August 24, 2021
On behalf of our nearly 5,000 member hospitals, health systems and other health care organizations, including approximately 3,000 post-acute care (PAC) members, and our clinician partners – including more than 270,000 affiliated physicians, 2 million nurses and other caregivers – and the 43,000 health care leaders who belong to our professional membership groups, the American Hospital Association (AHA) appreciates the work by the Centers for Medicare & Medicaid Services (CMS), the Department of Health and Human Services Assistant Secretary for Planning and Evaluation (ASPE), and their contractor RTI International (RTI), to develop a new payment model for PAC services. This policy development process, which was mandated by the Improving Medicare Post-Acute Care Transformation (IMPACT) Act of 2014, is extremely complex, particularly its unprecedented attempt to merge four Medicare prospective payment systems (PPS) into one.
AHA and its designees have engaged in RTI’s three technical expert panels (TEP) convened thus far, and plan to actively participate in the next TEP set for late summer 2021. These sessions are intended to help inform and obtain feedback from stakeholders on the ongoing development of a PAC PPS prototype model, which CMS and ASPE plan to convey to Congress in 2022. However, based on the preliminary work RTI shared during its most recent TEP meeting in June, we have substantial concerns about the model’s development. In particular, we are not confident that the eventual prototype will be capable of yielding accurate payments for the full array of patients treated across the four PAC settings. If a PAC PPS is to reliably operate in the post-COVID-19 world, it must ensure access to the proper care at the right time for all PAC patients — especially the severely ill patients who require substantial resources. Yet, payment accuracy, a key component to ensure access, is far from certain based on the data and policy work presented to the TEP so far.
In addition, we are concerned about the timing and transparency of the model. The work RTI presented during its June TEP meeting did not seem to be mature enough to yield a working model that could be submitted to Congress next year. Further, the information shared then as well as through RTI’s follow-up questions to the TEP came without comprehensive underlying data or policy rationales, leaving TEP participants largely unable to fulfill their role. In fact, the absence of key design elements has left us unable to address RTI’s technical follow-up questions. Rather, the questions seem best suited for a stakeholder that possesses a working PAC PPS model, which we, and even RTI, lack. The only entity that we know with a PAC PPS prototype is the Medicare Payment Advisory Commission (MedPAC), which has a framework that significantly differs from the preliminary design direction shown by RTI. Moving forward, TEP members need more detailed information if they, and we, are to meaningfully evaluate RTI’s work and proposals. The AHA strongly urges CMS, ASPE and RTI to address our concerns, which are discussed in more detail below, at the next meeting of the TEP.
UNDERLYING DATA USED
One of our key concerns is RTI’s use of data from 2017 through 2020 to design PAC PPS payment categories. Specifically, these data do not reflect many substantial reforms that recently have occurred in the PAC payment systems – especially the home health (HH), long-term care hospital (LTCH) and skilled nursing facility (SNF) PPSes. They also reflect pre-COVID-19 care patterns, as well as the anomalous care patterns of the pandemic itself. Thus, their use will affect the actual design and restrict the applicability of this model to the post-pandemic landscape. Indeed, RTI itself recognized the shortcomings of using these data to build a PAC PPS prototype. However, while RTI believes that its eventual prototype could simply be updated with post-reform and post-pandemic data — we firmly disagree. Specifically, RTI’s reliance on 2017-2020 data has undoubtedly already influenced and altered the design process, which will affect the eventual framework of the prototype. And it will continue to do so for any future iteration. In order for the model to be relevant for real-world use after the public health emergency (PHE), the PAC PPS design process itself must be based on data from the post-payment reform and post-pandemic era. If it does not, the RTI-recognized limitations of using these old data should be a prominent and comprehensively explained component of its report to Congress.
As an example, AHA analyses1 of inpatient PPS discharge data show that during the PHE period starting early 2020, patients transferred to each of the four PAC settings experienced unique operational and clinical shifts due to a combination of recent PAC payment reforms and the PHE, including:
- A decrease in total case volume;
- An increase in average case-mix index; and
- An increase in average payments per case;
In addition, all PAC settings, especially during a local surge, have experienced the following:
- An increase in personnel costs for many types of physicians, nurses and therapists;
- A shift in the mix of therapy modalities, in large part to augment existing infection control methods; and
- A significant increase in telehealth services.
These sudden and material shifts are expected to have long-lasting effects on PAC operations and must be accounted for in the design of any future payment models.
