Research Inquiry: Elder Care (GBMT-7)

Status: Phase 0 executed 2026-08-03 — findings, baseline, deviations. Verdict GO, with the §12 unpaid-caregiving-valuation kill condition fired as a headline, anchors 7 and 11 broken/reframed. Full execution pending. Scope: United States; long-term services and supports (LTSS) for older adults and adults with disabilities — home- and community-based care, assisted living/residential care, skilled nursing, unpaid family caregiving, and the financing that sits under all four (Medicaid, Medicare's narrow role, private long-term care insurance, out-of-pocket spend-down). Method: Imports method/gubment-method.md (M1–M8) in full. This document contains only domain content. Artifact: A feasibility assessment of the leading "fix elder care" theories, naming binding constraints in priority order and scoring candidate policy architectures — Whitepaper No. 7.

1. Definitional decisions to settle before fieldwork

1. Which elder-care problem? At least five travel under the phrase, with different mechanisms and different fixes: (a) direct-care workforce shortage and wages; (b) financing — spend-down to Medicaid impoverishment and the collapse of the private long-term care insurance market; (c) unpaid family caregiver burden — health, financial, and labor-market cost borne by relatives; (d) institutional quality and safety — nursing home staffing and neglect, private-equity ownership effects; (e) access — Medicaid home- and community-based services (HCBS) waiting lists and the institutional bias in how the entitlement is actually structured. Treat as linked but distinct, the way GBMT-1 treated age bands and GBMT-2 treated the four housing problems.
2. Long-term care is not the same system as acute medical care, and the funders are not who most people assume. Medicare covers acute care and only narrow, time-limited post-acute stays (skilled nursing up to 100 days, conditional home health) — it does not pay for custodial long-term care. Medicaid, not Medicare, is the payer of last resort for LTSS, reached only after a household spends down assets. Every workstream below states which system it is analyzing and does not let "elder care" default to the Medicare-shaped assumption most readers carry in.
3. Beneficiary of the policy — the older adult, or the family caregiver? Mirrors childcare's child-vs.-parent question (§1.2 of the childcare protocol). A caregiver tax credit or paid-leave mandate relieves the family but does nothing for direct-care workforce capacity or institutional quality; a wage-floor-and-staffing design does the reverse. Most current federal proposals are caregiver-support bills, not care-delivery bills — the deliverable states which target each candidate architecture actually serves.
4. Setting taxonomy — four different markets, unevenly counted:
   • Home- and community-based services (HCBS): personal care aides, home health aides, adult day programs, PACE. Delivered mostly through Medicaid §1915(c)/(i)/(k) waivers, capped and waitlisted in most states.
   • Assisted living / residential care: licensed and regulated at the state level only — no federal Conditions of Participation, unlike nursing homes. A large and growing share of paid care with the thinnest data.
   • Skilled nursing facilities (SNF): the only setting with a federal entitlement (Medicaid) and federal CMS regulation, and consequently the most-studied and most-legislated setting despite housing a minority of care recipients.
   • Unpaid family caregiving: by hours, plausibly the largest "provider" in the system, and — as with FFN care in childcare — the segment most likely to be undercounted by any baseline built from paid-market data alone. A protocol that only covers nursing homes, the way most elder-care journalism does, misses most of the actual care hours. Each of the four settings gets its own workstream below (§5 HCBS, §6 SNF/institutional quality, §7 assisted living, §8 family caregiving) rather than defaulting to comparable-rigor-in-principle only.
5. Objective function (per M6/GBMT-1 lesson): cost containment, institutional quality and safety, direct-care workforce sustainability, family-caregiver relief, and aging-in-place/autonomy rank candidate architectures differently. A caregiver-tax-credit-first design scores well on relief and poorly on workforce; a wage-floor-first design is the reverse. Score under explicit weightings; report rank stability, as GBMT-1 and GBMT-2 did.

2. Workstreams

§2 Baseline — population 65+ and 85+ by state (the 85+ band is the load-bearing denominator, since LTSS need concentrates there far more than raw 65+ counts suggest); current LTSS spending by payer (Medicaid, Medicare's narrow post-acute slice, out-of-pocket, private LTC insurance, and the imputed value of unpaid care) vs. utilization by setting. Microdata-first: ACS, HRS (Health and Retirement Study — the key longitudinal instrument for this domain, the NSECE-equivalent), CMS/Medicaid administrative data (MSIS/T-MSIS, HCBS waiver reports), NHATS. Seed: H2.1 unpaid family caregiving accounts for the majority of LTSS hours delivered nationally, mirroring childcare's FFN finding; H2.2 Medicaid, not Medicare, is the primary public LTSS payer despite Medicare being the program most associated with elder benefits in public perception; H2.3 no unified national HCBS waiting-list count exists — states report under incompatible waiver definitions and with unknown duplicate-listing rates (the H2.3-of-elder-care).

