Executive Summary

• The Center for Medicare and Medicaid Innovation (CMMI) was created by the Affordable Care Act and is housed in the Centers for Medicare and Medicaid Services; its mandate is to test payment and service delivery models that generate savings and maintain or increase quality for Medicare and Medicaid enrollees.
• Since its inception, however, CMMI has faced a range of criticism from lawmakers and industry representatives, including legislative circumvention, inefficient model design, and poor fiscal outcomes for Medicare.
• CMMI’s actions in 2025 and forthcoming priorities – including models focused on GLP-1 usage, Most-Favored-Nation drug pricing, and a range of Make America Healthy Again initiatives – further demonstrate the importance of understanding the agency’s mission and how its models work.

Introduction

The Center for Medicare and Medicaid Innovation (CMMI) – the Centers for Medicare and Medicaid Services (CMS) Innovation Center – was established by the Affordable Care Act (ACA) to test “payment and service delivery models” for Medicare and Medicaid beneficiaries with a specific objective: reduce program expenditures while preserving or enhancing quality, and prioritizing models that improve care coordination, quality, and efficiency. There have been over 100 models announced by CMMI since its establishment; administrations of both parties have attempted to use its waiver and demonstration powers to accomplish payment goals in Medicare and Medicaid. In 2025, the Trump Administration began its latest efforts to leverage CMMI to accomplish its goals, announcing models focusing on Make America Healthy Again, GLP-1 usage, and Most-Favored-Nation pricing.

CMMI has usually operated in relative obscurity, given the restrictive application of the models to participants (either through a voluntary status or mandatory rulemaking) but has earned disdain from lawmakers and industry alike. Since its establishment, there have been consistent questions and criticisms about CMMI because it sits at an unusual intersection of technical payment policy and unaccountable administrative power. On paper, it is a test-and-learn “innovation lab.” In practice, demonstrations can reshape how Medicare and Medicaid pay for care – and, under certain statutory conditions, CMS can expand a model beyond a pilot through rulemaking. Because of this divisiveness, it is noteworthy that the Trump Administration is using CMMI authority to effectuate sweeping changes in the way pharmaceuticals are reimbursed in Medicare and Medicaid.

What Is CMMI?

Structurally, CMMI is a component of CMS and functions as a portfolio manager for a set of time-limited model tests. CMS describes CMMI’s work as developing and testing payment and service delivery models to improve care and lower costs, supported by ongoing monitoring, independent evaluation, and public-facing transparency efforts (including sharing model data and evaluation reports). The “models” themselves can be organized around a condition (e.g., end-stage renal disease), an episode of care (e.g., joint replacement), a provider type (e.g., primary care), a community (e.g., rural), or innovations in Medicare Advantage/Part D – reflecting how CMMI builds its portfolio around discrete policy hypotheses rather than sweeping program rewrites.

CMMI is best understood as CMS’ “applied policy laboratory.” Its job is not to administer benefits day to day – that is the core mission of its sister centers. Instead, CMMI runs time-limited experiments that try to answer a narrow but consequential question: If we change incentives and care delivery rules in a defined way, can we improve quality and reduce spending compared to the status quo? This matters because the “status quo” in Medicare of fee-for-service payment can reward fragmentation: more tests, more visits, more procedures – without necessarily producing better outcomes. Medicaid, while structured differently across states, faces its own pressures: tight state budgets, variation in provider participation, and the complexity of serving high-need populations.

“Models” Are Really Payment Rules

CMMI models – while billed as pilots or demonstrations – are fundamentally new payment rules in support of improved quality, better outcomes, innovative care-delivery mechanisms, and more, for a specific population and set of participants over a specific period. The thesis behind the model enterprise is straightforward. Medicare and Medicaid are not just entitlements for enrollees; they are the largest purchasers of health care in the United States, and the way they pay for care shapes how the delivery system organizes itself. If payment primarily rewards volume, then even well-intentioned providers have limited room to invest in the less-visible work that prevents complications and can generate future health value: care coordination, medication management, post-discharge follow-up, and addressing high-risk patients before they deteriorate.

CMMI models exist to test whether changing that incentive structure can reduce total spending and improve outcomes. Sometimes that mechanism is an episode-based bundle, where providers are responsible for the total cost of care for a defined episode, such as a surgery plus the post-acute period. Other times it is a population-based arrangement, where an organization is rewarded for keeping a group of patients healthier and out of the hospital. In still other cases – particularly for drug-focused models – CMMI can test alternative rebate and pricing constructs that operate upstream of the bedside but still flow through to premiums and beneficiary cost-sharing.

Because these are live payment tests, the details matter. Models specify who is eligible to participate (or who must participate), what beneficiaries are included, what services are in scope, and what quality measures serve as guardrails. They also define the accounting rules – benchmarks, risk adjustment, attribution, and reconciliation – that determine whether the model changes behavior and whether observed savings are real.

How Models Affect Medicare and Medicaid in Practice

Models influence Medicare and Medicaid through three pathways: direct effects on participants, broader diffusion of lessons, and – most important – the possibility of expansion.

