By Jonathan W. Pearce, CPA, MBA, FHFMA and Jessica Walradt, MS

The moment has arrived. Your first BPCI episodes went live nine months ago. You carefully monitored your monthly analytics reports, honed your post-acute provider network, studiously reduced readmissions, and are cautiously optimistic as you view the icon for the Excel reconciliation workbook on your computer desktop. You’ve eagerly awaited this first CMS reconciliation, counting on it to validate your efforts. As you open the workbook, your heart rate increases slightly.

Success is with you today! You knocked it out of the park with your major joint replacement (MJR) episodes, showing a positive net payment reconciliation amount (NPRA) of 6% of the target price – a respectable return in the first quarter. Your performance in congestive heart failure (CHF) is slightly less positive – a surplus of only 1% of the target amount, with mixed surpluses and deficits across the underlying DRGs. As you ponder the results, you remember that this is only the first of four reconciliations to be performed for this performance quarter, and you're suddenly gripped with fear at the question "Is this real"?

This question will be on the minds of all BPCI participants who receive their first reconciliation next January 1, and is the subject of this article. Many current participants’ NPRAs changed significantly across subsequent reconciliations (“true-ups”), and those participants are concerned that the gains may be reduced or losses increased across the entire reconciliation period.  Unfortunately, the multitude of factors that impact reconciliation prevent conclusive projections; however an understanding of the elements that drive changes to NPRA across true-ups will help providers to assess the financial risks.  

The diagram below shows the factors that can affect the NPRA across true-ups. There are two central components:  1) episode cost, which is the cost to the Medicare program of payments to providers; and 2) target prices.  Claims lag, changes in payment regulations, BPCI episode precedence rules, and patient eligibility standards can change both episode costs and target prices. We’ll examine each of these factors individually.

Claims Lag

The most easily understood component (which shows how complex the rest of the issues are) of the reconciliation puzzle is the lag in claims payment. There's always a delay between the provision of healthcare service and recognition of the payment for that service. The extent of that lag depends on the type of provider billing for the service, the time period in which it was billed, and efficiency of the payer.

Timing of Services Within the Episode

The claims completion rate for a 90-day episode is dependent on many factors, but is largely driven by the type of clinical condition that defines the episode (the clinically-related group of DRGs or “episode family”). Claims completion rates differ across episode families primarily because of the timing of services and the corresponding claims within the episode. As shown in the graph below, about 75% of the episode cost of a MJR episode occurs in the first 30 days post-discharge, whereas about only 45% of CHF episode costs are incurred during the same period. This fact means that a higher portion of MJR episode claims will be generated and paid earlier in the episode, while few claims are generated during the last 30 days. By contrast, 25% of CHF claims are incurred in the last 30 days of the episode, and consequently many CHF episodes claims are not paid until after the episode is completed. Other factors equal, this trend dictates that a greater percentage of MJR episode claims will be paid at the initial reconciliation, which should lead to less variation across true-ups.

Claims Completion Rate

The timing of service delivery across a 90-day episode directly correlates to a given clinical condition’s episode claims completion rate. The graph below shows the claims completion rate for two episode families, CHF and MJR, throughout the episode period, and thereafter. The colors indicate the number of weeks from the start of the episode, while the vertical axis shows the percentage of episode claims that are complete at that point.

For example, at the end of a 90 day CHF episode, approximately 60% of claims have been paid and will be known for reporting and reconciliation. By contrast, about 80% of major joint replacement claims have been paid by the end of the episode. This difference arises from the distribution of cost within the episode; LEJR episodes have little cost in the last 60 days of the episode, whereas CHF episodes have a large amount cost during that period as described above. This is because MJR patients may have some post-acute care in the month after their surgery, but relatively little care after that, and also have relatively few readmissions. By contrast, CHF patients have ongoing chronic conditions that can require long periods of care from post-acute providers, and they also have a much higher rate of readmissions.  

Reconciliation occurs 6 months after the last day of a performance quarter. This time frame provides a minimum three-month claims run-out period for all episodes that occur during the quarter. That is, the first reconciliation occurs 90 days after the end of the last episode of the quarter. For example, episodes in the first quarter 2015 can begin on January 1, but also could start as late as March 31. These episodes would end on or before June 30. The reconciliation results are posted three months later, on October 1. This time period allows for somewhat less than three months of claims lag, as extracting claims from the Medicare data warehouse and preparing corresponding reports takes time.