RTI’S PRELIMINARY FRAMEWORK FOR A PAC PPS CASE-MIX SYSTEM
During the June TEP meeting, RTI did not present a PAC PPS payment model. Rather, it confirmed that its current focus is on the development of possible PAC clinical and payment groupings. Since RTI did not present an overall framework for a unified PAC case-mix system, the PAC PPS design elements that were presented were difficult to put into context and seemed somewhat fragmented. Further, many elements were not comprehensively explained and did not include adequate underlying data, analyses, and discussion of the variables that shaped RTI’s design choices. We describe these concerns in more detail below and urge RTI to address them at the next TEP meeting.
Draft Uniform Clinical Groups. RTI presented a multi-tiered payment system that was based on the structure of the inpatient rehabilitation facility (IRF) PPS and PAC data from 2017 through 2020:
- Twenty-eight Uniform Clinical Groups (UCG, which are a new payment category under development by RTI);
- Each UCG would have up to six subgroups, based on the patient’s function assessment;
- An additional adjustment for comorbidities that require extra resource use; and
- Additional non-clinical adjustments, such as for prior hospital procedures and PAC short stays.
Unfortunately, RTI did not present adequate information on the analyses underlying the UCGs. Stakeholders need more information on these technical design considerations, as well as their related data, to provide informed feedback on the model:
- How robust are the classification and regression tree (CART) splits across the subgroups in the UCG categories?
- How are the statistical cost splits determined to be clinically appropriate? Who made these clinical judgments? Were they generalized? How did they affect the final UCG parameters?
- What sensitivity analyses were conducted to determine if alternative cost splits affect resulting UCG weights, payments, and ultimately patient and provider winners and losers?
- What other variables were included in the CART model?
- What is the unit of observation for cases included in the CART model?
- Do they include only cases that transition directly from community to PAC?
Which Existing PAC Payment Models Provide the Best PAC PPS Template? Thus far, RTI has not provided a clear explanation, or any data, to justify its choice to base its UCG design on the IRF PPS. As such, we are left with many questions as to why it chose the oldest PAC case-mix system as its policy foundation. For example, why are the newest PAC case-mix systems, the “PDGM” for HH and the “PDPM” for SNFs, which include additional risk adjustment levers, and, therefore, seemingly promote greater payment accuracy, not preferable as a basis for the PAC PPS? Why does another PAC payment system, LTCH PPS, which uses the same risk adjusters as the inpatient PPS and divides patients into two sub-categories to attempt to distinguish the sickest patients from those with lower levels of complexity, not yield greater accuracy? Indeed, the highest attainable level of payment accuracy is essential for preserving access to care for medically-complex PAC patients, who have traditionally faced the greatest access challenges. We are especially concerned given that MedPAC’s’ PAC PPS prototype fell short with regard to the predictive power needed to cover the costs of treating this vulnerable patient subgroup.
Design of a PAC Patient Placement Process. We also have concerns that RTI has not, thus far, addressed which entity might be appropriate to assign patients to a PAC PPS UCG. Would that entity be the referring hospital, the patients’ primary care or other treating practitioner, the PAC provider itself, or — in the case of patients directly transitioning from the community to PAC services — another entity? What specific clinical and administrative capacity would be needed to administer the PAC patient placement process under the UCGs? These issues warrant attention from RTI if stakeholders are to understand their vision for actual UCG implementation — an understanding that will facilitate a more meaningful evaluation of RTI’s PAC PPS design.
RTI’s Use of a 50% Sample. RTI based some of its key analyses on a 50%-sample, or approximately 2.1 million PAC stays, rather than 100% of the population.2 Both the small size of this data sample and the sampling approach, as far as the methodology is understood, raise questions given current computing power for the relatively small size of the full, combined PAC datasets. In comparison, in 1998, CMS created its outpatient PPS model using 98 million hospital outpatient claims. Less than a month ago, it proposed the CY 2022 outpatient PPS rule using 93 million claims.3 Recent inpatient PPS rules have used over 9 million claims to group patients into Medicare Severity-Diagnosis Related Groups and set relative payment weights. In fact, many other examples exist, therefore, we not only seek a full description of RTI’s sampling methodology, we also request the rationale for this particular sample.
