Evaluating Opportunities in the Medicare Bundled Payment Program

The Medicare Bundled Payment for Care Improvement (BPCI) program offers hospitals and post-acute providers the opportunity to work collaboratively to coordinate care occurring during and after an inpatient stay, and to retain most of the savings that may be generated as a result of the elimination of unnecessary provider services. The Centers for Medicare and Medicaid Services (CMS) is allowing providers that are already participating in this program to evaluate expanding their participation to additional episode families. This article provides a structured approach to those providers for analyzing and evaluating new episode families for inclusion in BPCI.

Several terms will be used in this article that require definition. First, unless otherwise indicated, the term “cost” refers to the cost to the Medicare program of services provided to patients. These costs are generally payments to providers, paid at the respective Medicare payment rates. The other term is “episode family” which describes a group of related DRGs all of which must be included for BPCI participation if any are included. For example, to participate in the “Simple pneumonia and respiratory infections” episode family, DRGs 177-179 and 193-195 must be included.  We use the term “episode family” since an episode may also refer to a single instance of an inpatient stay followed by a post-acute period.


The BPCI program presents three major areas of opportunity for cost reduction and shared savings. These areas and the strategies to address them are not mutually exclusive and participants may decide to pursue any or all of them.

Reducing the incidence of high-cost episodes

Generally, the cost drivers for the most expensive episodes are readmissions or extraordinarily high-cost post-acute services. The primary indicator of significant opportunities here is the presence of a large number of these expensive episodes, which could be defined as the percentage of cases that exceed the BPCI published 90th and 99th percentile limits. The strategy is to identify the root causes for those high-cost outliers and eliminate them.  Some episode families are naturally attractive to this strategy; for example chronic disease DRGs such as heart failure or CHF, which have relatively high numbers of readmissions when compared to surgical DRGs such as joint replacements or CABGs.

Moving the overall cost curve downward

 Indicators of opportunities in this area are: a large number of post-acute providers having differing average episode costs; different post-acute pathways that drive variances in episode costs; or other similar metrics. The strategy here is to target reductions in total episode costs through revision of the treatment process for each and every episode. It generally requires developing relationships with a selected group of physicians and post-acute providers who can provide a higher degree of care coordination.

Reducing the providers' internal costs

Although many of these opportunities exist regardless of the bundled payment program, the ability to share savings with physicians through the BPCI's gain-sharing safe harbors may create a higher level of engagement and success. The strategy here is to target the internal costs to the provider (generally the inpatient acute care hospital) by reducing length of stay or the costs of medical supplies, drugs or implants.

Understanding the episode participation landscape

Because most providers have limited, if any, access to information (data) about the services provided to patients outside of that provider's setting, they have little understanding of the whole episode composition.  Therefore, in order for any BPCI participant to evaluate which, if any, of the above opportunities/strategies apply, it is important to obtain an understanding of the “landscape” of the services within an episode or episode family.

Total financial risk

Under BPCI, the contracting provider organization is at risk for the total episode cost, including all services by post-acute providers. A report such as the one shown below is necessary to understand the total financial commitment involved in a particular episode family. This table also shows the coefficient of variation for each DRG, which will be discussed below, as well as the percentage of cases that would cross the 99th percentile high-cost outlier threshold.

Variation in average costs per episode 

Each DRG-defined episode has its own cost distribution, which looks like the graph below, with each bar representing a single patient episode within that DRG definition. At the low end of the cost distribution are episodes that are closest to the minimum cost of the episode, i.e. the hospital’s DRG payment plus the RBRVS payments to the physicians. The DRG episode costs increase from there as post-acute services, including readmissions, occur and increase.

The cost curve graph is useful for several reasons. First, it can point out opportunities as listed in our first two areas for cost reductions by highlighting levels and variations in costs. Second, it will be extremely helpful in selecting the BPCI outlier threshold later on, if the provider decides to participate in this episode family.   In our “A Tale of Two Cities” article we discussed the different decisions that hospitals may make about outlier thresholds depending on how they view their ability to affect those high costs.

The metric used to measure the variation between episodes' costs is the coefficient of variation (CV).  The CV is the standard deviation of the costs among episodes divided by the mean cost of the episodes.  The higher the CV, the more variation.  Most surgical DRGs have CVs close to .5, while the DRGs for chronic diseases like heart failure and COPD have CVs closer to 1.0.


