Value-Based Payment As A Tool To Address Excess US Health Spending

Value-Based Payment As A Tool To Address Excess US Health Spending

For decades, the level and growth of US health care spending has diverged from both international and domestic norms, leading many to characterize rising health expenditures as “unsustainable.” Between 1970 and 2019, total US health spending grew from 6.9 percent of gross domestic product (GDP) to 17.7 percent of GDP, according to the Centers for Medicare and Medicaid Services (CMS). In 2020, amid unique strain on the health care system and a dramatic economic downturn due to the COVID-19 pandemic, health spending accounted for nearly one-fifth (19.7 percent) of US GDP. According to prepandemic analysis, health spending was not projected to reach this level until 2028, and it remains to be seen how the pandemic will affect the long-term trajectory of health spending. Meanwhile, the Organization for Economic Cooperation and Development (OECD) estimated that total health spending averaged 8.8 percent of GDP among member countries in 2019 compared with 16.8 percent in the US.

In 2019 Health Affairs launched the nonpartisan Council on Health Care Spending and Value to study excessive health spending in the US and recommend strategies to address it. The council, which plans to release its recommendations in early 2023, defines excessive spending as that which both diverges from a norm and is not commensurate with the health it produces. This research brief is one in a series of briefs that provides snapshots of key literature that informed the council’s inquiry into health spending drivers and interventions.

Why Focus On Value-Based Payment?

Value-based payment is a systemic intervention with the potential to affect all drivers of excess health spending and growth, whether those drivers are prices, administrative waste, clinical waste (including excess volume or inefficient mix of care), or other factors. See previous research briefs in this series for a discussion of several of these spending drivers. Importantly, under value-based payment models, decisions about which spending drivers to address—and how aggressively to address them—are left to the discretion of the entity placed under the spending constraint, such as a state, a health plan, or a delivery system.

Defining Value-Based Payment

In this brief, “value-based payment” is used interchangeably with “Alternative Payment Models” (APMs). These terms refer to a variety of arrangements, all of which are best defined by what they are not: open-ended fee-for-service payments, or straight pay for volume. Value-based payment models can exist at multiple levels within the health care system: the health plan, the delivery system (here used to include institutional providers, such as hospitals, physician groups, or combinations thereof), and the individual clinician. For example, Medicare may impose value-based payment arrangements on private health plans with which it contracts (as in Medicare Advantage), health plans may impose them on delivery systems, and systems may impose them on individual clinicians with whom they contract. This brief, as well as much of the research on value-based payment, focuses on arrangements imposed on delivery systems.

Conceptually, value-based payment models run the gamut from fee-for-service with bonuses for quality to models that are considered “advanced,” such as bundled payments, accountable care organizations (ACOs), and global capitation. They vary in the extent to which they maintain an element of fee-for-service payment and in the degree of clinical and financial risk imposed on the provider. There are many ways of categorizing value-based payment models according to these and other dimensions, but for this brief, we adopt the framework most recently promulgated by the Health Care Payment Learning and Action Network (the LAN) for APMs. This group has developed a widely used framework for categorizing such models (exhibit 1).

EXHIBIT 1 The Alternative Payment Model Framework

Source: Health Care Payment Learning and Action Network (the LAN), 2017, used with permission and adapted for this brief. Note: Subcategories labeled with “N” indicate “no quality considerations.” These models are not considered by the LAN to represent true payment reform and are not tracked as part of measuring the achievement of the LAN’s goals.

In the LAN’s framework, each payment model falls into one of four categories, distinguished by their reliance on fee-for-service, with the degree of clinical or financial risk imposed on the payee layered in as a secondary descriptor. Only categories 3 and 4 are considered advanced and are thus the focus of the LAN’s tracking and technical assistance efforts and of this brief. More precisely, the LAN primarily encourages two-sided risk models, which include some models in category 3 and all models in category 4. Notably, category 3 includes ACOs and bundled or “episode-based” payments, which are two of the most prominent value-based payment models used in Medicare. Category 4 includes models that do not involve any fee-for-service payments, such as global or per person capitation.

More Than Half Of Health Care Payments Are Still Based On Fee-For-Service

Medicare—and to a lesser extent Medicaid—is well known for its use of advanced APMs. This is largely a result of the efforts of the Center for Medicare and Medicaid Innovation (CMMI) within CMS, which has been home to many APMs since 2014. However, CMMI is not the only place to look for examples of these models. Another important source is California’s private sector, in which capitated-delegated medical groups have proliferated. California has approximately 200 medical groups that participate in what is called the “delegated model,” under which they are paid a risk-adjusted per member per month capitation from commercial payers (primarily health maintenance organizations), either for all professional (physician) spending or for the full continuum of care, including hospital and drug spending. Importantly, although these groups are themselves capitated by health plans, they may still pay individual physicians on a fee-for-service basis.

