Are bundled payments working? Yes and no.

In 2013, the Centers for Medicare & Medicaid Services (CMS) implemented a voluntary “bundled payments” program, to test whether this payment model could reduce costs and overuse. In this program, providers are rewarded for keeping costs under a target price for the total episode of care (defined as the inpatient stay plus professional services after discharge for 1-3 months).

Providers aren’t paid for cost of care over the target, but if they keep the cost under the target, they get to keep the difference. This gives clinicians at different institutions (hospitals, post-acute care, skilled nursing facilities, etc) incentives to work in teams and avoid unnecessary interventions. The impact of value-based payments of global budgeting in cases like Maryland’s all-payer system shows promise. However, value-based payment models do not always work as well in practice as they do in theory, as we’ve seen in evaluations of some population-based payment models and efforts to reduce unnecessary hospitalizations

For the CMS Bundled Payments for Care Improvement (BPCI), the evidence is also mixed, as recent studies have shown. Here’s a rundown: 

Bundles for joint replacements are promising

Joint replacement procedures are in many ways the low-hanging fruit for improvement through bundling: hip and knee replacements are the most common inpatient surgeries among Medicare beneficiaries, but the cost and quality of care for these procedures varies widely across providers in the fee-for-service system. 

About two years after the bundling program was implemented, hospitals that used bundles had lower Medicare payments without a decline in quality, compared to a control group of non-participating hospitals. The bundles save an estimated $1000 per episode, mostly because hospitals that bundled were more judicious about referring patients to institutional post-acute care. An evaluation of the Comprehensive Care for Joint Replacement program, a mandatory bundled payment model for joint replacement instated in 2016, also found that average total payments at participating hospitals declined without significant changes in quality of care. 

However, health professionals and policymakers were still concerned that hospitals in the bundling program would just do more surgeries to make up for lower payments per-surgery. A new study in JAMA explores the potential of unintended consequences of bundling, and found that fortunately, hospitals have not been increasing their volume of surgeries. These results are “incredibly reassuring for the policy” of bundled payments, said lead author Dr. Amol S. Navathe, in a Vox article about the study

Bundles for medical conditions are underwhelming

As promising as bundled payments seem for joint replacements, results for the 40+ other bundled medical conditions are less enthusiastic.

The CMS annual evaluation of BPCI, released a few months ago, compares cost, outcomes, and patient assessment information from participating hospitals with non-participating hospitals for each of the four “models” of bundling available. For the most comprehensive bundling model, researchers evaluated 32 episodes of care over 15 months and found that Medicare payments declined compared to control group hospitals, without a decline in quality. However, the decline was only statistically significant in two non-orthopedic conditions. Also, the decline in Medicare payments didn’t actually result in much savings for Medicare. From the report’s conclusion:

The analysis indicates that in only limited circumstances does BPCI generate savings to the Medicare program. After considering the NPRA paid to participants that reduced their episode payments below their target amount, the Medicare program likely achieved savings on only two clinical episodes. 

A recent study in the New England Journal of Medicine further suggests that cutting costs with bundling is not as simple for non-orthopedic conditions. Looking at just the five most commonly selected medical conditions in BPCI (congestive heart failure, pneumonia, chronic obstructive pulmonary disease, sepsis, and heart attack) from 2013-2015, there was no significant difference in cost reductions or quality changes between participating hospitals and control group hospitals. 

Why the discrepancies?

Why does bundling appear to work better for joint replacement than other bundles? Part of the issue is that patients with these conditions are generally older and less healthy than joint replacement patients. “We’re talking about people with multiple comorbidities and really quite a lot of need, a lot of nursing home use, presumably a lot of really complex disease,” said lead author Dr. Karen Joynt Maddox in an interview in TCTMD

The voluntary nature of BPCI also makes it difficult to evaluate. The CMS report notes that hospitals can withdraw if they start to lose money, complicating both program effectiveness and evaluation. Also, there are likely significant differences between the institutions that decide to participate or not, as Boston researchers note in a recent JAMA Cardiology report. They found that hospitals participating in BPCI cardiac care bundles tend to be larger, not safety net hospitals, and have access to cardiac catheterization labs. This may limit the how we can generalize the results of bundling to other institutions across the country.

Researchers are still optimistic that bundles can be effective with more time, although these studies are a sobering reminder that bundles aren’t a “silver bullet” for reducing health care costs. “In the absence of payment incentives that can get the CEOs and CFOs talking care redesign, a frontline clinician [alone] can’t change the system,” said Maddox.