Bundled Payments for Care Improvement Advanced (BPCI Advanced)

ValuesEdge Institutional Analysis | Strategic Intelligence Report

---

1. Executive Summary and Strategic Context

What Is This Model and What Problem Does It Solve?

BPCI Advanced was a voluntary Advanced Alternative Payment Model (Advanced APM) operated by the CMS Center for Medicare and Medicaid Innovation (CMMI) from October 1, 2018 through December 31, 2025 — spanning 8 model years across three participant cohorts. The model tested whether retrospective episode-based payment accountability could reduce Medicare expenditures while maintaining or improving quality of care for fee-for-service Medicare beneficiaries.

The structural problem BPCI Advanced addressed is well-documented: Medicare beneficiaries admitted to the hospital or undergoing outpatient procedures frequently encounter fragmented, uncoordinated care — physicians unfamiliar with the patient's full history, primary care physicians not notified of the hospitalization, and post-acute care (PAC) providers operating in a separate financial and clinical silo. The result is unnecessary institutional PAC utilization, preventable readmissions, and high per-episode spending that the standard fee-for-service payment system has no mechanism to constrain.

BPCI Advanced attacked this problem by designating a single accountable Episode Initiator (EI) — either an Acute Care Hospital (ACH) or a Physician Group Practice (PGP) — responsible for the total cost and quality of care during a 90-day Clinical Episode beginning at hospital discharge or outpatient procedure completion. Providers continued billing under standard Medicare FFS; accountability was imposed retrospectively through a semi-annual reconciliation process comparing actual Medicare expenditures against a pre-established Target Price. Participants that came in under the target price received a Net Payment Reconciliation Amount (NPRA) from CMS; those that exceeded it owed a Repayment Amount to CMS. Critically, there was no safe harbor: downside financial risk was required from the first day of participation, distinguishing BPCI Advanced from one-sided risk models.

The model covered 34 Clinical Episodes organized into 8 Clinical Episode Service Line Groups (CESLGs), encompassing both common acute medical conditions (sepsis, congestive heart failure, COPD, stroke) and high-volume surgical procedures (major joint replacement of the lower extremity, spinal fusion, coronary artery bypass graft). This breadth was deliberate: CMS sought to test episode accountability across the full spectrum of inpatient and outpatient specialty care, not just elective orthopedics.

Strategic Significance for the VBC Market

The final independent evaluation — conducted by The Lewin Group with partners Abt Global Inc., GDIT, and Telligen, published August 2026 — establishes BPCI Advanced as the most consequential voluntary bundled payment experiment in Medicare's history. The headline result: approximately $800 million in net Medicare program savings across all 8 model years. Savings were concentrated in Model Years 4 through 6 (2021–2023), which generated an estimated $915.8 million in net savings (90% confidence interval: $745.1 million to $1,086.5 million). Model Years 7 and 8 added an estimated $85.8 million in net savings (90% CI: $61.0 million to $110.6 million), though the evaluation team flags these estimates as likely overestimated by approximately 13.6% due to methodological differences in the final-period analysis.

The mechanism driving savings was primarily reductions in PAC utilization, not participant repayments. In Model Years 4 through 6, $886.9 million of the $915.8 million in total savings derived from episode expenditure reductions, with only $28.9 million attributable to net repayments from participants to CMS. This finding has direct implications for health system strategists: the model's value-generation engine was care redesign — specifically, the "why not home?" discharge philosophy, preferred PAC provider networks, and care navigator deployment — not financial clawbacks.

At its peak in Model Year 3 (2020), the model engaged more than 1,100 hospitals and more than 800 PGPs, reaching nearly 1 in 3 eligible hospitals and 1 in 4 eligible clinicians. Across all 8 model years, BPCI Advanced touched approximately 2.5 million hospital stays and procedures for 1.8 million patients. More than half of BPCI Advanced episodes were for patients not enrolled in a Medicare ACO, extending value-based care accountability to a population that would otherwise be entirely in unaccountable FFS. By December 31, 2021 — before the model's extension — over 1.2 million Medicare beneficiaries had received care from BPCI Advanced participants, and over 1,800 ACHs had engaged in care redesign activities alongside 69,867 physicians.

Surgical episodes outperformed medical episodes on every savings metric. In Model Years 4 through 6, BPCI Advanced reduced episode payments for surgical episodes by $1,835 per episode (90% CI: –$2,199, –$1,471), a 5.9% reduction from baseline. For medical episodes, the reduction was $756 per episode (90% CI: –$923, –$588), a 2.9% reduction. Participants consistently explained that elective surgical procedures allowed pre-operative optimization, patient and family expectation-setting, and more aggressive discharge-to-home planning — none of which is possible for emergent medical admissions.

