J U N E 2 0 2 3 Updated Simulation of a Prospective Payment System for Post-Acute Care A report by the Urban Institute for the Medicare Payment Advisory Commission ME AC The views expressed in this report are those of the authors. No endorsement by MedPAC is intended or should be inferred. ME AC Medicare Payment Advisory Commission 425 I Street, NW | Suite 701 | Washington, DC 20001 | (202) 220-3700 | www.medpac.gov HEALTH POLICY CENTER RE S E AR CH RE P O R T Updated Simulation of a Prospective Payment System for Post-Acute Care Doug Wissoker and Bowen Garrett January 2023 AB O U T T HE U R BA N I NS T I T U TE The nonprofit Urban Institute is a leading research organization dedicated to developing evidence-based insights that improve people's lives and strengthen communities. For 50 years, Urban has been the trusted source for rigorous analysis of complex social and economic issues; strategic advice to policymakers, philanthropists, and practitioners; and new, promising ideas that expand opportunities for all. Our work inspires effective decisions that advance fairness and enhance the well-being of people and places. Copyright © January 2023. Urban Institute. Permission is granted for reproduction of this file, with attribution to the Urban Institute. Cover image by Tim Meko. Contents Acknowledgments iv Updated Simulation of a Prospective Payment System for Post-Acute Care 1 Data and Methods for Cost Modeling and Analysis 2 Modeling the PAC PPS 2 Evaluating the Design of the PAC PPS 11 Findings 16 Findings for the PAC PPS: Immediate Implementation of a Budget-Neutral Policy 16 Distribution of Impacts on Payments 18 Findings: Immediate Implementation of PAC PPS Payments with a 5 Percent Reduction 19 Findings: Phased-in Implementation of PAC PPS Payments with a 5 Percent Reduction 20 Conclusion 20 Appendix A. Payment Models and Impacts 21 Notes 55 References 56 About the Authors 57 Statement of Independence 58 Acknowledgments This report was funded by the Medicare Payment Advisory Commission. We are grateful to them and to all our funders, who make it possible for Urban to advance its mission. The views expressed are those of the authors and should not be attributed to the Urban Institute, its trustees, or its funders. Funders do not determine research findings or the insights and recommendations of Urban experts. Further information on the Urban Institute's funding principles is available at urban.org/fundingprinciples. This work benefited greatly from the close collaboration of Carol Carter and Kathryn Linehan. We thank Lauren Lastowka for editorial assistance. Any errors are the responsibility of the authors. iv ACKNOWLEDGMENTS Updated Simulation of a Prospective Payment System for Post-Acute Care This report presents the methods used to develop and assess the potential for fee-for-service Medicare to pay for post-acute care (PAC) using a unified prospective payment system (PPS). Our initial work for the Medicare Payment Advisory Commission (MedPAC) on a unified PAC PPS used data from 2013 to demonstrate the potential to pay for post-acute stays based on administrative data available in all four settings-home health agencies (HHAs), skilled nursing facilities (SNFs), inpatient rehabilitation facilities (IRFs), and long-term care hospitals (LTCHs). The model in that initial PAC PPS work and its estimated impacts are described in detail in Wissoker and Garrett (2016) and updated in Wissoker and Garrett (2019). In this report, we update to our initial modeling of a PAC PPS and estimate impacts of a PAC PPS on PAC providers and beneficiaries. In keeping with the original design and findings from our earlier work, the PAC PPS design described in this paper would pay by stay. Wissoker and Garrett (2018a, 2019) assessed the feasibility of paying by episode rather than by stay. These studies found that an episode- based system would likely overpay short episodes and underpay long episodes and could have undesirable incentive effects.1 As a result, we model a stay-based payment system. The PAC PPS design in this report responds to our previous finding that a PAC PPS that sets payments without using patient-level functional status data yields profits that are substantially below average (defined as having a payment-to-cost ratio less than average) for low-functioning patients and substantially above average (defined as having a payment-to-cost ratio greater than average) for high- functioning patients. Garrett, Wissoker, and Skopec (2021) investigated whether proxies for functional status could be used that were not subject to systematic misreporting or gaming and concluded that effective proxies are not available. Therefore, in this work, our model uses function measures as predictors. We also present impacts from a payment model excluding function, allowing a direct comparison of the performance of the two versions of the model in explaining costs per stay. This analysis is based on fee-for-service post-acute stays that began between April and September 2019 and have functional assessment data reported. Current payments are measured using actual payments for IRF patients, actual payments adjusted for the site-neutral payment policy for LTCH patients, and simulated payments for SNF and home health patients under the new payment systems for those settings (introduced in fiscal year 2020). For our primary estimated impacts, we assume that the system is implemented immediately and is budget neutral. We also simulate two other scenarios for implementation: immediate implementation with