Status: Pass-1 scorecard post-red-team (research/ws09-scorecard.md; log research/ws09-red-team-log.md). Blind re-score next. Protocol drafted 2026-08-09. Phase 0 landed 2026-08-11. Workstreams §1–§8 landed 2026-08-11 (research/ws01–ws08; research/ws-source-register.md). Starred anchors 1–8 verified or corrected; anchors 9–10 verified in §5–§6; 11–12 verified in §3. KC1, KC2, and KC3 fire. H1–H7 Supported; H8 Not supported as written (awards plurality fails at 12.7%; stock ≈28.6% → GBMT-12). Post-red-team ranks: #5 crisis continuum leads W1; #2 CCBHC leads W2/W3 (all-three CCBHC lead withdrawn after O1/O2 cuts); #10 do-nothing last; #6 parity/ERISA swing narrowed (examined-band O4 only); asylum-scale #4 demoted per H7; measurement architecture named as missing scored row. Designation: Numbered filing GBMT-11. Directory mental-health/. Branch gbmt-11-mental-health per M9 after Phase 0 lands; claim with scripts/preflight-filing.sh mental-health and git fetch before any substantive work. Scope: United States; the behavioral-health delivery and financing system — who needs care (serious mental illness and common disorders as distinct bands), what capacity exists (acute, crisis, outpatient), who can actually get an appointment under the payer that covers them, and which instruments have measurably moved access or capacity. Method: Imports method/gubment-method.md (M1–M10) in full. This document contains only domain content. Proportionality (M10): Three headline findings maximum, at full M1 rigor. Supporting claims: one typed, dated source, root-traced where citogenesis is plausible. Scorecard cells: cited to a workstream finding or held neutral, no separate verification pass. Target ≤8 findings files. One red team, one structurally blinded re-score, reconciliation as a table, deviations log at one line per entry. Artifact: A feasibility assessment of the leading "fix mental health" architectures, naming binding constraints in priority order — Whitepaper No. 11.
1. Scope and definitional decisions to settle before fieldwork
1.1 The public fight, and the desk wager. The public mental-health argument runs on two numbers that rarely meet: how many people have a condition (prevalence) and how many clinicians are licensed (supply headcount), with a side fight over psychiatric beds. This filing's wager is that the binding constraints are usable Medicaid supply (who accepts new patients at the rates and paperwork the dominant payer for serious illness actually pays), acute and crisis geography under the IMD exclusion, and the absence of a national instrument for realized access — not the census of licensed therapists. That wager is falsifiable; H2 and KC1 are written to falsify it.
1.2 Which mental-health problem. At least five travel under the phrase, with different instruments and clocks: (a) prevalence and unmet need — who meets diagnostic criteria and who reports unmet need; (b) outpatient supply that takes the payer — therapy, psychiatry, prescribing, primary-care behavioral integration; (c) acute / inpatient capacity — psych beds, ED boarding, the Institutions for Mental Diseases (IMD) rule; (d) crisis continuum — 988, mobile crisis, stabilization; (e) parity and network adequacy — whether insurance products that claim behavioral coverage deliver usable access. The public argument leans on (a) and licensed headcounts under (b). This filing treats (b)–(e) as the load-bearing objects and scores (a) as baseline, not headline.
1.3 Serious mental illness vs. common disorders. SMI (schizophrenia-spectrum, bipolar, severe depression with functional impairment — definition to be locked against SAMHSA/NSDUH usage in Phase 0) and common disorders (anxiety, milder depression, ADHD, etc.) are different markets: different payers, different workforce, different bed demand, different disability-program pathways. Workstreams that mix them without saying so are defective. Where a claim only holds for one band, the finding says so.
1.4 Shared seam with GBMT-12 (Disability). Serious mental illness is a major pathway onto SSI/SSDI. This filing owns care delivery and capacity; GBMT-12 owns cash benefits, adjudication, and the work cliff. The shared seam — SMI award rates onto the rolls, representative-payee patterns, and whether the care system or the cash system binds for that population — gets a short dedicated treatment in §8 of each protocol. Neither filing re-litigates the other's object.
