Emily A. Gadbois, PhD; Shayla Durfey, BS; David J. Meyers, MPH; Joan F. Brazier, MS; Brendan O’Connor, BA; Ellen McCreedy, PhD; Terrie Fox Wetle, PhD; and Kali S. Thomas, PhD
Pay for Success (PFS) is a model whereby investors provide capital to fund an intervention to address social, health, or environmental needs.1,2
In particular, PFS allows for initial financing of evidence-based programs to address social risk factors3
for vulnerable groups.4,5
The success of the initiative is evaluated by an independent entity and the investor is repaid by the outcome funder if the predetermined, contracted outcomes are achieved.6
PFS reduces the financial risk of innovation for funders by shifting up-front payment responsibilities to more risk-tolerant investors. PFS also allows greater evidence and data to be collected to inform future programmatic decision making.7
PFS projects have been implemented in 20 countries to date (including the United States, Canada, the United Kingdom, the Netherlands, and several other European countries), targeting housing, behavioral health, foster youth, early childhood education, and chronic disease, among other issues.4
Twenty-two percent of PFS projects are US-based,4
and although these projects are still in progress, early reporting indicates that many are achieving their predetermined outcomes.8-10
Despite the recent growth of PFS initiatives in public health and their potential to address the historic underfunding of programs to target social risk factors in the United States,11
there has been limited adoption of these initiatives in the healthcare sector. One area of healthcare that could potentially benefit from PFS is the Medicare Advantage (MA) program. MA is the fastest-growing segment of the Medicare market,11
now covering more than one-third of all Medicare beneficiaries.12
MA plans are paid on a capitated basis, receiving a set annual amount to cover the comprehensive needs of their members each year. As a result, MA plans may have an increased incentive to improve outcomes and reduce costs. Unlike traditional Medicare, MA plans may offer supplemental benefits outside of standard medical care, including fitness memberships, vision care, nursing hotlines, and case management. The Creating High-Quality Results and Outcomes Necessary to Improve Chronic (CHRONIC) Care Act passed in 2018 expanded the definition of supplemental benefits to include nonmedical services (eg, caregiver support, in-home supportive services, adult day care) and newly allows plans to target these benefits to specific populations beginning in 2020.13,14
This new flexibility for supplemental benefits could be instrumental in addressing upstream social risk factors and presents an opportunity to consider the ways in which MA plans may use PFS as a financial risk mitigation tool to experiment with offering new services and benefits.
PFS may provide MA plans a valuable opportunity to offer expanded benefits and services to their members without assuming additional financial risk. In a PFS model, MA plans could test innovative programming that addresses nonmedical needs, and if program outcomes are successfully achieved, the MA plan could then incorporate those services into their benefits packages. Additionally, PFS projects with well-defined, targeted outcomes measures would contribute to the creation of a larger evidence base to further support the value of addressing nonmedical needs in this population. Despite this potential, little is known about MA plan leaders’ perceptions or awareness of PFS. The purpose of this research was to understand MA plan leaders’ interest in uptake of PFS and the barriers and opportunities that they see to PFS adoption.
This research was an exploratory aim of a larger project that sought to understand the experiences of 1 MA plan, henceforth referred to as plan A, that had considered implementing a PFS project. For more information, see the recent study by several of this paper’s authors.14
We conducted a preliminary interview with representatives of plan A with the goal of understanding the opportunities and barriers to implementing PFS in the MA environment. We then interviewed representatives of 16 additional MA plans in an effort to see if they shared the perspectives revealed by representatives of plan A about the attractiveness of PFS and potential barriers to implementing a PFS project.
We conducted semistructured interviews with representatives from 17 MA plans around the country that represent more than 65% of MA beneficiaries nationally. Following our preliminary interview with the representatives of plan A, we began with a convenience sample of known MA contacts and solicited contact information for representatives of other plans upon completion of each interview. We continued to recruit representatives of plans with varying characteristics (ie, plan enrollment, star rating, age of organization) and geographic locations until saturation was achieved.15
For each plan, we asked to speak with those knowledgeable about the interview topics, and representatives self-identified.
