Date: 2026-08-11 Scope: the protocol's starred anchors (1–8), SMI/AMI definition lock, the psychiatric-bed citogenesis root-trace, the MH IMD waiver inventory, and the KC1 realized-access instrument test. This is a gate pass, not a scorecard or a whitepaper. It changes no site page.
Session note: Work resumed from a Claude session that claimed the branch, spawned three verification agents, and stopped at a session-limit mid-fetch. This pass re-opens primary sources rather than trusting unfinished agent scratch.
Verdict: GO — and KC1 / KC2 both fire
The desk wager survives the cheap tests that can kill it.
- KC1 fires. No federal (or ≥80%-population) series publishes outpatient mental-health appointment wait or new-patient acceptance by payer. The circulating numbers are vendor surveys, professional-association surveys, and one-off secret-shopper studies. CMS has begun writing standards (Medicare Advantage 30-business-day routine behavioral wait; proposed Medicaid managed-care secret-shopper rules) — standards are not a measured national series. Under the protocol, that collapse is the filing's headline shape, not a caveat: the country argues about a shortage it does not measure at the point of use.
- KC2 fires. The circulating “~95% of psychiatric beds eliminated since 1955” / “558k → ~35k” figure is single-root advocacy arithmetic on state (and historically state/county) hospital beds. It is not a whole-system bed census, and the 1955 and modern counts are not the same construct once general-hospital psychiatric units, private psychiatric hospitals, and VA beds are in frame. §3 must drop absolute “beds since 1955” claims and score contemporary capacity plus IMD rules.
- Anchor 8 is corrected hard. Mental disorders are 12.7% of 2023 disabled-worker awards, not “roughly one-third.” The one-third prior does not survive the SSA awards table. Absolute stock on the rolls is still large; the precise stock share is handed to GBMT-12 with the diagnostic split.
No kill condition stops the filing. KC1 and KC2 reframe it: measurement infrastructure and definitional honesty become load-bearing, and bed-rebuild architectures enter the scorecard without a false historical baseline.
1. Definitions locked (AMI / SMI)
From SAMHSA's 2024 NSDUH annual national report (verbatim structure):
- AMI (adults 18+): any mental, behavioral, or emotional disorder in the past year of sufficient duration to meet DSM-IV criteria, excluding developmental disorders and substance use disorders.
- SMI: among adults with AMI, any such disorder that substantially interfered with or limited one or more major life activities.
- Production method: statistical prediction models trained on clinical interview data from a subset of adult NSDUH respondents (2008–2012), applied to the 2021–2024 adult samples; source variables are statistically imputed. A DSM-5 clinical calibration study is in progress; current published AMI/SMI estimates are still the DSM-IV model.
Band rule for this filing: SMI and common-disorder / AMI-minus-SMI markets are not interchangeable. Workstreams that mix them without saying so are defective (protocol §1.3). Adolescent MDE is a separate instrument and is not an adult AMI/SMI substitute.
Source: 2024 NSDUH Annual National Report (HTML); highlights PDF.
2. Gate results by starred anchor
| # | Protocol prior | Phase 0 result | Confidence / source type | Effect |
|---|---|---|---|---|
| 1 ★ AMI / SMI prevalence | Adult AMI ≈ 20–23%; SMI ≈ 5–6% | Verified (2024): AMI 23.4% (61.5M); SMI 5.6% (14.6M). No statistically significant change 2021–2024 for adults overall; SMI among 18–25 fell 12.0% → 9.4%. | Strong — SAMHSA primary. | Prior holds. Carry 2024 figures with as-of date. |
| 2 ★ National wait / acceptance by payer | No federal series | Confirmed absence as a series. GAO-22-104597 documents access problems without publishing a national by-payer median wait. HRSA workforce briefs cite a 48-day “national average wait” to the National Council for Mental Wellbeing (2025) — not CMS/SAMHSA/HRSA administrative data. AMN / Merritt Hawkins physician wait surveys are commercial samples. CMS has MA network wait standards and has moved toward Medicaid managed-care secret-shopper rules; those are compliance instruments, not a published national median series by payer. | Strong on the absence of a federal series; moderate that no ≥80%-population non-federal series fills the gap. | KC1 fires. O1 cannot be scored on realized access from a national instrument. |
| 3 ★ Medicaid acceptance <50% for psychiatrists | Under 50% in multi-state / national studies | Supported on the NAMCS root: Bishop et al., JAMA Psychiatry 2014 — psychiatrists' Medicaid acceptance 43.1% (95% CI 34.9–51.7) in 2009–2010 vs 73.0% for other physicians; private noncapitated 55.3% vs 88.7%. A later Psychiatric Services update (2007–2016) exists; HRSA's 2024/2025 workforce briefs still cite mid-40% Medicaid acceptance figures for psychiatrists. | Strong for 2009–10 NAMCS; moderate that the level remains <50% nationally without a new official series. | H2's acceptance leg is live. Still needs a second independent method (secret-shopper / claims) in §2 before full support. |
| 4 ★ “95% beds gone since 1955” citogenesis | Single-root / non-comparable | Confirmed single-root advocacy arithmetic. Treatment Advocacy Center (Torrey lineage): 1955 558,239–558,922 state/county public beds → 2005 52,539 (≈17/100k) framed as “95% no longer available”; 2016 TAC survey 37,679 staffed state hospital beds (11.7/100k). Roots are TAC surveys plus earlier CMHS/NRI public-bed censuses — not an independent whole-system inventory. | Strong on root identity and definitional mismatch. | KC2 fires. Drop “beds since 1955” absolutes. |