Political Context and Administrative Alignment

BPCI Advanced was launched under the first Trump administration and extended during the Biden administration. The model concluded its performance period on December 31, 2025 — under the current Trump administration, with Dr. Mehmet Oz as CMS Administrator and Abe Sutton as CMMI Director. The two-year extension (announced October 13, 2022) and the third cohort launch on January 1, 2024 were executed under the Biden-era CMS leadership, but the final two model years (Model Years 7 and 8) ran entirely under the current administration.

The model's conclusion is directly linked to the launch of its mandatory successor: TEAM (Transforming Episode Accountability Model), which began January 1, 2026. The final evaluation explicitly frames BPCI Advanced and the Comprehensive Care for Joint Replacement (CJR) Model as the two foundational evidence sources that informed TEAM's design. TEAM is mandatory and geographically randomized — a direct response to the evaluation's finding that voluntary model attrition undermined the generalizability and long-term sustainability of BPCI Advanced's results. For health system executives, this trajectory is unambiguous: the current CMMI leadership under Sutton is building on BPCI Advanced's demonstrated $800 million savings track record to justify mandatory episode accountability for high-volume surgical procedures.

Market Transformation Implications

BPCI Advanced demonstrated that episode-based accountability can generate material Medicare savings at scale without measurable quality deterioration on claims-based measures — readmission and mortality rates were statistically indistinguishable between BPCI Advanced participants and comparison groups. This evidence base is the single most important input into any assessment of where CMMI's specialty care strategy is heading.

The model also revealed structural constraints in voluntary program design that are instructive for current market positioning. Participant attrition accelerated after Model Year 4 — the same year CMS introduced retrospective trend adjustments that significantly tightened target prices. Experienced VBC participants — those who had already wrung efficiency gains from prior participation in the BPCI Initiative, CJR, or ACO programs — found themselves unable to reduce spending further below a ratcheting target price. The "you're fighting against yourself" dynamic described by BPCI Advanced PGPs during site visits is a key strategic data point for ACO leaders and health system CFOs evaluating participation in TEAM or future mandatory models.

The model's reach into safety-net hospitals deserves explicit attention. The pricing methodology was designed to produce higher target prices for hospitals with higher historical spending — which systematically included safety-net institutions. This design choice made BPCI Advanced one of the few CMMI models to successfully engage safety-net hospitals in two-sided risk arrangements, with safety-net hospitals receiving larger reconciliation payments than non-safety-net hospitals in Model Years 1 through 3. However, safety-net hospitals also became less financially successful after Model Year 4 design changes, a tension that CMMI and the current administration will need to address in future mandatory episode model design.

---

2. Model Architecture and Payment Structure

Participation Structure

BPCI Advanced used a two-tier participation structure distinguishing between entities that bore risk for themselves versus those that aggregated and apportioned risk across multiple downstream providers.

Convener Participants were entities that brought together at least one Downstream Episode Initiator (Downstream EI) — which was required to be either an ACH or a PGP — to participate in the model. Conveners bore and apportioned financial risk on behalf of their downstream EIs, entered into agreements with those EIs requiring compliance with all applicable model requirements, and often provided supporting services such as data analytics, case management, and performance monitoring. For the Model Year 7 (2024) application cycle, Convener applicants were required to be Medicare-enrolled providers or suppliers, or Medicare Accountable Care Organizations (ACOs). Existing Convener Participants active during Model Year 6 could sign an Amended and Restated Participation Agreement regardless of whether they met the new eligibility criteria.

Non-Convener Participants were Episode Initiators that bore financial risk solely for themselves with no downstream EIs. Only ACHs and PGPs could participate as Non-Convener Participants. Non-Convener applicants for Model Year 7 were likewise required to be either an ACH or a PGP.

Across all 8 model years, the evaluation found that 80.9% of BPCI Advanced hospitals and 76.7% of PGPs participated under a convener — reflecting the substantial operational and analytical support conveners provided and the complexity of managing two-sided episode risk independently.

Beneficiary eligibility required enrollment in Medicare Parts A and B for the full duration of the 90-day Clinical Episode. Specifically excluded from BPCI Advanced beneficiary status were: Medicare Advantage enrollees, Health Care Prepayment Plan enrollees, cost-based HMO enrollees; beneficiaries with End-Stage Renal Disease (ESRD); beneficiaries for whom Medicare is not the primary payer at any point during the episode; beneficiaries covered under United Mine Workers plans; and beneficiaries who died during the Anchor Stay or Anchor Procedure.