The graph shows that approximately 94% of CHF claims have been processed and paid at the initial reconciliation, while about 97% of MJR claims are paid.

Initially, it may appear that the understatement of claims may produce inflated NPRA results that will be wiped out when subsequent claims are paid, but that isn't how BPCI works. Since the trend factors used to calculate targets are based on national episode costs, and national episode cost are based on claims expense for the same period, there are unpaid claims in the national episodes costs, which will initially lower the targets As long as your claims are billed and paid at the same rate as the national average, there shouldn't be any net effect on NPRA. Unfortunately, it rarely works out that way. We'll talk about the net effect of claims lag later in this article.

Loss or Gain of Episodes

Another factor affecting the NPRA across reconciliations is changes in number of episodes in each DRG. This situation can occur more frequently than one might expect; changes in patient eligibility, reclassifications to different DRGs, and/or readmissions can all contribute to changing episode counts.  

Readmissions

Episodes can be eliminated if they are determined to be readmissions in pre-existing episodes. In most cases these situations are known during the first reconciliation, but in some cases they are corrected until subsequent reconciliations.

Readmissions can affect different episode families in different ways. A readmission in the MJR episode family for a subsequent MJR (e.g., a replacement of the left knee when the initial admission was for the right knee replacement) will cause a new episode to be triggered and the initial episode to be retroactively deleted. By contrast, a CHF patient is more likely to have unplanned readmissions to other hospitals, which may also be participating with CHF in BPCI. Thus, a readmission for CHF within an existing CHF episode may initially create a new episode, but would ultimately be dropped once it is determined that the readmission occurred within the 90-day period of a pre-existing episode.

Loss of episodes from readmissions may not be known until reconciliation occurs since CMS does not reconcile these readmissions in the monthly data. In the monthly data, the episodes initiated by the readmissions will appear, and will remain until reconciliation, at which time they’ll be removed from later monthly data. In most cases this adjudication is completed for the first reconciliation; however we’ve seen the adjudication occur in later reconciliations with the incorrect episodes showing up in the initial reconciliation(s).

DRG Reclassification

Another factor affecting the episode count within DRGs is the reclassification of episodes into different DRGs. This occurs on the inpatient claim but the change affects the classification of the BPCI episode as well. This can happen if CMS rejects the grouping of the diagnosis and/or procedure codes into the DRG placed onto the index admission claim and requires the claim to be resubmitted with a different DRG. This may occur, for example, in a DRG with a MCC classification if the diagnosis codes that create the MCC classification cannot be documented. In that case an episode having a MCC DRG with a high target rate may be reclassified to a "CC" or “w/o CC” DRG having a lower target rate.

Impact of Episode Count Changes

The loss or gain of an episode can affect the reconciliation different ways depending on the cost of the episode lost. Eliminating a high-cost episode whose cost exceeds the target will improve financial performance, while eliminating episode whose cost is below the target will decrease financial performance. The table below shows an example of the loss of a high-cost and low-cost episode on the average episode cost of a population of episodes. Initially the average episode cost is $25,921, but removing the lowest-cost episode increases the average to $26,040, while removing the highest-cost episode reduces the cost to $25,809.

Changes in Average Episode Costs

All of the above factors come together to influence the average episode cost (Medicare payments for services delivered during an episode), which is one of the two main factors used to compute the final NPRA.  Based on the claims completion graphs above, it might appear that the average episode cost would increase significantly from the first reconciliation to the second, and then increase more slowly in the following reconciliations. This pattern frequently occurs, but occasionally a different pattern is observed. The table below shows the relative episode costs for several participants in the CHF episode family. The expected pattern (low value followed by three approximately equal higher values, like this   ) frequently occurs but other patterns are annoyingly common. This indicates the difficulty in predicting the changes in average episode costs across reconciliations. The graphs below show the relationships among average episode costs across the initial reconciliation and three true-ups for several hospitals and DRGs.

As can be seen, some DRGs follow the expected patterns that would be predicted from the normal completion of claims, while other DRGs follow a very different pattern based on how claims are paid for that DRG across the reconciliation periods.

Changes in Targets

Now that we’ve covered changes in the average episode costs, let’s look at the other component of the NPRA – the target payment rate. The average episode cost is compared to the target and the excess or deficit of that calculation becomes the NPRA for that DRG in that reconciliation period.

