There is widespread agreement in the health policy community that reforming how Medicare pays providers for its services has the potential both to produce savings for the program and potentially for other payers as well and to encourage the provision of higher quality, more effective health care. The most recent example of how prevalent this view is can be seen from the overwhelming support given to the “Doc Fix,” where an otherwise deeply divided Congress passed the Medicare and CHIP Reauthorization Act of 2015 (MACRA) legislation by a vote of 392 to 37 in the House and 92-8 in the Senate. Admittedly, the vote reflected the strong desire of the Congress to have the perennial issue off the table and was able to pass it because the most contentious issue – how to avoid an increase in the deficit – was only partly addressed. Nevertheless, the substance of the bill indicates a strong belief in the potential of reformed payment approaches, such as accountable care organizations (ACO), bundled payments or medical homes, to reduce costs and improve care quality. It provided that, beginning in 2019, physicians who participate in alternative payment models (APM) for a large enough portion of their practice will be paid 5 percent more per service.
This belief – that a set of metrics can be developed or delivery systems specified that could lead to the delivery of care that would both increase quality and improve efficiency – has not been diminished by the fact that the early rounds of various demonstration and pilot projects have encountered multiple challenges and, so far, shown disappointing results. Many groups are still attempting to develop or refine the metrics that will capture both efficiency and relevant clinical outcomes and do so in a way that won’t unreasonably burden the clinicians who are providing these services.
1 Background on Medicare Payment
Medicare’s original provider payment approaches can be described as “highly passive.” Hospitals were reimbursed their costs. Bills from physicians and other practitioners were screened to make sure they were consistent with the practitioner’s customary charges for Medicare patients and with prevailing charges in the community. After applying various caps to these systems, Medicare switched to prospective payment approaches in the 1980s and 1990s. So hospitals are now paid a fixed amount for an inpatient stay. Physician payment was changed to a fee schedule in which the relative values for various services were based on research on relative costs.
Today’s payment reform goes beyond by focusing on broader units of service and including multiple providers in the payment. So in the case of a hip replacement, where hospitals have been paid an amount for the admission and the surgeon paid a global fee for the procedure and 90 days of post-operative care, the current notion of bundled payment not only combines the hospital and the surgeon but also includes other practitioners, rehabilitation services and any unplanned readmissions to the hospital. ACO payment is based on an attributed beneficiary’s spending over the course of a year.
Inspired by some initiatives by private insurers, the Affordable Care Act (ACA) authorized and funded the Centers for Medicare and Medicaid Services (CMS) to pursue a broad ranging program of pilot programs reflecting this approach. CMS to date has pursued its initiative vigorously.
2 What Have We Learned to Date?
Several types of payment reform strategies currently are being piloted. These include the patient centered medical home (PCMH), which began before the ACA was passed, an advanced payment version that combines public and private payers, and a bundled payment initiative that involves different payment strategies across different provider types (e.g. hospitals and physicians, hospital and post-acute care providers, etc.) The ACA also defines ACOs, allowing groups of physicians and physicians and hospitals that are not formally organized to take partial financial risk for how spending per attributed beneficiary compares to a benchmark. ACOs have had a choice of taking only one-sided risk, i.e. to sharing savings with the Government but not losses or to take two-sided risk where they can earn higher levels of savings but also be subject to losses. To date, most ACOs have chosen to share only savings. A more advanced model is also offered – called Pioneer ACOs – for groups that have had experience taking risk in providing care to populations. Pioneer ACOs can share in larger gains but are also subject to losses.
Most of the pilot programs are still early in their implementation phase and only a few have released formal independent evaluations, so at this stage any conclusions should be regarded as tentative. In addition, most PCMH results have been reported by the payers organizing the medical homes rather than by independent evaluators. The very few independent evaluations have reported no or very small savings. A recent study of the Pioneer ACOs suggested small savings in the first year (i.e. lower expenditure trends) and even smaller savings in the second year (Goldsmith and Kaufman 2015). In addition, almost half of the groups that initially agreed to become Pioneer ACOs have dropped out – either because they were losing money or thought the risk-reward options offered by CMS did not justify continued participation.