Direct effects are the most intuitive. If a hospital participates in a bundled payment model for joint replacements, it may redesign discharge planning, tighten relationships with post-acute providers, and focus more heavily on reducing complications. If a physician group participates in an accountable care model, it might invest in care managers, build analytics capacity, and standardize referral pathways. Beneficiaries may not “feel” the model as a new benefit, but they can experience different care patterns – more follow-up, more coordination, fewer avoidable emergency department visits.

Diffusion effects happen even when models do not expand. Participation often forces organizations to build capabilities – data systems, contracting expertise, care management – that persist after the model ends. In that way, CMMI can shift the market indirectly. A model becomes a forcing function for delivery-system modernization, and private payers sometimes borrow the same constructs (or align their contracts with them) to reduce provider burden and scale the approach.

Expansion effects are where CMMI becomes uniquely consequential. Under its statutory authority, CMMI is not limited to testing and publishing results. If a model is determined to reduce spending without reducing quality (or improve quality without increasing spending), CMS has authority to expand it – potentially broadly – through administrative action. That possibility should change how stakeholders view models. A pilot is not just a temporary demonstration; it can be a preview of the next era of Medicare payment policy.

Recent Model Announcements

As evidenced by the rush of CMMI model announcements at the end of 2025, it appears that the Trump Administration is relying on the Innovation Center’s amorphous accountability structure to effectuate significant policy change. Below are brief summaries of these recently announced models.

GUARD (Guarding U.S. Medicare Against Rising Drug Costs) Model
GUARD is a proposed mandatory Innovation Center model focused on Medicare Part D. In effect, it would “layer” an international price-benchmark concept onto the existing Part D Inflation Rebate Program by testing an alternative rebate calculation that uses international pricing information (rather than the current domestic benchmark) to determine when manufacturers owe additional rebates. CMS declares that the aim is to lower net Part D costs (and, by extension, premiums and beneficiary affordability pressures) while maintaining or improving quality. Operationally, manufacturers would be the model “participants,” and the test would run for a five-year performance period beginning January 1, 2027, with the model implemented in randomly selected geographic areas representing 25 percent of Part D enrollees; rebate invoicing and reconciliation would continue beyond the performance period.

GLOBE (Global Benchmark for Efficient Drug Pricing) Model
GLOBE is the companion proposed mandatory model for Medicare Part B drugs – CMS highlights categories such as oncology and immunology/rheumatology, among others. Like GUARD, it would test an alternative calculation under the Part B Inflation Rebate Program, using an international pricing benchmark to determine manufacturer rebates; CMS also indicates beneficiary out-of-pocket costs would be tied to the international benchmark, thereby lowering cost-sharing for included drugs. The model is designed to apply to a subset of Part B drugs, including single-source drugs/biologics and drugs with significant Part B spending. GLOBE would launch October 1, 2026, and run for a five-year period, implemented in selected geographic areas representing 25 percent of Medicare Part B beneficiaries, with reconciliation continuing after the performance period.

BALANCE (Better Approaches to Lifestyle and Nutrition for Comprehensive hEalth) Model
BALANCE is a voluntary model that blends drug-price negotiation with a coverage-and-care strategy around obesity/metabolic health. CMS’s core mechanism is direct negotiation: The agency would negotiate drug pricing and coverage terms with GLP-1 manufacturers on behalf of state Medicaid agencies and Medicare Part D plan sponsors, with the intent of expanding access to select GLP-1s for weight management while pairing drug use with structured lifestyle supports. Participation is voluntary for manufacturers, Medicaid agencies, and Part D plans. CMS indicates Medicaid participation can begin May 2026 and Part D participation January 2027, with testing concluding December 2031. CMS also describes a separate short-term Medicare demonstration beginning July 2026 – intended as a “bridge” to BALANCE – in which eligible Part D beneficiaries would pay $50 per month for GLP-1 medications furnished under that demo.

MAHA ELEVATE (Make America Healthy Again: Enhancing Lifestyle and Evaluating Value-based Approaches Through Evidence) Model
MAHA ELEVATE is structurally different from the drug-pricing and payment-rule models above: It is a voluntary Innovation Center model that uses cooperative agreements to build an evidence base for whole-person functional or lifestyle medicine interventions in Original Medicare. CMS describes funding of roughly $100 million to support up to 30 three-year cooperative agreements focused on interventions that are not currently covered by Original Medicare but have documented evidence of efficacy and are intended to support – not replace – conventional medical care. CMS’ emphasis is on generating cost and quality evidence that could inform later coverage decisions or future model design. CMS states it will release a Notice of Funding Opportunity in early 2026, and the model is slated to launch September 1, 2026.

ACCESS (Advancing Chronic Care with Effective, Scalable Solutions) Model
ACCESS is a voluntary Original Medicare model aimed at modernizing how Medicare pays for technology-supported chronic care. CMS’ theory is that fee-for-service payment rules can under-support longitudinal, tech-enabled care models. ACCESS would test Outcome-Aligned Payments – recurring payments to participating care organizations for managing qualifying chronic conditions, with payment tied to measurable health outcomes rather than completion of specified “billable” activities. CMS highlights a starting clinical focus that includes high blood pressure, diabetes, chronic musculoskeletal pain, and depression. CMS states the model will run for 10 years beginning July 5, 2026, with an initial application cycle for the first performance period and later start dates for subsequent applicants.

