Jonathan Pearce, CPA, FHFMA and Coleen Kivlahan, MD, MSPH

Many participants in “Phase I” of the Medicare Bundled Payment for Care Improvement (BPCI) program were surprised by their first look at the impact of CMMI’s BPCI precedence rules when comparing  two sets of baseline data recently provided by CMS. This data show the number of episodes and target prices for episodes under two scenarios:

• Assuming that all Phase 1 BPCI participants would continue in BPCI as of April 2015; and
• Assuming only the subject hospital would continue.

In some cases the episode counts were significantly lower under the first assumption, leaving many hospitals to wonder where all of those episodes went. This article addresses the issue of "episode precedence", describes the precedence determination process, and discusses several strategies that are being employed by participants to deal with this issue.

Episode precedence is important because it can affect the number of episodes and their costs. In some cases, a provider may lose so many episodes because of precedence rules that it becomes unfeasible to participate in certain episode families. In other cases participation may still be feasible, but significantly less attractive.

During the course of 2014, CMMI made a policy decision that determines which episodes are assigned to which providers.  On the surface, the policy was straightforward: physician group practices have precedence over hospitals, and hospitals who join BPCI before other hospitals have precedence over those hospitals.  In practice, the policy has created considerable risk of diminishing participation in BPCI.  In addition, revisions to the outlier assignment strategy were also made in 2014 that added the data of BPCI participation to the already-confusing logic of determining which outliers will create new episodes. 

A provider group may enter Phase II of BPCI, commit resources to care management and data analysis, and build strategic alliances with other providers, only to find that another organization preempts many of their episodes.

A participating organization may find many of its episodes preempted, leading to a population size that is too low to provide stability in episode costs. As we’ve previously written, lower episode counts result in higher variations in average episode costs, which can create wide swings in quarterly reconciliation amounts.

Consider the case of a hospital having two orthopedic groups that perform major joint replacement procedures. One group focuses on hip procedures (Group A), while the other performs primarily knee replacements (Group B). The average episode cost of hip and knee replacement episodes are significantly different from each other. In this case, if Group A participated as episode initiators it would own all of its episodes. This would leave the hospital with the remaining episodes, which were performed by Group B. This would significantly affect the average episode costs of the hospital. Fortunately this would not affect reconciliation because the target prices are computed taking into account the effect of participation of the physician group.  However, financial planning and potential gainsharing payments could be affected by this type of unexpected change in episode costs.

The BPCI initiative provides significant opportunities for providers across the episode timeframe to gainshare, cooperate in providing information and coordinate care management activities. The precedence policies have hampered the ability to develop these relationships by adding uncertainty as to the ownership of episodes. The types of relationships in the market may change based on participation of other parties; for example a Model 3 provider may choose to be an episode initiator in certain episodes where the index hospital is not participating in BPCI, but might instead develop a gainsharing relationship with the index hospital if the hospital serves as the episode initiator.

There are several circumstances in which an index admission at a hospital is not counted as a BPCI episode for that hospital, or in which an episode can retroactively terminate.

This occurs when a physician group practice (PGP) and hospital are participating in the same episode family. Depending on other conditions described below, a PGP may own almost all of the episodes in an episode family in which it is the admitting or operating physician. In some cases this situation may be known to the hospital; for example the hospital may know that the orthopedists are planning to participate on their own. In other cases some physician group may not make the hospital aware of their participation.

This occurs when a patient who is in an episode at one participating hospital is readmitted to another participating hospital. The final ownership of the episode depends on the readmission DRG and the participation status of the readmission hospital. The extent to which this will occur is not generally known in advance (although it may be predictable from historical data), but participants should recognize that readmissions to other hospitals and readmissions for included conditions will impact their case volumes, and should plan accordingly.

This occurs when a patient in an episode is transferred to another participating hospital during the index admission. The final ownership of the episode depends on the model of the participants and several other characteristics. Transfers are relatively rare for most hospitals, but may be significant for some hospitals like academic medical centers. For example, smaller community hospitals may frequently transfer certain types of cases to larger hospitals that may also be participating in BPCI.