| 5 ★ Contemporary bed supply ≈ 10–15 / 100k | Definition-sensitive | State-hospital staffed beds ≈ 11.7 / 100k (TAC 2016) — inside the prior band for that construct. OECD-style “psychiatric beds” comparisons TAC cites (~68/100k peer average) are a different construct. No Phase 0 claim is made for all-payer staffed psych beds including general-hospital units + private + VA. | Moderate — best contemporary public census located is advocacy-run; NRI/SAMHSA inventory still needed in §3. | Carry as state-hospital supply only, with definition footnote. |
| 6 ★ IMD still blocks most adult MH FFS stays | Outside narrow waivers | Verified. SSA Act IMD exclusion still bars FFP for adults 21–64 in IMDs (>16 beds) as the default. CMS SMI/SED §1115 opportunity (SMDL #18-011, Nov 2018) allows short-term IMD stays with milestones; CRS (IF10222, as of Jan 14, 2025) lists a patchwork of approved MH IMD waivers (AL, CA, CO, DC, ID, IN, KY, MD, MO, NH, NM, OK, UT, VT, WA; MA listed on CMS's opportunity page) and a pending queue — not a national repeal. Distinct from SUD IMD waivers (GBMT-8). | Strong — statute + CMS/CRS. | H3 stays live; score waiver vs non-waiver geography in §3. |
| 7 ★ 988 volume roughly doubled in two years | vs pre-988 Lifeline | Direction verified; “doubled” is roughly right for answered contacts, not a single official annual multiplier. Vibrant KPI extracts: Jan 2022 answered 161,267 network contacts (ex-VCL) vs Jan 2024 403,544 (~2.5× that month). KFF: by 2024 monthly contacts exceeded 0.5M, ~80% above May 2022. GAO-26-108114: ~19.1M calls/texts/chats routed Jul 2022–Sep 2025; call volume +87% over that window; answer rates improved. No multi-state causal ED/arrest diversion series located in this pass. | Strong on volume growth; weak on diversion. | H4's volume leg supported; diversion leg still thin → H4 leans supported pending §4. |
| 8 ★ Mental disorders ≈ 1/3 of working-age awards | SSA diagnostic tabulation | Corrected. SSA Annual Statistical Report on the DI Program Chart 10 equivalent: among 2023 disabled-worker awards, mental disorders were 12.7% (musculoskeletal 34.0%, neoplasms 13.6%). SSA breaks mental disorders into autism, developmental, intellectual, depressive/bipolar, neurocognitive, schizophrenia-spectrum, and other — intellectual disability is inside the mental-disorders umbrella, not a separate top-level share. December 2024 beneficiary stock tables show large absolute mental-disorder caseloads (e.g. depressive/bipolar ≈ 1.23M; intellectual disorders ≈ 1.23M among 18–64 in the published distribution). | Strong on awards share; stock % left for GBMT-12 once Table 69's denominator is archived offline. | Prior “one-third of awards” is false. H8's award-share leg must be rewritten; overflow-valve story can still use stock + local access links. |
3. Baseline skeleton — non-substitutions
- Prevalence is not access. 61.5M adults with AMI and 14.6M with SMI are need estimates from a prediction model. They are not appointment capacity.
- Licensed headcount is not usable supply. Bishop's NAMCS gap (Medicaid 43% vs other physicians 73%) is the existence proof that payer participation can dominate headcount. It is a 2009–10 survey of office-based physicians — not a 2024 secret-shopper census of therapists.
- State-hospital beds ≠ system beds. Any scorecard cell that cites “beds since 1955” without the TAC/state-hospital qualifier fails the record.
- 988 volume ≠ diversion. Answered-contact growth is real and well-documented by the administrator and by GAO. Downstream ED and arrest effects are not established at national tier in this pass.
- Awards ≠ rolls. 12.7% of new disabled-worker awards is not the same claim as “a third of the disability system is psychiatric.”
4. Kill / re-rank triggers after this gate
| Trigger | Status after Phase 0 | Filing consequence |
|---|---|---|
| KC1 — no national realized-access instrument | Fires | Whitepaper leads with measurement failure; O1 scored on proxies and disclosed as such. |
| KC2 — bed counts citogenic | Fires | §3 drops historical absolutes; bed-rebuild architectures scored on contemporary evidence only. |
| H2 early refute (Medicaid acceptance high) | Does not fire | Flagship §2 stays; still needs second method. |
| H7 (community > asylum rebuild) | Untested | Equal-effort steelman still owed in §7. |
| H8 award-share prior | Prior fails | Rewrite before any GBMT-12 seam prose uses “one-third of awards.” |
5. Prospective follow-up (not done here)
| Item | Why it matters | Owner |
|---|---|---|
| Second acceptance method (secret-shopper / claims, multi-state, post-2015) | Completes H2 | §2 |
| NRI / SAMHSA contemporary bed inventory with definitions | Replaces TAC as sole modern census | §3 |
| Archive CRS IF10222 Table 1 + CMS SMI demo list as-of date | Perishable waiver geography | §3 |
| Vibrant / SAMHSA annual 988 series with pre-2022 Lifeline baseline in one table | Tighten anchor 7's multiplier | §4 |
| SSA Table 69 full denominator for 18–64 stock % | GBMT-12 seam | §8 / GBMT-12 |
| NHIS / Household Pulse alternate prevalence definitions | H1's >1.5× cross-series test | §1 |
Freeze
Protocol at docs/research-inquiry.md is the commitment as of Phase 0 start (2026-08-11). Anchor table verified values above supersede the unverified priors for anchors 1–8. Deviations logged in research/deviations-log.md.