Among the standard complaints about the American health care system is that care is expensive and wasteful. These two problems are related, and to address them, Medicare has new ways to pay for care.
Until recently, Medicare paid for each health care service and reimbursed each health care organization separately. It didn’t matter if tests were duplicated or if a more efficient way of delivering care was available — as long as doctors and organizations were paid for what they did, they just kept providing care the way they always had.
But ordinary people do not think this way. We focus on solving our health problem, not which — or how many — discrete health care services might address it. New Medicare programs are devised to more closely align how care is paid for with what we want that care to achieve.
One of these programs is known as bundled payments. Instead of paying separately for every health care service associated with a medical event, you pay (or Medicare pays, in this case) one price for the entire episode. If health care providers can address the problem for less, they keep the difference, or some of it. If they spend more, they lose money. Bundled payment programs vary, but some also include penalties for poor quality or bonuses for good quality.
Medicare has several bundled payment programs for hip and knee replacements — the most common type of Medicare procedures — and associated care that takes place within 90 days. This includes the operation itself, as well as follow-up rehabilitation (also known as post-acute care). In theory, if doctors and hospitals get one payment encompassing all this, they will better coordinate their efforts to limit waste and keep costs down.
Do bundled payments work? They certainly appear promising, at least for some treatments. But it’s important to conduct rigorous evaluations.
Previous studies for Medicare by the Lewin Group and other researchers suggest that Medicare’s Bundled Payments for Care Improvement program has reduced the amount Medicare pays for each hip and knee replacement.
But that doesn’t mean the program saved money over all.
One possible issue would be if, despite saving money per procedure, health care providers wastefully increased the number of procedures — replacing hips and knees that they might not otherwise. A related concern is if hospitals try to increase profits by nudging services toward patients who may not need a procedure as much as patients with more severe and more expensive conditions. An average joint replacement costs $26,000, split almost equally between the initial procedure and post-acute care. But more expensive cases can be $75,000 to $125,000 — a costly proposition for hospitals.
A recent study published in JAMA examined whether the volume of Medicare-financed hip and knee replacements changed in the markets served by hospitals that volunteered for a bundled payments program, relative to markets with no hospitals joining the program. It found no evidence that the bundled payment program increased hip and knee replacement volume, and it found almost no evidence that hospitals skewed their services toward patients whose procedures cost less.
“These results suggest bundled payments are a win-win,” said Ezekiel Emanuel, a co-author of the study. “They save payers like Medicare money and encourage hospitals and physicians to be more efficient in the delivery of care.”
But Robert Berenson, a fellow at the Urban Institute, urges some caution. “Studying one kind of procedure doesn’t tell you much about the rest of health care,” he said. “A lot of health care is not like knee and hip replacements.”
Michael Chernew, a Harvard health economist, agreed. “Bundles can certainly be a helpful tool in fostering greater efficiency in our health care system,” he said. “But the findings for hip and knee replacements may not generalize to other types of care.”
Christine Yee, a health economist with the Partnered Evidence-Based Policy Resource Center at the Boston Veterans Affairs Healthcare System, has studied Medicare’s previous efforts and summarized studies about them. (I and several others were also involved in compiling that summary.) “Medicare has tried bundled payments in one form or another for more than three decades,” Ms. Yee said. “They tend to save money, and when post-acute care is included in the bundle, use of those kinds of services often goes down.”
One limitation shared by all of these studies is that they are voluntary: No hospital is required to participate. Nor are they randomized into the new payment system (treatment) or business as usual (control). Therefore we can’t be certain that apparent savings are real. Maybe hospitals that joined the bundled payment programs are more efficient (or can more easily become so) than the ones that didn’t.
Another new study in JAMA examines a mandatory, randomized trial of bundled payments. On April 1, 2016, Medicare randomly assigned 75 markets to be subject to bundled payments for knee and hip replacements and 121 markets to business as usual. This policy experiment, known as the Comprehensive Care for Joint Replacement program, will continue for five years. The JAMA study analyzed just the first year of data.
“In this first look at the data, we examined post-acute care because it is an area where there is concern about overuse,” said Amy Finkelstein, an M.I.T. health economist and an author of the study. “In addition, prior work suggested that it’s a type of care that hospitals can often avoid.”
