Administrative action on drug pricing: Lessons and opportunities for the Center for Medicare and Medicaid Innovation

Abstract

BACKGROUND:

In October 2022, the Biden administration issued an executive order to the Center for Medicare and Medicaid Innovation (CMMI) to develop new health care payment and delivery models to lower prescription drug costs and promote access to innovative therapies. In response, the agency proposed 3 novel drug payment models for testing.

OBJECTIVE:

To understand the impact that CMMI demonstration projects can have on the prescription drug market.

METHODS:

We examined each of the models listed on the CMMI website and searched the Federal Register and news articles for additional models that contained interventions related to patient out-of-pocket drug costs, Medicare drug spending, or Medicaid drug spending. We excluded models with indirect effects on drug costs (for example, bundled payments). We comprehensively reviewed all previous cases in which CMMI has attempted models addressing prescription drug costs and spending and evaluated the circumstances, impact, and lessons learned that may aid policymakers in the design and implementation of new models.

RESULTS:

We identified 9 CMMI models containing direct interventions related to drug costs. Among prior models addressing drug prices, nearly half (44%, 4/9) were not implemented because of their scope, voluntary nature, and procedural challenges. No implemented models met the CMMI standard for expansion, although key elements of the Senior Savings model limiting monthly insulin costs to $35 were later incorporated into the Inflation Reduction Act.

CONCLUSIONS:

In future CMMI implementation efforts, we suggest maximizing voluntary collaboration when selection bias concerns are minimal, using mandatory models when not, ensuring that the geographic scope is not overly ambitious, and adhering closely to statutory authority and established administrative procedure to minimize legal challenges and maximize model demonstration utility.

Plain language summary

The Center for Medicare and Medicaid Innovation (CMMI) is a federal agency that tests new ways to pay for and deliver health care. Its goal is to deliver better care at lower costs. We looked at past attempts to lower drug prices for patients or lower public drug spending. Most previous models were not implemented. We suggest that future models maximize voluntary participation, limit geographic scope, and learn from prior legal cases.

Implications for managed care pharmacy

Since 2010, CMMI has designed models specifically to address pharmaceutical cost, spending, and reimbursement, and, in 2023, proposed 3 new models designed to lower out-of-pocket costs for Medicare and Medicaid beneficiaries. We systematically reviewed past CMMI models to extract lessons to inform new CMMI models and other pharmacy benefit plan designs. Factors such as voluntariness, geographic scope, and alignment with established precedent will be critical as CMMI’s prescription drug models face legal and political challenges.

High brand-name prescription drug costs have been a key target of legislators and policymakers in the United States in recent years because they prevent patients from accessing key treatments prescribed by their clinicians. Over the past 40 years, prescription drug spending increased more than 11-fold and currently about 1 in every 7 health care dollars in the United States is spent on medications.¹

The Inflation Reduction Act (IRA) initiated historic Medicare drug price negotiation, with negotiated prices taking effect in 2026 and statutory criteria delineating which drugs are eligible for negotiation.² Recognizing the limited scope of the IRA, the Biden administration issued an executive order in October 2022 directing the Center for Medicare and Medicaid Innovation (CMMI) to test “new health care payment and delivery models that would lower drug costs and promote access to innovative drug therapies for beneficiaries enrolled in the Medicare and Medicaid programs.”³

CMMI, located within the Centers for Medicare & Medicaid Services (CMS), was created by the Affordable Care Act to pilot innovative payment and service delivery models within Medicare and Medicaid. The models are designed to address “a defined population for which there are deficits in care leading to poor clinical outcomes or potentially avoidable expenditures.”⁴ Models are first implemented through a testing and evaluation phase; then, if a model successfully reduces net expenditures while maintaining quality of care or improves quality of care while maintaining expenditures, CMS can choose to expand it to the entire Medicare program.⁴ Over the past 10 years, CMMI has launched more than 50 innovative payment models⁵ and expanded 4.⁶

In response to President Biden’s executive order, in February 2023, CMMI proposed 3 new models related to drug prices (Table 1).⁷ The first, the Medicare High-Value Drug List, aims to address different levels of cost-sharing and prescription drug coverage tiers faced by Medicare beneficiaries. Under this voluntary model, Medicare Part D sponsors would promote greater access to high-value generics that treat chronic conditions by providing a standard set of approximately 150 drugs for no more than $2 each per month.⁷ These generics would not be subject to utilization management.

TABLE 1.