1.5 Excluded, with reasons.
- Long-term services and supports / HCBS / nursing homes — GBMT-7 (elder care). Disabled adults' LTSS stays there; this filing does not re-open waiting lists or aide wages.
- General medical prices, hospital RCC, marketplace architecture — GBMT-3. MH-specific parity, network adequacy, and Medicaid behavioral rates are in; commercial hospital price levels are out.
- Substance-use disorder treatment as the primary object — GBMT-8. OTP/methadone, contingency management, and SUD-specific IMD waivers stay there. This filing owns psychiatric IMD / acute psych capacity and cross-cites drugs where dual-diagnosis or SUD IMD waivers are the comparison case.
- Disability cash design (SSDI/SSI) — GBMT-12, except the shared seam above.
- School counseling / K–12 behavioral health as primary — different decision-makers and funding streams; a future K–12 filing. In only as a measured youth-access line item if the baseline requires it.
- Criminalization as the whole filing — jails and prisons enter as a measured de-facto bed stock (§3), not as a criminal-justice architecture scorecard.
- Clinical comparative-effectiveness of specific therapies — different research discipline. This filing scores access and capacity, not which modality works.
1.6 Regulatory geography. Behavioral health is a federal–state mash: Medicaid (state plan + waivers + IMD), MHPAEA parity (federal, with state insurance enforcement for fully insured plans), ERISA self-funded plans (federal), 988 (federal funding + state implementation), involuntary treatment / assisted outpatient treatment (state). Which layer binds is an empirical question answered per problem-class, not a backdrop.
1.7 Perishability. 988 implementation metrics, IMD waiver statuses, and state Medicaid behavioral rates move on appropriations and CMS approval clocks. Findings touching those series carry an as-of date and are marked perishable where a single waiver decision or budget cycle could void them.
2. Objective function
Candidate architectures are scored against five objectives, each chosen because the evidence base can plausibly speak to it:
| # | Objective | What it measures |
|---|---|---|
| O1 | Realized access | Ability to get a timely appointment or crisis response under the payer that covers the person — not prevalence, not licensed headcount |
| O2 | Acute and crisis capacity | Psych beds, ED boarding, 988/mobile-crisis throughput, and the geography of who can be admitted where |
| O3 | Workforce that takes the payer | Clinicians accepting new Medicaid (and, separately, commercial) patients; vacancy/turnover in safety-net settings |
| O4 | Financial protection / parity | Out-of-pocket exposure, parity compliance in practice, network adequacy that survives a secret-shopper test |
| O5 | State-capacity and federalism load | New machinery required; whether an existing agency at the relevant level has demonstrated it can run it |
O1 and O3 are expected to move together if the desk wager is right; O2 can improve under instruments that leave O1 untouched (more beds, same outpatient desert). Rank stability across weightings is reported; the swing is the reader insight.
3. Three objective weightings (M10)
| Weighting | The reader it represents | Emphasis |
|---|---|---|
| The Medicaid enrollee with SMI | Someone who needs ongoing care and sometimes crisis care, covered by the payer that dominates serious illness | O1 and O2 heavy; O3 as the means; O4 second |
| The commercially insured parent | Seeking outpatient care for a common disorder; has a card that claims parity | O1 and O4 heavy; O2 discounted |
| The state behavioral-health commissioner / Medicaid director | Must stand up 988, manage IMD rules, set rates, and not lose the safety-net workforce | O3 and O5 heavy; O2 second |
4. Workstreams (8)
Seeds are seeds. Each workstream opens with a documented PRISMA-lite search (OpenAlex/Semantic Scholar; SAMHSA/CMS/NRI/NASMHPD grey literature) with strings, date ranges, and inclusion criteria logged. Findings files target: ≤8 (M10).
§1 Baseline — prevalence, spending, and who pays. Question: how large is the need under SMI vs. common-disorder definitions, what is spent, and which payer dominates each band? Evidence: NSDUH / SAMHSA mental-health prevalence series (definitions locked before comparison); CMS NHEA and Medicaid expenditure reports for behavioral line items; MEPS for household-side MH spend; state mental-health agency revenues (NRI Uniform Reporting System). Seed: H1.