The extent to which the data sample is representative of the field is a concern that, in part, is based on MedPAC’s prior work on the congressionally-mandated study, the PAC Payment Reform Demonstration – a precursor to the PAC PPS development mandate in the IMPACT Act. In that case, the sampling methodology raised significant concerns, given that it yielded a too-small sample, its composition did not align with the actual distribution of cases across the four PAC settings, and, in our view, it inadequately accounted for the clinical complexity of the sickest PAC patients.
To understand RTI’s analyses, we attempted to replicate its sample of 2.1 million PAC claims for their four targeted years, 2017 through 2020, recognizing that our understanding of its sample criteria is incomplete. Through this effort, we counted over 28 million PAC claims across the four settings and four years.4 Given this substantial gap, we not only ask RTI to clarify the specific parameters used to create its sample, including all exclusions, we also seek its take on the ability of its sample to represent the overall PAC field.
Further, we are concerned that RTI’s sample appears to be inappropriately over-concentrated in six out of the 28 UCGs. Specifically, 51% of cases fall in the following UCGs: Orthopedic (Other); Neurological; MMTA: Cardiac; Other; Major Joint Replacement w/o LE Fracture; and Cardiovascular. We note the possibility that the use of the partial sample may contribute to this concerning pattern.
Based on these multiple concerns, we ask RTI to explain to TEP members the origin and ramifications of this small and skewed distribution, as well as alternative sampling approaches that were considered and discarded. In addition, to understand the reliability and validity of RTI’s analyses, we request statistics on the predictive power of the model for the proposed and alternative approaches, along with coefficients of variation and observed-to-expected ratios for each of the UCGs and subgroups. We also request payment-to-cost ratios.
A PAC PPS PAYMENT UNIT
The future determination of whether a PAC PPS payment unit should be based on unique patient visits versus a defined time period received inadequate attention during the June TEP. This is a significant design choice that will highly influence PAC patient care management and other important operational considerations for PAC providers, and perhaps both average and aggregate PAC spending by Medicare. For example, per-visit payment units are easier to define, but may reduce care management relative to a bundled or episodic unit. As such, we recommend that RTI comparatively evaluate the impact of both the per-stay and episodic approaches. These data and RTI’s interpretation of the findings should be shared during the next TEP and in the pending report to Congress. In particular, the efficiency and clinical outcome tradeoffs for both approaches should be considered. In addition, the comparative payment weights for the per-visit versus per-bundle payment units, which would significantly differ, should also be explained to the next TEP and in the report to Congress.
DEFINING COSTS FOR A PAC PPS
Another important PAC PPS design element, which also deserves more attention and discussion with the TEP, is whether costs should be defined as total costs per case versus the minimum costs for medically-necessary care. The former could utilize post-pandemic reported fixed and variable costs, which would align with a budget-neutral approach in the aggregate and/or per PAC setting. Under the latter approach, a PAC PPS would alternatively incorporate a to-be-specified level of savings relative to aggregate PAC spending at that time. If adequate accuracy and reliability cannot be achieved, we can anticipate likely access challenges for complex patients with greater resource needs, based on the chronic nature of this particular challenge.
CODING AND ICD-10-CM ISSUES
The AHA has a host of substantial questions regarding the coding-related issues raised in RTI’s work and its follow-up questions to the TEP.
- Standardized Cross-setting Principal Diagnosis Data are Unavailable. RTI stated that “at this time there is not a standardized cross-setting approach for PAC providers to identify the underlying condition or procedure for which the patient is receiving PAC. Therefore, we use several sources to assign PAC patients into clinical groups.” Specifically, RTI indicates that they used these calendar year (CY) 2017 to CY 2020 data sources for their analyses:
- Medicare Provider Analysis and Review (MedPAR);
- HH agency claims;
- HH Outcome Assessment Information Set (OASIS);
- SNF Minimum Data Set (MDS);
- IRF Patient Assessment Instrument (PAI); and
- LTCH Continuity Assessment Record and Evaluation (CARE).
We are concerned about this approach because the patient assessment data above use diagnosis definitions and coding guidelines, which are, in part, inconsistent with the coding guidelines used for Medicare claims.
These inconsistencies will materially weaken all analyses using RTI’s current version of the UCGs, and any future iteration that relies on these incompatible data without a reliable crosswalk — which doesn’t currently exist — including the reliability of any resulting PAC PPS design. Given that all policy findings that flow from this approach will incorporate this major flaw, including the model that is slated for submission to Congress in 2022, we ask CMS and ASPE to immediately intervene to halt this policy path.
