Why Medicare's Bundled Payment is Better than Commercial Programs

Submitted by jonpearce on Sun, 2014-02-23 15:31

A professional colleague recently commented that some of his clients were considering bypassing the Medicare Bundled Payments for Care Improvement (BPCI) program and were focusing on pursuing bundled payments with commercial payers. As we’ve previously written, we think that BPCI is one of the best advanced payment system options available, and it worthy of consideration by almost every hospital. Here’s why:

Payment methodology – BPCI uses a “discount” arrangement, under which the participating provider agrees to accept a slight reduction (2  - 3%) from its own historical episode cost as a target price. The target amount does not vary because the Medicare program is taking its saving off of the top; hence the provider organization can retain all “savings” that occur when costs are below the target, and can use them to cover the costs of care management programs or for other purposes. These differences, and our reasons for believing that discount arrangements are preferable to shared savings arrangements, are described in “Emerging Payment Models for ACOs and Bundled Payment Participants” Most commercial payers appear to be focusing on shared savings arrangements, which incorporate the conflicting financial incentives for providers that are described in the abovementioned article.  In addition, from the payer’s point of view discount arrangements should also be attractive, because the payer would receive the discount on every episode, regardless of whether any savings have been achieved by the provider(s). Right now, only about half of ACOs are apparently showing “shared savings”, and hence only half of the payers involved with those ACOs are gleaning any benefit from the program. In a discount arrangement payers would benefit from each admission.

Availability of data – As part of the BPCI program, the Centers for Medicare and Medicaid Services (CMS) provide an extremely broad range of historical and ongoing data that allows interested provider organizations to explore the details of each and every episode. They can apply a wide range of analytical techniques to identify opportunities and risk before making a decision to participate, and can use that knowledge to develop the arrangements with post-acute providers that will be critical to success. CMS also provides the detailed data behind the target price calculations, creating full transparency. The CMS episode definitions are fully documented; there are no “black boxes” in this process. By contrast, commercial payer-based bundled payment programs generally do not make the historical data available, which may not allow applicants to understand what they are getting into in these programs. Commercial payers frequently use proprietary software to develop the episodes and manage the payments that don’t allow the providers to audit the outputs.  Our experience working with commercial payers' data has been that this data is fragmented, inconsistent, not well-documented, and difficult to use. Providers that are considering participating in commercial payer arrangements should seek assurance that they can load, analyze and understand the data that the payer will provide.

National participation – As a national program, BPCI has a large group of participants, all subject to the same rules and receiving the same set of data in the same format.  This allows for collaborations and efficiencies among and between participating organizations.  Along with our partners at DataGen, Singletrack provides analytic services to the Association of American Medical Colleges’ convened BPCI group and has seen how this arrangement facilitates communication among participating hospitals to enhance their understanding of the program, provide consistent high-value analyses and reports, and identify and facilitate the development of best practices in care redesign with hospitals across the country. By contrast, most commercial payers’ programs are local and specific, leaving hospitals essentially on their own in figuring out how to achieve good results.

Episode definition – Bundled payment programs are gravitating towards two different approaches for defining the episodes.  One approach is to slice the episodes thinly, excluding as many risk-producing and non-directly related elements as possible. In these programs, episodes are defined by ICD-9 codes (which brings up the question of what happens when ICD-10 is implemented), and carefully carve out characteristics (such as excluding patients with BMIs above a certain threshold) who might create high costs to create the bundle.  The episode is built from the bottom up, so to speak, and can be very narrowly defined.

The other approach is to use broad episode definitions, with limited exclusions for unrelated services. This top-down approach uses DRGs, or groups of DRGs, to define the episodes. Rather than attempting to slice out risk, these programs mitigate risk by increasing the population size which makes the effect of outliers less significant. As we noted in “Episode Choices and Bundled Payment Risk”, increasing the population included in the payment program, even by combining clinically-dissimilar DRGs, reduces statistical variation in the average episode costs. By allowing hospitals to participate in a wide range of DRG families, the BPCI program creates larger risk pools that should lower the overall statistical variation in average episode costs. We believe that this approach is more effective for reducing risk than the more narrow episode approach used by many commercial payers.

Population size – This issue is similar to the one described above, but is created from a different source. For virtually all non-pediatric hospitals, Medicare patients constitute the largest single payer group. In almost every clinical category except obstetrics, Medicare patients will have the largest number of cases and, as noted above, larger population sizes create lower risk. We examined that issue in “Assessing Participation Risk in the Medicare Bundled Payment Initiative”, in which we noted the effect of population size and episode variation on the overall average episode cost. For example, we pointed out that the average episode cost (which is the final measure of risk) may vary more than 30% for episodes such as congestive heart failure (that has a coefficient of variation of about .8) in populations of fewer than 100 patients. This means that the participant’s average episode cost will be affected far more by population randomness than it will by any clinical initiatives. Therefore, the key to reducing risk is to have a larger population, which generally isn’t possible with individual commercial payers. Within a smaller commercial population, some risk mitigation may be possible with lower-variation surgical DRGs, but rarely is the commercial population of patients with chronic diseases large enough to reduce statistical risk to an acceptable level.  Only Medicare has enough of these patients to make statistical risk manageable.

Sensible outlier limits – BPCI includes three risk tracks from which the participant can choose depending on their risk/opportunity perspective, and we’ve written several articles (here, here and here) on risk track selection. We believe that CMS made wise choices in the selection of these risk tracks; the 99%/1% track protects against the rare train-wreck outlier without eliminating opportunity for reducing high-cost cases, while the 75%/5% track eliminates the majority of variation (reducing risk, but also opportunity) for participants who plan to focus primarily on internal cost reductions and don’t see much opportunity in affecting post-discharge costs or readmissions. The 95% limit is a middle-ground option for participants whose risk aversion exceeds their expectation for achieving cost savings. Commercial payers may also implement risk tracks and outlier limits, but it’s difficult to see how they could improve on those selected in BPCI. 

Mandatory bundled payment adoption – Singletrack Analytics generally doesn’t prognosticate, but we’re willing to stick our necks out on this one – we believe that there’s a high probability that Medicare will move to a mandatory bundled payment program for certain procedures like major joint replacements within several years. The recent announcement opening the application process to new applicants across the country highlights CMS’ interest in and commitment to gathering data to develop these payment systems, which are far easier to implement than population-based programs like ACOs. If this is the case, it behooves providers to gather as much information as possible about the current care continuum of their patients, and to begin building care management and open information technology corridors with partner providers.

Conclusion – For many reasons, we believe that the Medicare BPCI program is the first bundled payment initiative that hospitals should explore, and one that almost every hospital should consider. The wealth of information about post-discharge care contained within the CMS data alone is extremely valuable, the size of the participation base (more than 500 hospitals) and ability to communicate within convener groups, along with optimal definition of episodes and a favorable payment methodology makes this program significantly more attractive than isolated programs with commercial payers.