The study found that bundled payments reduced the use of post-acute care by about 3 percent, which is less than what prior studies found. “Those prior studies weren’t randomized trials, so some of the savings they estimate may really be due to which hospitals chose to participate in bundled payment programs,” Ms. Finkelstein said. Despite reduced post-acute care use, the study did not find savings to Medicare once the costs of paying out bonuses were factored in. The study also found no evidence of harm to health care quality, no increase in the volume of hip and knee replacements, and no change in the types of patients treated.
“Savings could emerge in later years because it may take time for hospitals to fully change their behavior, “ Ms. Finkelstein said. In addition, the program’s financial incentives will increase over time; bonuses for saving money and penalties for failing to do so will rise.
On the other hand, Dr. Berenson said, health care providers could figure out how to work the system: “In three to five years, we may see volume go up in a way that offsets savings through reduced payments for a procedure. We’ll wait and see.”
Medicare put its best foot forward by using a randomized design. Not only were the markets selected in a randomized fashion, but providers in those markets were also required to participate. Though common in medical studies, randomization is rare in health care policy, as is mandatory participation. Nearly 80 percent of medical studies are randomized trials, but less than 20 percent of studies testing health system change are. Organizations that would be subject to the experiments often strongly resist randomizing health system changes and forcing providers to participate.
Unfortunately, the randomization of the Comprehensive Care for Joint Replacement program will be partly compromised in coming years. The Centers for Medicare and Medicaid Services announced last year that hospitals in only half of markets under the program would have to stay in it. Participation is voluntary in the other half, and only one-quarter of hospitals opted in.
Going to a partly voluntary program will make it harder to learn about longer-term effects, Ms. Finkelstein said, and to get at the answers we’re seeking.
As an alternative payment model, bundled payments hold the potential to improve the value of care by holding clinicians and organizations accountable for episode-specific quality and costs. Medicare has scaled bundled payments nationwide via several programs that define episodes based on hospitalization and up to 90 days of post-acute care.
However, the impact of bundled payments appears to differ between surgical and medical episodes. On one hand, Medicare has achieved promising results from bundling surgical care for lower extremity joint replacement. Medicare’s evaluation of its largest national bundled payment program, the Bundled Payments for Care Improvement (BPCI) initiative, has demonstrated that participation in joint replacement bundles is associated with a 3.8 percent decrease in per-episode spending with stable-to-improved quality. Other work evaluating the experience of high performers in BPCI demonstrates that bundled payments may reduce the costs of joint replacement episodes by up to 20 percent, with sizeable bonuses to physicians and hospitals and small improvements in quality – outcomes that, if scalable, would represent a win for patients, clinicians, organizations, and Medicare alike. On the other hand, recent evidence corroborates analyses conducted by Medicare and its contractor, suggesting that as designed, bundles for medical conditions such as congestive heart failure (CHF) and chronic obstructive pulmonary disease (COPD) are not associated with significant changes in quality or Medicare spending.
Therefore, one critical aspect of understanding the impact of bundled payments is evaluating how and why it differs for surgical versus medical care. This insight is particularly important given that surgical and medical episodes will be further expanded at a national scale in the forthcoming Bundled Payments for Care Improvement Advanced (BPCI-Advanced) program. In this post, we describe why the lack of episode savings in Medicare’s medical bundles may not be unexpected, why policymakers should not abandon medical bundles, and why existing evidence poses three important policy implications for the future of medical bundles.
By designing bundles that span hospitalization and post-acute care, Medicare has emphasized reductions in post-acute utilization and spending as major financial savings opportunities. While this approach suits surgical care in which a procedure triggers a cascade of acute and post-acute care, it may pose several challenges for episodes related to medical conditions. First, spending patterns for surgical versus medical care differ, more predictably spiking after surgical procedures but adopting a more cyclical pattern for chronic medical conditions. Accordingly, hospitalization may be more appropriate as an episode trigger for surgical episodes than for medical ones.
In surgical care such as joint replacement, hospitalization is a clear, distinct trigger before which there would be no expected episode-related utilization (e.g., little to no joint replacement-associated services prior to the surgery) and after which there is a distinct cascade of related utilization (e.g., physical rehabilitation, wound care, and post-surgical follow-up). In contrast, hospitalization only represents one aspect and phase of management for medical conditions such as CHF and COPD, which span outpatient, inpatient, emergency department, and post-acute settings over longer periods.