Cohort and Application Timeline

Milestone	Date
Model announced	January 8, 2018
First Cohort application open	January 11, 2018
First Cohort application deadline	March 12, 2018
First Cohort (Model Year 1) begins	October 1, 2018
Model Year 2 begins	January 1, 2019
Second Cohort application open	April 24, 2019
Second Cohort application deadline	June 24, 2019
Second Cohort (Model Year 3) begins	January 1, 2020
Model Year 4 begins	January 1, 2021
Model Year 5 begins	January 1, 2022
Two-year extension announced	October 13/14, 2022
Model Year 6 begins	January 1, 2023
Third Cohort application open	February 21, 2023
Third Cohort application portal closes	May 31, 2023
Third Cohort (Model Year 7) begins	January 1, 2024
Model Year 8 begins	January 1, 2025
Model period of performance ends	December 31, 2025

Clinical Episode Architecture

Clinical Episodes were initiated either by an Anchor Stay (inpatient admission, triggered by Medicare Severity-Diagnosis Related Group [MS-DRG] codes) or an Anchor Procedure (outpatient procedure in a Hospital Outpatient Department, triggered by Healthcare Common Procedure Coding System [HCPCS] codes). The Clinical Episode window was the Anchor Stay or Anchor Procedure plus 90 days, with the day of discharge or procedure completion counted as Day 1 of the 90-day period.

Starting with Model Year 6 (2023), the model included 8 Clinical Episode Service Line Groups (CESLGs) with 29 Inpatient, 3 Outpatient, and 2 Multi-Setting Clinical Episode Categories — totaling 34 Clinical Episodes. The addition of outpatient Total Shoulder Arthroplasty (triggered by HCPCS 23472) to the Major Joint Replacement of the Upper Extremity (MJRUE) category created the second multi-setting episode.

CESLG	Type	Selected Clinical Episodes
Cardiac Care	Medical	AMI, Cardiac Arrhythmia, Congestive Heart Failure
Cardiac Procedures	Surgical	CABG, PCI (Inpatient/Outpatient), Cardiac Defibrillator, Endovascular Cardiac Valve, Pacemaker, Cardiac Valve
Gastrointestinal Care	Medical	GI Hemorrhage, GI Obstruction, Inflammatory Bowel Disease, Liver Disorders
Gastrointestinal Surgery	Surgical	Bariatric Surgery, Major Bowel Procedure
Neurological Care	Medical	Seizures, Stroke
Medical and Critical Care	Medical	Cellulitis, COPD/Bronchitis/Asthma, Renal Failure, Sepsis, Simple Pneumonia, UTI
Spinal Procedures	Surgical	Back/Neck Except Spinal Fusion (Inpatient/Outpatient), Spinal Fusion
Orthopedics	Surgical/Multi-Setting	Double Joint Replacement (Lower Extremity), Fractures of Femur/Hip/Pelvis, Hip and Femur Procedures, MJRLE (Multi-setting), MJRUE (Multi-setting), Lower Extremity/Humerus Procedure

Clinical Episode attribution followed a strict precedence hierarchy at the Episode Initiator level:

1. PGP associated with the attending physician's NPI
2. PGP associated with the operating physician's NPI
3. ACH where the Anchor Stay or Anchor Procedure occurred
4. If none of the above apply: no BPCI Advanced Clinical Episode is triggered

The bundle included all Medicare FFS items and services during the episode window — inpatient and outpatient hospital services, physician services, clinical laboratory, durable medical equipment, skilled nursing facility (SNF), inpatient rehabilitation facility (IRF), long-term care hospital (LTCH), home health, hospice, and Part B drugs — with specified exclusions including organ transplant, trauma, and cancer-related readmissions; contralateral procedures with the same MS-DRG; excluded cardiac rehabilitation codes under 42 C.F.R. §410.49; new technology add-on payments under 42 C.F.R. §412.87 and §412.88; and transitional pass-through payments under 42 C.F.R. §419.62 and §419.66.

Target Price Methodology

Target Price = Benchmark Price × (1 – CMS Discount)

The Benchmark Price for ACHs was calculated using regression models that account for hospital-specific historical spending patterns relative to a peer group defined by census division, bed size, major teaching hospital status, safety-net status, and urban/rural classification. PGPs received hospital-specific Target Prices derived from the hospital Benchmark Price, with the hospital-wide Patient Case Mix Adjustment (PCMA) replaced by the PGP's own PCMA specific to episodes initiated at that hospital.