- Medicare Claims versus IRF-PAI Coding Criteria. As an example of the lack of compatibility between the IRF-PAI datasets and claims data, the coding definitions used for IRF claims and the IRF-PAI are not interchangeable. To explain, we note that the principal diagnosis criteria on a claim (“That condition established after study to be chiefly responsible for occasioning the admission of the patient to the hospital for care.”) differ from the criteria used on the IRF-PAI for “etiologic diagnosis” (The etiologic problem that led to the condition for which the patient is receiving rehabilitation.). While these two definitions appear similar on the surface, the actual ICD-10-CM codes used to satisfy these two distinct definitions are dissimilar. For example, for a patient admitted for hemiparesis due to a cerebral infarction, a Medicare claim calls for code I69.359 as the principal diagnosis, while the IRF-PAI would use a code from a different category, I63 under the etiologic diagnosis, and the hemiparesis would be captured as an impairment group code and not reported using a ICD-10-CM code.
- Medicare Claims versus OASIS Coding Criteria. Another example is the lack of compatibility between the OASIS dataset and claims data, for which the coding sequencing is not interchangeable. The OASIS dataset requires the listing of each diagnosis for which the patient is receiving home care with “sequencing of diagnoses that should reflect the seriousness of each condition.” The OASIS dataset also requires a rating of the degree of symptom control for the condition(s) listed under the primary diagnosis and other diagnoses. Coding rules for the claim do not have a corresponding instruction for either the sequencing of the conditions, nor a rating for the degree of symptom control.
- Inconsistent Comorbidity Data Should Not be Used in PAC PPS Design. In its analysis of PAC comorbidity data, RTI’s data sources use inconsistent criteria for coding comorbidities. Specifically, RTI is using MedPAR and HHA claims data, as well as patient assessment data from the IRF-PAI, MDS, LTCH CARE and OASIS databases. As an example, which AHA has raised in detail with CMS in the past, IRF-PAI and IRF claims utilize different coding requirements.5 As a result, some comorbidities are coded on an IRF claim, but not in the IRF-PAI. In addition, the IRF-PAI manual does not require the use of ICD-10-CM codes for all comorbidities: “Enter ICD codes which identify comorbid conditions that are not already included in the Impairment Group Code (IGC).”
- Cross-setting ICD-10-CM Coding Accuracy Must be Validated. There are likely different rates of coding accuracy across the PAC settings. Before proceeding with using ICD-10-CM data for cross-setting comparisons and/or as a key building block for PAC PPS policy development, CMS and ASPE must at least validate the rate of coding accuracy across the PAC settings. We note that AHA’s coding team has observed that LTCHs and IRFs are generally more likely to use credentialed coding professionals, while for SNFs and HHAs, ICD-10-CM coding may be performed by clinical or other staff whose primary job is not coding. In addition, LTCHs and IRFs tend to be more attuned to the Official Coding Guidelines and its annual updates. Further, until recently, HH and SNF coders were less likely to remain current with coding guidelines changes, unless they were affiliated with a health system that used health information management professionals responsible for the entire system’s ICD-10-CM coding.
- Accepted PAC Codes Should Not be Excluded. RTI indicates that “Factors Influencing Health Status and Contact with Health Services” (Z-codes) and “External Causes of Morbidity” codes were all grouped into the “Other” UCG due to the non-specific nature of the categories. This is a questionable approach as there are many Z-codes for aftercare, which are appropriate for use on PAC claims as the reason for the encounter. A few examples include:
- Z47.81: Encounter for orthopedic aftercare following surgical amputation
- Z47.32: Aftercare following explanation of hip joint prosthesis
- Z47.1: Aftercare following joint replacement surgery
- Proposed Diabetes Code Exclusions Rule Out Acceptable PAC Principal Diagnoses. RTI has classified diabetic conditions into their “Other” UCG because “Diabetes is not likely to be the primary reason a patient would require institutional PAC.” While that statement is true, in general, there are some combination codes that include the diabetes as well as the manifestation, such as an ulcer, which could be reasons for PAC. In addition, there are combination codes for diabetic manifestations that, as a set, are appropriate principal diagnoses for PAC, such as diabetes with diabetic foot ulcer or with diabetic neuropathy.
