November 13, 2013
OPTIONS FOR REDUCING THE DEFICIT: 2014 TO 2023

Mandatory SpendingOption 10

570 - Medicare

Bundle Medicare’s Payments to Health Care Providers

(Billions of dollars) 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2014-2018 2014-2023
Change in Mandatory Outlays                        
  Bundle payments only for inpatient care 0 0 0 -0.4 -1.1 -1.9 -2.8 -3.2 -3.5 -3.6 -1.5 -16.6
  Bundle payments for inpatient care and 90 days of postacute care 0 0 0 -1.2 -3.1 -5.4 -7.8 -9.1 -9.8 -10.2 -4.3 -46.6

Note: This option would take effect in January 2017.

Overview of the Issue

Although some steps have been taken to move toward other payment methods, most payments for health care—under the Medicare program and other forms of insurance—are made on a fee-for-service basis. In a fee-for-service system, separate payments are generally made for each office visit, lab test, surgical procedure, or other service that is delivered by doctors, hospitals, or other health care providers. The fee-for-service payment method tends to create incentives for providers to deliver more services (and more expensive services) but not to coordinate the care that patients receive. Many experts thus believe that the widespread use of fee-for-service payment has contributed significantly to the high costs and uneven quality of health care in the United States.

Those concerns have prompted considerable interest in the idea of bundling payments, in which single payments would be made for groups of related services. The broad concept of bundling could be applied in various ways, but one commonly discussed approach is to make fixed payments for each “episode of care”—that is, for all or most of the services that patients receive from various providers that are related to a particular disease or treatment over a defined period. Episode-based payment does not always involve multiple providers. For example, obstetricians often receive a fixed payment (or “case rate”) for all of the care they provide to a pregnant patient; that payment does not cover the costs of hospital care for a birth or prenatal care delivered by other providers. However, this discussion focuses on episode-based bundled payments that encompass services delivered by a range of individuals and organizations during the course of a patient’s treatment—an approach that offers more opportunities for savings but is more difficult to implement successfully.

In any system of bundled payments, the amount of the payments would differ depending on the diseases or treatments involved and would reflect the average costs of providing those treatments. In most proposals for bundling, however, payments would not vary with the number or mix of services provided to a particular patient. As a result, providers of care covered by a bundled payment would have an incentive both to limit the cost or reduce the number of services they provide and to coordinate care so as to avoid costly complications and the delivery of unnecessary services. At the same time, bundling payments could give providers an incentive to stint on care that is medically beneficial. And as with fee-for-service payment, episode-based payment would not encourage providers to keep patients healthy or to prevent episodes of care from occurring in the first place.

Medicare already bundles some of its payments, but they typically cover services provided by a single individual or organization. For example, hospitals generally receive a fixed payment for each admission to cover all of the discrete goods and services they provide during a patient’s stay. Likewise, home health care agencies receive a fixed payment to cover all of the visits they provide to a patient during a 60-day episode of care, and skilled nursing facilities (SNFs) are paid a per diem rate that covers all of the services they furnish to a resident in a day.

Nevertheless, a patient undergoing surgery typically generates a range of separate Medicare payments before, during, and after his or her hospital stay: to the hospital in which the procedure takes place; to the surgeon performing the operation; to the anesthesiologist; to other doctors providing care while the patient is in the hospital; to doctors and labs for follow-up visits and tests after the patient is discharged; and to SNFs, home health care agencies, or other organizations providing postacute care (recuperation and rehabilitation services after a hospital stay). All told, CBO estimates, Medicare spent more than $170 billion in 2013 on services provided during a hospital stay or within 90 days of discharge. That total accounts for at least half of all nondrug spending in Medicare’s fee-for-service program.