Based on our preliminary interview with representatives from plan A, we designed semistructured interviews to understand MA plan representatives’ knowledge of and receptivity to PFS initiatives. We drafted the interview guide (eAppendix
[available at ajmc.com
]) and reviewed it among the project team and advisors, including research and industry experts in PFS and MA. In advance of each interview, representatives were emailed the interview guide and information about PFS. This was done to ensure that the most appropriate leaders from each plan participated in the interview and that operational definitions when discussing PFS were universally understood. Interviews were conducted over the phone and recorded (with representatives’ consent), and lasted about 1 hour. This project did not require institutional review board review as it was deemed not to be human subjects research by the Brown University Institutional Review Board.
Interview transcripts were qualitatively analyzed using a modified grounded theory approach to identify overarching concepts and themes.16-19
We first developed a preliminary coding scheme based on the questions asked in our interview protocol, then adjusted the scheme in an iterative process to add codes and refine code definitions. Initially, all qualitative analysis team members (4 of this paper’s authors: E.A.G., T.F.W., E.M., J.F.B.) read and individually coded the first 2 transcripts. In subsequent team meetings, we discussed and refined the coding scheme and code definitions according to how well the codes fit the transcript data. We discussed preliminary patterns that we perceived in the data and reconciled our interpretations of the first coded transcripts. We continued this process, with at least 2 team members coding each transcript independently, then meeting to reconcile codes, discuss potential themes, track prevalence of these themes across transcripts, and search for competing interpretations. During analysis, we kept a comprehensive audit trail that recorded ongoing team decisions, including selection and definitions of codes and discussion of emerging themes.17,20-23
Coded data were entered into Nvivo (QSR International; Melbourne, Australia) to allow for comparative and relational analyses across themes.
Plan and Representative Characteristics
The 17 participating plans varied in their geographic coverage, enrollment, star rating, and organizational age. Plan characteristics are provided in Table 1
. Interviews were conducted with 1 to 6 participants per plan for a total of 38 representative participants. Representatives included those in upper-level management roles (eg, president, chief medical officer, vice president of medical affairs, director of health policy) who had been in their positions between 1 and 30 years.
Findings From the Interview With Plan A
Results from the preliminary interview with representatives of plan A revealed nuanced perspectives regarding PFS. Plan A made efforts to engage in a PFS model to address members’ nonmedical needs. Initially, they met with a community-based organization about providing an intervention to target food insecurity among their members. Representatives from plan A indicated that they “like the Pay for Success model” because it “mitigates [their] risk” in testing innovative services. Ultimately, however, plan A decided not to pursue the PFS approach, instead opting for a more familiar model (plan–vendor contract). Representatives from plan A noted that a major barrier to the PFS model was that the cost of a new benefit might potentially be incorporated into premiums and borne by the plan’s members, before it is known if the outcomes will be achieved: “In a [PFS] model, we don’t have confidence that if it’s not successful, then the cost to [the member] will be zero. And yet, we’ve already priced something into the plan for the end patient to pay. If we pay nothing and they have to pay something, the government’s not going to like that” (plan A).
Representatives from plan A also described hesitancy with regard to sharing data with project partners and having outcomes evaluated by an independent third party. Although such sharing is often a component of plan–vendor contracts, this was highlighted as a potential barrier to PFS. Additionally, representatives from plan A described that although they perceive PFS to be potentially valuable, they would like more of a “blueprint” for how it could be implemented before they engage in it themselves. Plan A representatives further discussed the concept of risk aversion and how uptake of PFS might be limited by the fear of “jeopardiz[ing]” their relationship with CMS—they were concerned about operating within CMS’ existing parameters. For representative quotes, see Table 2
Findings From the Interviews With Additional MA Plans
The interviews carried out with representatives from 16 other MA plans revealed that representatives were largely unfamiliar with PFS compared with those from plan A, although they were receptive to further exploring its potential value. When probed about specific components of PFS, including willingness to share data and taking on the risk associated with an alternative payment model, representatives had varying responses. These themes are discussed here.