Summary of Center for Medicare and Medicaid Innovation Models Proposed Following President Biden’s Executive Order

	Medicare high-value drug list a	Cell and Gene Therapy Access model a	Accelerating Clinical Evidence model a
Year proposed	2023	2023	2023
Drugs included	~150 unspecified high-value generic drugs	Eligible cell and gene therapies	Drug indications approved via the FDA Accelerated Approval pathway
Participation	Voluntary	Voluntary	Mandatory
Affected stakeholders	Part D plans	State Medicaid plans	Medicare Part B fee-for-service providers
Geographic scopeb	–	–	–
Length of demonstrationb	–	–	–
Beneficiary cost sharingb	$2 for listed drugs	–	–
Evaluation type	Matched beneficiaries	Trends over time	Trends over time
Evaluation metrics	Quality-of-care metrics and program expenditure metrics	Access to therapies over time, clinical and patient experience outcomes, drug spending and utilization, out-of-pocket spending, total program expenditures, and drug discounts	Rate of confirmatory study completion, total Medicare costs, beneficiary-focused metrics of access, safety, health outcomes, quality of care, and price trends of accelerated approval drugs and Part B drug spending
Risk sharing	Retrospective performance-based payments via increased premium subsidies	Outcomes-based payments to drug manufacturers	On manufacturers if confirmatory trials are negative

^aSee Becerra 2023.⁷

^bScope, demonstration length, and many beneficiary cost-sharing details have not yet been announced.

FDA = US Food and Drug Administration

The second model was the Cell and Gene Therapy (CGT) Access model. CGTs target a patient’s genome in a way intended to provide relief in a single treatment.⁷ CGTs are expected to expand rapidly in coming years. Complicating access, though, are high upfront costs and efficacy concerns. The CGT Access model would allow CMS to bargain with manufacturers on behalf of a pool of state Medicaid programs as well as to negotiate and administer “outcomes-based agreements” that would scale payment levels for drugs based on their efficacy in treating patients’ illnesses.

The third model was the Accelerating Clinical Evidence model,⁷ designed to address drugs granted accelerated approval by the US Food and Drug Administration based on promising, but unproven, surrogate measures awaiting required confirmatory clinical trials. Studies show that those trials have not been completed in a timely fashion^8,9; the model’s goal would be to pay less for accelerated approval drugs until confirmatory trial completion.

The success of these models may be informed by the outcomes of CMMI’s previous demonstrations related to prescription drugs. We sought to understand the circumstances that contributed to model implementation, evaluate their impact on the drug market, and find key lessons applicable to the current trio of models.

Methods

To identify all CMMI demonstrations related to drug costs proposed or implemented from 2010 to 2022 we examined each of the models and submodels listed on the CMMI website (n = 137).⁵ We also evaluated the first 2 rounds of 2 CMMI grant programs designed to spur innovation through funding innovative models: the Health Care Innovation Awards (n = 147) and State Innovation Models Initiative (n = 56), and searched the Federal Register and news articles for additional models. For each model, we extracted the model’s name, year of implementation, demonstration stage, intervention category (a broad, internal CMMI category, eg, episode-based payment initiatives), legal authority, and whether each model included a drug pricing component. Models were included if they used Section 1115A (Affordable Care Act Section 3021) authority, the same authority underlying the current models, as distinct from models that were congressionally mandated. For inclusion, models had to contain an intervention related to patient out-of-pocket drug costs or Medicare or Medicaid drug spending; we excluded models with only indirect effects on drug costs (eg, bundled payments). We reviewed all models discovered during our search even if they were proposed but not implemented. For each included model, we extracted the proposal year, which drugs/drug classes were included, whether participation was voluntary or mandatory, geographic scope, details on how the model would be evaluated, proposed demonstration length, and details of the pricing tools leveraged by each model. For included models that were implemented, we additionally extracted the number and nature of model participants (eg, Medicare Part D plan sponsors) and the actual years of implementation. We used all included models to extract key lessons learned that may aid future model design and implementation for policymakers. This review of CMMI demonstrations does not constitute human subject research and did not require institutional review board approval.

Results

Among 58 distinct models, 79 submodels, and 206 grant awardees using Section 1115A authority (57 from the CMMI website and Federal Register and 1 from the literature), 9 contained interventions related to drug costs. Three models were directed toward Medicare Part B, 5 to Medicare Part D, and 1 was a grant award applied to a particular state.