One rationale for bundling payments for those services is that Medicare’s current costs often vary significantly for a given type of episode, and in many cases that variation does not seem related to differences in patients’ illness or outcomes. One recent study found that Medicare’s average payments for several common surgical episodes (hip replacement, heart bypass, back surgery, and colon surgery) frequently varied among hospitals by 10 percent to 40 percent, even after accounting for disparities in patients’ health and for geographic differences in the prices that Medicare pays for specific services. A large share of that variation in costs stemmed from spending on postacute care, but in many cases differences in total payments for the initial hospitalization and for readmissions were notable as well.

Similarly, an analysis by the Medicare Payment Advisory Commission (MedPAC) concluded that the extent and types of postacute care that patients receive after being discharged from a hospital “vary widely for reasons not explained by differences in beneficiaries’ health status, indicating that, in aggregate, fewer services could be furnished to Medicare beneficiaries without necessarily compromising patient outcomes.” Examining 10 types of episodes that frequently involve postacute care, MedPAC found that spending on such care within 90 days of a hospital discharge commonly varied more than fourfold between higher-cost and lower-cost cases—a gap that averaged about $13,000 per case during the 2007–2008 period.

Several demonstration projects to experiment with bundling Medicare payments have been launched over the years—most recently, the Bundled Payments for Care Improvement initiative, which the Centers for Medicare & Medicaid Services (CMS) began developing in 2011 under provisions of the Affordable Care Act and which has just started to operate. That initiative is exploring four models of episode-based bundled payments; the models differ in their scope and payment methods, but in all four, an episode of care is triggered by a hospital admission. Participation in the initiative is voluntary, and so far more than 300 organizations (mostly hospitals) have expressed interest in taking part. Results from that initiative will not be available for some time, and the voluntary nature of the initiative raises questions about how broadly applicable those results will prove to be. Earlier (but more limited) demonstration projects about bundling yielded some estimated savings for Medicare, at least on a preliminary basis, but they were also voluntary. The main problem in evaluating such voluntary initiatives is that the hospitals that opted to participate were probably more capable of changing the ways they deliver care, and more likely to succeed financially, than hospitals that decided not to take part. Thus, participants’ experience with bundling seems likely to overstate the savings that would probably be achieved if all providers were required to adopt bundled payments.

In addition to Medicare’s demonstration projects, private insurers and state Medicaid programs are exploring episode-based payment. However, their efforts are generally at an early stage as well.

Key Design Choices That Would Affect Savings

Payment bundling is a broad concept that could take many forms. The federal savings that could result from greater bundling would depend on many design specifications, such as the types of bundles constructed and their scope, the duration of the services covered by a bundle, the levels at which bundled payments were set and the mechanisms used to set them, the method of payment used, the schedule for implementing the bundling policy, and the terms of participation (in particular, whether bundling would be voluntary or mandatory).

In general, more extensive bundles encompass more spending and may provide more opportunities to generate savings. But they also expose health care providers to more financial risk, particularly when the total costs of the bundle depend on services delivered by a variety of providers who are not affiliated. Bundling payments for different providers can also raise significant administrative challenges, and some solutions to those challenges may weaken incentives to control costs. In addition, aggregating payments while giving doctors, hospitals, and other providers greater leeway to share savings among themselves could encourage those providers to generate more episodes of care.

Among the many design issues that arise, the levels of bundled payments and the rate-setting and payment mechanisms are perhaps the most important. Fundamentally, reducing federal spending through bundled payments would require providers to be paid less overall than they are under current law—either because they would be delivering fewer or less complex services to enrollees or because they would be receiving less money per service.

Types and Scope of Bundles Constructed. Recent proposals for bundling payments generally involve grouping services that are provided during an episode of care, either to treat a patient with a particular disease or to provide a particular treatment (such as a surgery) and its related care. In principle, nearly all of the services that patients receive could be grouped into episodes of care, but in practice, the wholesale adoption of episode-based payment would face many obstacles. For example, ongoing efforts to create all-encompassing “grouper” software that assigns each of the services received by Medicare patients to specific episodes have been hampered by the fact that Medicare patients are more likely than younger people to suffer from multiple health problems at the same time, which makes it harder to determine which services should be assigned to which episodes.