Trend Factor

The target rate is calculated from the average episode cost in the baseline period multiplied by the “trend factor” computed for each reconciliation. The trend factor is computed by CMS from a national population of episodes by computing the average episode cost in that population for the baseline period and then for each performance period, and dividing the second amount by the first.

Note that the average episode cost in the performance period is based on the claims paid as of that reconciliation date, and will therefore increase slightly across reconciliations for the same performance period as additional claims are paid. Theoretically the target would increase at the same rate as the participant’s average episode cost if the participant’s claims were paid at the same speed as those of the national population. If the participant’s claims are paid faster than the national average payment rate, it would have lower performance in the earlier reconciliations and performance would improve slightly in later reconciliations. Conversely, if the participant’s claims are paid at a slower rate the participant's financial performance would steadily decrease as costs increased more quickly than the target rates. These differences in claims payment may cause targets to increase at a different rate than your episode costs, leading to differences in the NPRA across reconciliations for the same participation period.

Baseline Cost

The other contributor to the target is the baseline cost. Since the baseline occurred from 2009-2012, it may seem that those claims are “set in stone” and won’t change much. Unfortunately, there are several factors that can retroactively change your baseline episode costs.

Episode Precedence

The first factor is episode precedence. CMS created precedence rules to ensure that no patient is in two episodes simultaneously, as could occur if multiple providers in a given market are participating in BPCI. Your baseline costs were computed assuming that certain other providers were participating in BPCI, and that readmissions to or from those providers would affect the specific episodes in your baseline cost. No new participants can enter BPCI or add episodes; however participants can still drop episodes or withdraw completely from the program, and that withdrawal can affect your baseline. If a local provider was participating in CHF episodes, for example, and a patient in one of those episodes was admitted into your hospital, that patient wouldn’t be a new CHF episode to you; instead it would be a readmission to the other hospital.  However, if that provider withdraws from participating in CHF, CMS will recalculate the baseline including that readmission as a new CHF episode, which will change your baseline cost for that CHF DRG. This will probably be infrequent but it may occur occasionally.

Payment Policies

BPCI payment policy changes can also impact baseline costs. An example may help clarify this issue. In examining the reconciliations in the fourth quarter of 2014 we found that some of the target rates for our clients had dropped from the amounts in the third quarter 2014 reconciliation. We examined the claims data provided in each reconciliation and found that it was identical, but a few of the episode payment amounts computed by CMS were different in the two reconciliation periods. Digging into those claims we realized that they were for readmissions for heart transplants. This was the cause of the change in episode cost, since CMS revised the exclusion policy for heart transplants between the third and fourth quarters of 2014, excluding the transplants from all episode families.  One hospital had three transplants in the baseline, so excluding the costs of those transplants from those episodes reduced the average baseline episode cost for those DRGs by more than $2,000. This effect would be different among participating hospitals, who would have had different rates of heart transplant readmissions in the baseline period; consequently the effect on the target prices was not consistent across all participants.

Implications

A small bead of sweat has emerged across your brow. You mutter to yourself “Geez, what the heck do I tell the physicians about all of this? They’re all focused on gainsharing but I’m afraid to make any payments to them. All of this might go up in smoke at the next reconciliation. What should I do? Mom was right – I should have been a pharmacist instead of studying finance.”

Relax – it’s not all that bad, but you should make some accommodations for the potential changes in reconciliation. Since the largest changes in reconciliation (other than those resulting from CMS policy changes) generally occur between the first and second reconciliation, it’s probably prudent to wait for the second reconciliation before disbursing any gainsharing funds. And even in subsequent reconciliations, a withhold amount of about 5% of NPRA for surgical cases and 20% for medical cases is prudent. Changes in the BPCI payment rules, such as the exclusion of heart transplants described above, can significantly affect later reconciliations.

So, as the great Yogi Berra once said, "It ain't over until it's over", but with some knowledge of how reconciliations can change you can hopefully assess the risk of changes in your reconciliations, provide appropriate caveats and limitations on gainsharing and mitigate the uncertainty until the final true-up.

Jonathan Pearce is the Founding Principal of Singletrack Analytics and manages the analytics services provided to about 50 BPCI participants. He can be contacted at jon@singletrackanalytics.com.

Jessica Walradt is a Senior Payment Reform Specialist at the Association of American Medical Colleges. In this role, she supports academic medical centers’ involvement in the Bundled Payments for Care Improvement initiative using data and policy analyses to explain financial trends and provide strategic advice regarding episode selection and risk mitigation. She can be contacted at jwalradt@aamc.org.