The number of cases that a provider encounters for each episode family in a given year is also a significant factor in episode selection. Episodes having low volumes may not present sufficient opportunity to plan and implement care redesign strategies, gain physician interest and buy-in, and measure and evaluate results. They may not have sufficient historical data available to perform the analyses described in this article and to achieve sufficiently meaningful results to support a participation decision.

Another critical factor is that the amount of random variation increases significantly as episode volume decreases. As we’ve described in our article on Episode Choices and Bundled Payment Risk, the amount of variation in the average cost per episode, when calculated from all of the episodes occurring within a year, increases significantly as the episode volume decreases, even when all episodes are selected from the same population. In other words, a DRG having 50 cases per year at Hospital A will have a higher amount of random variation than the same DRG at Hospital B with 100 cases per year. The amount of that variation caused by the inherent variation in cost between individual episodes (the “coefficient of variation” described above) also influences the variation in average episode cost. The graph below, taken from the referenced article, shows the percentage in the average cost per episode across all episodes occurring in a year (on the Y axis) as a function of the number of episodes (on the X axis) and the CV of the underlying episodes (the size of the circles). As can be seen, DRG-defined episodes having lower CVs (such as joint surgery) can achieve lower overall average cost variations with smaller populations than the same population size in DRG-defined episodes with higher CVs, such as heart failure and CHF. Thus, it’s important to consider both factors (population size and between-episode variation) in evaluating the effect of randomness in episode family selection.

Variation in the average episode cost of groups of episode families

In our article on Episode Choices and Bundled Payment Risk, we describe how the statistical variation in average episode cost decreases as the number of episode families increases. For example, the coefficient of variation (CV) of participating in the highest volume episode family (major joint replacement) in our sample was .48, but it decreased to .31 when we added the four other highest-volume episode families – even though those families had higher individual CVs. This occurs because the larger total case volume allows lower-cost cases in one episode family to offset higher-cost cases in another family. Therefore, potential participants should evaluate the risk associated with the composite of all episode families under consideration, rather than the variation of each individual episode family.

Understanding the services provided

A major prerequisite to success under bundled payments is understanding the types of services provided and the care pathways involved. This information is especially enlightening for hospitals that have never had access to data about the post-acute services that are provided to their patients after discharge. Charts and tables such as those below display the costs of those services arrayed across the episode cost continuum and can provide significant insights as to the major drivers of episode costs.  The evaluator must then determine whether there is opportunity to improve efficiencies and reduce costs based on the current care setting landscape.

Cost trends by provider type

Another important factor in episode selection is the trend in overall episode costs, and in the components of those costs. The ultimate success under bundled payment programs occurs when costs decrease[1]; therefore an increasing overall cost trend in a particular episode family may make it problematic to achieve a cost reduction. In some episode families costs may already be decreasing, which may make them more appropriate candidates. It is also useful to look at trends by cost component (particularly Skilled Nursing care, Home Health care, and Inpatient Rehabilitation), since those individual costs may be manageable through the participating organization’s care management initiatives.  The graph below shows the trends in HHA costs for a specific provider by calendar quarter, applying a linear trendline to show the direction of the cost trends.

One important factor to be considered when analyzing cost trends is that episodes take significant time to complete and for all of the claims data to be processed through the claims payment system. For 90-day episodes, reliable episode cost data is not available until about six months after the episode initiation[2]; using data for a shorter period will result in undercounting episode costs and indicate an incorrect downward trend.

Understanding the provider landscape

Post-acute providers: Once the costs and trends of provider types (SNFs, HHAs, IP rehab facilities and physicians) are understood, it is helpful to understand the cost patterns of the individual providers within those classes. In many cases we’ve found significant differences between the per-episode costs of post-acute providers that may represent opportunities for developing coordinated treatment protocols for these services.  Therefore, it is useful to review the number of episodes, average cost per episode, and other factors for those higher volume post-acute providers, to understand those differences and identify potential areas in which coordination of services may provide some cost benefits. The table below shows the top 10 SNFs for a selected DRG-defined episode, and the average costs per episode associated with each provider.

