Completes the baseline leg queued since Phase 0 (deviations-log #4).
No single canonical "LTSS spending" number exists — every figure is a downstream analyst choice
Raw CMS National Health Expenditure Accounts (NHEA) categories don't map cleanly onto LTSS: the "nursing care facilities & CCRCs" line mixes short-stay Medicare post-acute rehab with long-stay Medicaid custodial care (Medicaid shows as only 30.4% of that line — an artifact of the mixing, not a real payer share for custodial care), and "home health" mixes skilled Medicare home health with Medicaid personal-care HCBS.
**Best available LTSS-specific compiled total: KFF, 415B(2022data). * *Medicaid61253B, split ~47pts HCBS/14pts institutional — up from a 52% split 29/23 in 2000, confirming the long-run HCBS shift independently of anything found in §5); out-of-pocket 17% (~71B); Medicare + privateLTCinsurancecombined 2187B, not separated by any current source found). By setting: HCBS $284B vs. institutional $131B of the $415B. Scope corrected on Phase 1 verification (2026-08-10): this sentence previously read "HCBS spending crossed institutional spending nationally in 2013 (KFF)" as though it described the all-payer $415B total. The 2013 crossover is a Medicaid LTSS event; KFF's own "10 Things About LTSS" reports the Medicaid series ("from only 10% in 1988 to 62% in 2020") and dates no all-payer crossover. Say "Medicaid HCBS spending overtook Medicaid institutional spending in 2013." So framed, consistent with §5's "architecture, not aggregate money" framing since the aggregate shift toward HCBS coexists with per-state capping/waitlisting.
Cross-check finds real disagreement, and it's diagnostic, not noise. Urban Institute's DYNASIM microsimulation shows a starkly different split (Medicare 9.9%, Medicaid 34.3%, OOP 52.3%, private insurance 2.7%) that a Health Affairs critique found diverges substantially from NHEA-based accounting for the same period. DYNASIM is a projection model, not current-spending accounting — treat it as a modeling exercise, not a competing measurement. CBO's 2013 report used yet another (broader, Medicare-inclusive) definition and produced a third split. The field has no single measurement convention; every "LTSS spending" figure embeds an unstated choice about which claims lines count — the same finding-type as housing's LIHTC-cost result and Texas's spending-transparency gap in §5.
Effect on the protocol
§2's baseline should report the $415B/61%/17% KFF figure as the primary current estimate, explicitly flag Medicare-alone and private-insurance-alone shares as not currently isolable from any source found, and state the DYNASIM/CBO divergence as a methodology note rather than a competing fact. This is the third instance this session of "the US doesn't measure this domain the way its own policy debate assumes" (after §5's Texas gap and §8's caregiver-valuation non-reconciliation) — worth naming as a cross-cutting theme in the whitepaper, not three unrelated footnotes.
Sources
CMS NHE Fact Sheet; KFF "10 Things About LTSS"; KFF "5 Key Facts About Medicaid's Share of National Health Spending"; KFF Medicaid & LTSS Primer; Health Affairs (Urban Institute DYNASIM critical review); AARP Valuing the Invaluable 2026; CBO, "Rising Demand for Long-Term Services and Supports" (2013); MedPAC July 2025 Data Book.
Confidence tier
Medium-high for the headline range (~$400–450B/yr, Medicaid ~60%, OOP ~15–20%). Line-edit applied 2026-08-10 (queued at deviations #19, missed by the #23 pass, caught on Phase 1 verification): this line previously justified the tier because "multiple KFF compilations converge." Red-team #6 established that those compilations share upstream NHEA accounting assumptions — they are restatements of one convention, not independent cross-checks. The tier stands on the quality of that one convention, not on convergence. Low-medium for Medicare-alone and private-LTC-insurance-alone shares specifically — no current source isolates these; any single-point figure for those two should be treated as provisional.