The full Preliminary Target Price formula was:

Preliminary TP = Standardized Baseline Spending (SBS) × Patient Case Mix Adjustment (PCMA) × Peer Group Historical Adjustment (PGHA) × Peer Group Trend (PGT) × CMS Discount

Final Target Prices added the PGT Factor Adjustment and reflected updated patient case mix and realized (retrospective) trends during the performance period — a critical change introduced in Model Year 4 that significantly tightened target prices relative to earlier model years.

Parameter	Model Years 1–5	Model Year 6 (2023) Onward
CMS Discount — Medical Episodes	3%	2%
CMS Discount — Surgical Episodes	3%	3% (unchanged)
PGT Factor Adjustment Cap	10%	5%

Baseline periods shifted forward each model year:

Model Year(s)	Baseline Period
MY1 and MY2	January 1, 2013 – December 31, 2016
MY3	October 1, 2014 – September 30, 2018
MY4	October 1, 2015 – September 30, 2019
MY5	October 1, 2016 – September 30, 2020
MY6	October 1, 2017 – September 30, 2021

This forward-shifting baseline is the mechanism by which participants who had already reduced spending faced declining target prices in subsequent years — the structural pressure that drove attrition and that participants described as being "penalized for success."

Reconciliation, Risk Limits, and Adjustments

Reconciliation occurred semi-annually (January–June and July–December performance periods), with two True-Up cycles per period to allow for claims run-out. Clinical Episodes were reconciled in the performance period in which they ended, not in which the anchor event occurred.

Stop-loss / stop-gain limits: Reconciliation payments in either direction were capped at ±20% of the volume-weighted sum of final Target Prices across all Clinical Episodes, aggregated to the EI level within the performance period. Over the full model, 31.8% of participants hit the stop-gain limit and 25.2% hit the stop-loss limit at least once.

Risk Cap: Applied at the 1st and 99th percentile of episode spending in both the performance and baseline periods.

Post-Episode Spending Monitoring Period: CMS monitored 30 days following the 90-day Clinical Episode. Any Medicare FFS expenditures during this 30-day window that exceeded an empirically titrated risk threshold were owed to Medicare by the Participant. This monitoring was performed semi-annually in the first True-Up Reconciliation.

COVID-19 Policy Shift in Model Year 6: From Model Years 3 through 5, beneficiaries with a COVID-19 diagnosis during the episode were excluded from reconciliation. Beginning Model Year 6 (2023), participants were held accountable for all Clinical Episodes regardless of COVID-19 diagnosis — a policy change that increased episode volume and was made in recognition of the availability of COVID-19 vaccines and treatments.

Quality Measures and Composite Quality Score

BPCI Advanced qualified as an Advanced APM under the Quality Payment Program (QPP) based on three criteria: (1) bearing of financial risk; (2) use of Certified Electronic Health Record Technology; and (3) tying payment to performance on quality measures.

Participants selected between two quality measurement frameworks:

• Administrative Quality Measures Set (available from Model Year 1): Seven exclusively claims-based measures collected directly by CMS.
• Alternate Quality Measures Set (available from Model Year 4): Up to five claims-based and registry-based measures per Clinical Episode, developed with specialty-specific clinical stakeholders and established registries.

Quality performance was aggregated into a Composite Quality Score (CQS) — a percentile ranking against peers across multiple measures — which adjusted the final reconciliation payment by up to 10%. The average CQS across the model was approximately 55 (just above peer median). To illustrate the financial magnitude: a participant at the 5th percentile of quality scores that owed CMS $500 per episode would see repayments reduced by approximately $13 per episode; one at the 95th percentile would see a reduction of approximately $41 per episode. Quality performance affected the direction of financial liability (owed versus received) but was not the dominant driver of reconciliation outcomes.

Model overlap rules governed situations where beneficiaries were subject to multiple CMMI models simultaneously:

• CJR episodes take precedence over BPCI Advanced and are excluded from BPCI Advanced reconciliation.
• MSSP ACO beneficiaries are included in BPCI Advanced (since January 2020).
• ACO REACH-aligned beneficiaries cannot trigger BPCI Advanced episodes, though BPCI Advanced participants may also participate in ACO REACH for non-REACH-aligned beneficiaries.
• Vermont Medicare ACO Initiative and CKCC options of the Kidney Care Choices (KCC) Model are not permitted to overlap with BPCI Advanced Clinical Episodes.