Published November 6, 2014
Since enactment of the Affordable Care Act (ACA) in 2010, much of the attention in the policy community has been on modernizing Medicare’s traditional fee-for-service (FFS) program. Through Accountable Care Organizations (ACOs), larger “bundles” of payments to fee-for-service providers for episodes of care, and tests of pay-for-performance models, the hope is that the traditional Medicare model can be remade through sheer force of bureaucratic will. The stated intent is to find a way to pay for value, not volume.
These efforts may or may not bear much fruit, but, over the longer term, it’s not likely to matter much. That’s because a more important transformation of Medicare is already well underway and is occurring despite more resistance than assistance from the program’s bureaucracy. According to the 2014 Medicare Trustees’ report, enrollment in Medicare Advantage – the private plan option in Medicare — has been surging for a decade. In 2005 there were 5.8 million Medicare beneficiaries enrolled in MA plans — 13.6 percent of total enrollment in the program. Today, there are 16.2 million beneficiaries in MA plans, or 30 percent of program enrollment. (See Table IV.C1) In addition, the Medicare drug benefit, which constitutes about 12 percent of total program spending, is delivered entirely through private plans. (See Table II.B1)
As MA enrollment has surged, so has recognition of its improved value. A recent, comprehensive review of the evidence conducted by Joseph Newhouse and Thomas McGuire of Harvard Universitymakes a compelling case that MA plans are providing higher value services at less societal cost than the traditional FFS program. Based on their findings, Newhouse and McGuire argue for policies that would provide incentives for even more beneficiaries to enroll in MA plans in the future.
The Newhouse-McGuire study is part of a notable transformation in views on the MA program. For many years, private plans were heavily criticized for costing too much while providing little additional value to beneficiaries. But those criticisms are now beginning to recede as the evidence mounts that MA plans can deliver more efficient and higher quality care than FFS. And as valid criticisms fade in relevance so too do the arguments against using MA as a foundation for a larger reform of the Medicare program.
The Relative Efficiency of MA Plans
There has been a long-running debate about the relative costs of MA plans compared to traditional FFS. That debate has often confused what Medicare pays MA plans with whether or not MA plans can deliver the same level of Medicare benefits less expensively than traditional FFS. A growing number of studies confirm that, in general, MA plans can operate more efficiently than FFS (and sometimes much more efficiently), even if the government’s payments to MA plans on behalf of the beneficiaries often exceed what would be spent if the beneficiaries were enrolled in FFS.
Recent data compiled by the Medicare Payment Advisory Commission (MedPAC) confirms the relative efficiency of MA plans compared to FFS. In 2014, MA plans of all types submitted premium bids to the Centers for Medicare and Medicaid Services (CMS) that came in at 98 percent of FFS costs. In other words, MA plans are able to provide the Medicare benefit package, as defined in the statute, for two percent less than what it costs FFS. When only MA HMOs are examined, the cost savings is 5 percent. (See Chart 9-6 and Table 1 below)
Of course, Medicare pays the MA plans based on benchmarks, which are set at levels that are generally above the MA bids. Since 2004, MA plans have been paid what they bid, plus 75 percent of the difference between what they bid and the benchmarks – if their bid is below the applicable benchmark. If a plan bids above its benchmark amount, the beneficiaries are required to pay the additional premium to make up the difference. More recently, this split in the share of the difference received by lower-bidding plans has been adjusted in the ACA based on quality ratings assigned to MA plans (See page 329); beginning in 2014, the rebated amount will be between 50 percent and 70 percent of this difference, with the highest rated plans receiving a 70 percent rebate. MA plans are also required to return the added payment above the bids to the beneficiaries, either in the form of benefits that go beyond what is required in Medicare law or through lower cost-sharing requirements, such as reduced premiums, deductibles, or copayments.
The relative efficiency of MA plans is highlighted in several recent assessments of the Medicare premium support model that find it would reduce program costs, and could reduce most beneficiaries’ premiums as well. (Various premium support reform proposals would create direct competition between MA plans and FFS by setting the government’s contribution toward coverage based on their respective bids and requiring the beneficiaries to pay higher premiums for more expensive coverage.) For example, the Congressional Budget Office (CBO) estimates that MA plans, even without a change in law, will be able to outperform FFS in 2020, with average bids that are 6 percent lower than average FFS spending.