a 5 percent reduction to payments and implementation of the PAC PPS with a 5 percent reduction phased in over three years. This report provides technical details supporting the discussion in MedPAC's forthcoming 2023 report to Congress (MedPAC, forthcoming). The MedPAC report will provide more of the implications of the findings for the design of a unified payment system, as well as the likely impacts of moving from the current setting-specific prospective payment system to a unified payment system. This report has two main sections. First, we detail the data sources and methods for the stay- and episode-based PAC prospective payment systems. Second, we report and briefly describe the results for the updated PAC PPS, the effects of inclusion of function in the payment model, and the effects on payments of the two modeling approaches. Data and Methods for Cost Modeling and Analysis In this section, we describe the data and methods used to model the prospective payment system for post-acute care. The data and methods are similar to those reported in Wissoker and Garrett (2016, 2018a) and for the "stays model" in Wissoker and Garrett (2019). The section concludes with a description of the groups used to evaluate the models. Modeling the PAC PPS ESTIMATING THE COST OF PAC STAYS The analysis in this report is based on a subset of stays for payment year 2019. We first describe the construction of the analysis file based on the full payment year and reasons for excluding certain stays. The data for the full year were used in early runs and then served to reassure us that the results are very similar when we focus on the subset of stays that began between April and September 2019 and have matching assessment items to measure function. The analysis file for the full payment year of PAC stays includes observations for 10.6 million stays across the four PAC settings. The institutional stays, which correspond to FY 2019, began between October 2018 and September 2019. Home health episodes ended during the calendar year 2019. 2 UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE Definitions of stays depend on setting. A stay is defined as a discharge in IRFs and LTCHs, a 30-day episode in HHAs, and days on Medicare-covered claims within a SNF stay. 2 Note that the unit of analysis for home health is a 30-day episode rather than a 60-day episode, to simulate the 30-day payment period begun in 2020. The shorter payment periods were simulated from the 60-day episodes in 2019 by assigning costs of visits in the first 30 days to the first payment period; if there were visits in the second 30 days, their costs were assigned to a second payment period. We constructed the file using post-acute claims from the Medicare Standard Analytic File (SAF). The claim files were first processed by Abt Associates and Acumen LLC to simulate the new payment systems for home health agency and skilled nursing facility stays. In total, we received records for 11.3 million stays in the year. Of these stays, approximately 4 percent of home health episodes, 16 percent of SNF stays, 7 percent of IRF stays, and 15 percent of LTCH stays and were dropped (table 1). TABLE 1 Distribution of 2019 Stays across Settings Percent of 2019 stays dropped Number of stays in Number of 2019 because of 2019 PAC PPS stays (entire year) exclusion rules analysis File Home health agencies 8,744,171 3.6% 8,425,034 Skilled nursing facilities 2,051,631 15.7 1,729,668 Inpatient rehabilitation facilities 408,354 7.0 379,845 Long-term care hospitals 94,538 14.6 80,731 Total 11,298,694 6.0 10,615,278 Sources: 2018–2019 Medicare acute hospital and post-acute care claims, Medicare 2019 risk score file, and Medicare cost reports for 2019. Note: Skilled nursing facility, inpatient rehabilitation facility, and long-term care hospital claims are for stays beginning between October 2018 and September 2019; the home health claims are for 60-day episodes that ended between January and December 2019. These drop rates reflect decisions made in preparing the files obtained for the analysis. As detailed in Appendix A, table A.1, stays were dropped for the following reasons: ◼ patients having health maintenance organization/Medicare Advantage coverage during the year ◼ missing provider data from cost reports, such as cost-to-charge ratios for institutional providers UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE 3 ◼ missing data on charges ◼ missing data on simulated payments for SNFs ◼ facilities outside the 50 states and DC (e.g., facilities in Puerto Rico) ◼ other issues (such as missing risk scores, missing an area wage index, multiple stays with the same start date for a beneficiary, SNF stays of over 101 days, and IRF and LTCH stays longer than three standard deviations above the mean of the logged distribution) The relatively small drop rate of home health episodes reflects that the source file for HHA claims (created by Abt Associates) included only episodes with cost data. The high drop rate found in the 2019 SNF file simulated by Acumen reflects that the file included many cases for which payments under the new PDPM model could not be simulated. In all four settings, the files before exclusions included some patients with health maintenance organization coverage. The stays include all health conditions, reflecting the assumption that the PAC PPS would be used to pay for all stays regardless of the principal reason for treatment or the patients' comorbidities. The stays were from 9,685 HHAs (39 percent of PAC providers); 13,925 SNFs (56 