GENEROUS (GENErating cost Reductions fOr U.S. Medicaid) Model
GENEROUS is a voluntary Medicaid drug-pricing model built around CMS-led negotiation for supplemental rebates in exchange for more standardized coverage terms across participating states. Conceptually, it seeks to align Medicaid net prices for included drugs with prices paid in certain other countries by having participating manufacturers provide supplemental rebates to participating states. CMS frames the objective as improving Medicaid affordability for states while preserving access and quality – creating fiscal room for other Medicaid priorities – while also improving the predictability and uniformity of coverage criteria across participating states. CMS states the model would launch in January 2026 and run for five years, with participation voluntary for both manufacturers and states and accompanied by formal application processes.

Why Understanding CMMI Matters

The end of 2025 demonstrates the importance of understanding CMMI model dynamics. Over a few short weeks, several different models were announced that advance longstanding priorities for the administration and could fundamentally change how Medicare and Medicaid pay for drugs. These CMMI models can move quickly from “pilot program” to system-shaping payment architecture and show how broad “model” authority has become across Medicare’s benefit silos. Thus, CMS is moving forward with models that look less like limited demonstrations and more like national policy levers – particularly when they are mandatory and applied through geographic selection.

A voluntary model can be treated as a bounded pilot – important for learning but limited in operational exposure. A mandatory model, especially one imposed through geographic selection, functions more like a governance instrument: It can compel compliance with new financial rules, force rapid operational buildout, and introduce program-wide uncertainty about who pays, who bears risk, and how access will shift – all without the kind of legislative cycle that typically accompanies major Medicare policy changes. In that environment, the decisive questions stop being philosophical (“is this a good idea?”) and become practical (“what must we change, by when, and with what downstream consequences?”). Entities that operate nationally – manufacturers, pharmacy benefit managers, plans, pharmacies, and multi-state provider systems – may have to run parallel administrative and contracting regimes, recalibrate forecasting and revenue assumptions, and manage beneficiary and provider expectations under compressed timelines. They can also generate spillovers and edge effects – behavior changes outside selected areas, shifts in site-of-care patterns, or strategic responses that blur the distinction between “model” and “non-model” markets – complicating both implementation and evaluation.

The bottom line is that CMMI’s authority is not merely an experimentation tool; in mandatory form, it becomes a mechanism for reshaping Medicare’s economics in real time, making attention to rule design, guardrails, and operational feasibility just as important as the policy intent.

CMMI Controversy

Since its establishment, there have been consistent questions and criticisms about CMMI, which attracts outsized controversy because it sits at an unusual intersection of technical payment policy and real governing power. On paper, it is a test-and-learn “innovation lab.” In practice, demonstrations can reshape how Medicare – and sometimes Medicaid – pay for care. And under certain statutory conditions, CMS can expand a model beyond a pilot through rulemaking without further authorization from Congress.

Whether CMMI is delivering on its core promise is also often debated: Have models lowered spending while maintaining or improving quality? Independent reviewers have repeatedly described a portfolio with uneven results and have pushed CMMI to run fewer, more coherent models with consistent benchmarking and evaluation approaches, arguing that overlapping model “tracks” can dilute incentives and create noise that makes true effects hard to measure. The Government Accountability Office has also noted long-running challenges around evaluation design and translating findings into clear, scalable policy choices.

Budget politics have consistently shaped debates about the Innovation Center. Models are intended to reduce spending – or at least maintain spending with documented increases in care quality or value. In a notable reassessment, the Congressional Budget Office estimated that CMMI’s activities increased net federal spending in its first decade (even while acknowledging that some individual models generated savings), which has fueled arguments that the Innovation Center’s mandatory funding and broad authority deserve tighter constraints or clearer performance thresholds. Because of these budgetary issues, providers and plans often criticize the operational footprint of CMMI – reporting requirements, attribution rules, and downside risk – arguing that “innovation” can become another layer of administrative burden, particularly when model design changes midstream or when participation expectations are perceived as quasi-mandatory.

Conclusion

In theory, CMMI’s mission is straightforward: test models, evaluate them, and expand what works. In practice, model outcomes have been uneven. That is not surprising. Payment reform is hard. Results depend on design choices that sound technical but are decisive: benchmarks, risk adjustment, quality measurement, beneficiary attribution, and how aggressively downside risk is phased in.

A model can fail not because “value-based care doesn’t work,” but because incentives were too weak, participation skewed toward low-risk entrants, or the comparison methodology could not cleanly isolate true savings. Conversely, a model can look successful in early years and fade as participants adapt and easy gains are exhausted. This is why observers often emphasize portfolio discipline: fewer overlapping models, clearer incentives, and more consistent evaluation methods.