A more feasible approach to bundling may be to group only those payments that are related to a hospital admission—the approach being taken in CMS’s demonstration project and in several private-sector initiatives. Under the demonstration, the scope of the bundles varies: One model (labeled “Model 4” by CMS) covers services that physicians and hospitals provide during an inpatient stay, another model (“Model 3”) covers only postacute care provided after a hospital discharge, and yet another model (“Model 2”) includes all care provided during an admission as well as postacute care. Even with those distinctions, defining which services provided after a discharge are “related” to a hospital admission can be difficult. Excluding certain services from the bundle could give providers an incentive to deliver more of those services. But including more services and more types of providers in the bundle would add to the administrative complexity of the payment system.

Duration of Each Bundle. The amount of spending encompassed by a bundle—and the financial risk that providers would face under a bundled-payment policy—would also depend on the length of time that the bundled payment would cover. For chronic health problems that generally are not cured, such as diabetes or hypertension, episodes of care may extend for a full year. With episode-based bundles that center on a hospital admission, proposals that include postacute care generally cover services provided over periods that range from 30 days to 90 days after discharge. According to MedPAC’s analysis of 10 common episodes that usually involve extensive postacute care, 84 percent of the spending that would be included if a bundled payment covered 90 days of services would also be included in a 30-day bundle. Similarly, CBO’s analysis of payment data for a broader set of episodes, which CMS generated for the bundled-payment demonstration, found that about three-fourths of the spending incurred during a 90-day episode was captured by a 30-day episode. (Both findings reflect the fact that hospital payments usually constitute a majority of costs for such episodes.) Thus, extending the duration of bundles from 30 days to 90 days would capture more spending, but far less than three times as much.

MedPAC also examined the variability of the resources used to care for patients. That variability indicates the extent to which providers’ costs for delivering care might deviate from the fixed payment they would receive and thereby sheds light on the degree of financial risk that providers might face under a bundled-payment policy. MedPAC found that the variability of resources used per episode of care was only slightly greater for 90-day episodes than for 30-day episodes—and was comparable to the variability of hospitals’ own costs per admission under Medicare’s current payment system for hospitals. (That system generally makes a fixed payment per admission that is based on the diagnosis-related group to which the patient is assigned.) Those findings suggest that providers would not bear undue financial risk under such a bundled-payment policy. But the degree of risk would also depend on how the rates for bundled payments were determined and on whether the payment system incorporated additional mechanisms to limit providers’ financial exposure.

Payment-Setting Mechanism and Level. Once the scope and duration of bundles had been defined, a central question would be how to set the payment rate for each bundle. The federal savings generated by a bundling policy would largely depend on how those rates compared with Medicare’s total payments to treat the same medical conditions under current law.

Two broad alternatives for rate setting are administered pricing and competitive bidding. Under the former approach, CMS could set payment rates for bundles using information about past Medicare costs or other factors (an approach that is common in fee-for-service Medicare). Such administered prices could be set below currently projected spending levels to generate savings for the federal budget, but those prices might initially overstate or understate the average savings that providers could actually achieve. Prices would need to be rebased periodically to keep payments in line with the costs of efficient providers. However, if the bundled payment rates for each group of providers reflected their average costs per episode (rather than a national or regional average of those costs), rebasing could undercut incentives to control costs per episode because providers would know that higher current costs would translate into higher bundled payment rates in the future.

Under a competitive bidding system, hospitals might submit bids in advance indicating the payment they would accept for each type of episode. CMS could then exclude high bidders from Medicare or use an average of the bids to set its payment rate. In theory, bidding systems can quickly reveal the costs that efficient providers incur. In practice, however, providers that are not already integrated to deliver the full spectrum of patients’ care during an episode might have trouble determining an appropriate bid. As experience with bundled payments grew, those challenges could become more manageable; thus, one option might be for the payment-setting mechanism to evolve over time from administered pricing to competitive bidding. Even then, however, many hospitals and some medical specialists might not have strong incentives to bid their true costs, partly because of limited competition in their markets.