This occurs for Model 3 participants when the preceding admission is with a participating Model 2 participant. The extent to which this will occur is generally knowable in advance, and savvy Model 3 participants will structure relationships with their primary Model 2 episode initiators to participate in gainsharing programs in episodes in which the Model 3 provider would be preempted from becoming an episode initiator themselves.

In some cases the episode precedence is based on the BPCI model of the participants. Model 4, which is a less popular model, can capture an episode occurring at a Model 2 participant if certain readmission or transfer characteristics exist. Model 2 captures episodes that might become Model 3 episodes in almost all circumstances.

Model 4 versus Model 2

It’s often said that “Model 4 trumps Model 2”, but this is a somewhat oversimplified characterization. A Model 2 and Model 4 hospital couldn’t each originate the same episode at the same time since a patient is only admitted at a single hospital. (By contrast, a PGP and a hospital CAN originate the same episode at the same time, which necessitates the precedence rule described later.) Therefore, the only time a comparison between these models can occur is in a readmission or a transfer that would initiate a new episode for a Model 4 participant but not for a Model 2 participant. For readmissions, this happens if a readmission occurs from an index admission at a Model 2 hospital. If the readmission occurs at a Model 2 hospital, the previous episode will terminate and a new episode will start only if the readmission is unrelated to the index admission. By contrast, if the readmission is at a Model 4 hospital, a new episode starts regardless of the relationship between the index admission and the readmission.

Transfers to a Model 2 hospital do not create a new episode; the episode remains with the original provider. If the transfer is to a Model 4 hospital that is not subject to the “less than full MS DRG transfer payment rules”, the original episode is terminated and a new Model 4 episode begins. In this case the Model 4 participant’s rules are different from those of Model 2 participants.

Thus, in readmissions the Model 4 hospital may gain episodes that a Model 2 hospital would not gain only in the case in which a related admission occurs. If the readmission is unrelated, there’s no difference in treatment between Model 2 and Model 4. In transfers, most transfers to Model 4 participants will create a new episode for that participant, whereas transfers to Model 2 participants will not create a new episode.

Of course, Model 4 participants are far less numerous than Model 2 participants, so these rules rarely take effect.

Model 2 versus Model 3

When the index admission for a potential Model 3 episode occurs at a Model 2 or Model 4 participant, the Model 2 or 4 participants will always own the episode.

If during a Model 3 episode, a patient is admitted to a Model 2 participant in a DRG in which the Model 2 participant is participating, the determination is based on the relationship of the readmission DRG to the index admission DRG. If the readmission DRG is excluded (unrelated) to the index DRG, the Model 3 episode retroactively ceases to exist, and a new episode initiates for the Model 2 participant. Conversely, if the readmission DRG is in a related DRG, the Model 3 episode will continue and no new episode will occur at the Model 2 participant.

For example, assume a patient is discharged after a knee replacement in DRG 470 from a hospital that is not participating in BPCI in that DRG, and then moves to a SNF that is participating in Model 3 in that DRG. This will initiate a Model 3 episode for the SNF. During that episode the patient was readmitted in DRG 66 (stroke, which is a related DRG for DRG 470) to a Model 2 participant participating in DRG 66.  In this case the Model 3 episode will continue and the stroke episode will not occur.

However, if the patient is readmitted to a Model 2 participant in DRG 26 (craniotomy), which is an unrelated DRG for the DRG 470 index admission, the Model 3 episode will be retroactively eliminated and the episode for the craniotomy will start at the admitting hospital.

Post-acute providers in Model 3 are at the bottom of the BPCI food chain, in that they can only own episodes that originate in a hospital that is not participating in the respective episode family. Unfortunately, the episodes that are most attractive to Model 3 participants are the same episodes in which Model 2 participants frequently participate. For example, post-acute providers play a large role in major joint episodes, but these episodes are also the most popular for hospitals and PGPs. This is unfortunate, since post-acute providers may have the best opportunity to create cost savings in these types of episodes.