The dominant response to these limited results has been to focus on the need to refine the models rather than abandoning the idea of payment or delivery system reform. It is also clear that timely, independent evaluations are critical to assessing current strategies so that timely adjustments can be made as appropriate.
The wide range of reforms being undertaken simultaneously and their rapid evolution is a challenge for evaluation. Earlier feedback will be important in making refinements to the models. Probably the best source of early information is provider decisions about participating in the programs offered, along with CMS discussions with providers. Although the process took far too long, the CMS response to the many providers who dropped out of the Pioneer program was to create a new option – Next Generation – thus providing some encouragement about the existence of such a feedback loop.
3 Changes Needed in Current Medicare Models
In contrast with the broad enthusiasm in the payer and policy communities and provider leaders, for payment approaches that diminish the role of fee for service, the actual models that CMS has been implementing in pilot programs have not been well received by providers. In some cases, this has led to providers participating as a bet that the models will soon evolve, in others only as a transition to a Medicare Advantage initiative and in others the providers are not participating. To us, the strong resistance of providers to taking two-sided risk is as likely to reflect a lack of confidence in the models and metrics as it is to being an aversion to risk.
The design issues needing the most attention include:
- Beneficiary engagement
- Quality measurement
- Development of bundles for a wider range of episodes
- Incorporating post-acute services into more payment approaches
A benchmark is the target rate of spending per beneficiary or per episode for the provider entity that is contracting with the Medicare program, and is the basis for calculating savings or losses to be shared. Medicare to date has used benchmarks that are based on the provider’s historical spending per beneficiary per year or per episode. This spending is trended forward by recent and projected program experience. CMS has tweaked this by using a trend expressed in dollars rather than a percentage change, so that lower-cost providers get a relatively more generous trend.
An advantage of provider-specific benchmarks is that large numbers of providers may perceive an opportunity to succeed under a reformed payment approach. But the providers that are most efficient will find it more difficult to achieve further savings. A provider-specific benchmark rewards improved performance but not good performance.
As long as provider participation is voluntary, provider-specific benchmarks appear to be essential, at least as a first stage in a transition. Uniform regional or national benchmarks would create an adverse-selection phenomenon where low-cost providers would volunteer in order to get paid more for continuing their efficiency, while high-cost providers would not participate since they would have to increase efficiency a substantial amount to avoid losses.
But provider-specific benchmarks have serious disadvantages and should not continue very long. Recalculating benchmarks using more recent spending data avoids their getting out of date, but substantially undermines the business case for participation. Increasing efficiency often requires significant investment. If the benchmarks are recalculated after 3 years, it means that only 3 years of potential shared savings could be achieved by the investment. It is possible that investments could lead to gains in efficiency that increase year after year for a long period of time, but we have not heard about such models. This is not an assumption you would want underlying a program. CMS appears to be making this mistake in its bundled payment for care improvement (BPCI) initiative, where benchmarks are updated very frequently based on each provider’s spending.
Any undermining of the business case could be mitigated by the continued use of the original benchmark, using national trend factors to adjust the benchmark for an additional few years. But few would advocate continuing to set benchmarks based on what a provider’s spending was 6, 8 or more years ago. Since the initial Medicare ACO contracts are coming up for renewal, it is important for CMS to make and publicize the decision on benchmarks for the second round as soon as possible. Until that happens, uncertainty is likely to limit provider investments in improving delivery.
For types of episodes where bundled payment policies are most developed, Medicare ought to make the change compulsory. Many analysts believe that for some types of episodes of care, such as joint replacement or colonoscopy, bundled payments are already promising enough to mandate them. For some time, Arkansas’ Medicaid program has mandated bundled payment for selected episodes and established benchmarks based on statewide data. Under a mandate in Medicare, the benchmark could transition over a period of 4 years from provider-specific to a national benchmark with adjustment for input prices. This was the approach that Congress used in 1983 when it enacted the Inpatient Prospective Payment System. If a voluntary approach had been used, Medicare might never have progressed beyond cost reimbursement! It is highly encouraging that the recently proposed joint replacement initiative from CMS would be mandatory in selected geographic areas and specifies a relatively rapid transition to a regional benchmark – although other parts of the proposal give us concerns (CMMI 2015a; Miller 2015).