Method of Payment. The concept behind bundling payments is generally to make a fixed payment per bundle, so that providers collectively bear all of the excess costs if total spending exceeds the fixed payment and get to keep all of the savings if their costs are lower than that payment. One way to implement that approach would be to make a single, prospective payment to one individual or organization—such as the hospital responsible for the initial admission—and require that recipient to arrange payments to other providers delivering the care covered by the bundle. For bundles that applied only to services provided during a hospital stay (including physicians’ services), that approach would seem relatively easy to administer; it is the payment method that CMS adopted for Model 4 of its current demonstration. For bundles that included services provided after discharge from a hospital, however, a single prospective payment to the hospital could prove complex to administer: The hospital would need to have payment arrangements with—and oversee—all of the various providers that might be involved in delivering care after a patient was discharged.

As an alternative to prospective payments, CMS could continue to make fee-for-service payments to providers (perhaps with a portion withheld) and later reconcile those total payments with the target payment rate for each bundle. In that case, CMS would have to distribute bonus payments or recoup overpayments if the total costs of the bundle were below or above the target. (A similar approach is being used in Model 2 of the current demonstration, which includes both inpatient and postacute care.) CMS would probably have to prorate the bonus payments and recoupments for all of the providers delivering services that were covered by the bundled payment, because the agency could not determine which providers were responsible for generating any savings or excess costs. Providers could develop selective arrangements among themselves to reallocate those bonuses and penalties (a process called “gain sharing”), but they would not be required to do so. (As with other provisions of a bundled-payment system, payments could evolve over time—in this case, from a prorated system to prospective payment.)

For a simplified example of how that prorated method might work, suppose that a given episode of care typically cost Medicare $20,000 per patient ($12,000 for services provided by a hospital and $8,000 for postacute care) and that CMS set the spending target for that episode at $19,000. If the payments to the hospital remained unchanged but the payments for postacute care fell by half, to $4,000—perhaps by reducing the length of a stay in a skilled nursing facility or shifting to home health care instead—the episode would initially cost Medicare $16,000. In that case, the hospital and the postacute care provider would divide a bonus payment of $3,000, the difference between the initial cost and the $19,000 target. Of that bonus payment, $2,250 would go to the hospital (because it would account for three-fourths of the $16,000 cost) and $750 would go to the organization that provided postacute care. The outcome would be different if Medicare’s fee-for-service payments rose instead of fell. If the hospital’s payments remained at $12,000 but payments for postacute care increased from $8,000 to $12,000, the episode would initially cost Medicare $24,000. In that case, the hospital and the postacute care provider would each account for half of the episode’s initial cost and thus would each owe Medicare $2,500, or half of the $5,000 difference between that initial cost and the $19,000 target.

As the example illustrates, the way in which bundled payments, bonuses, and penalties were distributed would affect both providers’ incentives to reduce costs per episode and the extent of financial risk that providers faced. In particular, prorating bonuses and penalties would mean that savings on payments to one provider might be shared by other providers and that higher initial payments to one provider might translate into penalties for other providers. Those features would weaken each individual provider’s incentive to control costs per episode, but they might also reduce the risk that providers would face if their patients used above-average levels of care. Whether higher or lower costs incurred by providers would translate into changes in Medicare’s initial payments would depend on the types of services involved. For example, higher costs for hospitals to coordinate patients’ care would not trigger higher Medicare payments initially (although they could generate bonus payments if the use of other services for which Medicare pays individually, such as days in a skilled nursing facility, was reduced as a result). Similar issues can arise with bundled payments that are made prospectively, depending on how those payments are subsequently allocated among the providers delivering care during an episode.

