Implicit Bias Is CMMI’s Latest Problem to Tackle

Advancing health equity is one of the five strategic objectives the CMS Innovation Center (CMMI) outlined in its Strategy Refresh back in November 2021.  It’s no surprise that health equity is becoming an increasingly important goal at CMMI as the Biden administration has made health equity a major priority since day one

However, CMMI has its work cut out for itself.  According to a July 2022 report, implicit bias is pervasive in at least three payment models, which signals challenges if CMMI is serious about advancing equity. 

Implicit bias is an involuntary bias that a person is unaware of.  Often times, negative attitudes and stereotypes can play a powerful role in shaping implicit bias, or “unconscious prejudice.”  Two top CMMI officials authored the report on implicit bias to analyze how implicit bias may be impacting beneficiary groups in payment models. 

In its analysis, CMMI only reviewed three models: the Comprehensive Care for Joint Replacement (CJR) Model, Kidney Care Choices (KCC) Model, and Million Hearts® Cardiovascular Risk Reduction Model.  CMMI selected these models because they represent a microcosm of CMMI models that vary by mandatory/voluntary status, financial methodology, and risk stratification. 

Overall, the report found that use of certain risk-assessment and screening tools, provider tools, and payment design and risk-adjustment algorithms has led to the exclusion of some beneficiaries from these models.  Here are some key findings:

  • The CJR Model tests bundled payment plans for participating providers that perform knee and hip replacements.  The report found beneficiaries receiving joint replacements were “less medically complex” than those receiving joint replacements at the same hospitals before model participation began.  The report also found that beneficiaries in the model are less likely to be dual-eligible, which indicates a lower socio-economic status.  It is also worth noting that Black Americans are likely to receive knee and hip replacements than White Americans. 
  • The KCC Model encourages nephrologists, dialysis facilities, and end-stage renal disease (ESRD) practices to focus on the total care of their patients and incentivizes kidney transplants for chronic kidney disease beneficiaries.  The analysis found that the model’s medical eligibility criteria may have inadvertently excluded Black American beneficiaries, despite the fact that Black Americans are over three times more likely to have ESRD. 
  • The Million Hearts Model provides financial incentives for practices to lower 10-year cardiovascular disease risk for the 30% of high-risk Medicare beneficiaries.  The evaluation found that despite being developed specifically for Black and White populations, the risk calculator used to predict risk scores underestimated risk among patients in other racial and ethnic groups that do not identify as White or Black as well as patients in lower-income households.   

CMMI acknowledged that its findings are “troubling” and underscore a need for a “more systematic evaluation of implicit bias in current and new models.”  As a next step, CMMI says it is working on a “step-by-step guide” to detect and address bias in current models and prevent bias from forming in future one.

While it’s encouraging to see CMMI has a plan to address implicit bias, the revelation that unconscious prejudice is prevalent in three key payment models only adds to the list of challenges CMMI needs to address.  According to an August 2021 report, only a handful of CMMI’s 40-plus models have met the center’s statutorily required goal of reducing costs or improving quality.  In addition to the Biden administration’s emphasis on health equity, the report’s findings likely catalyzed CMMI to lay out its strategic refresh in November 2021, which makes reducing costs and improving quality its “overarching goal.”  On top of CMMI’s difficultly of meeting its statutory obligations, the center now faces the challenge of addressing implicit bias, which may also be prevalent in additional payment models.

All in all, CMMI’s issues with reigning in costs, boosting quality, and stopping implicit bias could signal larger structural problems within the center.  Of note, CMMI’s problems are not lost on lawmakers, in May 2022, Sen. Cory Booker (D-NJ) and Rep. Terri Sewell (D-AL) introduced bicameral legislation to require the center to work with experts on health equity when developing and reviewing payment models.  

At least CMMI has acknowledged its challenges and is laying out plans to address them, including its strategic refresh and a forthcoming “step-by-step guide” on tackling implicit bias.  However, CMMI won’t be able to solve its problems overnight, and the center has a long way to go if it not only wants to achieve its statutory goals of bringing down costs and enhancing quality, but also take on new priorities like improving health equity.

After 10 Years, How Is the CMS Innovation Center Doing?

The Center for Medicare and Medicaid Innovation (CMMI), also known as the CMS Innovation Center, just celebrated its tenth birthday last year.  Tasked to address growing concerns about rising costs, quality of care, and inefficient spending, CMMI is a powerful tool for innovation in the US health care system.  After a decade, is CMMI delivering on its promise to innovate health care, or does the young agency still have much to accomplish?

All About CMMI

Created upon enactment of the Affordable Care Act (ACA) in 2010, CMMI is statutorily mandated to design, implement, and test new health care payment and delivery models for Medicare and Medicaid.  Managed by the Centers for Medicare and Medicaid Services (CMS), CMMI has launched over 40 new payment models since its inception, including accountable care organizations, medical homes models, and bundled payment models.  CMMI separately awards grants to state agencies, researchers, and other organizations for projects to design and implement new payment models with the same goals of improving care and lowering costs, and some of CMMI’s work includes multi-payer alignment models that impact patients with commercial insurance. 

2020 Report to Congress

Released on August 4, the 2020 Report to Congress provides an in-depth look at the performance of CMMI models and serves as a key indicator of how the center is doing in its effort to address rising costs and boost quality.  While the report focuses on CMMI’s activities from October 1, 2018, to September 30, 2020, it also highlights some actions taken from September 30 to December 31, 2020. 

In the 2020 Report, CMS estimated that over 27.8 million Medicare and Medicaid beneficiaries plus enrollees in commercial insurance plans have received care from over 500,000 health care providers or plans participating in alternative payment models under CMMI.  The 2020 Report analyzed a total of 38 active models within CMMI, including 11 new models announced since the 2018 Report and 27 active models that were launched prior to October 2018.