Awareness of and receptivity to PFS.
Representatives had a range of familiarity with PFS. Some had never heard of PFS or first discovered it as a result of our sending introductory materials for the interview, some had “read about it” but did not have “personal experience” (plan 1), and others were more familiar with PFS, indicating that PFS had been discussed at recent board meetings (plan 7) or that they were collaborating with consultant groups to think through the possibility of implementation (plan 16). Representatives recognized the potential of PFS to address members’ social risk and mitigate the financial risk of a new intervention. They suggested that PFS was “right up our alley” (plan 13) and that it “fills a gap” (plan 3). Representatives also expressed interest in receiving more information about this payment mechanism (plan 1). Other representatives highlighted potential barriers to PFS, including a desire for additional evidence. Representatives were hesitant to be among the first to implement PFS (plan 8) and expressed concerns about both incorporating this model “within the confines of CMS” (plan 11) and achieving a balance of risks and rewards (plan 9). For representative quotes, see Table 3
Receptivity to data sharing and working with an independent evaluator.
Representatives discussed their willingness to share data with external partners, as well as their receptivity to having outcomes assessed by an independent evaluator. Representatives described their plans as being “very open and hav[ing] a desire to continually evaluate the efficacy of our work” (plan 11), but they also highlighted that they must “know at the beginning” of the relationship with service providers and evaluators what they will be collecting and measuring and “build that into the work effort” (plan 7). Representatives also highlighted the need to be confident that they are working with “appropriate partners” (plan 5). Such representatives were also willing to work with an independent evaluator for the purposes of assessing whether or not predefined outcomes were achieved. Some representatives reported working with third-party evaluators “many times in the past” (plan 6) and employing “external partners” to assess outcomes (plan 16). However, other representatives expressed hesitancy to share data with outside evaluators, preferring instead to conduct evaluations in-house (eg, plans 3, 4, and 9), or they highlighted barriers to sharing data outside of their organization. Some representatives described their plans as “very hesitant to share our data outside of our organization” (plan 8), citing “a really high bar…dictated by federal and state law” (plan 2), as well as the competitive environment in which they work (plan 1). See Table 4
for representative quotes.
Willingness to innovate/test new services.
When asked about their willingness to consider alternative payment models like PFS, representatives responded by commenting more broadly on their willingness to innovate. Some representatives were very open to testing new services and programs. These were representatives of plans that tended to be smaller and less established in their markets, and they seemed to view innovation as a way to differentiate themselves from more well-known, larger plans. Such representatives described their plans as having the need to pilot interventions “in our DNA” (plan 12), liking to “test and learn” (plan 13), viewing “excellence in innovation” as a “pillar” of their plan (plan 9), and being a “giant intervention machine” (plan 5). These plans often had informal methods and acceptance of quickly testing innovative ideas. Plans that were more risk averse tended to be larger, with more formal, established systems around innovation and the testing of new services or programs. Such representatives described requiring “extensive research” examining “what’s already been proven in the literature” (plan 16), wanting to be “pretty sure of the outcome that we want” before committing resources (plan 3), desiring to “build on already-gathered evidence” (plan 1), having thorough processes in place for vetting new ideas (plan 4), and the importance of ensuring return on investment (plan 15). For representative quotes, see Table 5
The purpose of this research was to understand MA plan representatives’ perspectives and interest in PFS as a mechanism to develop new initiatives targeting members’ social risk factors. MA plan representatives in this study were largely unfamiliar with PFS and were generally interested in learning more. Similarly, those who were familiar with PFS expressed receptivity to exploring it further. Although some representatives reported willingness to share data and measures with project partners and to work with an independent evaluator to assess if predetermined outcomes had been met for the purposes of repayment, others were more hesitant and voiced concerns or expressed their preference to validate analyses internally. Lastly, although most representatives described a mission of innovation and goals of piloting new programs and services, some were more risk averse and described preferring to use tried-and-true methods to deliver new programs and services.