Proposals for bundling payments may also include features designed to compensate providers for costs that are beyond their control or to encourage providers to treat high-cost cases (which they might otherwise be reluctant to do); such features would influence both the incentives and risks for providers. For example, payment targets could be risk adjusted to reflect predictable differences in the costs of treating patients who were healthier or sicker than average. Also, episodes that were extremely costly could generate additional “outlier” payments (as happens for Medicare’s hospital payments under the current payment system). Finally, some proposals would have Medicare and providers share savings and losses when initial payments were below or above the payment target (for example, with a 50-50 split) rather than having providers keep all savings and bear all excess costs.

Implementation Schedule. Savings from a bundled-payment system would depend partly on how soon the new system began, how quickly it was phased in, and how comprehensive it ultimately became. Implementing a bundled-payment system and preparing to operate under it would probably take the government and health care providers a few years following enactment of the policy—in part because CMS would still be in the midst of implementing and learning from the current demonstration.

In that demonstration, CMS has designated 48 types of episodes encompassing treatments that seem most amenable to bundling and that together span about 25 percent of Medicare’s diagnosis-related groups (DRGs). (Because the DRGs included are more common than other DRGs, those bundles would encompass about two-thirds of all DRG payments in fee-for-service Medicare if they were applied nationwide.) Most participants in Models 2 and 4 of the demonstration are adopting bundled payments for only a few of the 48 types of episodes, which suggests that broader implementation should proceed gradually. However, adopting such bundling for only a limited set of episodes could expose providers to random fluctuations in costs if they delivered services for relatively few episodes of care, whereas with a larger range of episodes, random variations in costs would be more likely to average out. Those considerations might argue for implementing bundled payments in a more rapid and extensive way.

Terms of Participation. The budgetary effects of bundling would depend significantly on whether participation was voluntary or mandatory and on which providers (if any) were required to participate. Indeed, if participation was voluntary and the bundled payment was set to equal the national average of Medicare’s costs, federal spending would probably rise because providers that expected to increase their total payments under that system would be much more likely to participate than providers that faced a cut in payments. In its demonstration, CMS avoids that problem by basing the payment targets for each participating hospital on Medicare’s past costs for episodes of care initiated at that hospital; as a result, hospitals with below-average costs per episode would have to reduce their costs further to gain financially. Still, because the CMS demonstration might be expanded if it proves successful, some of the savings from bundling payments may be generated under current law—so enacting a bundling program under Medicare that was similar and voluntary might not save the federal government additional money. Legislation specifying a mandatory shift to episode-based bundled payment over the next several years, however, could generate federal savings because such a shift would probably represent a more aggressive approach than CMS will pursue under its current authority.

Another factor affecting federal savings is whether hospitals that Medicare currently pays on the basis of their own costs (rather than making fixed payments) would have to participate in the bundling policy. Such hospitals, which are designated “critical access hospitals,” account for about 5 percent of Medicare’s hospital payments.

Specific Alternatives and Estimates

To illustrate the budgetary effects of bundling Medicare payments, CBO examined two alternative approaches. In each, Medicare would set a target payment amount for specified episodes of care triggered by a hospital admission. The two approaches differ in several ways:

  • In the first alternative, a bundled payment would cover services provided by hospitals and physicians during a patient’s initial hospital stay and any related hospital readmissions occurring within 30 days of discharge. For each admission, the hospital would receive a prospective payment that was 3 percent lower than Medicare’s projected average payments per episode for those services under current law.
  • In the second alternative, the bundled payment would cover the same inpatient and physicians’ services but would also include any postacute care (such as SNF, home health, or rehabilitation services or outpatient physical therapy) that was delivered within 90 days of discharge. Other services provided after discharge, including physician visits and lab tests, would be excluded from the bundle (on the grounds that payments for those services would generally constitute a small share of the total payments for each bundle and might represent unrelated services). In this alternative, CMS would pay claims on a fee-for-service basis, withholding 10 percent pending reconciliation of actual payments with the spending targets. Those targets would be 5 percent lower than Medicare’s projected average payments per episode under current law.