§2 Licensed ≠ available (flagship). Question: of the clinicians counted in shortage designations and workforce tallies, what share accepts new patients under Medicaid vs. commercial insurance, and at what wait? Evidence: HRSA HPSA mental-health designations and their methodology limits; NPI/NPPES and Medicaid provider enrollment files where usable; secret-shopper and audit literature (Medicaid acceptance rates); APA / insurance-regulator network-adequacy exams; any national or multi-state series on appointment wait by payer. Two-method rule on "shortage": HPSA designation counts against a realized-access measure. Seed: H2. This is the kill-condition test for O1.
§3 IMD and acute capacity. Question: how many psychiatric beds exist, where, under what payer rules, and what does the IMD exclusion do to Medicaid's ability to buy adult inpatient capacity? Evidence: AACAP/APA or NASMHPD bed inventories (root-trace the circulating "95% of beds closed since 1955" claim — citogenesis risk is extreme); CMS IMD definition and state §1115 IMD waivers for MH (distinct from SUD IMD waivers in GBMT-8); ED boarding studies with stated methods; jail/prison mental-health census as de-facto capacity. Seed: H3.
§4 Crisis continuum — 988, mobile crisis, stabilization. Question: what did 988 actually change in connection volume, local answer rates, and downstream diversion from ED/law enforcement — and what remains unmeasured? Evidence: Vibrant/988 Lifeline performance metrics; SAMHSA grant reporting; state implementation dashboards; mobile-crisis and crisis-stabilization evaluations; law-enforcement diversion programs with measured outcomes. Seed: H4.
§5 Medicaid rates, take-up, and the safety-net market. Question: do Medicaid behavioral rates and administrative burden explain non-participation better than "stigma" or "shortage" narratives? Evidence: state fee schedules for psychotherapy and psychiatry codes; Medicaid managed-care behavioral carve-outs; Certified Community Behavioral Health Clinic (CCBHC) demonstrations and evaluations; FQHC behavioral expansion. Seed: H5.
§6 Parity and network adequacy in practice. Question: after MHPAEA and the 2024 parity rules, can a covered person actually use the benefit? Evidence: DOL/HHS/Treasury parity rulemakings and enforcement actions; state insurance exams; litigation outcomes (e.g. Wit v. United Behavioral Health lineage — hold to holdings, not press); GAO/OIG parity reports; ERISA vs. fully-insured enforcement split. Seed: H6.
§7 Precedents and jurisdictions. Question: which instruments have a measured effect, and does it transfer? Domestic: state IMD waivers (MH); assisted outpatient treatment / Laura's Law / Kendra's Law natural experiments (methods discipline — selection is the whole game); CCBHC expansion states; Massachusetts and NY parity enforcement records; VA mental-health access standards as an existing measurement precedent. International: peer bed-ratio and community-care mixes (UK, Netherlands, Australia) with transferability assessment — financing and civil-commitment law travel badly. Seed: H7 (steelman lives here).
§8 Political economy and the GBMT-12 seam. Question: why does the legislative record favor awareness months and 988 over rate floors, and what share of SSI/SSDI working-age awards are psychiatric? Evidence: lobbying and appropriations for SAMHSA vs. Medicaid behavioral rate policy; OpenSecrets / LDA; SSA disability award statistics by diagnostic group (cross-cite GBMT-12); representative-payee prevalence for psychiatric awards. Seed: H8.
5. Hypotheses with pre-registered adjudication criteria (M3)
Written before evidence collection. Adjudication is by these criteria and no others; changes go in the deviations log.
H1 — Prevalence is not the scarce number. Circulating "crisis" prevalence figures are definition-sensitive and are not the binding constraint on access. Supported if ≥2 authoritative series (different definitions) for adult AMI or SMI differ by >1.5× for a comparable year or unmet-need measures move materially less than prevalence headlines imply across a decade. Refuted if a single federal definition underpins the public figures and unmet need tracks prevalence within 20%. Indeterminate if definitional crosswalks cannot be built from published docs.