Summary of Findings

Unfortunately, only a handful of CMMI models met either goal of reducing costs or improving quality.   Furthermore, only the five following models delivered “statistically significant savings” to the Medicare Trust Fund according to the report:

Additionally, a few models led to improvements in quality but did not yield any noteworthy savings, including the Comprehensive End-Stage Renal Disease (ESRD) Care (CEC) Model and the Comprehensive Care for Joint Replacement (CJR) Model.

Saving money and raising quality aren’t the only metrics for a model’s success.  For CMS to consider permanently expanding a model for the federal health care entitlement programs, models must meet several additional criteria, including assurance from the CMS Office of the Actuary that a model’s expansion would not deny or limit coverage or provision of benefits under Medicare, Medicaid, and CHIP.   According to the report, only three models met the criteria:

How Can CMMI Improve?

While the 2020 Report to Congress does not explicitly offer recommendations on how to improve model performance, it does identify four issues that contributed to lower-than-expected model performance:

  • Selection bias created by voluntary models.
  • Benchmark inaccuracy.
  • Quality measure misalignment.
  • The need for greater data transparency.

These four issues identified in the report suggest a few ways CMMI models could produce better quality and provide for lower costs, mainly through mandatory model participation and more data transparency.  The idea of making more payment models mandatory is not a new idea.  In a July 2021 interview with Health Affairs, CMMI Director Liz Fowler explained that a shift towards mandatory models, which had already begun during the previous administration, will continue under the Biden administration and are likely to play a greater role in CMMI’s future.

The fact that CMMI models are underperforming is not lost on CMS leadership.  In a recent Health Affairs blog post, a few top agency officials including CMS Administrator Chiquita Brooks-LaSure and Fowler acknowledged only a handful of models have incurred savings and met the requirements to be expanded.  In addition to recounting a few recommendations from outside experts, such as MedPAC’s endorsement for streamlining and harmonizing models, Brooks-LaSure and Fowler offered several takeaways to inform how model performance could be improved.

  • CMMI needs to reevaluate how it designs financial incentives in order to boost meaningful provider participation.
  • Challenges in setting financial benchmarks have undermined models’ effectiveness, underscoring a need to ensure models are not resulting in overpayment and explore ways to improve or replace the current risk adjustment methodology. 
  • Since providers find it hard to accept downside risk if they lack the tools to change care delivery, CMMI should help ensure providers have options for managing risk, such as support in transforming care, waivers, and data. 

Despite dozens of underperforming models, CMS recognizes that the Innovation Center has room for improvement, and the agency’s leaders are keen on delivering a strategy that works.  Hopefully, through streamlining current models, implementing more mandatory models, boosting participation in voluntary models, improving financial incentives, and ensuring model participants have the tools they need to succeed, CMMI models could hopefully be in a better position to  both reduce costs and improve quality in time for the center’s 20th anniversary.

CMS Puts Geo Model on Hold

On March 1, the Center for Medicare and Medicaid Services (CMS) announced its Geographic Direct Contracting Model is “currently under review” until further notice.   While CMS has yet to provide a reason for the pause, criticism from industry stakeholders and the new Center for Medicare and Medicaid Innovation (CMM) Director may explain the pause.

Announced in December 2020, the Geographic Direct Contracting Model, or the “Geo Model,” envisions using direct contracting entities (DCEs) to build integrated relationships with health care providers and coordinate care for Medicare beneficiaries in designated geographic regions.  The model, which had been under development for two years, builds on lessons learned from the Next Generation ACO model, Medicare Advantage, and other initiatives.

However, the Geo Model is the only CMMI model that has been paused by the new Administration so far.   In the absence of an official justification from CMS, one of the reasons why the model may have been put on hold was criticism from health care stakeholders.  In a December 2020 letter to then-CMMI Director Brad Smith, the National Association of ACOs warned the model could generate confusion among beneficiaries over who is compelled to participate.  Additionally, the Center for Medicare Advocacy urged the then-Biden Transition Team in a December 2020 letter to halt the Geo Model due to uncertainty over how the model would work with other forms of insurance such as Medigap.

The move to suspend the Geo Model is not entirely without precedent.  The Biden Administration paused or withdrew numerous actions by the previous Administration since assuming power, which is typical for new presidential administrations.  For instance, CMS has withdrawn proposed rules on oversight for accrediting organizations, revisions to dialysis coverage requirements, and changes to Medicare Part A enrollment requirements since January 20. 

It should be noted that not all stakeholders oppose the Geo Model.  In a March 4 letter, America’s Physician Groups urged CMMI to restart the model as soon as possible due to “extensive financial investments” participating providers have already made in preparation for the model’s launch, as well as a genuine belief that the model represents a meaningful shift away from the fee-for-service model.

In addition to external criticism, pressure to put the Geo Model on hold may have come from within CMMI itself.  Recently, the Biden Administration tapped Liz Fowler to serve as CMMI Director.  Prior to her new leadership role at CMS, Fowler served as Executive Vice President for Programs at the Commonwealth Fund, where she co-authored a December 2020 blog post about the Geo Model.  While the blog post largely served as an explainer for the model, it raised several questions, including how CMS would be able to produce savings CMS had projected and whether CMS is adequately tracking beneficiaries’ use of services under the model.

Thus, pressures from both inside and outside CMS have likely led to the Administration’s decision to suspend the model for the time being.  That said, while not knowing yet where CMMI may take the geographic-based care and payment model, it certainly seems possible that the Administration could choose to pause other CMMI models in the future.