More limited degrees of compulsion are possible as well. Alternative payment models could continue to be voluntary but providers given stronger incentives to choose them. Under MACRA (described above) physicians will have an opportunity to be paid rates that are 5 percent higher if the proportion of their services paid for under APMs exceeds a threshold. We expect substantially increased interest in these approaches on the part of physicians and political pressure on CMS to develop additional payment approaches that would qualify. For example, bundled payment approaches for important services in some specialties, such as oncology, need to be distinct from ones using inpatient episodes such as heart surgery or joint replacement. Incentives to participate in alternative payment models could be created for hospitals and other providers as well.
Once broad provider incentives to volunteer for alternative payment approaches are in place, there is more of an opportunity to diminish the role of provider-specific benchmarks. This would permit regional or national benchmarks to be blended with provider-specific benchmarks, though not as aggressively as under mandatory participation. While national benchmarks would be appropriate for bundled payment, regional benchmarks, such as counties or metropolitan areas, may be more suitable for population-based approaches, such as ACOs or medical homes. It is important to align benchmarks for episodes and ACOs to approaches used for inpatient hospital payment and Medicare Advantage (MA), respectively.
The Congress addressed these issues for MA payment by limiting the variation in rates across counties in relation to variation in spending. This approach could be used for ACOs and perhaps for advanced medical home models in which primary care physicians share savings based on overall spending by their panel of Medicare beneficiaries. Under the current situation, delivery systems have strong incentives to choose between an ACO strategy and a Medicare Advantage strategy depending on how their costs per beneficiary compare to others in their region. Those with high costs are likely to focus on ACOs, while those with low costs have an incentive to not pursue ACOs but instead to create a Medicare Advantage product that is built around its delivery system, either on their own or in partnership with an insurer. Such a selection process, which is clearly underway, could be costly for the Medicare program.
3.2 Beneficiary Engagement
The Medicare Shared Savings Program, which was created through specific legislative provisions in the Affordable Care Act, did not provide any incentives for beneficiaries to join an ACO or respond to ACO efforts to coordinate care. Indeed, beneficiaries attributed to an ACO may not even know they are in one, and providers may not know in advance which beneficiaries they are accountable for. The only significant beneficiary provision created an option for beneficiaries who are attributed to an ACO to block the ACO from accessing data on services received from providers outside of the ACO. Many believe that ACOs can accomplish the most through enhanced management of chronic disease, providing support to patients to encourage greater compliance and coordinating care provided by multiple specialists and facilities. ACOs need to be able to engage beneficiaries to realize this potential.
Through comments to CMS, many ACOs have made suggestions to foster greater beneficiary engagement. Beneficiary attribution to ACOs could be done prospectively instead of retrospectively, allowing ACOs to know in advance whose care they are responsible for so they can better target outreach to beneficiaries to provide support. Some have suggested that beneficiaries be given an opportunity to attest to the fact that they have a primary care physician in an ACO. The ACO might offer those beneficiaries a waiver of Medicare cost sharing for visits to their primary care physician. The provision in MACRA that prohibits Medigap insurers from filling in the Part B deductible will help make this more effective. CMS, in its proposed “Next Generation” ACO model, would incorporate some of these suggestions.
A network approach to ACOs, which would require legislation, would strengthen beneficiary engagement further and address some other shortcomings as well. Under a network approach, ACOs would create a network of physicians in different specialties and potentially other provider types, such as skilled nursing facilities, that they believe provide care that is more efficient and at higher quality. Beneficiaries using these providers would get reduced Medicare cost sharing. Those using providers outside of the network would pay additional cost sharing. These variations in cost sharing could be made by the ACOs directly or by Medicare. Full choice of provider in traditional Medicare would continue, as would current limits on balance billing for those providers not accepting assignment. But ACOs would have another mechanism to steer those who have signed attestation forms to providers believed to be superior. Beneficiaries might also be offered a reduction in their Part B premium in return for agreeing to this arrangement. Again, this payment could either be made by Medicare or by the ACOs.