The savings target of 3 percent in the first alternative equals the discount required of participants in Model 4 of the CMS bundling demonstration. Nationwide, less than 10 percent of hospitals chose to participate in any of those bundling models, which indicates that many hospitals and associated health care providers would face challenges in meeting such a target. The larger savings target of 5 percent in the second alternative reflects CBO’s judgment that more opportunities would exist to economize on spending if postacute care was included. That judgment partly reflects the findings that spending on postacute care varies widely for reasons not explained by differences in patients’ health, as well as studies indicating that the transition period after a hospital discharge presents substantial opportunities to improve the quality and efficiency of care. According to MedPAC’s analysis of 10 common episodes, reducing spending on postacute care and on hospital readmissions within 90 days of discharge by an average of 10 percent would decrease the overall costs of those bundles by 5 percent.

In both alternatives, the bundled-payment system would apply to all short-term acute care hospitals beginning in 2017 and would be phased in over four years, at which point it would cover the 48 types of episodes specified in the CMS demonstration. Admissions for other DRGs would remain exempt from bundling. CMS would have discretion about which bundles to implement first but would have to phase in the policy so that roughly equal increments of affected Medicare spending were added each year (thus covering 25 percent of that spending in 2017, 50 percent in 2018, 75 percent in 2019, and 100 percent in 2020). Once bundling began, the capitation amount or target payment—which would initially be based on an extrapolation of Medicare’s past payment levels—would be updated using a weighted mix of the update factors that apply to the types of services included in each bundle. (Medicare’s payment rates are generally updated each year to reflect increases in providers’ input costs, which can vary for different services, and those updates may also be modified by statute.) Medicare’s extra payments for graduate medical education and for hospitals that treat a disproportionate share of low-income patients would not be included in the target payment (or counted as part of the bundle’s costs) and would continue as under current law.

CBO estimates that the first alternative—bundling payments only for inpatient care—would reduce Medicare spending by $17 billion through 2023. The second alternative—bundling payments for inpatient and postacute care—would produce larger savings: $47 billion through 2023. By that year, with the changes fully phased in, the savings from the first alternative would represent 0.5 percent of Medicare’s net outlays for all nondrug services, and the savings from the second alternative would represent 1.4 percent of those outlays.

A primary factor determining the savings under this option is that the spending that would be bundled accounts for about one-fifth (for the first alternative) or one-third (for the second alternative) of gross nondrug outlays in Medicare’s fee-for-service program. Savings would be greater if all DRGs were included; limiting bundling to the 48 types of episodes specified by CMS excludes about one-third of spending connected to hospital admissions. Savings would also be greater if the reductions used to determine the payment targets were larger than 3 percent and 5 percent, respectively, but achieving greater savings by economizing on services would become increasingly difficult for most providers.

Another factor affecting the estimated savings from both alternatives is that a bundled-payment policy would overlap or interact with several initiatives being pursued under current law, including CMS’s latest bundling demonstration; penalties for hospitals with high rates of readmission for certain conditions (which would, in this option, be phased out for affected DRGs as bundled payments were phased in); and accountable care organizations, or ACOs (groups of providers that accept responsibility for managing the quality and total costs of patients assigned to them). ACOs are allowed to share savings with Medicare if the total costs of treating their patients are below certain targets; thus, those organizations might capture some of the savings generated by the broader application of bundling. CBO’s estimates for the two bundled-payment alternatives take those overlaps into account. In addition, savings under Medicare’s fee-for-service program would translate into lower federal payments for Medicare Advantage plans (private insurance plans that provide Medicare benefits); that effect is also included in the estimates above.