- Communication and Cognitive Function Patient Assessment Data. RTI proposed not to use patients’ communications and cognitive functional status as case-mix adjusters based on its finding that patients with cognitive or communication impairments have lower costs than patients with no such impairments. Since the reason for this finding could be that all PAC providers do not capture these domains well, it would be appropriate for RTI to utilize input from clinical practitioners to try and make clinical sense of this finding. As it is our general understanding that patients with these limitations typically require additional clinical resources, RTI’s conclusion should be re-evaluated, and, if warranted, incorporated in any future PAC PPS case-mix system.
In response to RTI’s question regarding how costs associated with treating patients with cognitive impairment and/or limitations should be better accounted for, we point out that ICD-10-CM diagnosis codes are available to describe aphasia and several other cognitive or communication impairments, especially post stroke. Those codes already are in use on claims and as patient assessment data. Available ICD-10-CM diagnosis codes can be used to describe aphasia and several other codes for cognitive deficits, especially post stroke. Clinically, we believe that providers would need to spend more time with these patients, which would drive up costs.
QUALITY MEASUREMENT AND REPORTING CONCERNS
As mentioned above, the TEP follow-up questions that pertain to quality measurement and reporting are difficult, if not impossible, to answer without a PAC PPS model and access to the relevant data used by RTI. As an alternative to those questions, we share our questions and concerns related to this important topic.
PAC Patient Assessment Data are Inconsistent Across the PAC Settings. We note that in their respective PAC PPS work, RTI and MedPAC are not aligned in their reliance on patient assessment data in their payment system design process. On the one hand, MedPAC intentionally avoids the use of patient assessment data because the Commission views these data as unreliable for use in setting payments. In stark contrast, RTI not only relies upon these data, but would expand upon them by creating “new, interoperable” patient assessment elements that attempt to translate performance in similar (but not identical or standardized) elements across the multiple PAC patient instruments into a common scale. This raises questions on validity and value. For example, RTI would convert both “supervision or touching assistance” and “substantial assistance” to the rescaled metric “some assistance,” which seems contrary to capturing the level of detail needed to differentiate clinical needs in a way that helps facilitate payment accuracy, but RTI argues is necessary to account for the different scales across assessment instruments. AHA’s take, as we shared at length in our comment letters to CMS on all four PAC proposed rules for 2018 and 2020, is that despite the planned adoption of several “standardized” patient assessment data elements (SPADEs), the PAC field still lacks truly standardized patient assessment metrics. After delaying adoption of the SPADEs in order to test them more thoroughly, CMS and its contractors presented elements with middling reliability and questionable utility; the COVID-19 pandemic further delayed implementation, so we still have not seen whether these elements will yield useful assessment information. RTI did not address this issue, or whether they plan to use these standardized elements in their design, during the TEP. RTI’s contrasting perspective on the use of patient assessment data, relative to that of MedPAC, warrants further discussion and explanation for the TEP, and should be comprehensively addressed in the pending report to Congress.
PAC PPS Risk Adjustment is the Key to Achieving Payment Accuracy. For any PAC PPS design, the risk adjustment methodology is of top importance as it is the key to achieving the highest attainable level of payment accuracy and fairness. However, because RTI has not yet provided an actual model for discussion and evaluation by the TEP, it remains unclear which portion, if any, of the PAC SPADEs they would use to inform their risk adjustment methodology. Until RTI can demonstrate the design and reliability of their PAC PPS risk adjustment methodology, stakeholders will remain in the dark about what their payment system would mean to PAC patients and providers. Thus far, we know that RTI plans to use functional status data for risk adjustment purposes — another stark departure from MedPAC — which is intentionally avoiding any PAC PPS reliance on functional status data gleaned from patient assessments because they are viewed as unreliable. In addition, we assume that RTI also would incorporate additional patient assessment domains into their risk adjustment scheme, such as special treatments, services, and interventions, as these are meaningful indicators of patient resource needs. AHA would have strong concerns about any PAC PPS risk adjustment reliance on the special services, treatments, interventions and impairments domains in particular, because many of the elements are irrelevant to IRF and LTCH settings. Finally, RTI did not provide details on how social risk factors or other non-clinical patient-level characteristics would influence their design; while CMS recently adopted several SPADEs that address social determinants of health, these elements have not yet been implemented and, as with several other SPADEs, may not produce reliable information.
