twice checked, not final. Red team plus one structurally blinded re-score. Per M10, one re-score is not settlement — crypto's dual-blind precedent moved disjoint cell sets. Say "twice checked," not "final." Verification Protocol Phase 1 has not yet run on this filing.
KC1 proxy classes are incommensurable. O1 ranks compare unlike instruments. Do not narrate "CCBHC beats parity on access" without naming the proxy each cell used.
Withdrawn headlines. Pass-1 "CCBHC leads all three weightings." AOT's former W2 near-top (hospitalization OR mis-scored as O1). #4 O2=4 as contemporary need. #8 O5=4 as demonstrated CoCM delivery. #6 O5=2 as pure EBSA failure. #10 O1=2 as pure outpatient failure. Awards ~⅓.
What this filing still cannot claim. A national by-payer wait median. Whole-system bed counts comparable to 1955. ERISA-majority commercial realized access. Multi-state causal 988→ED/arrest diversion. That rate floors alone open new Medicaid panels at H2 scale. That IMD waivers have already rebuilt Construct B supply.
Measurement row still missing. Architectures 1–10 are delivery/financing instruments. The protocol's own KC1 consequence — score measurement infrastructure as a row or exclude it explicitly — remains owed (deviation #19 / red-team Attack 7).