Tests H7.1 (incident undercounting) and H7.2 (growing Medicaid AL footprint) with real data.
H7.1 — SUPPORTED: undercounting is a persistent, government-documented gap, not a one-off
Two GAO reports eight years apart document persistent federal blind spots in this setting — the first on incident counts, the second on spending: GAO-18-179 (2018) found only 22 of 48 studied states could report critical-incident counts (abuse, neglect, medication errors, unexplained deaths) in Medicaid-funded AL settings — 26 could not, citing inadequate tracking systems, and only 34 of 48 made any incident data public. GAO-26-107884, Assisted Living Facilities: Information on Federal Spending and Medicaid Coverage (June 2, 2026) documents a different undercount: "federal Medicaid and Medicare spending for services provided in assisted living facilities totaled at least $12 billion in 2024. This amount is likely an undercount because of data limitations" ($3.5B Medicaid + $8.5B traditional Medicare). Corrected on Phase 1 verification (2026-08-10): this sentence previously read that GAO-26-107884 "confirms the gap persists," meaning the incident-reporting gap. It does not — that report is about federal spending and Medicaid coverage and makes no critical-incident-reporting finding. Two reports, two undercounts, of different things. The eight-years-apart framing survives; "both found most states can't count incidents" does not. CMS finalized a rule (May 2024) requiring annual state critical-incident reporting, but it doesn't take effect until 2027 at the earliest — the fix is real but not yet operative, so the gap this hypothesis describes is current, not historical.
Where real reporting infrastructure exists (Washington's mandatory online incident portal, Florida's public FloridaHealthFinder.gov), incident volume looks substantial — a Washington Post investigation (paywalled, secondhand only) found 2,000+ elopement incidents since 2018 with ~100 deaths, fatality rates rising sharply with time-missing. This is consistent with "undercounted elsewhere, not actually safer," but no rigorous state-by-state AL-vs-SNF per-resident incident-rate comparison exists in public research — the undercounting claim is well-supported: the "not actually safer" inference is plausible, not itself measured.
H7.2 — CONSISTENT WITH, BUT NOT EVIDENCE FOR, de facto institutionalization
Medicaid HCBS-waiver coverage for AL services now exists in most states (sources disagree 41–46 depending on survey date/definition — not a real conflict, a measurement-consistency issue consistent with this domain's recurring pattern). Medicaid HCBS spending overall surpassed institutional LTSS spending nationally in 2013 and has kept growing. NCAL/AHCA data puts ~18–20% of AL residents as Medicaid-supported (services only, never room-and-board). No clean multi-year time series exists for AL-waiver state-count growth or for how the Medicaid-supported resident share has moved. The national HCBS trend is consistent with, but does not discriminate for, the de facto-institutionalization hypothesis: home-based HCBS could account for the whole shift. This is a genuine data gap, not a null result.
Effect on the protocol
H7.1 is supported by two independent GAO reports rather than advocacy framing. H7.2 remains a named data gap, not a directional finding. The whitepaper can cite the 2027 CMS reporting rule as a concrete, dated marker: assisted-living data quality has an actual (if distant) fix already in motion, which is a different and more useful claim than "the data doesn't exist and nothing is being done."
Sources
GAO-26-107884 (2026); GAO-18-179 (2018); Washington DSHS Online Incident Reporting; KFF, "Medicaid Home Care (HCBS) in 2025"; KFF, "What Services Does Medicaid Cover in Assisted Living Facilities?"; NIC, "Assisted Living Medicaid Waivers Expand Access and Choice"; Washington Post elopement investigation (paywalled, secondhand only).
Confidence tier
Medium for H7.1. GAO figures are government-primary and cross-validated across two reports eight years apart. The Washington Post state-comparison evidence rests on secondhand characterization of a paywalled source. Low for H7.2: no clean longitudinal dataset identifies the assisted-living-specific growth or payer mix needed to test it.