CBO expects MA plan bids to fall by an additional 4 percent under a switch to a premium support model, due in part to the intensified competition the reform would bring about. (See pages 36-37) The CBO analysis projects that MA plan bids will be an additional four percent below the average FFS spending for a beneficiary in 2020 — under either an average-bid or second-lowest bid option for premium support.
The growing differential between MA plan costs and FFS would result in program-wide cost reductions that could be shared with Medicare’s beneficiaries. CBO expects that a premium support plan which specifies that the government’s contribution will be tied to the weighted average bid in a region, would cut Medicare spending by about 4 percent and beneficiary spending on health expenses by about 6 percent.
CBO’s findings are echoed in other assessments of the premium support model, including assessments by those generally critical of the concept. Zirui Song, David Cutler, and Michael Chernew wrote an assessment of a similar Medicare reform plan in 2012 that argued that premium support would result in higher premiums in FFS, and thus would be harmful to beneficiaries. That assessment, though critical, was also based on the assumption that MA plans would, in many parts of the country, provide the Medicare benefit package at costs well below FFS. Specifically, the authors of this study estimated that the second-lowest MA plan bid in a region would, on average, be about 9 percent below FFS costs. Consequently, beneficiaries who elected to stay in the more expensive FFS option would pay more.
The study presumed that many millions of beneficiaries would be resistant to switching out of FFS, even if staying in FFS costs them much more each month in premiums, but that appears to stretch the limits of plausibility. More importantly, this study supports the conclusion that, in many parts of the country, MA plans will be able to bid far below projected FFS costs.
The Spillover Benefits of MA Enrollment
The positive impact of private plans is not limited just to what is delivered to MA plan enrollees. Several studies continue to show that higher MA enrollment in a community is beneficial for all Medicare beneficiaries, including the non-MA participants.
A 2008 study by Michael Chernew, Philip DeCicca, and Robert Town found that, for every 1 percent increase in MA enrollment in a community, there was a corresponding 0.9 percent reduction in FFS spending per enrollee. Another study in 2013 by Katherine Baicker, Michael Chernew, and Jacob Robbins found that when MA penetration rates rose in a county, costs per hospital admission and average lengths of stay fell system-wide, including a spillover reduction in FFS hospital costs and treatment intensity.
The implication of these studies is that higher MA plan penetration has a positive effect on the practice patterns among the physicians participating in the MA plan, as well as the hospitals with which they work, and that these physicians carry these improved practice patterns over into their care for FFS enrollees as well.
Minimizing Risk Selection
Newhouse and McGuire devote significant attention to the question of risk adjustment. For many years, critics of private-plan risk contracting in Medicare have argued, with some evidence, that the private plans benefitted from favorable selection and a rudimentary risk adjustment methodology. Until 2000, payments to the private plans were only adjusted based on the age, sex, and the eligibility categories of the beneficiaries, not their underlying health conditions.
Beginning that year, however, CMS administrators started to phase in additional payment adjustments to MA plans based on inpatient diagnoses discernible from FFS claims data and MA plans’ “encounter” data. In 2004, CMS introduced a more comprehensive risk-adjustment model based on physician and outpatient claims data from the FFS population as well. Since 2007, risk scores for MA plans have been based entirely on this model. These scores are used to adjust the government’s capitated “benchmark” payments to MA plans to reflect the risk scores of their enrollees.
In addition, legislative amendments enacted in 2003 have phased in a “lock in” feature for MA plans that limits the ability of beneficiaries to disenroll between annual open enrollment seasons. (Previously, beneficiaries could disenroll from MA plans on a monthly basis, and move back and forth between FFS and MA as often as they wished.) Having reviewed the most recent data, Newhouse and McGuire conclude that these two features of current MA policy have greatly reduced favorable risk selection for the participating private plans, to the point where accusations of skewed selection are no longer a strong argument against the efficacy of the program.
A 2011 study by Jason Brown, Mark Duggan, Ilyana Kuziemko, and William Woolston challenges this conclusion by contending that risk selection has now moved from avoiding those with expensive health conditions to finding the least costly patients within a risk adjustment category. In other words, their study suggests that the private plans have ways of finding the least costly diabetic among those who get coded with diabetes in the new risk adjustment system. Those enrollees would appear to be cheaper to care for than their official risk scores predict.
The authors estimate that the amount by which MA private plans were paid above the cost of covering them directly through FFS — so-called “differential payments” — increased after risk adjustment and totaled $30 billion (almost eight percent of total Medicare expenditures) in 2006.