MEDICARE PART B PRESCRIPTION DRUG MODELS

Medicare Part B covers drugs administered by clinicians in outpatient settings, such as injections and infusion-based therapies. Clinician reimbursement for most Part B drugs is determined using the average sales price—the average price pharmaceutical manufacturers sell a drug to all nonfederal purchasers after discounts and rebates—plus 6% (or 4.3% during sequestration).¹⁰ Between 2015 and 2020, CMMI proposed 3 models to address paying for prescription drugs, but none were implemented (Table 2).

TABLE 2.

Summary of Center for Medicare and Medicaid Innovation Models for Part B Reform

	Part B drug payment model a	IPI model b	Most Favored Nation model c
Year proposedd	2016	2018	2020
Drugs included	Almost all Part B drugs	Single-source drugs and biologics	Top 50 Part B drugse
Participation	Mandatory	Mandatory	Mandatory
Affected stakeholders	Medical providers and suppliers	Physician practices and hospital outpatient departments	Medical providers and suppliers
Geographic scope	75% of United Statesf	50% of United Statesf	Nationwide
Evaluation details	Randomized 4-arm trial comparing: 2.5% flat fee add-on, VBP tools, 2.5% fee and VBP, and control group	Randomized 2-arm trial comparing: IPI-adjusted prices with flat fee add-on and control group	Interrupted time series analysis without direct comparison group
Length of demonstration	5 years	5 years	7 years
Part B pricing
Base payments	ASP	ASP scaled by average price paid by 14 countries	Lowest GDP-adjusted price paid by select countriesd
Add-on payments	2.5% + flat fee	Flat fee	Flat fee
Other tools	Reference pricing by drug class Indication-based pricing Outcomes-based risk-sharing agreements Coinsurance discounts for high-value drugs Voluntary provider clinical decision support tools	Private vendors to negotiate with manufacturers and distribute to physicians and hospitals	N/A

^a See “Medicare Program; Part B drug payment model.” ¹¹

^b See “International Pricing Index (IPI) model.” ¹³

^c See “Most Favored Nation model.” ¹⁶

^d No models were implemented.

^e Based on overall spending in 2019, with exclusions for certain drugs classes (eg, vaccines).

^f Geographic areas randomly assigned.

^g Twenty-two Organization for Economic Cooperation and Development member counties with at least 60% of the US GDP per capita.

ASP = average sales price; GDP = gross domestic product; IPI = International Pricing Index Model; N/A = not applicable; VBP = value-based pricing.

In 2016, the Obama administration proposed testing a novel, two-phase Part B payment model called the Medicare Part B Drug Payment model.¹¹ The first phase would have redistributed Part B’s 6% add-on payment for physician-administered drugs as a 2.5% add-on plus a flat fee. The second phase would have implemented clinical decision support and value-based purchasing tools, such as drug class reference pricing (ensuring that Medicare Part B did not pay more for new members of a drug class with similar efficacy than already existing members), pricing based on safety and cost-effectiveness, outcomes-based risk sharing (reducing drug spending for patients who do not respond clinically to a drug), and reduced patient coinsurance. Both proposals would have been implemented together as a mandatory 5-year randomized intervention across geographic areas. Almost all Part B drugs would have been included. Its evaluation was proposed to be a randomized, 4-arm trial comparing 2.5% flat fee add-on, value-based purchasing tools (eg, reference pricing, indication-based pricing, and clinical decision support tools), 2.5% flat fee plus value-based purchasing tools, and a control arm.

The proposal received criticism from patient and physician groups, particularly oncologists.¹² They argued that the administration did not adequately engage stakeholders and failed to show that the proposal addressed “deficits in care,”¹² a statutory requirement for model testing.⁴ CMS abandoned the proposal in December 2016; it was formally withdrawn in 2017.

In 2018, the Trump administration proposed a new International Pricing Index (IPI) model.¹³ This model offered a 5-year randomized trial of reduced Part B drug prices that would be closer to the average price paid by 14 selected countries. Additionally, it would have allowed private vendors to act as middlemen between manufacturers and providers and redistributed the 6% add-on as a flat fee. Like the Obama administration’s proposal, the trial component required mandatory participation by providers and hospitals in randomly selected geographic areas. The proposal received mixed reactions, with pharmacy benefit managers supporting the model, hospitals cautiously optimistic, and the main pharmaceutical industry lobbying group opposed.¹⁴

Hospitals and pharmacy benefit managers, as 2 of the main purchasers of Part B drugs, were enthusiastic about the possibility of spending less to acquire pharmaceuticals via reference pricing. However, lower prices for drugs would have reduced revenues for pharmaceutical manufacturers, who were critical of the proposal and claimed that lower revenues would ultimately “hinder patient access” to Part B drugs.¹⁵ The model’s evaluation plan was proposed as a randomized 2-arm trial comparing IPI-adjusted prices with a flat fee add-on to the control group.