Advanced APMs remain the nondominant form of payment in the US. The most recent analysis by the LAN represented spending by seventy-nine public and private payers, accounting for approximately 80 percent of the covered US population in 2020. Overall, the percentage of all payments in category 3 or 4 APMs has grown steadily since the LAN began measuring it, up from 23 percent in 2015 to approximately 41 percent in 2020. A subset of the latter, only 6.7 percent of all payments was population-based (that is, these payments included no fee-for-service payment at all).

Although the growth in category 3 and 4 APMs is notable, nearly 40 percent of payments in 2020 remained pure fee-for-service—not being tied to any type of quality yardstick—and approximately 20 percent of health care payments were fee-for-service tied to quality or a measure of value.

Medicare Advantage had the highest uptake of category 3 and 4 APMs in 2020, at 58 percent, followed by traditional Medicare at 42.8 percent, Medicaid at 35.4 percent, and commercial payers at 35.5 percent.

Savings Attributable To ACOs Range From Just Under 1 Percent To Just Over 6 Percent

A large body of research has examined savings attributable to one prominent value-based payment model, the ACO. This model was first codified in the Affordable Care Act as a new payment model under Medicare. The model has since spread beyond Medicare to Medicaid and the private sector. The actual details of the arrangements vary considerably on the basis of the payer, who leads the model (hospital versus physicians, or a combination of the two), and a variety of other factors. In general, ACOs are groups of doctors, hospitals, and other health providers who come together voluntarily in a formal organization that is separate from the constituent parts or members to give coordinated high-quality care to their patients. ACOs are incentivized to lower costs for a defined population of patients while also achieving measurable quality improvements.

Several challenges have bedeviled research on savings from ACOs. One is the lack of a true counterfactual (what would have happened in the absence of the ACO) against which to compare ACO-derived savings. Researchers can establish a plausible “counterfactual,” using rigorous quasi-experimental methods. However, many reports of ACO savings compare actual ACO spending with benchmarks representing the targeted amount of savings that program administrators believe a given ACO should be able to achieve. Michael Chernew and colleagues (and others) have argued that these benchmarks are imperfect comparators because they are set according to many confounding factors beyond historical spending and because they do not take geographic variation into account.

A 2018 study of the early ACOs in the Medicare Shared Savings Program by J. Michael McWilliams and colleagues used a rigorously designed counterfactual, finding that for Medicare Shared Savings Program cohorts entering the program in 2012, 2013, and 2014, the average ACO reduced spending by 3.1, 1.4, and 0.4 percent per beneficiary, respectively, by the end of 2015. A key finding was the identification of superior financial performance by physician-led ACOs compared with those led by hospitals.

Exhibit 2 summarizes additional selected studies of ACOs, most with a longer time frame than McWilliams and colleagues, showing modest savings, ranging from under 1 percent to just over 6 percent. Notably, some researchers have found that it takes some time and experience for organizations to make the changes necessary to achieve savings, suggesting that more significant savings might be observed as the programs mature.

EXHIBIT 2 Selected studies of savings from accountable care organizations (ACOs)

Source: Authors’ analysis of selected research. Notes: ^aPercent savings calculated by authors on the basis of savings of $673 per beneficiary per year, reported in the source study, divided by total per beneficiary spending per year of $11,000–$11,500, as shown in exhibit 9 of the source study. ^bPercent savings calculated by authors on the basis of savings of $169 per beneficiary under the Pathways to Success program and $106 per beneficiary under legacy programs, divided by total Medicare (fee-for-service and Medicare Advantage) per person spending in 2019 of $11,582, as report by the Centers for Medicare and Medicaid Services (CMS).

Effects Of Bundled Payments Vary Across Procedures And Patient Populations

Bundled payments are a form of prospective payment in which prices are set on a per diagnosis, per procedure, or per episode basis. Similar to ACOs, the locus of recent bundled payment activity has been at CMMI. The Bundled Payments for Care Improvement (BPCI) initiative of CMMI, and its newer iteration, BPCI Advanced, are the broadest bundled payment programs, in terms of the number of conditions and procedures covered. Results have been mixed, with joint replacement procedures showing the greatest promise for savings. Additional studies of joint replacement bundles, both within and outside the BPCI, have generally, but not universally, found slightly more positive results (see exhibit 3).

EXHIBIT 3 Selected studies of savings from joint replacement bundles

Source: Authors’ analysis of selected research. Notes: CJR is Comprehensive Care for Joint Replacement. MSA is Metropolitan Statistical Area. BPCI is Bundled Payments for Care Improvement.

