It’s Time to Advance Payment Reform Using the Principle of Policy Equipoise

We are at a critical juncture in value-based payment reform. In the decade since lawmakers created the Center for Medicare and Medicaid Innovation, we have seen the benefits of some new payment arrangements, but not others. To ensure that these results inform future policy agendas, policy makers can apply the principle of policy equipoise to existing evidence. 1 Policy equipoise will help prioritize which payment models to test and when to do it using randomized designs to generate rigorous evidence.


Pay For Performance: Absence of Policy Equipoise
For decades, pay for performance (P4P) has been a cornerstone of payment reform.Numerous health care providers and insurers have used these arrangements to measure and financially reward (eg, via bonuses) or penalize (eg, via salary withholdings) providers based on performance on quality or utilization metrics.As a prominent example, Medicare continues to feature P4P at a national scale via the Merit-based Incentive Payment System (MIPS).
Unfortunately, strong data supporting P4P are lacking.Under the Hospital Value-Based Purchasing Program, the payment arrangement did not improve a range of outcomes-an overall conclusion underscored by other inpatient and ambulatory programs. 3,4Though the effect of P4P can vary by context, the preponderance of data argues against the benefit of P4P in many settings.
Given this absence of policy equipoise, policy makers can pursue several alternatives.First, they could test P4P in new settings where there is equipoise.For instance, there is less evidence for using P4P in specialty care, and more evaluations would clarify the benefit of P4P in this setting.Take cancer care as an example.Although early studies evaluating P4P in cancer care did not show expected results, a recent examination of the largest national cancer P4P program showed increased evidence-based prescribing. 5Second, policy makers could redesign P4P around the elements that are most likely to improve care.Currently, programs such as MIPS aggregate performance across a number of specific care processes-in the context of MIPS, across quality, electronic health record use, cost, and improvement activities.
Future initiatives could involve creating programs that focus on specific elements that have not been well evaluated under the P4P construct, such as use of electronic health records to engage patients in self-care and self-management.Short of exploring these strategies, policy makers would be better served by sunsetting P4P in well-studied settings, such as hospitals, and by redirecting resources toward others.

Population-Based Payment: Presence of Policy Equipoise
In contrast, policy equipoise applies to comprehensive population-based payment models, such as accountable care organizations (ACOs), which require providers to assume accountability for the quality and costs of care across discrete patient populations.Despite evidence showing benefits for individual programs, the overall evidence for the ACO model remains mixed, creating continued uncertainty about the effect of ACOs compared with existing payment arrangements. 6e presence of policy equipoise suggests that policy makers should continue testing population-based models (eg, ACOs and direct contracting) against fee-for-service and should do so in view of other payment models (eg, bundled payments) given the potential for providers to coparticipate in multiple models. 7Another related and potentially promising solution would be to test population-based payments in combination with other strategies.For instance, as we have begun to do at the Value & Systems Science Lab and Payment Insights Team, policy makers could compare the impact of population-based vs combined (population-based and episode-based) incentives on patient outcomes.
Policy equipoise can also help policy makers prioritize randomized trials.Randomization can provide strong evidence that overcomes issues that have plagued policy evaluations, such as insufficient statistical power, lack of valid comparison groups, and limited adjustment for confounding.By doing so, randomization can help policy makers navigate uncertainty about the benefits and harms of different payment incentives and minimize time spent on ineffective policy; large-scale randomized trials may have helped obviate years of ineffective P4P policy.However, because RCTs are also time consuming and costly, policy equipoise can clarify where the bang justifies the buck.

Implementing Randomized Payment Model Experiments
In the context of payment models, a key challenge arises from voluntary participation, ie, the ability for providers to voluntarily enroll in payment programs.This approach may be desirable for multiple reasons, eg, encouraging participation among organizations most prepared for and likely to succeed in payment reform.
However, voluntary participation also increases the risk for selection and cherry-picking-ie, differences in the types of providers that decide to participate and, upon participating, differences in how they care for patients.The implications can be considerable.For instance, we previously demonstrated that nearly one-third of observed cost savings under voluntary joint replacement bundled payments may be due to selection rather than true program effect. 8quiring participation, such as Medicare has done in its mandatory joint replacement program, can counterbalance these issues.But this approach also has limitations that must be weighed against voluntary participation. 9One alternative would be for policy makers to encourage providers to voluntarily enroll in a payment model before randomizing that group to different incentives or even randomize regions to be offered the voluntary program or not.
To be fair, well-conducted randomized trials are not costless, requiring policy makers to provide adequate administrative support and partner with researchers on trial design choices.Nonetheless, we should strive to use policy equipoise to set our payment agenda-efforts that would help our country effectively reboot value-based payment after COVID-19.