§12 PACE and sequencing — late evidence pass

Purpose. Close the scorecard's declared PACE gap and test the protocol's sequencing claim without silently rewriting the published board. This is a research record, not yet a scorecard or site update: adding PACE to the board requires re-running the three rankings and reconciling any changed headline against the existing red-team record.

M3 limitation. The frozen protocol named PACE but did not pre-register PACE-specific adjudication criteria, and this late pass began before such criteria existed. It is therefore exploratory evidence and a provisional scorecard recommendation, not a final M3 adjudication. The administrative directory enumeration below is descriptive, not a causal hypothesis test. A future integration pass must write support/refute/indeterminate conditions before seeking additional comparative or causal PACE evidence.

Criteria frozen for the next PACE integration pass

These criteria apply prospectively; they cannot cure the sequencing defect in this late pass.

Any future score uses a neutral cell for a criterion still indeterminate and must cite the resulting finding, rather than importing generic HCBS evidence.

M2 search record

Run 2026-08-06. Searches: “PACE cost nursing home admissions mortality 2006-2011”; “PACE integrated care systematic review dual eligible 2024”; “PACE CMS application interdisciplinary team service-area expansion”; and “CMS PACE enrollment plan directory 2026”. Inclusion: U.S. PACE evidence that compares outcomes, spending, or operational requirements; federal program documents; and a recent systematic review. Excluded: provider marketing, uncited assertions about national enrollment, and studies of adjacent integrated-care models without a PACE-specific result.

This is a bounded late pass, not a new systematic review. It updates the unscored architecture from its own evidence and leaves claims it cannot support neutral.

What PACE is — and why its delivery model matters

PACE is a Medicare program and Medicaid state-plan option for community-based adults who meet their state's nursing-home level-of-care standard. It is not a voucher, a waiver-rate increase, or a generic managed-care label: the organization accepts prospective Medicare and Medicaid capitation and must provide the full covered-service package. CMS requires services at a PACE center, the participant's home, and inpatient facilities; a new organization also undergoes federal application and state-readiness review.

That design carries a practical scale constraint. The required PACE center has an interdisciplinary team, and CMS's posted PACE manual specifies eleven roles (including a physician, registered nurse, social worker, therapists, dietitian, home-care coordinator, personal-care-attendant representative, and driver). Each center needs a functioning team for its participants; expanding PACE therefore creates a demand for the same direct-care and clinical labor that the filing has not shown can expand on command. This is a design fact, not a claim that PACE cannot scale.

Sources: CMS overview; CMS 2025 application and service-area expansion materials, pp. 4–5 and 74–75; CMS PACE manual, interdisciplinary-team requirements, pp. 3–4.

What the outcome evidence can and cannot say

1. There is favorable outcome evidence, but it is not a clean national causal estimate

The early CMS/Abt evaluation compared PACE enrollees with people who declined PACE, adjusted for observed characteristics. In the first six months, it found lower hospital admissions and fewer nursing-home days for PACE enrollees; it also found more days in the community and initially higher reported life satisfaction. The differences narrowed at later follow-up points, and the report itself flags selection bias and small samples late in follow-up. This is positive, partial evidence for care coordination and autonomy — not an RCT, and not proof that a new PACE center will reproduce the same effect.

The 2024 JAMA Health Forum systematic review searched 2010–2023 studies of integrated care for full-benefit dual eligibles. Its PACE subset found associations with reduced long-term nursing-home stays in three of four studies. That independently synthesizes the direction of the evidence, but it does not erase the selection and model-variation problems in the underlying studies.

Verdict: moderate, non-causal evidence for community-time and selected utilization outcomes; no basis for calling PACE a universally proven quality/safety intervention.

Sources: CMS/Abt PACE demonstration evaluation, pp. 26–30, 34–35, 44–45; Roberts et al., 2024 systematic review.

2. The cost-saving story fails the evidence floor

The most relevant federal comparative evaluation (Mathematica for ASPE) matched new PACE enrollees in eight states during 2006–2008 to HCBS-waiver enrollees and nursing-home entrants, then followed them through 2011. Medicare spending was mostly similar to predicted fee-for-service spending, while actual Medicaid capitation exceeded predicted Medicaid spending in every reported interval. Results varied materially by state, and mortality differences may reflect unobserved health and functional-status differences. The study does not support the simple claim that PACE saves Medicaid money.

ASPE's literature review reached the same guarded bottom line for the older evidence: no demonstrated savings for either program and higher overall cost through Medicaid expenditure, while hospital and nursing-home-use findings were mixed. These are not two independent causal estimates — the review partly synthesizes the same evidence family — so they should be treated as corroboration of the caution, not as a two-source proof of a universal cost effect.

Verdict: PACE has direct evidence against assigning a positive cost-containment score. Its cost effect is state- and comparison-dependent; the honest cell is weak/negative rather than “unmeasured.”

Sources: ASPE/Mathematica matched evaluation; ASPE literature review.