H2 (flagship / KC1 test) — Licensed ≠ available, and Medicaid is the wedge. A majority of the apparent outpatient shortage is non-participation in Medicaid (and narrow networks), not an absolute absence of licensees. Supported if ≥2 independent methods show Medicaid new-patient acceptance for psychiatrists or therapists below 50% in a multi-state or national frame and commercial acceptance is materially higher in the same frame. Refuted if acceptance rates are high (>70%) for Medicaid in the same frame, or licensed supply is below a stated panel-size benchmark even at 100% participation. Indeterminate if no multi-state acceptance series exists — which itself feeds KC1.
H3 — IMD is a binding Medicaid inpatient constraint for adults. The IMD exclusion (or its incomplete waiver patchwork) measurably limits Medicaid-financed adult psych capacity relative to need. Supported if bed-supply or boarding metrics worsen at the Medicaid-adult margin in non-waiver states relative to waiver states in ≥2 independent evaluations, or CMS/state documents state capacity cannot be purchased because of IMD. Refuted if waiver and non-waiver states show no material capacity difference after case-mix adjustment. Indeterminate if bed inventories are not payer-resolvable.
H4 — 988 grew calls without proving diversion. Connection volume rose; measured ED/law-enforcement diversion is thin or single-root. Supported if national answer-volume gains are documented and no multi-state causal estimate of ED or arrest diversion survives an independence audit. Refuted if ≥2 independent evaluations show material diversion. Indeterminate if only vendor-reported metrics exist.
H5 — Rates and paperwork bind before "stigma." Medicaid fee levels and administrative burden predict participation better than provider-attitude surveys. Supported if participation elasticities or natural experiments on rate changes show material supply response in ≥2 studies and attitude-only explanations fail when rates are held constant. Refuted if rate increases produce no participation response in well-identified settings. Indeterminate if fee schedules cannot be linked to participation data.
H6 — Parity on paper ≠ access in practice. Post-MHPAEA, network and medical-necessity practices still produce systematically worse behavioral access than medical/surgical for comparable plans. Supported if ≥2 enforcement actions, exams, or secret-shopper studies in different jurisdictions find material parity gaps after 2021. Refuted if recent multi-state exams find substantial compliance and wait/access gaps are explained by workforce alone. Indeterminate if ERISA plans are invisible to the available exams (report the visibility gap).
H7 — STEELMAN, unfashionable direction: the bed story is dated, and community capacity was the right bet. Built with equal effort per M3. Deinstitutionalization's bed decline is real but the binding deficit today is outpatient and crisis continuum, not a return to asylum-scale inpatient stock; jurisdictions that invested in community systems show better access without proportional bed restoration. Supported if (a) per-capita bed ratios in high-access peer regions are not the main correlate of boarding/access once outpatient capacity is controlled, and (b) ≥2 evaluations of community-crisis systems show access gains without large bed rebuilds. Refuted if boarding and unmet acute need track bed supply after outpatient controls. Indeterminate if bed and outpatient series cannot be jointly observed. §7 adjudication (2026-08-11): Supported. CCBHC + mobile/BHCC continuum meet (b); OECD bed dispersion + peer cases meet (a) directionally (no fitted international model — deviation 15). Asylum-scale rebuild demoted; narrow acute/forensic beds remain on the board. See ws07-precedents.md.
H8 — Psychiatric disability awards are a care-system overflow valve. A large share of working-age SSI/SSDI awards are mental disorders, and award pressure correlates with local treatment scarcity more than with fraud narratives. Supported if SSA diagnostic tabulations show mental disorders as a plurality or near-plurality of working-age awards and ≥1 credible study links local treatment access to award rates. Refuted if mental disorders are a small minority of awards or access–award links fail replication. Indeterminate if SSA public tables do not support the crosswalk (hand to GBMT-12). §8 adjudication (2026-08-11): Not supported as written. Awards share 12.7% (plurality fails; MSK 34.0%). Stock ≈28.6% handed to GBMT-12. Swenson & Urban link more MH establishments to more applications (pathway), not scarcity overflow. See ws08-political-economy-seam.md; deviations-log #16.
Tilt audit: H1–H6 and H8 lean toward "measurement and Medicaid supply bind." H7 is the counterweight and gets equal effort in §7.