The way in which savings targets were set would affect the amount of savings that particular hospitals and other providers would need to achieve under a bundled-payment system. Those effects can be seen by comparing two approaches to implementing the second alternative that would yield roughly the same overall savings to Medicare but that would have very different implications for different providers. One approach—used in CMS’s bundling demonstration—would set the payment target for a given episode of care at a different level for each hospital, reflecting Medicare’s average historical payments for that type of episode initiated in that hospital. Another approach—which would more closely resemble the DRG payment system—would set the payment for each bundle of services using the national average of Medicare’s payments for that bundle, adjusted only for geographic differences in Medicare’s payment rates (which reflect geographic differences in providers’ input costs, but not differences in the average quantity or intensity of services delivered).

The first approach (using hospital-specific targets) might make it easier for providers with high-cost practice patterns to achieve the target level of savings but might make it harder for providers that were already operating at a lower cost to achieve the specified savings goal. The second approach (using national-average targets) would create greater challenges for high-cost providers, whereas low-cost providers could receive bonus payments even if they did not change their practice patterns under the new system. Specifically, data from one of the studies cited above indicate that, with national-average targets, about one-fourth of hospitals and associated providers would have to reduce their costs to Medicare for specific episodes of care by more than 10 percent to achieve a target that was 5 percent below the national average. At the same time, about 40 percent of hospitals and associated providers would not have to reduce their average costs to Medicare at all to meet that target (and could see their payments increase). By contrast, with hospital-specific targets, all providers would need to reduce their average costs per episode by 5 percent to keep their costs in line with their payments. (As with other parameters of the option, a transition process could be specified that would shift the targets over time from hospital-specific to national-average amounts.)

Other Considerations

Bundling Medicare’s payments for episodes of inpatient or postacute care, or both, would represent a significant change to the program’s current payment system. That change would have myriad effects on health care providers, on Medicare beneficiaries, and on patients and programs outside Medicare. Many of those effects are difficult to predict precisely.

Effects on Medicare Providers. Adapting to a bundled-payment system would create both challenges and opportunities for affected health care providers. If Medicare’s payments encompassed services delivered by a range of providers, those providers would probably want to enter into new organizational arrangements to manage patients’ care and to allocate payments equitably. Prospective payments would effectively require the affected providers to contract with each other about payment terms and responsibilities, and providers would need to structure those contracts carefully so that participants’ incentives were properly aligned with the overall goal of delivering high-quality care at a lower total cost. Making fee-for-service payments, reconciling them afterward, and distributing bonuses or penalties on a proportional basis would not require such arrangements to exist and thus would be easier to implement nationwide than prospective payments. But, as noted above, those payments might not match well with each provider’s costs, and the proportional sharing of bonuses or penalties among participants would weaken their individual incentives to control the total cost of an episode of care. Consequently, hospitals would still be encouraged to make selected arrangements with doctors and postacute care providers to coordinate care and reallocate its financing.

Given those complexities, the effects of broadly bundling Medicare payments for services delivered by a range of individuals and organizations are uncertain. Under the first alternative described above, hospitals and physicians might collaborate to reduce input costs (for example, by consolidating purchases of medical devices and seeking volume discounts from their manufacturers) and then share the gains from doing so. Under the second alternative, hospitals would probably aim to reduce the quantity and intensity of postacute care that their patients received and to economize on the use of physicians’ services during a hospital stay, but they would have flexibility about how to pursue those efforts.

The extent to which hospitals and other providers would be ready to undertake such changes, and ways in which they would react to a bundled-payment system, would naturally vary. Providers that were able to reduce their costs per episode could see meaningful improvements in their profit margins, whereas providers that were not able to reduce costs could see those margins decline significantly. In some cases, providers might respond by increasing the number of admissions and episodes of care that occurred; doctors and hospitals might have stronger incentives to do so than under current law because they could share savings on low-cost cases. Providers might also change their coding practices or take other steps to deliver more services that would be paid for outside the bundled payments.