Newhouse and McGuire (See pages 373-375) make a couple of points to refute this argument. First, they conducted their own analysis and found that the distribution of high-cost and low-cost cases within a risk adjustment code had not changed substantially from the years preceding introduction of the new inpatient-and-outpatient-diagnosis-based risk adjustment method. In fact, the amount of risk bias favoring the plans had fallen, rather than risen. Therefore, whatever is happening is not the result of adjustments in plan behavior since introduction of the new method.
In addition, they note that there is no evidence of favorable selection at all across the risk adjustment coding system, such as from Medicare beneficiaries with cancer to those with diabetes. They point out that it would be far more profitable for an insurance plan to employ methods to skew their enrollment in this way than to try and find the least costly patients within a diagnosis code. And yet there is no evidence this is happening.
There are also more benign explanations for within-diagnosis risk selection. Some beneficiaries in MA plans may disenroll temporarily in order to get a procedure done out of their MA plan network. It is also possible, and perhaps likely, that the MA plans are more aggressively and efficiently managing the care of their more expensive cases, and that is showing up as “risk selection” in after-the-fact analyses that assume FFS and MA are no different in this regard. The Brown et al. model assumes away the possibility that MA plans might be better at controlling costs than FFS; nor does it account for the effects of insurers identifying the enrollees for whom their managed care practices would be particularly effective at reducing costs below FFS.
The stylized approach by Brown et al. only looks at beneficiaries switching between FFS and MA, but not enrollees choosing MA in their initial year of Medicare eligibility. However, in more recent years,younger Medicare beneficiaries are more likely than older ones to enroll in MA plans. They also appear to be coming at random from the entire risk distribution. (See page 379) Not surprisingly, any efforts by MA plans to selectively enroll the healthiest beneficiaries will become increasingly difficult as their larger plan populations grow inherently more representative of all Medicare eligibles.
In its assessment of the “premium support” reform model, CBO notes that MA plans would be able to bid well below FFS costs in many counties around the country, and that improved efficiency among MA plans in a more intensively competitive environment would be one of the reasons for the lower private plan bids. The agency states that the lower bids would also be due to some additional selection bias – that is, the risk adjustment methodology would not fully correct for the healthier enrollees that MA plans would attract in a direct competition with FFS. But CBO does not provide a precise estimate of what it expects the additional selection bias to be, so it is possible that the estimated effect is small.
Moreover, under premium support, MA enrollment would increase to an even larger percentage of total program participation, thus making it more difficult for MA plans to serve a beneficiary population that is atypical of the Medicare population in general. In addition, there is no reason to believe that the substantial progress that has been made in recent years in reducing selection bias in the MA program, through policy changes as well as better data, would halt if the program moved toward the premium support model.
Higher Value Services in MA Plans
In the past, MA plans and their predecessors were often dismissed for not providing any added value for the beneficiaries. This was always more of an assertion than a proven fact, given the paucity of clear quality comparisons between MA and FFS. Indeed, one of the great ironies of the MA-FFS debate is that it is far easier to secure measures of quality for plans with organized networks of providers than it is from the unaffiliated and disorganized providers receiving FFS payments. Moreover, given the documented benefits of well-organized and integrated systems, there was never a very good empirical basis to assume that the beneficiaries were better off in FFS than MA.
In fact, more evidence has begun to emerge in recent years indicating that MA plans can and do provide higher quality care than FFS in many instances. A 2013 study in Health Affairs found that MA HMOs outperformed FFS on breast cancer screening, diabetic care, and cholesterol testing for heart disease. A separate study, in the American Journal of Managed Care, found that MA plans have hospital readmission rates that are 13 to 20 percent below those of FFS.
Several recent studies have attempted to debunk the value of MA plans to beneficiaries by arguing that the plans have failed to pass through to the beneficiaries a sufficient amount of the payments that are made to them above their bid amounts. For example, a 2014 study by Mark Duggan, Amanda Starc, and Boris Vabson claims that only about one-fifth of the additional payments provided to MA plans find their way to the enrollees in the form of “better coverage.” The implication is that a larger share accrues to private insurers in the form of higher profits and possibly a large increase in advertising expenditures to attract MA enrollees.
The study relies on a 2001 Medicare payment policy change that raised the minimum payment floor in larger urban counties. It focuses on the effects of those payment hikes during the 2007-2011 period. The study uses a process of elimination (little evidence of much lower premiums, substantial additional medical benefits, higher MA plan ratings, greater utilization of services, better health outcomes, changes in beneficiary selection) to imply that only a small share of the increased payment levels were passed on to consumers.