§3 Workforce (treat as the primary binding constraint, per the childcare and housing rhyme) — direct-care workforce (home health aides, personal care aides, CNAs, nursing home staff): headcount, wages, turnover (among the highest of any major occupation group), credential/training-hour requirements by state, and reliance on foreign-born labor. Seed: H3.1 direct-care turnover and vacancy rates constrain HCBS waiver expansion more tightly than waiver funding does — the same shape as childcare's H4.2 and housing's H4.1; H3.2 immigration restriction is a first-order supply constraint given the sector's foreign-born worker share, and its interaction with any restrictive immigration policy should be modeled explicitly, not gestured at.

§4 Financing & the private LTC insurance collapse — Medicaid asset/income spend-down rules and estate recovery; the near-total exit of carriers from the private long-term care insurance market and the premium-shock litigation that followed for policyholders who stayed; Washington State's WA Cares payroll-tax public LTC benefit as the one live domestic social-insurance natural experiment; the CLASS Act's 2013 repeal before it ever launched (a fired kill condition worth dissecting for why, since a federal LTC social-insurance program has already been tried and failed at the design stage, not the implementation stage). Seed: H4.1 private LTC insurance is not viable at a price point most households will pay, at any level of medical underwriting (test against carrier-exit and premium-increase data, not assumption); H4.2 WA Cares is the load-bearing precedent for whether a payroll-tax social-insurance design survives contact with a real legislature and a real opt-out fight.

§5 HCBS waiting lists & the institutional bias — nursing home care is a mandatory Medicaid entitlement; HCBS is delivered mostly through optional, capped waivers with enrollment caps and waiting lists in most states — the reverse of what "home care is cheaper, so of course it's prioritized" intuition suggests. Money Follows the Person program history (lapses and reauthorizations); the ADA's Olmstead integration mandate as the legal lever forcing state rebalancing toward HCBS. Seed: H5.1 states with the largest HCBS waiting lists are not those spending the least on LTSS overall but those with the most institutionally-biased spending mix — a "binding constraint is architecture, not money" finding echoing housing's zoning result.

§6 Institutional quality & ownership — nursing home staffing adequacy relative to the 2024 CMS federal minimum-staffing rule (verify current status — implementation, litigation, and rescission risk under a change of administration are all live); CMS Care Compare and the Special Focus Facility program; private-equity ownership of nursing homes and its documented effects on staffing levels and resident mortality (there is real peer-reviewed literature to search here, not just advocacy claims — locate it via the §0.2-style method rather than recalling headline figures). Assisted living's regulatory gap is real but is its own workstream (§7), not a footnote here — the two settings should not be scored against the same literature base. Seed: H6.1 PE-owned facilities show a measurable staffing/outcome delta versus non-PE facilities in the published literature (size it, do not assume the magnitude commonly cited in press coverage).

§7 Assisted living — the setting with the thinnest federal data footprint (per §1.4): licensed and regulated at the state level only, no federal Conditions of Participation, no CMS Care Compare equivalent, and no uniform staffing-ratio or incident-reporting requirement comparable to nursing homes. State licensure-category proliferation (assisted living, residential care, personal care home, memory care — inconsistent naming, scope, and acuity limits across states) is a first-order measurement problem that has to be resolved before any cross-state comparison is possible. Private-pay dominance: Medicaid rarely covers room-and-board here, only services layered on via HCBS waivers, which sorts access by income in a way nursing-home Medicaid eligibility does not. NIC MAP (National Investment Center) is the best available industry census — a private, subscription, real-estate-investor-facing source, which is itself a source-type worth flagging in the register per M1 rather than treating as neutral administrative data. Seed: H7.1 the absence of federal reporting requirements means assisted-living safety incidents are systematically undercounted relative to nursing homes, not because the underlying incident rate is actually lower; H7.2 Medicaid's assisted-living footprint (services-only, via HCBS waivers) is a growing share of state HCBS spending — a de facto institutionalization-by-another-name that the §5 "architecture, not money" framing should be tested against, not assumed to be untouched by it.