6. Kill conditions (M6)
KC1 (fires as headline if confirmed) — no national realized-access instrument. If §2 finds no national (or ≥80%-population) series for outpatient MH appointment wait or new-patient acceptance by payer, objective O1 cannot be scored on realized access and collapses to prevalence plus clinician headcount. That collapse is the filing's headline, not a caveat — the country is arguing about a shortage it does not measure at the point of use.
KC2 — bed counts are citogenic. If the circulating historical bed-decline figures trace to a small number of non-independent roots and no payer-resolvable contemporary inventory exists, §3 drops absolute "beds since 1955" claims and scores only contemporary capacity and IMD rules.
KC3 — parity enforcement cannot see ERISA. If §6 finds that the plans covering most commercially insured lives are outside the exams that generate compliance evidence, O4 for commercial populations is scored as a band-only / unknown axis and the filing says so.
KC4 (re-rank, not kill). If H2 is refuted — Medicaid acceptance is high — §2 demotes and §3/§4 (acute/crisis) become the flagship. If H7 is strongly supported, the whitepaper leads with "rebuild outpatient and crisis, not asylums," and bed-rebuild architectures fall in the ranking.
7. Anchor Table (M4)
Every value below is an unverified prior stated to orient the search — not a finding. ★ = verify first in Phase 0.
| # | Anchor (unverified prior) | Used in | Verify against | Verified value / delta |
|---|---|---|---|---|
| 1 ★ | Adult any-mental-illness prevalence ≈ 20–23%; SMI ≈ 5–6% (NSDUH recent year) | §1, H1 | SAMHSA NSDUH latest detailed tables; definition footnotes | Holds (2024 NSDUH): AMI 23.4% (61.5M); SMI 5.6% (14.6M). AMI/SMI locked to DSM-IV model + functional impairment (see Phase 0). |
| 2 ★ | No federal series publishes national median wait times for outpatient psychiatry/therapy by payer | §2, KC1 | CMS, SAMHSA, HRSA, GAO search; Merritt Hawkins / AMN as the circulating non-federal root | Confirmed absence as a series. Circulating waits are vendor/association; CMS has standards, not national by-payer medians. KC1 fires. |
| 3 ★ | Medicaid acceptance for psychiatrists is under 50% in multi-state secret-shopper or claims-based studies | §2, H2 | Peer-reviewed acceptance studies; state AG/insurance exams | Supported (two methods): Bishop NAMCS 43.1%; Brahmbhatt JAMA 2024 secret-shopper 17.8% appointment success (4 cities). H2 Supported. |
| 4 ★ | The "95% of psychiatric beds eliminated since peak deinstitutionalization" figure is single-root or definitionally non-comparable across eras | §3, KC2 | Root-trace circulating citations to primary bed censuses | Confirmed single-root TAC/Torrey public/state-hospital arithmetic; 1955≠modern whole-system. KC2 fires. |
| 5 ★ | Contemporary US psychiatric bed supply ≈ 10–15 per 100k (definition-sensitive: staffed vs. licensed; include/exclude VA) | §3 | NASMHPD / SAMHSA / OECD health-care resources | Two constructs locked (§3): state-hospital staffed ~10.8/100k (TAC 2023; 11.7 in 2016); CMS HCRIS hospital IPBs ~28.4/100k (2023, McBain). Do not blend. |
| 6 ★ | Medicaid IMD exclusion still blocks FFS payment for most adult MH IMD stays outside narrow waiver exceptions | §3, H3 | SSA Act §1905(i); CMS SPD / waiver inventory | Holds. Exclusion still default; SMI/SED §1115 patchwork (CRS IF10222 as of 2025-01-14). Distinct from SUD IMD (GBMT-8). |
| 7 ★ | 988 answered-volume roughly doubled vs. the pre-988 National Suicide Prevention Lifeline baseline within two years of transition | §4, H4 | Vibrant/988 performance reports; SAMHSA | Sharpened (§4). Prefer Jan answered ~2.5× (161k→404k, 2022→2024) or GAO routed contacts ~+90% / 19.1M (Jul 2022–Sep 2025). No single official “2×” headline. H4 Supported — diversion unproven. |
| 8 ★ | Mental disorders account for roughly one-third of working-age SSDI/SSI awards (diagnostic tabulation) | §8, H8, GBMT-12 seam | SSA Annual Statistical Report on the DI / SSI programs | Corrected. Disabled-worker awards mental disorders 12.7% (2023), not ~1/3. Stock % → GBMT-12. |