Physicians: It is important to identify the physicians who will be involved in providing care to BPCI patients and to bring those physicians into the bundled payment team. Average episode costs, treatment protocols, use of post-acute services and even readmission rates may vary by physician and understanding and resolving these differences may be a key strategy for care improvement and cost reduction. As episode families are evaluated for bundling the ability to work with the major physicians in each episode family will be a key factor. The table below shows the costs per episode, readmission statistics and outlier counts for patients categorized by the Admitting Physician specified on the inpatient claim.

First post-acute setting: The post-acute provider setting (SNF, HHA, IP rehab, etc.) to which the patient is referred after discharge is often a significant factor in episode costs. Some bundled payment participants have discovered that the episodes in which patients are routinely referred to SNFs have lower post-acute provider costs but higher readmission rates and, hence, higher total episode costs.  In such instances the additional cost of referring patients to inpatient rehabilitation facilities may be recouped in lower readmission costs.  The first step towards improving efficiencies is to identify where they occur.

Understanding readmission rates

The strategy of reducing readmissions is frequently utilized by bundled payment participants as a way to reduce both episode costs and boost quality scores that are based on readmission rates. There are many ways to look for opportunities to reduce readmissions:

·         The readmission DRG as related to the index DRG –in many cases it useful to relate the readmission DRG(s) to the index (episode) DRG. When the readmission DRG is related to the index DRG (or is in the same DRG family), changes to the treatment protocol and care hand-offs may be effective in reducing readmission rates.

·         The readmission DRG without regard to the index DRG – in some cases the readmit DRG(s) may not be clinically related to the episode DRG, or the readmission DRG may be part of a larger facility-wide issue. For example, readmissions for septicemia may indicate a unit or facility-wide problem to be addressed procedurally.

·         The hospital in which the readmission occurs – This is of concern to the participating hospital because the “cost” of a readmission back to the index hospital is the hospital’s internal marginal cost of the readmission. The Medicare cost of the readmission, which is charged against the episode target cost, is offset by the DRG payment that the index hospital receives for the readmission.  However, a readmission to a different hospital results in a charge against the episode target cost with no corresponding payment to the index hospital for the readmission. While no readmission is desirable, readmissions back to the index hospital are preferable to readmissions to other hospitals.

·          The timing of the readmission – The potential participant should assess the aging of their readmissions, for any episode family under consideration, to determine whether there is an opportunity to reduce those readmissions in a meaningful way.  Readmissions that occur shortly after discharge are more likely to be related to the care received during the index admission, or during the initial post-acute period and therefore may be more controllable through care management processes.  Readmissions that occur later in the episode are less likely to be directly related to the index stay or the care transition plan, and are not as easily controlled. 

·         The post-acute setting or provider from which the readmission occurred  – Readmission rates may be affected not only by the post-acute provider type, but also by the provider itself.  If a post-acute care provider does not have the clinical resources to evaluate and/or manage complications, their protocol may require sending the patient to a hospital ED for evaluation. Examining the patterns of readmission sources may provide insights into the opportunities to partner with certain post-acute providers and avoid readmissions.

One example of readmission analysis is shown below:

Putting it together

We have highlighted some specific areas of risk and opportunity to look for when evaluating new episode families for bundled payments. Once these analyses have been completed, an overall selection strategy can be developed. The strategy should be to select episodes in which:

·         There is enough episode cost variation to create opportunities for shared savings, or where other specific cost-saving targets have been established;

·         Enough of the drivers of those variations have been identified as actionable; and

·         The clinicians and providers who are significant participants in the episode are part of the team that will be coordinating and managing the care re-design.

Other considerations for episode selection are:

·         Increasing the number of episode families decreases the statistical risk and may not decrease the opportunities for cost savings; therefore, selecting a larger number of episode families may be a good risk-reduction strategy.

·         Physicians are the drivers of care (and therefore cost) change within the care continuum. Episode families in which a “physician champion” can be identified have the greatest opportunity for success.

[1] In the BPCI program the actual target calculation that will be performed by CMS will include a component that factors in the nation-wide cost trends for that episode family. Therefore, targets could increase or decrease based on how costs across the country change over time.

[2] One month to allow for episodes beginning at the end of the month, plus three months of the episode period, plus two months for claims runout.


Evaluating Opportunities in the Medicare Bundled Payment Program.pdf392.31 KB