These precedence rules relate to the initial ownership of the episode and are based on the type of the participant and the date on which the participant went “at risk”.

Participation date

The participation date refers to the date at which the participants joined "Phase 2" of BPCI, and were at risk for the episodes. This criterion is applied in the case of a tie in episode ownership from the above criteria; for example when two Model 2 participants may be eligible to own the same episode. There are three possible participation dates: October 1, 2013, January 1, 2014, and April 1, 2015 (which has now been extended to July 2015 but will have the same precedence as April 2015). If the potential episode owners are in the same model, the participant having the earlier participation date will own the episode. This issue occurs between PGP and hospital participants, both of whom might be initiators of the same episode, and can also affect the assignment of new episodes to participants.

Participant type

If there continues to be a tie among the participants (which will usually only occur when a hospital and PGP are both participating in the same DRG), the PGP will own the episode.

This is one of the most pernicious situations in BPCI, which occurs when a hospital and a PGP group are participating in the same episode families. This may occur, for example, if the hospital and an orthopedics group both participate in the Major Joint Replacement of the Lower Extremities episode family. Because of the precedence rules, the PGP will own all of the episodes in which a PGP physician was the operating physician. The hospital may have invested in significant physical and personnel infrastructure, developed a care management team, and built robust analytical capabilities, only to have no opportunity to recover those costs through savings generated in those episodes. If multiple orthopedics groups practice at the hospital, the non-participating group’s case mix may be different; for example if one group specializes in elective hip replacements and the other specializes in hip fracture and hip replacement revisions. Since the hospital’s base period costs will be computed only with the non-participating group’s cases there will be no immediate bias between the target prices and the episode costs; however if the participating group trended towards the lower cost elective hip procedures and the non-participating group performs primarily the hip fractures and revisions, the target rates would not change commensurately and the hospital would almost always lose money.

Many hospitals who are in Phase I, but whose physicians are considering BPCI participation, are waiting to see what the PGPs decide to do. If the PGPs participate, the hospital won’t participate. This situation may end up being problematic for PGP's, since it is unlikely physician groups can create the type of infrastructure required for success without cooperation and additional resources of the hospital.

This is one of the most confusing precedence areas, since an episode can be lost if it IS a readmission or if it CONTAINS a readmission. This is most likely to occur in a market area in which multiple providers participate in BPCI, but it can also occur within an individual hospital. The key issue is the episodes in which each participant is involved, since an episode can only be lost to a participant who’s participating in the “capturing” DRG. Hospitals in larger metropolitan areas in which BPCI is popular among providers may be at greater risk for losing episodes than in suburban hospitals where there aren’t other local BPCI participants. Another factor is the participation date of the participant, as described above.

Admissions into a hospital might not create episodes if they are readmissions from the same or another facility. This can occur if:

• The index admission is at a BPCI-participating provider participating in the DRG of the index admission, and
• The DRG of the admission to the second hospital is a related DRG to the DRG of the index admission, OR
• If the second hospital participated in BPCI earlier than the first hospital (in which case the relationship of the index and readmission DRGs is irrelevant).

For example, a patient is admitted to Hospital A in COPD DRG, in which that hospital is participating. That patient is then readmitted to Hospital B in a CHF DRG, and Hospital B did not enter BPCI before Hospital A. The admission to Hospital B will not create a new CHF episode, because the CHF admission is a related readmission to the COPD admission and will be included in the COPD episode at Hospital A. (Note that Hospitals A and B can be the same hospital.)

However, if Hospital B had entered BPCI before Hospital A, Hospital B would have ended up with the CHF episode and the COPD episode at Hospital A would be retroactively eliminated.

The key to understanding this issue is to focus on the readmission, which can potentially become a new episode in the hospital in which the readmission occurs. If that readmission is a related DRG for many episodes and is a frequent DRG for readmissions, it has a relatively high probability of being included in a previous episode.

For example, the CHF DRGs are included (related) conditions in most other episode families. Therefore, if a CHF readmission occurs, there is a high probability that it is related to the index DRG, and will therefore not start a new episode. (This presumes that the index provider is participating in BPCI in the index DRG.)