With the passage of the CHRONIC Care Act and recent industry changes,13
managed care organizations, like MA, are increasingly looking to address members’ social needs in addition to their medical needs.24
The findings of this study align with those of previous work that has also found that although MA plans may be interested in expanding the types of services they are offering, hesitation about regulation and the true potential of programs that address social needs remains.14
In 2019, when MA plans were first granted flexibility in offering new supplemental benefits, only 12.7% of plans offered any newly available service.25
PFS could be one way to increase uptake of new supplemental benefit offerings in future contract years, as it may help mitigate plan financial risk.
However, in addition to the barriers revealed in our interviews, a number of other barriers may be potentially problematic in employing PFS in MA. First, as highlighted by representatives of plan A, CMS may not allow a plan to layer the costs of an intervention onto members’ premiums, costs that they will be required to pay whether or not the plan ultimately achieves the intervention’s anticipated outcomes. Further regulatory guidance from CMS is needed to help plans understand the extent to which this arrangement would be allowable. Second, many PFS projects are multiyear, whereas MA contracts are annual. This introduces risk to the investors if the plan changes vendors, decides not to rebid, or is not awarded the contract from CMS, or if other regulatory changes disrupt the project. More work is needed to understand the degree to which plan reserves or other funding pools can be tapped to repay investors in the case that projects are prematurely shut down due to contracting or CMS policy changes. Third, if an MA bid is submitted in June, accepted in October, and expected to be deployed on January 1, but the project is unable to secure an investment to fund the service delivery, the service provider, typically a community-based organization (CBO) partner, will still be expected to provide the services, but without the capital required to do so effectively. To mitigate this risk, plans should consider how investors could be conditionally secured before CMS approval of the bid, or ways in which MA plans can act as a bridge financing vehicle for performance-based contracts with CBOs that require startup and bridge funding.
In the standard PFS model in which investor repayment flows from the plan back to the investor through a special purpose vehicle, a legal entity created for a limited business transaction, there may be regulatory barriers to the adoption of PFS programs by MA plans. Guidance from CMS is needed on how this cash flow model may or may not align with regulations. This may be further complicated in Dual Eligible Special Needs Plans and Medicare–Medicaid plans within MA, both of which are additionally subject to state Medicaid regulation. However, more traditional forms of performance-based contracting between MA plans and providers could be applied to CBOs, opening up the ability of CBOs to partner with plans and participate in risk sharing. It is likely that variations of the original PFS model will be needed to give plans confidence that they are structuring projects in a compliant manner. Future studies could explore how plans that utilize value-based contracting apply similar models to nonprovider vendors and CBOs. With the passage of the CHRONIC Care Act13
and the expansion of the MA Value-Based Payment Insurance Design pilot,26
MA plans will be given more flexibility in addressing social risk factors. Plans may want to test the provision of services that address social risk factors, and PFS may allow them to do so with less financial risk. In order for PFS to be a viable model, however, further guidance from CMS will likely be needed to assuage these regulatory concerns.
This research had several limitations. Because we began with a convenience sample, we may have initially selected plans known to our organization that were proactively working on alternative payment models or innovative practices. Additionally, a convenience sample means that our findings may not be generalizable to the entire set of MA plans offered in the United States, despite our sample representing 65% of MA beneficiaries nationally. Our snowball strategy added plans with which we were not familiar, and although we selected plans of varying sizes, geographic locations, and quality, participating plans might differ from plans that did not participate. For example, plans or representatives with a particular interest in PFS may have self-selected to participate in our study. Nonetheless, results from this research provide initial evidence of how plans are approaching innovative alternative financing models like PFS.
Our results indicate that MA plan representatives may be interested in PFS as an option for expanding their supplemental benefits offerings and that some may be more willing or able to address challenges with PFS than others. Allowing PFS programs to be disconnected from member premiums may help plans incorporate PFS into the complex cost structure in MA as long as concerns about the regulatory environment can be addressed. Future work is needed to better understand the types of PFS arrangements that may be most efficacious to MA plans and their members. As the healthcare system continues to evolve to address upstream health issues, PFS may have the ability to impact the lives of Medicare beneficiaries.