A network approach would also provide a mechanism to engage more physicians and other providers in ACOs. ACOs must have primary care physicians, but many specialties are not typically included by ACOs. This could potentially pose a problem for implementation of the APM provision of MACRA, where physicians need to meet thresholds for activity in APMs to qualify for payment rate bonuses. Under the network approach, ACOs would have to include physicians in each specialty in their network, even if they chose not to offer involvement in the leadership of the ACO and a sharing of savings and losses to those specialists.
For the network approach to be effective, further changes in rules for Medigap coverage would be needed. Beyond the MACRA provision concerning the Part B deductible, restrictions on filling in coinsurance would be needed so that reductions associated with using network providers or increases for using non-network providers would not be fully offset.
3.3 Quality Measurement and Alignment across Payers
Quality measures should be an important part of alternative payment approaches. The public no longer assumes that medical care quality is uniformly high; the Medicare program needs to measure quality, make it publicly available and incorporate it into payment. Although FFS has not insured good quality, concerns with alternative payment and its incentives to provide less care make incorporation of quality measures particularly important.
To date, quality has been incorporated into Medicare payment by requiring that providers meet high quality standards before they can receive shared savings payments. This direction should continue because the ability to potentially trade off costs and quality appears to be a long way off. It may never happen, at least for public payers, which would have difficulty stating explicitly that they have sacrificed some quality in order to save money.
Ultimately, beneficiaries are best positioned to make these tradeoffs. So priority needs to be given to developing quality measures that are meaningful to beneficiaries and aggregating them in a way that can make complex data more accessible. After years of evolving star ratings for Medicare Advantage plans, CMS recently began to issue star ratings for the patient-focused aspects of hospital quality. While this is a positive trend, there is always the danger that aggregating different measures produces a number that is not very meaningful. An approach like the star ratings could ultimately be applied to ACOs, assuming a meaningful metric can be developed. This could be especially useful in an environment in which beneficiaries actively choose ACOs.
Providers have expressed serious concerns about the current situation of different payers using different quality measures in their alternative payment initiatives. Not only does this increase the expense of developing the data required by each payer but it means that providers lack clear directions about what aspects of quality are most important. Private insurers and Medicaid programs using different measures from Medicare is at odds with longstanding history of alignment in other areas. In coverage decisions and hospital and physician payment, the record has been one of other payers following much of what Medicare does. Why has this not happened so far in quality measures or, more broadly, in design of alternative payment approaches? Presumably it would be easiest to follow Medicare. Perhaps the problem is that Medicare to date has not done the best job.
While more alignment would certainly be desirable, we believe it would be a bad idea to require others to follow Medicare. What is needed instead is meaningful communication between the various payers. A promising process is the collaboration of America’s Health Insurance Plans with CMS and the National Quality Forum on aligning quality measures. The collaboration has had input from physician specialty societies and plans to bring employers and consumers into the process as well (AHIP 2015). Such collaborative processes could also be used to enhance other aspects of designs of various alternative payment approaches. CMS’ recently announced Learning Action Network might also contribute to this goal.
3.4 Expanding Range of Services Included
Although some of earliest initiatives for reformed payment included only hospital and physician services, inclusion of additional services would expand the opportunities to reduce costs and increase quality. As the recent IOM report on geographic variation in Medicare spending showed, post-acute care, which includes skilled nursing facilities, rehabilitation facilities and home health services, is by far the most important factor behind variation in overall spending (Institute of Medicine 2013). Fortunately, Medicare ACO models currently do include post-acute care. But many of the bundled payment models do not. For example, with rehabilitation such an important part of joint replacement, excluding it from a bundled payment limits what can be accomplished by providers in terms of increased value.
Part D prescription drugs have also been excluded to date, presumably due to the substantial administrative challenges. Claims systems, which are run by the Part D insurers, are separate. In addition, a mechanism would have to be developed to transfer a portion of the savings accruing to the Part D plan to the providers that are taking risk including drug spending. The savings remaining with the insurers would be expected to flow to Medicare through lower future bids, which in turn would reduce the Part D benchmarks and premiums.