By contrast, Major Joint Replacement DRGs are not related readmissions for most index DRGs. Therefore, admissions for these DRGs are most likely to create new episodes regardless of whether there is a preceding index admission.

Note that the first consideration is the participation date of the hospitals, followed by the relationship of the readmission DRG to the index DRG. If one hospital has an earlier participation date, it will always own an episode in which it is participating, regardless of the relationship between the index and readmission DRGs.

The issue of related and unrelated readmissions also can affect an episode that’s not a readmission from a previous episode. This can happen in the following instance:

• A readmission occurs in a provider that’s participating in the readmission DRG,
• The readmission occurs in a DRG that is unrelated to the index admission, OR
• If the second hospital participated in BPCI earlier than the first hospital (in which case the relationship of the index and readmission DRGs is irrelevant).

For example, a patient is admitted to Hospital A in a CHF DRG. That patient is then readmitted to Hospital B in a Major Joint Replacement DRG, which is unrelated to the CHF index admission. Hospital B is participating in the Major Joint DRGs and Hospital A did not enter BPCI before Hospital B. In this case the CHF episode would retroactively be cancelled and the Major Joint Replacement episode would begin at Hospital B. Again, Hospitals A and B can be the same hospital.

However, if Hospital A had entered the BPCI program before Hospital B, Hospital A would have retained the CHF episode and the Major Joint episode at Hospital B would not occur.

Undoubtedly the most significant precedence problem for hospitals is the PGP precedence over hospitals that started BPCI at the same time. Participation of a major PGP will almost always eliminate the hospital from participating in that episode family, since there would be no opportunity for savings except through possible gainsharing payments from the physicians. Consequently, the hospital would have little opportunity to recover the costs of care management activities, and therefore is unlikely to engage in those activities. The physicians and their convener group (if any) will need to provide the care management and analytics resources necessary to achieve cost savings.

Possible strategies here may include participating in a gain sharing program between the PGP and hospital, negotiating a sufficient share to offset the costs of developing and maintaining personnel and care management infrastructure.

In markets in which several large hospitals are in Model 2 simultaneously, there are several opportunities for effective execution of BPCI.  These include collaborative approaches to episode selection based on physician staff expertise and market strength in key episodes (cardiac valve surgery, major joints, CHF). This actually creates an opportunity for right-sizing services and volume distribution in a community: not every health system needs to offer major joint services as a primary service line, for example.  In addition, only about 50% of overall readmissions in this population return to the index hospital; this inefficiency in management, repeat testing and additional procedures could be reduced by collaborative strategies that promote readmission back to the index hospital when clinically indicated.  Finally, BPCI fundamentally anticipates that providers compete against themselves to reduce cost and improve quality. There is nothing proprietary about evidence-based care pathways or care manager training or community care models. The BPCI hospitals in Model 2 could collaborate on care management resources that reduce costs for all and reduce duplication currently inherent in having each hospital build its own care management unit.   

Model 3 providers have a community view of patients

Relationships between hospitals and Model 3 post-acute care (PAC) providers (SNFs, HHAs and rehab providers) can be of great benefit in BPCI. When the bundled payment-related gainsharing payment can supplant their current fee-for-service payments, PAC providers can provide additional resources to coordinate and improve post-acute care, reduce length of stay and handle medical conditions internally that would otherwise cause readmissions. Understanding the heightened scrutiny on their services under bundled payment, many of these providers are becoming proactive, reaching out to participating BPCI hospitals in an effort to retain or increase market share, and also to be included in gain sharing programs. Since a Model 2 participant will generally own an episode instead of a Model 3 participant in the same episode, PAC providers are frequently choosing to gain share with the hospital for episode families in which the hospital is participating, and participate in episode families in which the hospital is not participating.