Effects on Medicare Beneficiaries. The effects of payment bundling on Medicare beneficiaries are also uncertain. With an episode-based payment system, beneficiaries who were hospitalized could benefit from greater coordination of their care, particularly during the transition from the hospital to postacute care. The incentive to avoid hospital readmissions would exist under both of the alternatives described above, but the incentive to limit other costs for postacute care would clearly be stronger under the broader bundling alternative. At the same time, hospitals might reduce the use of physicians’ services or postacute care that was medically beneficial, which could have a negative effect on beneficiaries’ health (although providers would still want to keep their patients from developing complications that generated additional costs for which they would not ultimately be reimbursed).

To address those concerns, implementation of a bundled-payment system could be accompanied by greater monitoring of the quality of patients’ care, or the payment of any bonuses could be made conditional on achieving certain standards for care quality. Currently available measures of care quality are limited, however: They focus mostly on specific processes of care or on whether patients develop certain complications or need to have a surgery redone, but they generally do not reflect patients’ health outcomes (such as improvements in health or avoidance of new medical problems). Although quality measurement is improving over time, developing new quality measures is generally a multiyear process. Achieving agreement about outcome-based measures can be especially challenging because poor outcomes may reflect both the performance of providers and the severity of patients’ health problems—and disentangling those effects is difficult.

Beneficiaries’ cost-sharing requirements would not change under this option, but their out-of-pocket costs could decline if episode-based payments reduced the use of services that require cost sharing (such as visits with physicians or stays in skilled nursing facilities that last longer than 20 days). Reductions in Medicare’s payments for physicians’ and outpatient services covered by Part B of the program would translate into lower Part B premiums for enrollees.

Broader Effects. Widespread application of a bundled-payment policy in Medicare could have a range of spillover effects on care and spending for other patients, but those effects could work in different directions. On the one hand, because Medicare is such a large payer, changing its payment methods could lead providers to adopt lower-cost practice patterns for all of their patients. (Medicare currently accounts for about one-fifth of national health expenditures and about one-fourth of all payments to hospitals.) In turn, those changes could reduce federal spending on the Medicaid program and the costs of federal tax subsidies for private health insurance. Moreover, private insurers and state Medicaid programs could find it easier to implement bundling policies of their own, which would tend to reinforce providers’ incentives to limit the cost of episodes of care.

On the other hand, if providers could not reduce the cost of their care for Medicare patients to the target amounts, the policy change would hurt their financial situation, which they might respond to by trying to shift some of their costs to other payers. Similarly, payment bundling could lead to greater consolidation of providers—in an effort to deliver more integrated care and control the full range of episode costs more directly—which in turn could give providers more bargaining power to secure higher payments from private insurers. Higher private payment rates would translate into higher insurance premiums and would raise the costs of federal tax subsidies for health insurance. And if other payers did not adopt similar payment models, it might not be feasible for providers to change their practice patterns, because reducing the use of services would harm their finances overall.

More broadly, a concern about bundling payments for episodes of care is that—as with fee-for-service payment—providers would still lack clear financial incentives to prevent episodes from occurring and would have only limited incentives to provide less intensive forms of treatment. The amount of the bundled payment would depend mainly on the type of treatment provided; thus, it would be much larger for, say, a heart bypass operation than for an angioplasty to treat a blocked coronary artery. By itself, then, adopting a bundled-payment policy might not slow the development and spread of new medical treatments and technologies, which have historically been key drivers of the overall growth of health care costs. For those reasons, some experts question whether bundled-payment policies are a useful bridge to broader reform of health care payments or instead are a diversion from the efforts needed to develop broader payment models.

Incentives to keep patients healthy and to control total costs for care would be stronger with even broader bundles that encompassed all of the services that a patient receives during a month or a year—such as capitation payments or shared-risk arrangements with accountable care organizations or similar entities. (In shared-risk arrangements, ACOs not only retain part of the savings if they reduce their patients’ total costs for health care below a target amount but also are penalized for part of the added costs if total spending for their patients exceeds the target amount.) Many providers are not ready to accept such degrees of financial risk, however, so bundling payments for episodes of care and encouraging providers to control the costs of those episodes might constitute a useful step, at least for the interim.