In November 2020, CMS proposed a new Most Favored Nation model to replace the IPI.¹⁶ Instead of scaling by average international prices, this model linked Medicare Part B payments directly to the lowest gross domestic product–adjusted price paid by an Organization for Economic Cooperation and Development member country with at least 60% of the US gross domestic product per capita. The vendor proposal from the IPI was dropped, and the add-on payment was standardized for all drugs and providers. The Most Favored Nation model would have been a 7-year nationwide demonstration, with mandatory participation by almost all providers and suppliers, and phased-in pricing for the top 50 Part B drugs over the first 4 years. The model’s evaluation was planned to be an interrupted time series analysis without a direct comparison group. Similar to the IPI model, the proposal elicited concerns from some organizations about reductions in patient access to drugs and financial burdens on hospitals and physician practices, respectively, if manufacturers did not lower prices to align with the proposed payment amounts.¹⁷ These concerns were amplified by the model’s near-universal proposed intervention and mandatory nature, which increased the magnitude of effects on stakeholders. The merits of this proposal were never formally addressed; the rule was blocked because of CMS’s failure to follow standard administrative procedures.¹⁸ The Biden administration withdrew the rule at the end of 2021.

MEDICARE PART D PRESCRIPTION DRUG MODELS

Medicare Part D was implemented in 2006 to address the high out-of-pocket prices for pharmacy-dispensed medications felt by Medicare enrollees. Under Part D, CMS contracts with private insurers to provide prescription drug coverage for enrollees who pay monthly premiums. Plans vary, but there are requirements for what the plans must cover and how much patients can pay.¹⁹ Although Medicare Part D is optional, around 75% of Medicare beneficiaries purchase Part D coverage every year.¹⁹

In Part D, 5 models have been introduced, 4 of which have been implemented (Table 3). In 2015, CMS announced the Enhanced Medication Therapy Management model.²⁰ The model was designed to incentivize greater use of medication therapy management services—medication reviews, refill reminders, and physician education—to improve adherence and reduce adverse events. Under Part D, prescription drug plans provide basic medication therapy management services to limited groups of patients. The Enhanced Medication Therapy Management model allowed participating prescription drug plan sponsors to provide more services and target them to specific patients—for example, providing copay assistance to address financial barriers to filling prescriptions. This model ran from 2017 to 2021 with 6 prescription drug plans choosing to participate. Two plans offered copay support by directly waiving copays or connecting patients to external financial support. However, copay supports were reportedly challenging to implement, and isolating their effects from other services in the model was not possible with the model’s matched evaluation design. Overall, the Enhanced Medication Therapy Management model produced no clear effect on total medical expenditures or medication adherence.²¹

TABLE 3.

Summary of Center for Medicare and Medicaid Innovation Models for Part D Reform

	Enhanced Medication Therapy Management model a	Part D Payment Modernization model b	Part D Senior Savings model c	Part D Medication Adherence Program demonstration d	Medicare Advantage Value-Based Insurance Design model e
Year proposed	2015	2019	2020	2020	2015
Years implemented	2017-2021	2020-2021	2021-2023	Not implemented	2015-2030
Drugs included	All Part D	All Part D	Insulin	Any drug	Drugs for certain chronic conditions
Participation	Voluntary	Voluntary	Voluntary	Voluntary	Voluntary
Affected stakeholders	Part D plans	Part D plans	Manufacturers and Part D plans	Beneficiaries	MA plans
Geographic scope	5 Part D geographic regions	Nationwide	Nationwide	Nationwide	6 states
Model participants	6 PDP sponsors	2 PDP sponsors	5 manufacturers and 106 PDP sponsors	N/A	11 MA sponsors
Evaluation details	Matched beneficiaries	Not found	Not found	Compare voluntary card users with nonusers	Matched beneficiaries
Length of demonstration	5 years	5 yearsf	5 years	One-time	15 yearsg
Part D pricing
Beneficiary cost sharing	Optional copay support	Optional flexibilities	$35/month OOP maximum for insulin	$200 OOP discount card	Variable, plan dependent
PDP risk sharing	Retrospective performance-based payments in the form of increased premium subsidies	Two-sided risk for federal reinsurance in catastrophic phaseh	Optional reduced risk, with first threshold at 2.5% if high proportion of insulin-dependent enrollees	N/A	N/A
PDP flexibilities	Tailored MTM services, such as: Medication reconciliation Medication review Social and cost-sharing support	Enrollee rewards and incentives Reduced cost sharing for generics and biosimilars Tailored MTM services Care management programs Installment payments for cost sharing	Enrollee rewards and incentives	N/A	MA plan flexibilities, including: Reductions in cost sharing for covered Part D drugs Part D Rewards and Incentives Programs
Manufacturer contribution	N/A	N/A	Coverage gap manufacturer discount applied before PDP supplemental benefit	N/A	N/A