| 9 | CCBHC demonstration states show improved access metrics in SAMHSA/CMS evaluations | §5, §7 | Official evaluation reports | Verified (§5). Mathematica/ASPE: adult time-to-eval 9.0→5.4 days (DY1→DY2); clients +~9%; open-access common. ED/hospital impacts heterogeneous. |
| 10 | MHPAEA final rules (2024) meaningfully changed comparative-analysis obligations for plans | §6 | Federal Register; DOL enforcement fact sheets | Verified (§6). Six-element NQTL analyses + data-evaluation / meaningful-benefits duties. Perishable: May 2025 nonenforcement of new 2024 provisions; CAA 2021 statute remains. |
| 11 | ED boarding of psychiatric patients is measured in hours-to-days in multi-site studies, with wide state dispersion | §3 | Peer-reviewed boarding literature; ACEP surveys (flag advocacy) | Verified (direction): Pearlmutter 10-ED MA study — boarding drives LOS; Medicaid/uninsured >2× odds of ≥24h stay. Hours-to-days band holds; national point estimate not locked. |
| 12 | Jail mental-health caseloads exceed state hospital censuses in aggregate (BJS / state hospital surveys) | §3 | BJS; NRI state hospital census | Direction holds with footnotes: BJS 2011–12 jail 44% MH history / 26% SPD vs ~36k state-hospital beds (TAC 2023). Prevalence≠caseload census; construct footnotes mandatory. |
8. Candidate architectures to score
Scored per M6 against O1–O5 under the three weightings; each cell cited to a workstream finding or held at neutral (M10).
- Medicaid behavioral rate floor + administrative simplification — pay enough that panels open; cut prior-auth friction for routine outpatient codes.
- CCBHC expansion as default safety-net model — prospective payment, crisis required, measured access.
- IMD repeal or broad MH IMD waiver — let Medicaid buy adult inpatient capacity where clinically needed.
- Bed rebuild / state hospital reinvestment — the asylum-adjacent architecture, scored honestly against H7.
- 988 + mobile crisis + stabilization continuum completion — fund the middle tier so ED/law-enforcement are not the default.
- Parity enforcement with ERISA teeth — federal exams that reach self-funded plans; network adequacy secret-shopper standards.
- Assisted outpatient treatment expansion — civil-court leverage into treatment; selection effects front-and-center.
- Primary-care behavioral integration / Collaborative Care — scale evidence-based integration billing (CoCM codes) rather than specialty-only supply.
- Workforce liberalization — interstate compacts, supervision-ratio reform, peer specialists with Medicaid billing — markets as terrain per M7.
- Do-nothing comparator — current 988 funding path, patchwork IMD waivers, parity-on-paper trajectory.
9. Phase 0 execution notes
Days, not months. The gate pass does five things:
- Verify starred anchors 1–8, with anchor 2 / KC1 first — does any national realized-access instrument exist? A confirmed absence is a productive kill.
- Lock SMI vs. AMI definitions against NSDUH/SAMHSA documentation so §1 does not mix bands.
- Root-trace the bed-decline citogenesis claim (anchor 4) and locate the best contemporary bed inventory with definitions.
- Inventory IMD waiver status for MH (distinct from SUD) as of verification date.
- Pull SSA diagnostic award shares for the GBMT-12 seam (anchor 8) and hand off detail to the disability protocol.
Re-rank triggers. If KC1 fires, the whitepaper becomes a measurement-infrastructure filing with supply architectures scored on proxies. If H2 is refuted early, acute/crisis (§3–§4) becomes the flagship. If H7 is strongly supported, bed-rebuild architectures are demoted before full scoring waste.
Freeze. This protocol at execution start is the commitment. Report structure, scorecard scales, and the three weightings are fixed before synthesis; deviations logged one line each — what changed, why, effect.