Second, physicians’ and hospitals’ ability to impact post-acute care utilization and spending may differ between surgical and medical episodes. This difference is not simply a reflection of the proportion of total episode spending paid to institutional post-acute care providers. For example, spending on skilled nursing facilities and inpatient rehabilitation facilities was only marginally higher for joint replacement compared with five medical conditions (26 versus 24 percent, respectively).
Rather, differences in the ability to impact post-acute care utilization may relate to the types of services provided in institutional post-acute settings for surgical versus medical patients. For surgical episodes, care at skilled nursing facilities often involves discrete, time-limited activities such as physical rehabilitation to achieve post-surgical recovery (e.g., strengthening, functional improvement). In contrast, given the natural history of diseases such as CHF and COPD, institutional post-acute care services for medical patients generally involve complex tasks such as medication management(e.g., diuretics) and multifaceted occupational therapy to promote self-care and activities of daily living. Consequently, hospitals in surgical bundles have achieved savings without compromising quality by shifting discharges from skilled nursing facilities and inpatient rehabilitation facilities towards home, with either home health or self-care. However, it remains unclear if similar efforts are possible or appropriate for the types of post-acute care that are often required as part of medical bundles. In turn, discharge patterns in medical bundles may reflect the less predictably defined roles of institutional post-acute care providers.
Another reason that shifting discharges away from institutional post-acute care providers may prove challenging under medical bundles is that they involve different types of patients than those often involved in surgical bundles. As noted recently, patients in medical bundles tend to be older and at higher risk for poverty and disability than patients in joint replacement bundles. In turn, patients receiving care for medical conditions may have greater clinical needs during and after hospitalization than patients undergoing surgical procedures.
Collectively, these dynamics offer insight into why clinicians bundling care for medical conditions have not achieved savings in BPCI. They also have implications for the design of medical bundles going forward.
First, Medicare could consider modifying when and how medical episodes begin. Rather than being a necessary pre-condition for an episode, hospitalization itself may be a modifiable element of variation in medical conditions. Consequently, unlike in surgical procedures, using hospitalization as a medical episode trigger may miss the opportunity to include cost and utilization variation across the care continuum. As an alternative, if medical episodes were triggered in the outpatient setting – for example, after two specialty office visits within one month — provides might be better able to coordinate medical bundles with other efforts to improve value (e.g., payment models such as accountable care organizations and policies such as the Hospital Readmissions Reduction Program).
Second, Medicare could design medical bundles so that the emphasis on improvement is not restricted to care delivered in the post-discharge period. While variation reduction is not an absolute requisite for performance in bundled payments, care standardization remains an important organizational strategy for improving episode-based care. Creating incentives to focus on outpatient and pre-discharge elements may be particularly fruitful for medical bundles given the complexities of ongoing (in the ambulatory setting) and acute (in the hospital setting) management, and the possibility that practice redesign may require more time and greater effort than in surgical episodes.
Third, more data are needed to understand the impact of medical bundles and how best to design them in the future. To date, we have only early evidence about the impact of medical bundles in BPCI (the mean number of months of BPCI participation was 7 months for these hospitals). Given that other alternative payment models such as accountable care organizations have required three or more years before participants achieved savings, medical bundled payment policy should be guided by longer-term evaluations. Such evaluations should also closely monitor the programs for unintended effects: while it may be reassuring that medical bundles have not appeared to inadvertently lead to more readmissions or emergency department visits, vigilance is nonetheless required given the history of racial disparities in access that stem from quality- and value-based policies. Finally, future work can speed progress towards improvement by providing more detailed descriptions of the utilization and spending patterns of patients involved in medical bundles, as well as highlighting the experiences of high-performing providers.
While existing evidence suggests that medical bundles may not improve the value of care, these findings are not necessarily unexpected, and policymakers should not abandon the effort to bundle the care of medical conditions. Instead, in addition to more long-term evaluations, the design of medical bundles may be improved in the future by modifying how they are triggered and which phases of care they capture.
The CMS has finalized its decision to toss two mandatory bundled-payment models and cut down the number of providers required to participate in a third.
Only 34 geographic areas will be required to participate in the Comprehensive Care for Joint Replacement Model, or CJR, according to a rulemaking released Thursday. Initially, 67 geographic areas were supposed to participate.
Up to 470 hospitals are expected to continue to operate under the model. That includes the CMS’ estimate that 60 to 80 hospitals will voluntarily participate in CJR. Originally, 800 acute-care hospitals would have participated under the program.