This study draws a broad policy conclusion — that MA plans are not providing substantial additional services to their enrollees for the higher payments they are receiving — based on an inexact methodology and questionable assumptions. More importantly, the premise is flawed. Even if accurate, the study is only a critique of payment adjustments that were made for certain counties (low-cost urban areas) for political reasons. There is no basis to connect this with the dynamics of a genuine premium support approach based on competitive bidding among all Medicare options, including MA plans and FFS.
Another 2012 study by Zirui Song, Mary Beth Landrum, and Michael Chernew similarly purports to demonstrate the limited benefits to consumers from competitive bidding and competition in Medicare. It finds that a $1 increase in Medicare’s payment to private HMO plans led to a 49 cent increase in plan bids, with 34 cents going to beneficiaries in the form of extra benefits or lower cost sharing. It adopts the rather unrealistic assumption that MA plans in a “simple” model of competition should not raise their bids when Medicare raises its payments to those plans, but instead should transfer the bulk of those extra payment amounts into expanded benefits for the plans’ enrollees.
The study uses a rebasing of MA benchmark payments in 2007 and 2009 as an experiment in changes in levels of benchmark reimbursement for plans that were largely uncorrelated with changes in their underlying costs in any specific county. Its analysis is limited to MA local plans, does not include employer plans or special needs plans, involved data from 2006 to 2010, and does not observe actual plan costs. It could not observe the impact of more recent Medicare payment changes.
While claiming to show the limited value of MA plans to consumers, what the unremarkable study actually shows is that competition in the Medicare market is not perfectly competitive under current rules and that bidding systems in which administratively determined benchmarks are known in advance are less effective than competitively determined benchmarks at revealing true costs. Administratively-determined benchmarks may even lead to higher provider fees. The authors acknowledge that administratively set prices have their own flaws, including political manipulation that may lead either to excessive spending or less access to care. This study, like the earlier one by Duggan et al., failed to address how a Medicare premium support reform based on competitive bidding across a level playing field for all Medicare options would actually perform.
A third study in 2009 by Steven Pizer, Austin Frakt, and Roger Feldman confirms this point. It is sometimes cited for the proposition that MA plans return little value to consumers when they receive higher payments from the Medicare program. But, among other things, the study found that the Medicare drug benefit structure for paying private plans was far more effective than the MA program in terms of delivering value to customers. The authors estimate that the drug benefit produced nine times as much value per government dollar for the beneficiaries compared to an increase in payments to MA HMOs. Why? Because the authors conclude that the solution is to more closely tie MA payments to “competitive bids,” rather than to administratively determined benchmark rates.
Improving Competition Between MA and FFS
The Newhouse and McGuire assessment of the state of MA policy is particularly interesting in its look at the playing field for MA-FFS competition.
For starters, they make it clear that default options matter. Under current policy, if beneficiaries make no overt choices, then they are presumed to select FFS. Further, like many other markets, Medicare displays “status quo” selection bias. That is, once a beneficiary is in a plan, he or she tends to stay there, even when switching would make sense. So the current system, with FFS as default coverage, is biased toward more FFS enrollment.
A possible remedy would be to change Medicare’s default rules. For newly eligible beneficiaries who do not make an overt selection of coverage, the Medicare program could provide random assignment among MA plan options instead of automatically placing them in FFS. To minimize undue financial hardship and unexpected surprises, those default options might be limited to the two low-cost Medicare plan options (under one version of competitive bidding) or to those plans with bids equal to or less than the average enrollment-weighted bid in a particular county (under another version of competitive bidding). This change would not apply to current enrollees in FFS.
Newhouse and McGuire also make it clear that the current bidding system is flawed and needs improvement. All beneficiaries, including those in MA plans, must pay the Medicare part B premium, which is generally withheld from the amounts otherwise due to the beneficiaries in their Social Security checks. MA plans are permitted to provide premium rebates to the beneficiaries as a way of attracting enrollment, but current policy requires those rebates to come in the form of adjusting the part B premium withheld from Social Security checks. This is a very non-transparent way of encouraging direct price competition between FFS and MA because a beneficiary choosing a plan offering a rebate would not see the change in any bill they owe to the MA plan. The adjustment would come in form of an adjustment to their net Social Security benefit, which is indirect and less visible. The result is that very few MA plans compete with FFS in this way; instead, they charge no premium above the part B premium and then give away whatever else they can in the form of expanded benefits, including adjustments in deductibles and other copayments.
