Methodology

• MSSP PY2024 Financial & Quality Results PUF — 476 Medicare Shared Savings Program ACOs (file revised Jul 2026).
• ACO REACH PY4 (2024) Core PUF — 115 ACO REACH entities (PY2024 was the program's final year).
• Service area / operator / contacts from the CMS SSP Organizations directory. Medicare fee-for-service only — Medicare Advantage and Medicaid are not in these files.

591 ACOs, 12.8M beneficiaries, $168B in addressable Medicare spend. PY2024 is the latest CMS-reconciled vintage; PY2025 publishes ~fall 2026 and loads automatically.

Savings come from reducing avoidable utilization — not a flat percentage of spend. For each ACO, using its own CMS utilization and unit costs:

acute_admits   = admits_per_1000/1000 × person_years
readmissions   = readmit_rate × acute_admits        (carved out of admits)
index_admits   = acute_admits − readmissions        (never double-counted)

savings_admits    = index_admits  × pct_admits_reduced   × avg_cost_per_admit
savings_readmits  = readmissions  × pct_readmits_reduced × avg_cost_per_admit
savings_ed        = ed_visits     × pct_ed_reduced       × ed_cost_per_visit
savings_snf       = snf_admits    × pct_snf_reduced      × snf_cost_per_stay

total = savings_admits + savings_readmits + savings_ed + savings_snf

Default levers (all adjustable on every ACO page):

Unit costs (per-admit, per-SNF-stay) are derived from each ACO's own CMS spend, so the model is grounded in that ACO's real cost structure. Readmissions are carved out of acute admits before the levers apply, so the two are never double-counted.

Each ACO page shows two tabs, because operators see the economics differently. MSSP is fee-for-service shared savings — the ACO keeps a share (40% BASIC → 75% ENHANCED) plus separately billable care-management codes (CCM/RPM/RTM). Medicare Advantage is capitation — the risk-bearing entity keeps ~all avoided cost, and revenue rises via risk-adjustment (RAF) and Star quality bonuses rather than FFS codes. MA has no CMS public data, so the MA view uses the ACO's FFS lives as a membership proxy plus editable model inputs (PMPM, RAF, Star bonus).

The national band computes the avoidable-utilization saving per beneficiary live across all 591 ACOs at the chosen levers (total modeled savings ÷ total person-years), then scales it:

per_life          = Σ ACO savings(levers) / Σ person_years
program_savings   = beneficiaries × reach × per_life × applicability
national_savings  = Σ enabled programs

Base population and spend figures (Medicare ~68M / ~$1.0T; Medicaid ~90M / ~$0.8T) are illustrative, editable estimates — replace them with official CMS figures. The Medicaid applicability factor (default 0.6) scales the Medicare-derived per-life effect to a different population. At 30% reach and default levers this yields ~$26B/year (~$261B over 10 years, undiscounted). It is a transparent sizing exercise, not a scored budget estimate.

• Real CMS data: lives, spend, benchmark, generated & earned shared savings, utilization rates (admits/ED/SNF/readmits per 1,000), unit costs, quality score.
• XSPAN estimates (adjustable): the reduction levers, care-management code revenue (CY2024 national-average rates), the entire MA view, and the national extrapolation.
• Star ratings are an XSPAN convention from the quality score — CMS does not star-rate ACOs; 35 ACOs carry a CMS default score and are shown "Met standard — unrated."

This is a deterministic planning tool on descriptive data — not causal inference. It does not adjust for confounding, selection, or regression to the mean, and effect sizes are the modeled levers rather than measured outcomes. Establishing the real effect requires a matched, risk-adjusted, intention-to-treat evaluation with a concurrent comparison group (difference-in-differences against secular trend). Medicaid and MA figures are modeled, not CMS-sourced.