Claims Completion and Lag in Bundled Payment Episodes

As we produce monthly BPCI Advanced reports for our clients, we are frequently asked how "complete" the data is for certain episodes at a particular point in time. Participants understandably want to know when an episode’s claims are “complete”, at which point the episode’s cost can be compared to the target price. This is especially critical in the BPCI Advanced program, in which participants often must make participation choices based on limited data. We'll explore these answers, but first need to cover some basic information affecting the speed of completion of Medicare bundled payment claims.

The CMS Bundled Payment Cycle

The critical points in the claims payment cycle are shown in the graphic below. The timeframe begins with the index admission and the associated inpatient claim for that admission. Throughout the episode (generally 90 days) services are provided at various points in time. The timing of the services differs among various types of episodes, as will be shown below. When the service is complete, it is billed by the provider. That bill is then sent to CMS, where it is paid. In Medicare bundled payment programs, CMS extracts claims data from the Medicare claims payment warehouse, processes that data and forwards it to providers or analytics teams who prepare analyses and reports for users. Finally, that data is made available to participant teams who analyze and utilize it for care management, financial projections, and other purposes.

Cost Timing During Episodes

An important differentiator in the speed with which bundled payment episodes are complete is the distribution of costs throughout the episode. In episodes involving elective surgery such as major joint replacement (MJR) most of the services are provided during the index admission and relatively early in the 90 day post-discharge period, which means that those services will be processed and paid sooner. However, in medical episodes such as congestive heart failure (CHF), services are provided throughout the entire 90 day post-discharge period. Services provided near the end of the episode will not have sufficient time to be paid during the episode, and sufficient time after the end of the episode must be provided to allow sufficient claims lag for these claims to be paid.

The graphs below show the periodic and cumulative percentage of claims incurred by week in MJR and CHF episodes. As can be seen, a large percentage of MJR episode cost occurs early in the episode because of the relatively high DRG payment and associated orthopedist surgical payment. The remaining costs are generally for short-term post-discharge SNF or HHA services that rarely extended into the second and third months of the episode. By contrast, CHF episodes have a relatively low DRG and physician payments during the index admission, but have significantly higher costs (many of which are caused by a readmission rate approaching 50%) throughout the episode; therefore the cost of these episodes is distributed throughout the entirety of the episode period.

Each of these total costs is driven by the components of the cost that occur throughout the episode. As shown below, MJR episodes have a significant component of physician and inpatient cost early in the episode (reflecting the DRG and physician payments for the surgery) with significantly less cost later in the episode. In contrast, CHF episodes have lower costs during the index admission because the DRG payment is lower and physician payments for hospital visits are lower than those for orthopedic surgery, but have significantly higher readmission and post-acute costs later in the episode.

These effects are also shown in the cumulative costs, in which about 50% of episode cost in MJR episodes is complete shortly after the index discharge, whereas only about 15% of CHF episode cost is complete at the same point in time. And the point at which costs are incurred has a significant impact on the completion rate; it's apparent at this point that MJR claims will complete much more quickly than CHF claims.

Claims Payment Timing

The above analysis shows the rate at which claims are incurred during a 90 day episode period. But incurred claims are not paid claims – after the claim is incurred it must be billed by the provider and paid by Medicare, and different claim types appear to be paid at different rates as shown in the graph below. As can be seen, carrier (non-institutional Part B claims, primarily physician) and DME claims are billed and paid relatively quickly after the service date, whereas hospital, SNF, and home health claims have a longer billing or payment cycle. Therefore, episodes having larger amounts of slower-paying provider types will have a longer lag from service date to payment, and therefore a slower completion rate.

All of these factors come together in the graph below, which shows the percent of claims that are paid (according to the paid date on the claim) by each week after the start of the episode. This graph shows, for example, that MJR episodes are about 95% complete about 17 weeks after the episode start date, whereas CHF episodes require an additional four weeks to be 95% complete. If all of a provider’s episodes completed at the same rate, the graphs below show the factors that could be applied to each episode to estimate its fully-completed cost.

Episode payment lag variation

But before placing too much confidence in the graph above, we need to be aware of the significant variation in completion rates of individual episodes. While the lines above represent an aggregate of several thousand episodes, the actual monthly episode count in most hospitals is significantly lower, and therefore completion rates of individual episodes can vary significantly from the average shown above. The graphs below show the variation in average completion rates of 100 groups of approximately 50 episodes in each group, with each bar showing the average and two standard deviation range in completion rate for each post-episode week. The completion rates of these groups of episodes varies significantly, often plus or minus ten percent from the average. Thus, it is extremely difficult to predict accurately the remaining episode costs even within several months of the end of the episode.

This graph shows that between about 80% and 90% of MJR claims are complete at the end of the episode (12 weeks after the start date), but only about 60% to 80% of CHF claims are complete at that point. Even six months (26 weeks) after the episode start significant amounts (5%) of CHF claims may remain unpaid.

CMS Claims Processing Timing

One step remains before participants can utilize the CMS data – CMS's data contractor must extract, process and forward the episode and claims data to participants. In the BPCI program this process appears to take about three weeks. BPCI claim feeds, which are generally transmitted on the last Friday of the month, generally include claims paid around the 10th day of that same month, so that process requires about three weeks. We've only received several months of BPCI Advanced data, but it appears that the current CMS BPCI Advanced data contractor requires about five weeks for this process, two weeks longer than the BPCI contractor. So while the claims may be paid in accordance with the graphs above, participants will not actually have the data for evaluation until about a month later.

Conclusion

While it’s highly desirable to estimate episode costs in advance of the CMS reconciliations, the accuracy of such a process is significantly questionable. Small amounts of variation in episode cost exist long after the episode has ended, and these differences can significantly affect episode surplus or deficit amounts. The error in such estimates is often greater than the amount of a surplus or deficit and would render such estimates unusable.  This is why CMS reconciles episodes many months after they have ended, and also why the reconciliation true-ups may be large.

The variation in completion rates appears to decrease about 20 weeks (about 5 months after episode start; 2 months after episode end) after which claims for surgical episodes such as MJR are approximately 95% complete and medical episodes are approximately 90% complete. At that point the completion error rate may be sufficiently low as to allow for relatively accurate comparisons of episode cost with target prices. However, the remaining error should still be considered when utilizing this data for financial reporting.