^a See Beringer et al, 2023. ²¹

^b See “Part D Payment Modernization model.” ²²

^c See “Part D Senior Savings model.” ²⁷

^d See “Medicare Part D medication adherence program (MAP) demonstration.” ²⁹

^e See “Medicare Advantage Value-Based Insurance Design model extension fact sheet.” ³²

^f Planned for 5 years but stopped early because of low participation.

^g Expanded and extended until 2030.

^h PDPs responsible for savings/losses compared with benchmark: 30% of savings less than 3%, 50% of savings greater than 3%, and 10% of losses.

MA = Medicare Advantage; MTM = Medication Therapy Management; N/A = not applicable; OOP = out-of-pocket; PDP = prescription drug plan.

In 2019, CMS announced the Part D Payment Modernization model, which launched in January 2020.²² This voluntary model imposed 2-sided risk on federal reinsurance costs in the catastrophic phase,²³ with prescription drug plans eligible to share in 30%-50% of savings below a target benchmark while also being responsible for 10% of excess costs. The model also gave prescription drug plans more flexibility in pharmacy benefit design; the proposal did not include details on how its evaluation would be conducted. On the last day of President Trump’s term, the administration changed the model to allow prescription drug plans to cover fewer drugs per therapeutic class, including protected drug classes.²² This change was opposed by cancer advocacy organizations, which raised concerns that limited formularies would negatively impact patients with cancer,²⁴ and was later reversed by the Biden administration.²⁵ The model was designed as a 5-year demonstration, but it was terminated after 2 years because only 2 health plans participated nationwide, presumably because of concerns of excess costs or logistical complexity.²⁶ However, no evaluations were conducted and sponsors did not make public comments about their lack of participation.

In 2020, CMS announced the Senior Savings model, aimed at lowering prices for insulin by facilitating voluntary cooperation among Medicare, prescription drug plans, and manufacturers through 2025.²⁷ The model incentivized Part D plans to provide supplemental coverage limiting beneficiary copays for insulin to $35 per month. Traditionally, plans rarely provided supplemental benefits in the coverage gap phase because brand-name manufacturers provided a 70% discount directly to beneficiaries. The Senior Savings model allowed prescription drug plans to share in the manufacturer discount in exchange for limiting insulin cost sharing throughout all coverage phases. Prescription drug plans also received greater risk protection if their plans attracted more insulin-dependent beneficiaries. The model was popular with more than 1,600 prescription drug plans covering more than 13.8 million enrollees and all 3 major insulin manufacturers participating in 2021, the plan’s first year.²⁷ An evaluation of the model’s first 2 years conducted by the RAND Corporation described participation as “robust,” and noted that effects on “key outcomes, such as costs, medication adherence, and improved health” would be quantified in future evaluations.²⁸ However, the model was later eclipsed by the IRA. Like the Senior Savings model, under the IRA cost sharing for each insulin product covered under any Part D prescription drug plan is capped at $35 per month, and Part D deductibles will not apply to insulin. Because of the passage of the IRA, CMMI elected to end the Senior Savings model at the end of 2023.

In 2020, the Trump administration proposed the Medicare Part D Medication Adherence Program Demonstration, which was not implemented. In a draft version, the proposal would have sent Part D beneficiaries a one-time $200 drug discount card to pay for out-of-pocket costs.²⁹ This model would have been a nationwide voluntary participation, comparing beneficiaries who chose to use their cards with those who did not. Experts argued that the proposal would likely have violated election law.³⁰

Only one Medicare Part D model demonstration is still ongoing. Originally announced in 2015, the Medicare Advantage Value-Based Insurance Design (VBID) model began in early 2017 as a voluntary 5-year demonstration. The VBID model allowed Medicare Advantage sponsors to restructure insurance plans to incorporate principles of value-based insurance, such as lowering out-of-pocket payments for high-value services, to improve outcomes for enrollees with certain chronic conditions. Participating sponsors were given broad implementation discretion, and 3 of 11 sponsors chose interventions designed to decrease drug costs for enrollees through Part D Rewards and Incentives Programs.³¹

One plan sponsor implemented Part D Rewards and Incentives for enrollees with hypertension by eliminating all cost sharing for hypertension drugs; another sponsor did the same for coronary artery disease drugs. For patients with chronic heart failure, a third plan decreased copays for tier 1 generic drugs from $7 to $0 while also eliminating copays for primary care and cardiology visits.

