The analyses presented in this report demonstrate the differences in episode composition and payments associated with beneficiaries initiating their PAC episode with an acute hospitalization versus the community entrants who initiate PAC use without a prior hospitalization. This work is an important complement to earlier work by ASPE and RTI looking at PAC utilization following an acute hospitalization. The current work highlights differences in the types of diagnoses and patterns of service use within episodes for those initiating services with an acute hospitalization compared with community entrants. The diagnoses of community entrants are in general more characteristic of chronic conditions and the need for ongoing care while the diagnoses for beneficiaries with acute hospital-initiated episodes are more acute in nature. The longitudinal cohort analysis also provides additional detail on the longer term use patterns for beneficiaries initiating PAC service use in different settings beyond what we observe in the first episode of care. Together, these analyses demonstrate the shorter term, higher cost nature of service use for beneficiaries initiating a PAC episode with an acute hospitalization compared with the longer term, relatively lower cost service use of those entering PAC directly from the community. Although these community entrants are a small proportion of total PAC users, their service use patterns are unique and indicative of longer term service use.
As in earlier work, the results looking at the impact of different episode definitions reveal substantial differences in the services included and the associated episode payments across the definitions examined. Because LTCH and IRF are most often the first sites of PAC for beneficiaries using these services, use and payments for LTCH and IRF are less sensitive to the shorter fixed-length episode definitions than HHA and SNF services, which are also often first sites of PAC but are often used following discharge from LTCH and IRF. Although a higher proportion of beneficiaries may use SNF and HHA under the longer, variable-length episode definitions, this proportion decreases when looking at the fixed-length definitions or definitions that exclude readmissions and subsequent PAC user. The inclusion or exclusion of readmissions and subsequent PAC use has a significant impact on total episode length of stay and episode payments when comparing across definitions. The illustration of these differences and their impact can help inform policy discussions on whether a readmission is part of an episode or the start of a new episode.
Another important contribution of this work is in the ability to track PAC use patterns over time using the cross-sectional analytic samples. In examining beneficiary episode utilization in 2006, 2007, and 2008, it is possible to see the slight changes in use. For example, there has been a slight increase in the percentage of beneficiaries discharged to PAC nationally (from 37.1 percent in 2006 to 38.7 percent in 2008). But of particular note is the slight change in the proportion of beneficiaries discharged to the different PAC settings over the 3-year period—specifically, the percentage of beneficiaries discharged to IRF. In 2006, 9.7 percent of beneficiaries were discharged to IRF, but this decreased to 8.6 percent in 2008, and MS-DRG-specific analysis highlighted the significant decrease in the proportion of beneficiaries with joint replacement (MS-DRG 470) discharged to IRF. These small changes are likely a result of CMS's phasing in changes in compliance criteria associated with IRF payment, but the ability to detect these changes is an important benefit of the analytic file.
The increased sample size of this work has also provided a valuable opportunity to look at differences in PAC use and payments at smaller geographic levels. The data provided here allow for more detailed analysis of MS-DRG level utilization at the state and CBSA level than were possible using the 5 percent files in earlier work. The results of this geographic analysis again highlight that provider supply and geography are significant drivers of PAC utilization and spending and that policy discussions related to PAC episode payment must recognize these issues and include discussions of the implications of benchmarking given different practice patterns and provider supply. This concept is also highlighted in the use per user, use per PAC user, and use per hospital discharge calculations of mean payments per service type. In areas of the country with LTCHs, a beneficiary using an LTCH may have mean payments of over $33,000 although the mean LTCH use per hospital discharge can be closer to $250 (depending on the episode definition examined). This raises important policy considerations if an episode payment were to be set based on national per-discharge use patterns.
The results of this work are meant to inform the larger discussion of PAC episodes and bundled payment policy. The information presented provides additional context to what episodes of care look like by different types of initiating events (acute versus community entrant) and what the implications are for setting an episode one way versus another. Although there has be much discussion of episodes definitions that include service use for 30 days following hospital discharge, it is important to point out that "30 days following hospital discharge" can mean different things, and more precise language may be necessary. Does this include any service initiating within 30 days of acute hospital discharge? Or is it prorated to exclude service use after day 30? Are readmissions included or excluded? What about episodes that do not start with an acute hospitalization? The work presented here shows the impact of these different dimensions and the importance of these considerations as policy makers continue to consider bundled payment.
This work also demonstrates that the services in a beneficiary's trajectory included or excluded from an episode do vary by MS-DRG. For example, over three quarters of PAC payments for the 30-day variable episode (episode definition A) are also captured by the 30-day fixed-length episode definition where any claim initiating in the 30-days following acute discharge is included in the episode (episode definition C) for MS-DRG 470, "major joint replacement or reattachment," but this decreases to less than two thirds of PAC payments for beneficiaries in MS-DRG 194, "simple pneumonia & pleurisy w CC," and MS-DRG 690, "kidney & urinary tract infections." However, under the 30-day fixed-length episode definition where services are prorated to reflect 30 calendar days of use, less than half of the PAC payments associated with the 30-day variable episode definition are captured for four out of the five top MS-DRGs by volume of PAC use. This indicates that in considering an episode definition, it is important to consider the proportion of services (and dollars associated with this use) in a clinical trajectory that the episode definition is actually capturing. It is also necessary to consider how service use will be paid for after an episode is complete, particularly if an episode includes 30 calendar days and payment for a service may be only partially covered under a bundle. The MS-DRG-specific analysis for acute hospital-initiated episodes also indicates that the decision to pay a bundle per hospital discharge versus per PAC user will have different implications for different MS-DRGs. A per-discharge bundle means something different for an MS-DRG with a high proportion of beneficiaries discharged to PAC services compared with an MS-DRG with a lower proportion of beneficiaries going on to use PAC services.
Next steps in ASPE and RTI's exploration of PAC episodes include work related to episode risk adjustment. This work is in collaboration with CMS and uses data from the Post Acute Care Payment Reform Demonstration, including the uniform patient assessment instrument and the Continuity Assessment Record and Evaluation (CARE) data. The goal of this work is to learn more about how patient assessment data can be used to predict episode utilization and payments. Additional work with ASPE will also examine the potential to use data from current assessment instruments and claims data. This work is part of ongoing research at ASPE on PAC episodes.