Cut through the noise, find the signal: Intelligence for serious healthcare leaders
Analysis updated November 9, 2026
BPCI Advanced's eight-year arc reveals a CMS that learned—sometimes painfully—how to close the loopholes that voluntary episode models inevitably generate. The model's evolution falls into three distinct strategic phases, each with explicit design lessons now baked into successor models.
Phase 1 (MY1–3, 2018–2020): Generous target pricing creates adverse selection. In the first three model years, CMS set target prices that systematically underestimated how much participants would reduce spending. The result: net losses to Medicare of $179.4 million ($65.7M in MY1–2 plus $113.7M in MY3), despite genuine episode expenditure reductions. Participants—particularly those with histories of high PAC utilization—gamed episode selection, concentrating in the most profitable clinical episodes. The median participant earned nearly $200,000 in reconciliation payments, with the top quartile clearing $850,000 or more. CMS was, in effect, paying above-market rates for efficiency gains that would likely have occurred anyway.
Phase 2 (MY4–6, 2021–2023): The methodology correction triggers a participation collapse. CMS introduced retrospective peer group trend adjustments in MY4, standardized episode overlap rules, and required participants to accept CESLGs (service line groupings) rather than cherry-picking individual clinical episodes. These weren't incremental tweaks—they fundamentally altered the economics. The MY4 design changes are the single most consequential policy event in the model's history. The median reconciliation amount swung from a gain of ~$200,000 to a net repayment of ~$4,000 owed to CMS. Participation fell sharply from its MY3 peak. The 31.8% of participants who hit the stop-gain limit at least once and the 25.2% who hit the stop-loss limit tell the story of a model with high variance—but the stop-loss cap at ±20% of aggregate target price provided meaningful downside protection.
Phase 3 (MY7–8, 2024–2025): Stabilization with self-selection bias. The two-year extension (announced October 13, 2022) opened a third application cohort (January 1, 2024) and incorporated pricing concessions—most notably reducing the CMS discount for medical episodes from 3% to 2%, and capping the Peer Group Trend Factor Adjustment at 5% (down from 10%). These changes were explicit retention incentives, not quality improvements. The MY7–8 participant pool skewed toward for-profit hospitals, safety-net hospitals, and facilities with lower HCAHPS scores, signaling that late-stage voluntary models attract organizations either in financial distress or with specific PAC arbitrage opportunities remaining.
The TEAM precedent. BPCI Advanced's voluntary design produced generalizable direction of effect but not generalizable magnitude. CMS explicitly cited this limitation in designing TEAM (mandatory, randomized, five surgical episodes, effective January 2026). BPCI Advanced is, in retrospect, the policy R&D platform. TEAM is the scaled deployment.
The $800 million headline requires disaggregation. Total net Medicare savings of approximately $800 million across 8 years and ~2.5 million episodes averages to roughly $320 per episode—a thin margin that was entirely dependent on MY4–6 performance ($915.8M in net savings) offsetting early-year losses. The true operational lesson: BPCI Advanced was financially viable only after CMS corrected its pricing methodology, and even then, primarily through PAC reduction rather than acute care efficiency.
Episode-level economics by type:
Break-even calculus: Participants needed to reduce spending by at least 3% below benchmark (2% for medical episodes in MY6+) before earning any reconciliation payment. Approximately 15% of participants fell below their benchmark price but above the discounted target—they reduced costs meaningfully yet still owed CMS money. For a hospital with a $12,000 average episode target price doing 150 episodes semi-annually, the stop-loss cap of ±20% limits maximum exposure to $360,000 per reconciliation period—a figure that fits within most health system risk tolerance, but not within that of a standalone community hospital or independent PGP.
Cash flow structure is operationally punishing. Under BPCI Advanced, providers continued billing FFS normally. Reconciliation occurred semi-annually with two true-ups for claims run-out. This means organizations could operate for 6–12 months without knowing their net financial position. CFOs needed working capital reserves to absorb potential repayments—a structural disadvantage for smaller organizations that made convener relationships financially rational despite margin sharing.