§8 Family caregiving — unpaid caregiver headcount, hours, and imputed economic value (the AARP/National Alliance for Caregiving Caregiving in the US survey series is the standard instrument here — treat its headline valuation as a claim to verify, not a fact to import); caregiver health and financial effects, including labor-force exit and retirement-savings impact; the interaction with paid-family-leave policy; caregiver tax credit proposals (e.g., the Credit for Caring Act) and their static-cost estimates; self-directed care programs that pay family members directly (Medicaid's Consumer-Directed/Cash-and-Counseling models — the clearest existing natural experiment in paying family caregivers rather than merely crediting them). The Older Americans Act's existing federal caregiver-support infrastructure — Title III-E's National Family Caregiver Support Program (NFCSP: respite, counseling, and support-service grants routed through the Aging Network of State Units on Aging and Area Agencies on Aging) and Title III-C home-delivered and congregate meals — predates every current legislative caregiver-support proposal by decades and should be treated as the existing baseline any new architecture (Credit for Caring Act included) is measured against, not as a novel intervention entering a vacuum. Seed: H8.1 the imputed value of unpaid family caregiving exceeds total paid LTSS spending nationally — verify the actual estimate and its methodology rather than repeating the commonly cited headline figure uncritically, mirroring childcare's H2.1 treatment of FFN care; H8.2 OAA Title III-E (NFCSP) national appropriations are orders of magnitude below the imputed value of unpaid caregiving (row 4) and below Medicaid HCBS waiver spending — verify the actual ratio rather than assuming "token" or "substantial" a priori, since existing federal caregiver support is a floor for any new architecture's cost-effectiveness comparison, not a null baseline.

§9 Federalism & legal mechanics — Medicaid waiver architecture (§1915(c)/(i)/(k), §1115 demonstrations) and the resulting state-by-state variation in HCBS generosity; Olmstead as the federal integration mandate forcing state rebalancing; the 2024 CMS HCBS access rule and nursing-home minimum-staffing rule as live, contested federal interventions with real litigation and reversal risk — good anchors to verify current status on, not assume settled. Seed: H9.1 (the GBMT-2 rhyme, tested rather than assumed) — is the binding lever federal (CMS rule-making, Medicaid federal match rate) or state (waiver design, waiting-list management, provider rate-setting)? Housing found the local/state lever dominant; elder care's Medicaid match structure may invert that.

§10 Precedents — Domestic: WA Cares (the live payroll-tax LTC benefit), the CLASS Act (federal LTC social insurance repealed pre-launch — why), PACE (Program of All-Inclusive Care for the Elderly — small, high-evidence, historically hard to scale), Money Follows the Person, the Older Americans Act (1965 nutrition/support-services base, NFCSP added 2000 — the oldest live federal caregiver-support precedent, and the natural baseline for scoring any new caregiver-support architecture's marginal value). International: Germany's Pflegeversicherung (social LTC insurance since 1995 — the direct financing analog), Japan's Kaigo Hoken (LTC insurance, and the most relevant international case given Japan's demographic pressure is furthest along), the Netherlands' Wlz. Each with a transferability assessment per the GBMT-1/GBMT-2 template.

§11 Candidate architectures (seed list, to be scored per M6): public LTC social insurance at federal scale (Germany/Japan model, or WA Cares nationalized); Medicaid HCBS waiver expansion/de-capping; direct-care wage floor tied to funding (the childcare-workforce-fund rhyme); family caregiver tax credit or direct payment (Cash-and-Counseling expansion); PACE expansion; federal minimum standards for assisted living; OAA/NFCSP expansion as the lowest-disruption incremental option; private LTC insurance market reform (reinsurance/risk-pool models); cash as the comparator, per M7. Seed: H11.1 (the GBMT-1/GBMT-2 rhyme) — instruments that add demand-side money without addressing direct-care workforce capacity inflate wages/prices in the paid-care market without adding care hours, the same failure mode found in childcare and implied in housing.

§12 Sequencing — workforce wage floor and pipeline first (slow-building and the likeliest binding constraint per §3); HCBS de-capping and waiver reform second; family-caregiver payment and tax-credit instruments (including OAA/NFCSP expansion) last, since they are the fastest to legislate but do least to solve the underlying capacity problem. To be tested, not assumed.

3. Anchor Table (seed — verify per M4; ★ = candidate for Phase 0)

Figures below are approximate priors stated to orient the search, not findings — the same discipline as GBMT-1/GBMT-2. Expect several to be wrong; a wrong anchor is a finding about this protocol, not an error to quietly correct.

4. Execution notes

Phase 0 first (M6): verify ★ anchors, pull the §2 baseline skeleton (ACS + HRS + a first T-MSIS/CMS administrative pull), scan WA Cares + the CLASS Act repeal + one HCBS waiting-list state case. Kill conditions: if §2 shows the unpaid-caregiving baseline cannot be reconciled to within a defensible range across sources, that is the headline, mirroring housing's §4 shortage-estimate result; if §3 shows direct-care wages are not the binding constraint on HCBS capacity once waiver funding is held constant, the workforce-first theory of the case (§12's sequencing) needs rebuilding before architecture scoring. Effort shape: §2, §3, §4, and §10 carry the most information per hour on current priors — but per M3, those priors are exactly what Phase 0 should try to break.