This limitation of the current, flawed price competition between MA and FFS can also be seen from the perspective of what the beneficiaries must pay to remain in FFS. Under current law, the premium for FFS is always the uniform, national part B premium, regardless of the relative cost of FFS to the available MA plans in the region.
The solution is to ensure clear, transparent price competition between MA and FFS that is visible and tangible for the beneficiaries during annual open enrollment seasons. The most straightforward approach would be to base the government’s contribution toward coverage on the average, enrollment-weighted bid in a region, with measured FFS costs included in the calculation. Under this approach, the government’s contribution toward coverage in a region would be the total, weighted-average bid less the part B premium that beneficiaries would pay under current rules for their FFS coverage. Beneficiaries would pay the entire additional premium for plans that cost more than the weighted-average bid, and they would get to keep 100 percent of the savings for enrolling in plans with below average premiums.
The government’s benchmark could also be set based on a lower-cost option, such as the second-lowest bid in a region. Setting the government’s contribution in this way would likely put even more downward pressure on bids from MA plans, and thus reduce total costs even more than a reform based on weighted-average bids. The downside of this approach is that it could be more volatile because one or two plan bids could substantially alter the benchmark, even if the plans making the bids had low enrollment. Basing the government’s contribution on the weighted-average bid in an area (at least during the initial years of implementing a premium support reform) ensures payments are closely tied to the actual cost experiences of the most popular plans.
Importantly, in any reform plan based on competitive bidding, the premium owed by the beneficiaries should be paid separate and apart from their Social Security checks. Under this approach, if FFS was an above-average cost plan, the beneficiary would owe the additional premium, and it would be visible. MA plans should also be required to offer to the beneficiaries a plan with an actuarial value equivalent to the statutory benefit. Added benefits would be offered in supplemental coverage, for an added premium. (Separating bids and premiums for standard Medicare benefits from supplemental coverage was originally recommended by the National Bipartisan Commission on the Future of Medicare.)
Other concerns about the impact of such Medicare payment reforms on more vulnerable beneficiaries can be addressed by additional premium support subsidies for low-income individuals and enhanced navigational support services for those who may have difficulty making plan choices without assistance.
The success of Medicare Advantage in recent years is changing the conversation on Medicare reform. It is now possible to envision genuine bipartisan support for fair competition between MA plans and FFS. The “premium support” concept still engenders highly politicized opposition in some quarters. But support for the idea has also begun to cross ideological divides. At various points, Senator Ron Wyden (D-Oregon), former Clinton administration budget director Alice Rivlin, and Austin Frakt of Boston University (and prolific defender of the ACA) have all embraced the idea. Unlike others, they support the idea that a truly level playing field for MA-FFS competition requires full FFS participation in the bidding process and transparent beneficiary choices based on the resulting premiums that must be charged for coverage.
Serious Medicare reformers on both sides of the political aisle also increasingly recognize that private plans are far more flexible than FFS, and that is a major advantage. They can adapt and improve their networks, their care management protocols, their customer support systems, and other features of their plans far more rapidly than can FFS. Moreover, in premium support, Medicare’s beneficiaries get to choose the kind of coverage that suits their needs. This is in stark contrast to other reforms, like ACOs, which attempt to place beneficiaries in new delivery models without their full consent.
Premium support is of course a complex reform. It requires risk adjustment of payments and regulation of plans to ensure fair competition. But the risk adjustment system and the regulation need not be perfect for the reform to work; indeed, the system already in place today for MA plans should provide sufficient confidence that a competitive reform model would be beneficial for the program’s participants.
The main obstacle to more intensive competition in Medicare has been distrust. MA advocates believe the bureaucracy will tilt the playing field toward FFS; and FFS defenders believe private plans will find new ways to risk select and game the system or to influence policymakers to provide them with overly generous terms on their payment rates. Only a more transparent competitive bidding system between MA and FFS has a chance of overcoming this mutual distrust and ensuring that both beneficiaries and taxpayers receive the most value for each Medicare dollar.
James C. Capretta is a senior fellow at the Ethics and Public Policy Center and a visiting fellow at the American Enterprise Institute.
Thomas Miller is a resident fellow at the American Enterprise Institute who studies health care policy and regulation.