For situations in which drugs are an important part of spending, such as chemotherapy episodes, making the effort to include drugs would be well rewarded. Indeed, CMMI recently created an oncology bundle for chemotherapy that does include both Part B (physician-administered) and Part D drugs (Center for Medicare and Medicaid Innovation 2015b). Indeed, under this initiative, claims for drugs trigger an episode, which runs for 6 months.
3.5 Additional Bundles
To date, most bundled payment has been for episodes triggered by a stay in an acute hospital. As care continues to shift from inpatient to outpatient settings, there likely will be substantial opportunities for bundled payment for episodes of care that do not require an inpatient stay. In contrast to acute inpatient episodes, where CMS could provide very general rules through offering four possible bundled payment models, outpatient episodes will require more detailed judgment to create attractive bundles.
For outpatient episodes, CMS should specify those eligible for bundled payment, starting with high-volume episodes. Characteristics of promising episodes would include substantial variation in spending from provider to provider, a relatively small number of providers to coordinate, especially those who already have significant interactions, where there are relatively clear objective clinical guidelines that trigger the episode and where the treatment is not especially supply sensitive (Bipartisan Policy Center 2015). A screening colonoscopy is one example of an episode with a number of these characteristics.
Opportunities include selected chronic episodes, where the episode is defined by a period of time, as well as acute episodes. The major challenge in this area is conditions where variation in severity has important implications for resource use. Chronic diseases that can be “staged” (e.g. stages 1 through 4 of congestive heart failure) are more likely to lend themselves to bundled payment and not likely to be subject to “upcoding,” assuming the staging characteristics are unambiguous). It is also likely to be necessary to differentiate bundled payments involving a chronic disease that includes co-morbidities that could affect treatment choices as for example would be the case for complex diabetes with CHF or other co-morbidities.
4 Path Forward to Reformed Payment
If payment models can be improved along the lines discussed above, how can policymakers chart a path so that the reformed payment models become the norm in traditional Medicare, leading to a greatly reduce role for FFS? This section sketches a number of components of such a strategy. It includes mandating reformed payment for services where confidence in the model is high, incentives for providers to voluntarily choose reformed models and resolving situations in which providers participate in more than one model that affects services for a beneficiary.
4.1 Mandate Some Payment Reforms for Medicare
Once there is sufficient confidence in the merits of specific reformed payment models, their use should be mandated. Mandating reformed payment is what Medicare has done traditionally and provides several important advantages relative to a system of voluntary payment reforms. First, the gains in efficiency and quality will affect more beneficiaries. Second, opportunities for federal savings will be greater, both because more services will be included under the model and because the federal government can garner more of the savings over time. Third, the problems with benchmarks discussed above can be resolved so that at the end of a transition, rewards can go for good performance by providers rather than only to improved performance. And fourth, the problems of self selection that plague all voluntary programs can be avoided.
We believe that such a mandatory approach can be applied at this time to at least selected episodes. Policymakers already know a lot about bundled payment for joint replacement and certain cardiac procedures through the ACE demonstration and the CABG demonstration from the 1990s (Health Care Financing Administration 1998). There is a lot of experience with these episodes in BPCI and many private insurers have been contracting for these services on a bundled basis for years. As CMS develops additional episodes, bundled payment could be mandated, using a transition payment model to move from the present fee for service payment to the bundled payment. We agree with CMS that joint replacement bundles can be mandated, but see little reason to limit it to selected geographic areas as opposed to implementing nationally.
4.2 Incentives to Encourage the Adoption of Reformed Payment
MACRA offers a 5 percent bonus payment to physicians with a high enough portion of revenues from APMs. This does not begin until 2019 and a lot of work will be required by CMS to establish which APMs should be eligible. The potential for bonuses is likely to increase participation in reformed payment strategies. If particular types of strategies dominate, Congress should consider whether to mandate them – or give CMS the authority to do so. A potential problem with the specific approach authorized by MACRA is that a wide range of APMs might be approved, making eventual consolidation into one or a small number of approaches far more challenging. It is one thing to have a distinct bundle for chemotherapy but another to have five different bundles for cardiac surgeries. However, as long as the various approaches produce auditable savings, it is not clear this represents a “bad outcome.”
