GUBMENTPlain talk · policy frontier
Filings / Mental health / Sources / §3 Findings: IMD and acute capacity
GBMT-11 · Research record · No. 11

§3 Findings: IMD and acute capacity

mental-health/research/ws03-imd-acute.md
This is a working research document from the mental health filing, published as written — including the parts later corrected. It is the underlying record for Whitepaper No. 11, not a summary of it.

Date: 2026-08-11. Extends Phase 0 anchors 4–6 (KC2 / beds / IMD) and non-starred anchors 11–12 (ED boarding; jail MH). KC2 carried: drop absolute “beds since 1955” claims.

PRISMA-lite search log

Field Logged
Strings CMS HCRIS inpatient psychiatric beds McBain; NRI state psychiatric hospitals beds 2025; TAC staffed state hospital beds 2023; CRS IF10222 IMD exclusion 1115 SMI; CMS SMDL 18-011 SMI SED demonstration; psychiatric ED boarding multi-site Medicaid; BJS jail mental health SPD census; AHA Annual Survey psychiatric beds
Date range Prefer 2015–2025 contemporary inventories; historical TAC only for citogenesis context already settled in Phase 0
Inclusion Whole-system or multi-setting bed inventories with definitions; statute/CMS/CRS on IMD; peer-reviewed boarding with methods; BJS inmate MH indicators. Exclude 1955→present percentage claims as scorecard inputs.
Searched 2026-08-11

1. KC2 carried — no “beds since 1955” absolutes

Phase 0 confirmed: circulating “~95% of beds eliminated since 1955” / “558k → ~35k” is single-root TAC/Torrey public/state-hospital arithmetic. 1955 ≠ modern whole-system construct (general-hospital psych units, private psych hospitals, VA). This workstream scores contemporary capacity + IMD rules only.

2. Bed inventory constructs — LOCKED

Carry two constructs; never blend them in a scorecard cell.

Construct A — State-hospital staffed beds (public/asylum-adjacent)

Source Vintage Estimate Definition
TAC Going, Going, Gone 2016 37,679 staffed / 11.7 per 100k State hospital staffed beds
TAC Prevention Over Punishment 2023 survey 36,150 / 10.8 per 100k; 52% forensic occupancy Online/staffed state-hospital beds for adults with SMI; excludes pediatric DD and geriatric dementia beds
NRI Use of State Psychiatric Hospitals, 2025 2025 Profiles 90% of responding states (43/48) report inpatient psych bed shortage; forensic shortage most cited (37 states); 38 states report non-state hospital shortages State mental health agency survey — shortage reports, not a substitute national bed census

Use Construct A for: state-hospital / forensic / civil-commitment capacity debates; H7 steelman contrast later. Do not call this “US psychiatric beds.”

Construct B — CMS-certified hospital inpatient psychiatric beds (whole-system hospital frame)

McBain et al., PLOS Medicine 2025 — CMS HCRIS 2011–2023, psychiatric hospitals + short-term acute-care hospital (STACH) psych beds:

Year Total IPBs / 100k Psych hospital beds / 100k STACH IPBs / 100k
2011 28.1 16.8 11.2–11.3
2023 28.4 19.5 8.9–9.06

Additional geography: 846 counties (≈244M residents) saw IPB rate declines; 1,449 counties (≈59M) never had IPBs in-sample. Share of STACHs with an inpatient psych unit: 24.9% (2011) → 20.6% (2023) (related JAMA Network Open analysis).

Limits (authors): excludes VA / military / IHS (undercount); self-reported cost reports; not payer-resolvable; IMD §1115 waiver states did not show statistically significantly more beds in their models.

Locked filing language: “~28 per 100k CMS-certified hospital psychiatric beds (HCRIS, 2023)” or “~11 per 100k staffed state-hospital beds (TAC 2023)” — always with the construct name.

AHA Annual Survey is used in regional studies (e.g., HRR shortage papers) as another hospital-survey root; for this filing, HCRIS/McBain is the preferred contemporary whole-system hospital inventory because it is CMS administrative, national, and definitionally explicit.

3. IMD exclusion and §1115 MH waiver patchwork

Default law (holds): Social Security Act IMD exclusion bars federal financial participation for adults 21–64 in Institutions for Mental Diseases (>16 beds) outside narrow statutory exceptions. Distinct from SUD IMD waivers (GBMT-8).