After 3 years, 1 study matched comparators in VBID insurance plans with and without Part D Rewards and Incentives interventions, finding that beneficiaries in programs with interventions showed a 1.6% increase in 30-day medication fill rates and decreases in the probability of hospitalization (8.8%) or emergency department admission (5%).³² However, this association was not observed for the same outcomes when matched comparators were compared with individuals in non-VBID plans, casting some doubt on whether the CMMI intervention was the cause of the differences observed.

Phase I of the VBID demonstration ended in 2019, but the program was expanded to encourage innovation beyond traditional VBID principles and to include hospice in 2020; the demonstration is currently scheduled to run through 2030.³² Current iterations of VBID do not appear to be testing dedicated Part D interventions.

HEALTH CARE INNOVATION AWARD MODEL

Through its Health Care Innovation Awards program, CMMI funded 1 external program with components explicitly designed to lower prescription drug prices (Table 4).

TABLE 4.

Summary of a Center for Medicare and Medicaid Innovation–Funded External Innovation Model

	Wyoming Medication Donation Program expansion a
Proposed or announced	2009
Implemented	2012
Drugs included	Noncontrolled prescription medications
Participation	One central pharmacy and voluntary participation from multiple pharmacies statewide
Scope	Statewide in Wyoming
Evaluation	Pre/post statewide outcome data
Length of demonstration	Indefinite
Beneficiary cost sharing	$0 for qualifying residents
Manufacturer contribution	None

^a See “Health Care Innovation Awards: Memorial Hospital of Laramie County DBA Cheyenne Regional Project profile.” ³³

In 2012, CMMI distributed a 3-year grant to the Wyoming Institute of Population Health for a 5-component program designed to transform rural health care delivery in the state through the creation of medical neighborhoods. One arm of the project expanded Wyoming’s Medication Donation Program, which provides mail order pharmacy services at no cost to low-income, uninsured, or underinsured Wyoming residents.³³ The program was positively received.³⁴ But because the program occurred as part of a 5-prong intervention, disaggregated effects on individual spending and outcomes were not ascertained.³⁵

Discussion

CMMI’s previous efforts to rein in Medicare drug prices have had limited success, highlighting some key considerations for future model design.

VOLUNTARY VS MANDATORY PARTICIPATION

The voluntariness and scope of models were substantial sources of controversy. One particularly controversial aspect of recent CMMI proposals has been mandatory participation, which could explain why some Part D models were successful but all Part B models were not. When CMS proposed mandatory participation in all Part B models, these models were criticized widely, even by members of Congress.^12,36

Only a few mandatory CMMI models have been successfully implemented to date, all coming outside the pharmaceutical market.⁶ The Comprehensive Care for Joint Replacement model initially required participation from hospitals in 67 counties, but the mandatory component has been gradually scaled back. The End Stage Renal Disease Treatment Choices model and the Radiation Oncology model (not yet implemented) both require the participation of clinicians and treatment facilities in randomly selected geographic areas. Although participation requirements are not popular, voluntary models are impacted by selection bias, which makes it difficult to evaluate performance, draw conclusions across the market beyond nonparticipants, generate system-level savings, and ensure equitable participation.

GEOGRAPHIC SCOPE OF MODELS

The geographic scope of CMMI models has generated controversy. The CMMI statute requires that models address a “defined population,”⁴ and most successful CMMI models have been tested on a small scale, applying only to certain geographic areas or clinical interventions. But every unsuccessful Part B proposal would have applied to at least half the country, leading to speculation that models might not have passed legal muster.³⁷ A nationwide proposal risks broad pushback and legal challenges that it goes beyond the scope of a test as intended in the CMMI statute. For example, in January 2019 the 2 top lawmakers in the House Ways and Means Committee sent a letter to CMS Administrator Seema Verma to ensure that CMMI “tests and promotes [innovative] delivery system models, but does not bypass Congress and the public to rewrite Medicare law for beneficiaries.”³⁸ A similar letter was sent by a group of Senators in early 2023.³⁹ Already, at least 1 legislative proposal threatens to limit CMMI’s authority.⁴⁰ Still, a broad model may best demonstrate system-wide savings and avoid cost shifting to nonmodel participants. The Senior Savings model has successfully been made available nationwide despite focusing on a single clinical condition.