Who won: Large PGPs with orthopedic or cardiac procedural volume, operating under a convener structure (80.9% of hospitals and 76.7% of PGPs participated via convener). These entities had high baseline PAC utilization, established preferred PAC networks, and could absorb reconciliation volatility. Safety-net hospitals received higher target prices due to patient case mix and peer group adjustments, creating a genuine upside opportunity in MY1–3 that eroded post-MY4.
Who lost: Hospitals or PGPs that had already achieved PAC efficiency through prior BPCI Initiative or CJR participation. Their low historical spending produced low target prices, eliminating reconciliation upside while preserving downside risk. The "fighting against yourself" problem—where efficiency in one year depresses the subsequent year's target price baseline—is structurally inherent to any rolling baseline model. This is BPCI Advanced's most important design flaw, and one that TEAM attempts to address through mandatory participation that eliminates selection exit.
MSSP/ACO REACH interaction: Since January 2020, BPCI Advanced no longer excluded MSSP-assigned beneficiaries, creating dual accountability. However, beneficiaries aligned to ACO REACH cannot trigger BPCI Advanced clinical episodes—a policy that reduced surgical episode volume for participants, particularly in MY7–8, as ACO REACH enrollment grew. This interaction meaningfully compressed available episode populations for late-stage participants who had also lost healthier surgical patients to ambulatory surgical centers (ASCs).
MA plan positioning: Medicare Advantage beneficiaries are explicitly excluded from BPCI Advanced clinical episodes. MA plans observed the model as a proof-of-concept for PAC reduction strategies—particularly the 5 percentage point reduction in PAC facility discharge rates for surgical patients (MY4–6) and the ~3 fewer SNF days for patients with SNF stays. These findings are now standard inputs for MA plan PAC network design and prior authorization criteria, regardless of whether the plan itself participated in any bundled payment model.
Physicians: BPCI Advanced's Advanced APM status was a material recruitment tool. Participating clinicians who qualified as Qualifying APM Participants (QPs) escaped MIPS reporting requirements and earned incentive payments. For high-volume orthopedic or cardiac surgeons, QP status alone justified participation regardless of reconciliation outcomes. The attribution precedence rules (attending physician NPI → operating physician NPI → ACH) created real tension: PGPs took first priority for episode attribution, meaning hospitals could lose credit for episodes occurring within their walls if a participating PGP's physician was the attending. This drove strategic negotiations between hospitals and physician groups about participation structure.
Health systems: Systems with owned IRF capacity faced an explicit conflict: BPCI Advanced incentivized reducing IRF admissions, but IRF daily census drove a separate revenue stream. The evaluation confirmed that IRF-owning participants had measurably worse PAC reduction outcomes. Systems need to model the net IRF revenue impact before entering any successor episode model—a calculation that BPCI Advanced made visible for the first time at scale.
Hospitals: Community hospitals without convener support were structurally disadvantaged. The data infrastructure required to manage 90-day episode spending in real time—claim-level attribution, PAC utilization tracking, preferred network management—requires technology investment that most independent hospitals cannot self-fund. The convener model partially addresses this, but conveners typically take 10–30% of reconciliation upside as their fee, compressing hospital economics further.
Payers and MA plans: BPCI Advanced provided seven years of evidence that the "why not home?" discharge protocol—combining risk stratification tools, caregiver support screening, and preferred PAC networks—reduces institutional PAC utilization by 5 percentage points for surgical patients without increasing readmission rates. MA plans should treat the BPCI Advanced final evaluation as the most rigorous available evidence base for SNF authorization policy. The ~$500 per-episode SNF reduction and 3 fewer SNF days per stay are directly translatable to MA plan utilization management benchmarks.
Bottom Line: BPCI Advanced proved that voluntary episode-based payment can generate Medicare savings—but only after CMS corrected for adverse selection and pricing generosity that characterized the first three years. The model's $800M in net savings is real but fragile: concentrated in MY4–6, dependent on a narrow set of high-performing participants, and undermined by the structural penalization of early efficiency gains. Organizations evaluating TEAM or future mandatory episode models should treat BPCI Advanced's MY4–6 operational playbook—PAC network development, "why not home?" protocols, care navigation during the 90-day window—as the minimum viable capability set. The participants that found BPCI Advanced financially untenable in MY7–8 are not well-positioned for mandatory successor models.