CMS opportunity: SMDL #18-011 (Nov 2018) — SMI/SED §1115 demonstration for short-term IMD stays with milestones.

Patchwork (perishable): CRS IF10222 (as of 2025-01-14, Phase 0 S10) lists approved MH IMD waiver geography (AL, CA, CO, DC, ID, IN, KY, MD, MO, NH, NM, OK, UT, VT, WA; MA on CMS opportunity page) plus a pending queue — not national repeal. Re-check before whitepaper ship.

Implication: Medicaid’s ability to buy adult inpatient capacity in IMDs remains waiver-contingent. That is a financing constraint even where Construct B beds physically exist.

4. Anchor 11 — ED boarding

Source Frame Finding
Pearlmutter et al., Ann Emerg Med 2017 10 Massachusetts EDs; 871 MH evaluations Median boarding/LOS by disposition: admitted 5.63h, transferred 9.32h, discharged 1.23h. Medicaid and uninsured: significantly longer total LOS; >2× odds of ≥24h ED stay vs privately insured
Nolan et al. / NHAMCS analyses (e.g., West J Emerg Med 2015 lineage) National ED sample Psych visits board at higher rates/durations than non-psych (e.g., ~21.5% psych vs ~11% all ED under a >6h boarding definition in one national analysis)
ACEP Advocacy / policy Documents hours-to-days boarding; flag as advocacy — use for problem recognition, not as the measured multi-site estimate

Verified for anchor table: ED psychiatric boarding is measured in hours to >24 hours in multi-site peer-reviewed work, with insurance-linked dispersion (Medicaid/uninsured worse in the MA multi-site study). National point estimate with tight confidence not locked; state dispersion is real.

5. Anchor 12 — Jail MH vs state-hospital census

Source Construct Figure
BJS Indicators of Mental Health Problems… 2011–12 Jail inmates with history of mental disorder (told by professional) 44%
BJS same Jail inmates meeting SPD (K6≥13) past 30 days 26% (vs ~5% standardized general population)
TAC 2023 Staffed state-hospital beds ~36,150

Comparison rule: Applying BJS history share to a mid-2010s jail custody population on the order of ~700k+ implies hundreds of thousands of jail inmates with a reported mental-disorder history — orders of magnitude above contemporaneous state-hospital bed census (~36k).

Caveats (mandatory in any scorecard cite): (1) prevalence among inmates ≠ a “mental-health bed” or treatment caseload census; (2) SPD/history ≠ SMI clinical diagnosis; (3) BJS vintage is 2011–12. Directional claim for de-facto capacity / criminalization overflow is supported; precise “jail caseload > hospital census” arithmetic must carry the construct footnotes.

6. H3 adjudication

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.

Criterion Result
Bed-supply or boarding metrics worsen at Medicaid-adult margin in non-waiver vs waiver states in ≥2 independent evaluations Not met cleanly. McBain (HCRIS) finds no statistically significant higher bed counts in MH IMD waiver states. Comparative waiver/non-waiver boarding evaluations not located as a pair this pass.
or CMS/state documents state capacity cannot be purchased because of IMD Met. Statute + CMS SMDL #18-011 + CRS IF10222: FFP for adult IMD stays is barred as default; states must obtain §1115 authority for short-term coverage. That is an explicit purchase constraint.
Boarding at Medicaid margin (supporting, not the formal criterion) Pearlmutter: Medicaid/uninsured face longer MH ED LOS / higher ≥24h odds — acute-system stress at the public-payer margin
Refute path (no material capacity difference after case-mix) Comparative bed null in McBain is a caution for architecture scoring, not a full refute of the financing constraint
Indeterminate if bed inventories not payer-resolvable Inventories remain not payer-resolvable (Construct B limit) — disclosed

Adjudication: Supported — on the statutory/CMS documentary path (Medicaid cannot buy most adult IMD capacity without waiver). Comparative waiver-vs-non-waiver bed evidence is weak/null so far; do not overclaim that waivers have already rebuilt Construct B supply.

Implications for scorecard

Confidence: Strong on IMD law and construct split; Moderate on boarding/jail absolute comparisons

Strong: KC2, Construct A vs B lock, IMD default + waiver patchwork. Moderate: boarding hours generalization beyond MA/multi-site samples; jail-vs-hospital comparison direction vs precise caseload math. Weak: payer-resolved bed inventory (not located).

← All Mental health research documents Sources digest Read the whitepaper