This situation is currently playing out among two of our clients, a health system and a large SNF chain, along with the orthopedists at the hospital. The orthopedists, who are affiliated with a large national convener group, are evaluating participating in the major joint replacement episode families. If this occurs, the remaining volume in these episodes is too small to provide a stable statistical base within which the hospital can participate. It also provides insufficient opportunity to create savings to cover the additional expenses that the hospital would incur in providing care management, data analysis, and other related activities. Therefore, the hospital is delaying the participation decision until they know whether the physicians will participate.

Meanwhile, the SNF chain, which was looking to partner with the hospital for these episodes, can't make the participation decision either. The SNF chain has the opportunity to allocate resources to better manage lengths of stay, and also to utilize physicians in the SNF facilities to reduce the number of readmissions. However, there is no economic model for this unless they participate in a gainsharing program with the episode initiator, whomever that may be in this case. The physicians’ convener group, which holds the BPCI purse strings for the physicians, has indicated that there is no room in their financial model for gainsharing payments to be made to post-acute providers; therefore the SNF chain probably won't participate if the physicians do. (One wonders how physicians believe they will achieve savings without cooperation of post-acute providers and the hospital, but that's a question for a future article.)

So at this point the major players in the entire episode are at a standstill waiting for the physician’s participation decision. If the physicians participate, they’ll essentially be on their own working with their convener group, and possibly with some analytical and care management support for the hospital if the physicians provide enough gainsharing opportunity to cover the hospital’s cost. They’ll have no support from the major post-acute provider, however. If the physicians decline to be the episode-initiating BPCI participant, they can be part of the hospital’s gainsharing program, which can also include the SNF provider. The amounts that the physicians can receive is limited to 50% of their fee-for-service payments, but the possibility of creating savings that reach that amount may be higher working with the hospital and SNF as partners than it would be working alone.

This hospital is in a major metropolitan area in which many other hospitals are in “Phase I” of BPCI. When the hospital received its baseline data sets from CMS, confusion resulted from a lower than expected volume of cases.   CMS provided two sets of data – an “all participants” set and an “only you” set. The “all participants” data set assumes that all participants in Phase 1 (the non-risk phase) of BPCI will continue to participate in BPCI, and for all of the episode families in which they receive data. The “only you” set assumes that only the subject hospital participates, and that no other participants will gain ownership of episodes.

The hospital found that about 40% of the episodes in many episode families that were present in the “only you” data were not present in the “all participants” data. This was initially confusing but can be explained by the fact that readmissions in most medical DRGs are related to almost all index DRGs. Since CMMI assumes (in the all participants data set) that the local hospital participants will be participating in ALL 48 episodes, many of the medical DRGs were eliminated as readmissions under a previous episode at one of the hospitals. Unfortunately this didn’t provide much guidance for episode selection, since most hospitals will eventually participate in only a few episode families.  After deeper analysis, we found that about half of the lost episodes were attributable to index admissions in popular DRGs, which provided a basis for making assumptions about ultimate episode volumes.

Many participants participate in the Major Joint Replacement of the Lower Extremities episode family because it has large volume, significant post-acute cost, relatively few readmissions and a long history of bundling, and thereby provides a relatively stable statistical and financial platform with a relatively good potential to achieve savings. This allows participants to add other episode families where the financial results are less certain, but for which the cost of participation may not be offset by the potential for savings. In this case, the hospital wished to participate in the Major Joints family and also in some other episode families, with the Major Joints family providing the stability for participation in the other families.

Unfortunately, an orthopedic PGP at the hospital is considering participating in BPCI as episode initiators under a different convener, and the precedence rules would assign all of the Major Joint episodes to the PGP. This leaves the hospital without the stability of those episodes, and may preclude the hospital from participating in the additional episodes that would expand the capabilities of the hospital and its care management team to deal with more varied but less voluminous episodes (such as medical condition like CHF, COPD). The PGP can skim off the episodes most likely to achieve rapid savings, and the hospital can’t participate alone in those conditions which are more costly and require larger up front investments.

Jonathan Pearce is a Principal with Singletrack Analytics. Dr. Kivlahan is a family medicine physician and a Senior Director at the Association of American Medical Colleges, where she directs their BPCI convener program.