Results related to other BPCI procedures and conditions have been less positive. Karen Joynt-Maddox and colleagues examined results for the five most common medical bundles in the program (congestive heart failure, pneumonia, chronic obstructive pulmonary disease, sepsis, and myocardial infraction), finding no statistically significant savings related to any of them.

Notably, the impact of bundled payment in commercially insured populations, as opposed to Medicare, has not been well studied. One recent study by Christopher Whaley and colleagues examined a BPCI-like program implemented among self-insured employers for four surgical procedures. After implementation, episode prices for these procedures saw a 10.7 percent relative reduction. This reduction was seen across all providers performing those services for the firms’ employees, not just those participating in the bundles.

Several important considerations should inform interpretation of the research on bundles. First, it is hard to extrapolate from any of these studies. As Michael Chernew and colleagues noted in 2020, “the impact of episode-based payment models may depend on the details of the specific episode.” In other words, as the saying goes, if you’ve seen one bundle—you’ve seen one bundle. Another important caution comes from Elliott Fisher in a JAMA editorial published during the early phases of the BPCI evaluation. Fisher cautioned that even if bundles do begin to significantly reduce payments for specific types of episodes, providers may respond by increasing volume, thus reducing total savings. Finally, a payment landscape in which many or most procedures or conditions were covered by bundles would become administratively cumbersome for all parties and would, at some point, invite the question of whether global capitation for all care would be preferable.

Research On Savings From Capitation Is Limited

Capitation, which the LAN calls “population-based payment,” has been defined in many ways. The American College of Physicians describes capitation as “a fixed amount of money per patient per unit of time paid in advance to the physician for the delivery of health care services.” Of course, capitation is not unique to physicians and can also be applied to other types of providers or to entire delivery systems. Here the focus is on capitation for the full (or nearly full) range of clinical services, as capitation for just a specific procedure or episode is more commonly considered “bundled payment.”

There is surprisingly limited research regarding potential savings from capitation as a payment model in the US, likely because there are so few delivery systems that receive capitation as their primary reimbursement mechanism. In a 2012 survey of twenty-one large multispecialty medical groups, which is arguably the type of physician organization most likely to be capable of receiving capitation, Robert Mechanic and colleagues found that only three groups had no fee-for-service revenue whatsoever. For all twenty-one groups combined, fully or partially capitated revenue accounted for about 30 percent of revenue, on average. Such partial penetration of capitation in any single provider group makes it difficult for researchers to tease out its effects.

The most plentiful evidence on savings from capitation at the delivery system level comes from California. Stephen Shortell and colleagues analyzed data from the Integrated Healthcare Association’s California Regional Health Care Cost and Quality Atlas to compare quality and total cost of care per member per year for California physician organizations, assuming no risk, professional risk only, and full risk for both professional and hospital or facility care. The authors found that average clinical quality scores rose and total casts of care fell with the assumption of more financial risk. The difference in the total cost of care between the no-risk physician organizations and the full-risk organizations is $161 per member per year, or about 3.6 percent higher for the no-risk physician organizations.

Although estimates of saving available from capitation vary considerably, as do the methods used to develop estimates, a few common threads emerge in the writings of experts on this topic. In particular, several factors appear to be crucial to the success of capitation. These include risk adjustment of capitated payments, robust quality measurement, the broadest clinical scope of capitation possible, a large enough patient population to spread risk, and the ability of clinicians themselves to share in savings resulting from the payment model.

Conclusion

The relatively modest results of CMMI’s value-based payment models may be related more to challenges in design and implementation than to the fundamental approach. Notably, there is strong CMMI support for addressing many of the criticisms that experts have identified in Medicare’s rollout of these models, including the voluntary nature of the models, “program fatigue” related to there being too many models available at once, inadequate support for delivery systems to accept downside risk, and technical challenges related to benchmark setting and risk adjustment.

Along with a commitment to address these challenges, CMS has indicated that the agency intends for anyone with Medicare coverage to be under a value-based payment arrangement by 2030. In light of a growing body of research analyzing the ability of value-based payment models to serve traditionally underserved communities and to support improvements related to social needs, CMS has also highlighted ACOs as a potential mechanism for improving health equity and has made health equity an area of focus for other value-based payment programs. To ensure that ACOs are narrowing, rather than exacerbating, disparities, CMS has identified improvements to avoid disadvantaging rural populations and populations with older and sicker individuals.

Acknowledgments And Disclaimers

Briefs are produced by Health Affairs staff based on research conducted to support the Health Affairs Council on Health Care Spending and Value. Each research brief is intended as a snapshot of especially salient published work on a given topic, rather than as a systematic literature review. Health Affairs thanks Elizabeth Keating and Claire Ewing-Nelson for their contributions to this brief and Grace Kim from New York University’s Robert F. Wagner Graduate School of Public Service for her review of a draft.

1669941453