With so many hospitals getting a reprieve, the CMS estimates the model will save $106 million less over the next three years versus what it would have saved if CJR had remained mandatory for all 67 geographic areas. The model is now expected to save $189 million over those years instead of $295 million.
The rule comes weeks after the CMS finalized a proposal to allow knee-replacement surgeries to take place in outpatient settings. When the proposal was released in July, some questioned if it was an attempt to undermine the CJR model.
The CMS has also finalized plans to cancel the Episode Payment Models and the Cardiac Rehabilitation Incentive Payment Model, which were scheduled to begin on Jan. 1, 2018. Eliminating these models gives the CMS greater flexibility to design and test innovations that will improve quality and care coordination across the inpatient and post-acute-care spectrum, the agency said.
These cardiac pay models were estimated to save Medicare $170 million collectively over five years.
The agency acknowledged that some hospitals wanted the models to continue on a voluntary basis, as they had already invested resources to launch them, but said those arguments were not detailed enough for the agency to do so.
“We note that commenters did not provide enough detail about the hiring status or educational and licensing requirements of any care coordinator positions they may have created and filled for us to quantify an economic impact for these case coordination investments,” the CMS said.
On average, hospitals have five full-time employees, including clinical staff, tracking and reporting quality measures under value-based models, according to the AHA. They are also spending approximately $709,000 annually on the administrative aspects of quality reporting.
More broadly, the average community hospital spends $7.6 million annually on administrative costs to meet a subset of federal mandates that cut across quality reporting, record-keeping and meaningful use compliance, according to the trade group.
Ultimately, the CMS decided to not alter the design of these models to allow for voluntary participation since that would potentially involve restructuring the model, payment methodologies, financial arrangement provisions and quality measures, and it did not believe that such alterations would offer providers enough time to prepare for the changes before the planned Jan. 1, 2018 start date.
The CMS acknowledged that hospitals and other stakeholders have voiced concerns that the Trump administration may not be as committed to value-based care as the Obama administration, but it insists that’s not true. The CMS said the Trump administration just believes voluntary models are the better way to go.
“We take seriously the commenters’ concerns about the urgency of continuing our movement toward value-based care in order to accommodate an aging population with increasing levels of chronic conditions,” the agency said in the rule. “We continue to believe that value-based payment methodologies will play an essential role in lowering costs and improving quality of care, which will be necessary in order to maintain Medicare’s fiscal solvency.”
Several authors from the Brookings Institution recently argued in favor of making Medicare’s Bundled Payment for Care Improvement (BPCI) initiative mandatory. While the principles guiding their recommendations are sound, the recommendations themselves fail to acknowledge five fatal methodological flaws within the BPCI program. Their analysis also overlooks the most logical and reasonable alternative: a physician-focused episode-of-care payment model.
Another report suggest value-based payment models will continue even, if in a different form, under the new administration’s governance of the U.S. Department of Health and Human Services, according to a Health Affairs report.
“The election of Donald Trump might change the strategy of advancing healthcare reform, but the movement toward value-based care both preceded the Affordable Care Act and has bipartisan support,” the authors said.
If Tom Price is confirmed as secretary and Seema Verma administrator of the Centers for Medicare and Medicaid Services Administrator, the agencies will support new value-based payment models said authors David Muhlestein, Natalie Burton and Lia Winfield.
But Price has already voiced his opposition to mandatory models such as bundled payments.
CMS, which has 74 healthcare initiatives and programs in different stages of research, testing, and adoption, recently proposed to make its cardiac care bundle mandatory and said opportunities exist for bundles that consider multiple chronic conditions.
While payment innovation may continue, the agency needs to articulate its overall strategy in four focus areas, the authors said.
The first is the expansion of the population-based model and disease-specific model.
A recent study linked Medicare’s Bundled Payment for Care Improvement (BPCI) focusing on lower extremity joint replacements to lower costs of care. An accompanying editorial disputed the findings claiming that a reported rise in the volume of procedures between baseline and performance years for the intervention cohort, relative to a comparison group, indicated that total spend may have increased instead of decreasing.
Our analysis of Medicare fee-for-service data from 2010 through 2015 by hospital, for all lower extremity joint replacements in the entire United States, indicates no evidence that participation in the BPCI was responsible for any increase in the volume of procedures between baseline and performance years, thus reaffirming the original paper’s findings. We also find that trends in regional demographic and market characteristics explain the change in volume over time.