3. Current availability is countable; the cause of its growth rate is not

The previous scorecard carried “historically hard to scale” without a source. CMS's July 2026 plan directory supplies a current administrative baseline. Filtering its CSV to Organization Type = National PACE yields 206 active contracts. 201 report visible enrollment, totaling 79,758; five contracts have blank enrollment because CMS masks values of 10 or fewer. The reported total is therefore a lower bound rather than all participants. Contract mailing addresses fall in 34 states.

Those figures should not be read as center count, service-area coverage, or a national access rate: the file is a contract directory, and address state is not service-area state. Nor is a cross-section a growth evaluation. Fifty-two listed contracts became effective in 2024–2026 (eight in 2026), but this does not measure net growth, closures, center openings, or the reason for any of them. The operational requirements above make a workforce-and-center bottleneck plausible; they do not establish that those requirements caused any observed growth rate.

Extraction: download CMS's July 2026 ZIP; read MA_Contract_directory_2026_07.csv; retain rows whose Organization Type is National PACE; count rows and sum numeric Enrollment values; separately count blank enrollment values; do not infer centers or coverage from legal-entity addresses.

Verdict: current contract and enrollment availability is MEASURED; “historically hard to scale” remains NOT MEASURED as a causal claim. Retire the latter shorthand unless a future pass compares directory vintages and launches with local workforce, capital, and state-readiness conditions.

Source: CMS MA, Cost, PACE, and demonstration plan directory — July 2026.

Provisional scorecard treatment — not yet entered

The existing anchors require evidence, not mechanism, for every non-neutral cell. On this record alone, the defensible provisional row is:

• CC 2: the matched ASPE study reports consistently higher Medicaid capitation than predicted; that is adverse evidence, though too dated and heterogeneous for a stronger claim.
• IQ 4 / AA 4: the CMS evaluation gives positive, adjusted but non-randomized evidence on acute-care use, nursing-home days, community time, and short-run satisfaction.
• WS 3 / CR 3: no included source identifies a net direct-care-workforce effect or a directly measured family-caregiver effect. The required-team design is not evidence of either effect, so both remain neutral.

This row must not be added piecemeal. Entering it requires recomputing all three objective weightings, checking whether it changes a headline, and updating the whitepaper/site only after that reconciliation. In particular, PACE must not borrow Cash & Counseling's RCT evidence or the generic “aging-in-place” intuition to score caregiver relief or workforce sustainability.

Sequencing decision

The protocol's prior order — workforce pipeline first, HCBS de-capping second, caregiver payment last — remains directionally sound, but it needs a PACE qualification:

1. Build and retain the workforce before treating PACE as a capacity expansion. A PACE center requires a multi-profession team and direct-care capacity. New capitation or a new federal authorization does not supply those roles. This is a feasibility gate, not evidence that a wage floor alone solves the shortage.
2. Use PACE as a targeted delivery model after local readiness is proven. State-readiness review, a center, an eligible service area, and an interdisciplinary team are hard prerequisites. A launch grant or technical assistance package should be sequenced only after a locality can demonstrate those inputs and report them publicly.
3. Do not sell PACE as a cost-containment first move. The best available federal comparison finds higher Medicaid payments, while outcome evidence is most favorable for community time and coordinated care. Its policy case is targeted autonomy/quality, not a generic budget fix.
4. Keep HCBS de-capping separate. PACE is one integrated provider model, not a substitute for improving the waiver system available to people outside a PACE service area. The filing's workforce and waiver findings still govern the broader capacity question.
5. Gate any national expansion on a measurement pass. Before ranking PACE above alternatives, count current PACE organizations, service areas, and enrollment from the CMS directory; pair that with local vacancy and recruitment data and a transparent comparison population. Without that, national scale is a conjecture.

What would change this finding

• A current, well-controlled comparison of PACE with matched HCBS alternatives that separately reports Medicare and Medicaid spending, long-term institutional use, quality outcomes, and caregiver outcomes.
• A longitudinal CMS-directory analysis linking actual PACE launches and enrollment to local workforce, capital, and state-readiness conditions.
• A direct measure of caregiver burden or labor-force outcomes attributable to PACE, rather than a general claim about keeping someone in the community.

Confidence: moderate on program design and the direction of the cost-containment caution; low-to-moderate on comparative outcomes; low on scalability and caregiver/workforce effects. The most important conclusion is the last one: those two latter effects are not measured by the material located here.

Prospective M3 evidence check — 2026-08-06

The criteria above were written before this focused follow-up search. The follow-up did not find evidence that clears any of the four stated thresholds. That is a finding about the present evidence record, not evidence of no effect.

One further 2026 qualitative study of Korean-American community readiness is useful for identifying awareness and transition barriers, but it is not a capacity panel or a comparative caregiver/workforce evaluation. It therefore does not alter a scorecard cell. The original provisional row remains unentered: the criteria correctly preserve neutral cells where the evidence is not yet capable of supporting a non-neutral result.