The notion of offering higher payments to providers for participation in APMs could be expanded to institutional providers. This might not be as easy as with physicians, who had the threat of the SGR hanging over them and ultimately agreed to a long period of very low updates along with the opportunity of a 5 percent bonus and a share of savings under the APMs. The experience from the quality reporting program a decade ago indicated that hospitals changed their behavior in response to relatively small bonus payments.
4.3 Coexistence of Different Payment Approaches in Medicare
As payment reform proceeds with multiple pilot projects, it will become more common for treatments to a particular beneficiary to be covered by multiple payment mechanisms. For example, an orthopedic surgeon could be part of a group participating in a bundled payment for joint replacement. Some of the beneficiaries having a joint replacement with this physician may be attributed to an ACO, where the surgeon may or may not be a member of the ACO. Under current policies, CMS will subtract the savings from the joint replacement that are shared with the surgeon’s group from the savings calculated for the ACO. This avoids double payment for savings under an episode of care.
But this appears to undermine a potential strategy for ACOs to achieve savings – to encourage beneficiaries attributed to the ACO to go to surgeons who are more efficient per episode of care. Indeed, we might ultimately want to evolve the ACO model so that the ACOs share savings per episode achieved with the physicians that treat ACO patients; in other words, implant approaches like bundled payment and medical homes into a new model of ACOs or at least make sure that policies are not adopted that would discourage such behavior.
Or we could move in an opposite direction, which would involve a medical home model that places primary care physicians at risk for overall spending for beneficiaries attributed to them, perhaps on an upside only basis, and a bundled payment model for episodes of care. This could make the ACO model superfluous.
The point for now, when trying out numerous payment reforms simultaneously makes sense, is to make sure steps are taken so that achieved savings are not rewarded more than once, but to do so in a way that does not inadvertently undermine promising models. Indeed, CMS might have to prioritize the approaches that it is developing and make sure that rules governing overlaps in approaches do not unduly disadvantage the higher priority ones. Unless CMS has a strong preference for bundled payments over ACOs, the current rules on overlaps between those approaches which directs all of the overlapped savings to bundled payment, do not strike us as wise.
Realizing the potential of reforming payment in the traditional Medicare program will require improvements to the payment models offered to providers. This will be the key to realizing HHS Secretary Burwell’s ambitious goals to expand the role of these approaches in Medicare. Realizing those goals will require extensive provider participation. Achieving this not only allows more beneficiaries to gain better outcomes of care, but also permits the critical step of moving away from provider-specific historical benchmarks to those based on regional or national experience. Voluntary participation is not a viable long-term strategy.
Prepared for Strengthening Medicare for 2030, Brookings Institution, June 5, 2015. The authors are grateful to Alice Rivlin for her comments on an earlier draft.
America’s Health Insurance Plans (AHIP). (2015) Moving Ahead on Transparency in Health Care Quality: A Public-Private Collaboration Frequently Asked Questions. March 2015.
Bipartisan Policy Center (2015) Medical Homes, Payment Bundles and the Role of Fee for Service. January 2015.
Center for Medicare and Medicaid Innovation (2015a) Comprehensive Care for Joint Replacement Model. Available at: http://innovation.cms.gov/initiatives/ccjr/ (accessed August 21, 2015).
Center for Medicare and Medicaid Innovation (2015b) Oncology Care Model. Available at: http://innovation.cms.gov/initiatives/oncology-care/ (accessed August 21, 2015).
Goldsmith, Jeff and Nathan Kaufman (2015) “Pioneer ACOs: Anatomy of a Victory,” Health Affairs Blog, June 18, 2015. Available at: http://healthaffairs.org/blog/2015/06/18/pioneer-acos-anatomy-of-a-victory/.
Health Care Financing Administration (1998) Medicare Participating Heart Bypass Center Demonstration. Available at: http://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports/Reports/downloads/oregon2_1998_3.pdf.
Institute of Medicine (2013) Variation in Health Care Spending: Target Decision Makers Not Geography. July 2013.
Miller, Harold D. (2015) “Bundling Badly: The Problems with Medicare’s Proposal for Comprehensive Care for Joint Replacement,” July 2015. Available at: http://www.chqpr.org/downloads/BundlingBadly.pdf.