METHODOLOGIC DESIGN OF EVALUATION

By statute, all models must evaluate how the model affects program costs and care quality.⁴ Most evaluations use a comparison group, comparing costs and quality between model participants and the rest of Medicare. Except for the Most Favored Nation proposal, the unsuccessful mandatory Part B models generally proposed randomized trials as their evaluation components. The voluntary Part D models, meanwhile, primarily relied on matched beneficiaries as control groups.

The Most Favored Nation models were unprecedented in proposing a nationwide demonstration with no comparison group. It remains unclear whether a comparison group is statutorily required, although it is the gold standard for a rigorous evaluation—as even comparison-matched results may experience residual confounding by unmeasured characteristics. The issue of selection bias for voluntary models will continue to challenge CMMI administrators as they balance risks and benefits for affected beneficiaries. For plans sponsored by private organizations, voluntary models will continue to be affected by firms’ financial incentives to minimize costs and maximize revenue. Therefore, despite past Part B models failing to reach the implementation phase, it remains important to prioritize mandatory CMMI models to increase generalizability for occasions when demonstrations are successful; future models must continue to balance system transformation with evaluation design.

PROCEDURAL STRUCTURE AND LEGISLATIVE AUTHORITY

Adhering closely to statutory requirements for model structure was an important factor in withstanding legal challenges. Many substantive design decisions are precluded from judicial review,⁴ so opponents of potential models have tended to challenge them on procedural grounds.

CMMI’s authorizing statute gives the agency authority to test models without using the formal notice and comment rulemaking process⁴; thus, many voluntary models are announced via memoranda. However, CMMI has used rulemaking procedures for all proposals with mandatory participation.

Public agency rulemaking is regulated by the Administrative Procedure Act (APA).⁴¹ Following established APA precedent, especially for nonvoluntary models, includes proposing models through the notice and comment process by issuing a formal Notice of Proposed Rulemaking and accepting and responding to public comments on that rule before issuing a Final Rule. The mandatory Most Favored Nation proposal bypassed the notice and comment process by attempting to use the APA’s Good Cause exception, which allows federal agencies to waive such rulemaking when it would be “impracticable, unnecessary, or contrary to the public interest.”⁴¹ The administration argued that the “need for affordable Medicare Part B drugs” was “particularly acute…in the midst of the COVID-19 pandemic,”⁴² but this argument proved unsuccessful.¹⁸

None of the prescription drug–specific models included in this review were expanded beyond the testing phase. Of more than 50 CMMI models, only 4 have met the actuarial bar required for expansion: the Home Health Value-Based Purchasing Model, Medicare Prior Authorization Model for Repetitive Scheduled Non-Emergent Ambulance Transport, Medicare Diabetes Prevention Program, and the Pioneer Accountable Care Organization model.⁶ Even if not ultimately expanded, a CMMI demonstration may be a helpful temporary solution and could also provide valuable information to lawmakers, like the Senior Savings model and the IRA.

IMPLICATIONS FOR CMMI’S NEWLY PROPOSED MODELS

Past models offer useful lessons for implementing the 3 innovative drug payment models recently proposed by CMMI.

First, in cases in which CMMI feels that selection bias concerns are minimal, voluntary collaboration between Medicare and its model participants should be utilized to smooth model implementation. In the Senior Savings model, CMMI removed a financial disincentive to align the interests of the agency with drug manufacturers and prescription drug plan partners, resulting in wide participation starting the model’s first year. For the High-Value Drug List model, the standardized price intervention should theoretically increase medication adherence, improve outcomes, and decrease expenditures because of lower and more predictable out-of-pocket costs—all financial incentives for plans to participate. If participation rates are low, CMMI has the flexibility to add a small financial award (for example, paying 0.5% of the Initial Coverage period to reduce the percentage plans pay from 75% to 74.5%) to increase voluntary participation, thereby accelerating data collection and model evaluation. Although CMMI has the power to create mandatory models, the advantages of continuing with a voluntary approach may be worth the added costs. Additionally, accelerating evidence collection may allow earlier evaluation and increased savings by halting costly models earlier or hastening the expansion of cost-effective models. Although no prior CMMI model has been struck down or not implemented because of its evaluation design, voluntariness should be considered along with a proposed model’s evaluation design to ensure that the results of a model are generalizable.