PAC PPS Value-based Program (VBP). It appears that RTI is not considering recent PAC developments in quality measurement and patient assessment in their PAC PPS design. For example, we await RTI’s update to the TEP regarding whether their design would incorporate a VBP element, or whether RTI is using any lessons learned from PAC VBP programs to inform its design. For example, the HH VBP program was recently proposed for national expansion; we ask whether RTI is considering incorporating any of the design elements into their PAC payment model. We note that HH VBP performance is assessed on measures informed by HH Consumer Assessment of Healthcare Providers and Systems (CAHPS®), claims-, and OASIS-based measures, with the latter including two normalized composite measures on functional status, which are unique to the HH VBP program. Our general take on these composite measures is that they provide more relevant and actionable information for providers and patients, facilitate logical comparisons of scores across disparate items through the normalization process, and clearly tie related assessment items together, making it easier to account for differences in severity of illness. Similarly, both CMS and MedPAC have acknowledged the shortcomings of the SNF VBP program, which is based upon a single measure and is not budget-neutral. Based on its September 2019 public meeting, we note that MedPAC already has conducted substantial work on this topic.6 Considering the amount of analysis that has come out of the field — including from MedPAC as well as from CMMI contractors evaluation the HH VBP model — we ask if RTI has reflected on any of this information, either for a potential VBP component of the PPS or for ideas for risk adjustment.
We note that in addition to RTI’s policy work developing a unified PAC payment model, in real terms, any future rollout of a PAC PPS would need to be paired with extensive companion regulatory, and likely some statutory, reforms. For example, unified PAC patient placement criteria, conditions of participation, and other regulatory infrastructure would be needed. These essential policy development steps, which are themselves substantial undertakings, would greatly magnify the current IMPACT Act scope of the work related to a unified payment system. The availability of these companion policies is necessary before a PAC PPS could comprehensively be evaluated and tested for real world applicability following the pandemic. We note that this additional policy work is not a part of the current CMS/ASPE contract with RTI.
In order to move forward, we strongly implore CMS, ASPE and RTI to address these questions, issues and concerns at the next meeting of the TEP. If some portion of these design challenges is found to be insurmountable, or cannot be adequately mitigated, our current ask of Congress to reconsider and refine the IMPACT Act timeline related to PAC PPS development would become even more urgent.
Thank you for the opportunity to comment on RTI’s latest round of PAC PPS development work. Please contact me if you have questions about our comments, or feel free to have a member of your team contact Rochelle Archuleta, AHA’s director of policy, at email@example.com.
Executive Vice President
1 An analysis of short-term acute care hospital cases discharged to each of the four PAC settings, during the pre-PHE and PHE periods.
2 We arrived at the 2.1 million figure by summing the counts in the table on PAC stays and mean wage adjusted cost on the last page of RTI’s June 16, 2021 letter which contained their follow-up questions.
3 These sample sizes were after exclusion criteria and other methodological considerations.
4 Our analysis counted unique combinations of claims by beneficiary, claim admission date and provider, and excluded claims for which Medicare did not make a payment.
5 According to the IRF-PAI Manual "Comorbidities that are identified on the day prior to the day of the rehabilitation discharge or the day of discharge should not be listed on the discharge assessment, since these comorbidities have less effect on the resources consumed during the entire stay." On a claim form, it doesn’t matter when the comorbidity was identified, so long as it meets the definition of a reportable secondary diagnosis: “For reporting purposes, the definition for “other diagnoses” is interpreted as additional conditions that affect patient care in terms of requiring: clinical evaluation; or therapeutic treatment; or diagnostic procedures; or extended length of hospital stay; or increased nursing care and/or monitoring.” An example could be a condition, sign or symptom reported on the day prior to discharge that required a diagnostic procedure, but the results indicate that it can be followed up as an outpatient or by another provider and therefore it does not prevent the patient from being discharged.
6 In September 2019, MedPAC presented a preliminary approach for a PAC PPS value incentive program (VIP) design, including a call for a small set of claims-based outcomes and resource use measures, which would be uniform and risk-adjusted across the four PAC settings. Specifically, all-condition, in-stay hospitalizations (including observation stays); discharge to community; and MSPB (for Parts A and B for the stay and subsequent 30 days) were suggested as initial measures. To recognize patients’ social risk factors, the VIP also would account for Medicaid eligibility for each PAC setting-specific groups of peers. In addition, an explicit position was stated against using functional status and other patient assessment data collected by the provider. Further, MedPAC noted that in the future, a patient experience and possibly other elements may be added.