CMMI should also carefully consider the geographic scope of its models. As was illustrated by the letters written by senior House lawmakers to the head of CMS, larger models (in terms of geography or beneficiaries) have faced steeper political hurdles to implementation. Although most successfully implemented CMMI models were of modest scope and focused on a limited geographic area, all unsuccessful Part B proposals would have applied to at least half of the United States. Thus, although CMMI models should be large enough to properly evaluate an intervention in a generalizable way, they should not be larger than is necessary, and models proposed to exceed 50% of the United States geographically should be carefully justified.

Adhering closely to legal precedent will be important for CMMI’s newest models to avoid the fate of the Most Favored Nation model. In particular, CMMI should take care to strictly follow the APA. For controversial proposals likely to face legal challenge, CMMI may consider issuing an Advance Notice of Proposed Rulemaking to elicit input from stakeholders before issuing a Notice of Proposed Rulemaking. CMMI should also clearly articulate how high drug prices are linked to clinical outcomes, which is required under CMMI’s authorizing statute.⁴ Justifying high drug prices as “potentially avoidable expenditures” is straightforward, but the “deficits in care” component could be an issue for litigation. As one strategy, CMS could explain how high prices decrease medication adherence, as it did with the Part D Senior Savings model. Additionally, CMS might draw from current CMMI strategic objectives, outlining a focus on health equity and the social determinants of health.

When considering previous CMMI prescription drug demonstration projects, the CGT model has the least amount of institutional precedent. CMMI has never tested a drug pricing model in Medicaid. Although CMMI is well positioned among federal agencies to represent states in pooled negotiations and contract agreements, it may consider learning from the interstate purchasing pools used by some state Medicaid programs that have leveraged higher volumes and formulary placement while consolidating the technical knowledge to negotiate competitive rates with manufacturers.⁴³

Finally, the Accelerating Clinical Evidence model is likely to be the greatest source of tension among drug manufacturers, insurers, and the federal government. Unlike the Most Favored Nation model (which was never implemented), CMMI announced that it may issue an Advanced Notice of Proposed Rulemaking to gather stakeholder input; this would be prudent to ensure the leading criticisms are acknowledged early in the process.

According to CMMI, the Accelerating Clinical Evidence model will have “[m]andatory participation for applicable Medicare Part B fee-for-service providers.”⁶ This leaves open a crucial design question: what are “applicable” providers? To incentivize pharmaceutical firms to complete confirmatory trials in a timely fashion, the proportion of participating providers must be substantial. If the model only applied to a small subset of plans or providers, it would not create sufficient financial leverage. But excessive financial pressure might invoke “irreparable harm” arguments against it.

The Accelerating Clinical Evidence model also faces at least one important set of challenges to implementation: indication-specific pricing. Many drugs have been approved for multiple indications by the US Food and Drug Administration, and current Part B fee-for-service drug payments are not tied to indication, meaning that implementation would require a new system to track and reimburse by indication with a plan for how to reimburse off-label use. Commentators have also expressed concern that lowering Medicare reimbursement rates may unfairly target the physician’s offices that supply clinician-administered drugs, mirroring concerns expressed over past Part B models.⁴⁴

LIMITATIONS

Our study has 2 primary limitations. First, although we attempted to analyze every proposed or implemented CMMI model using Section 1115A statutory authority, some models (eg, Medicare Part D Medication Adherence Program) were announced but never formally proposed, with documentation of the models incomplete. Second, many models or demonstrations may have substantial indirect impacts on drug pricing for both payers and beneficiaries (eg, bundled payments models). Because these effects are not direct and occur simultaneously with other interventions, disaggregated effects are challenging to ascertain and were excluded from our analysis. However, their cumulative impact may be substantial and would not be accounted for in this review.

Conclusions

CMMI has the potential to identify new prescription drug payment and care delivery models that may, if successful, lower Medicare spending and improve patient access to affordable treatments. For future models, we suggest maximizing voluntary collaboration when selection bias concerns are minimal, using mandatory models when they are not, ensuring that geographic scope is not overly ambitious, and adhering closely to statutory authority and established administrative procedure to minimize legal challenges and maximize model demonstration utility. Although nearly half of the models in our study were not implemented, by carefully designing new models that incorporate lessons from previous demonstrations, CMMI may still provide useful innovations for the prescription drug market.

Funding Statement

Arnold Ventures provided funding for this study. The funder had no role in the design or conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.