percent of PAC providers); 1,061 IRFs (4 percent of PAC providers); and 339 LTCHs (1 percent of PAC providers). Overall, 9 percent of stays were with hospital-based providers.3 We base the analysis in this report on the six-month sample of stays that began between April and September 2019. Because we wanted to include functional status in the risk adjustment, we had to limit the analysis to stays that had matching patient assessments with uniformly defined measures of function. Institutional providers were required to collect this information (the "GG" items) beginning on October 1, 2018. HHAs were not required to collect this information until January 1, 2019. As a result, there is considerable missing data in the early months of collection from HHAs. Our use of the April– September window helps ensure that the measures were collected for a consistent share of stays in each setting across the entire window. Table 2 reports the counts and shares of stays and episodes with usable function data during the six-month period.4 The share of episodes with usable function data is much lower for home health episodes than for stays in institutional settings. 4 UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE TABLE 2 Number and Share of Stays Included in the Function Analysis File, by Setting Number of stays in final PAC PPS Number of analysis file stays started started in April Share of stays Number of stays in April to to September in 6-month file in 2019 PAC September 2019 with CARE with CARE PPS analysis file 2019 function items function items (1) (2) (3) (4) Home health agencies 8,425,034 4,203,989 2,639,025 0.628 Skilled nursing facilities 1,729,668 843,856 840,922 0.997 Inpatient rehabilitation facilities 379,845 189,800 176,755 0.931 Long-term care hospitals 80,731 38,103 35,362 0.928 Total 10,615,278 5,275,748 3,692,064 0.700 Sources: 2018–2019 Medicare acute hospital and post-acute care claims and assessments, Medicare 2019 risk score file, and Medicare cost reports for 2019. Notes: PAC = post-acute care. PPS = prospective payment system. CARE = Continuity Assessment Record and Evaluation. The PAC PPS analysis file (column 1) contains institutional stays beginning in FY 2019 and home health episodes ending in calendar year 2019 after excluding problematic stays as described in Appendix A table A.1. Column 2 restricts the 2019 PAC PPS analysis file to those stays and episodes that began between April and September 2019. The final PAC PPS analysis file (column 3) restricts the 2019 PAC PPS analysis file to stays begun between April and September 2019 for which CARE function items were reported on a matched assessment. Overall, the average cost per stay is $5,495 for those with CARE function measures versus $4,266 for all stays. This occurs because of non-random reporting of the CARE function measures, both within and across settings. Home health episodes with usable function measures have higher average costs than all home health episodes ($1,685 versus $1,492 for all episodes). And, as can be calculated from table 1, SNF, IRF, and LTCH stays combine to make up a larger share of the observations of those with usable function measures (28.6 percent of stays in column 3) than among all observations (21.3 percent of stays in column 1). Restriction of the sample to stays with usable function data that began between April and September 2019 reduced the underlying number of HHA providers by 4 percent and the number of institutional providers by 1 percent. The remaining sample includes stays from 9,285 HHAs (38 percent of PAC providers); 13,868 SNFs (57 percent of PAC providers); 1,047 IRFs (4 percent of PAC providers); and 335 LTCHs (1 percent of PAC providers). For this sample, 10 percent of stays in the final sample were with hospital-based providers. Costs per stay include both routine and ancillary costs (including overhead costs), and for IRFs, the costs associated with teaching programs and treating low-income patients.5 UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE 5 For institutional stays, we estimated routine costs as the average routine cost per day from the 2019 Medicare cost report times the stay's covered length of stay from the claims. For free-standing SNFs, the cost report figure is adjusted upward by 16.4 percent to account for the higher nursing costs associated with treating Medicare beneficiaries compared with other patients, particularly long-stay nursing home residents (the cost report includes a facility's total costs for treating all patients and residents). We estimated both therapy and nontherapy ancillary costs by converting charges on the PAC claims to costs using facility- and department-specific cost-to-charge ratios from each provider's 2019 Medicare cost report. For HHAs, routine and ancillary costs are calculated by aggregating the estimated cost for the 30- day episode over six types of visits. Routine costs are the sum of the costs of the three nontherapy visit types (skilled nursing, home health aides, and medical social services). 6 Therapy costs are the sum of the costs of the three therapy visit types (physical therapy, occupational therapy, and speech language pathology services). The cost of each type of visit is the product of the number of minutes of that visit type from claims and the cost per minute from the 2019 Medicare cost report and was provided on the data file prepared by Abt Associates. Nontherapy ancillary (NTA) costs for HHAs are not calculated because they are not covered separately for HHA episodes. All costs were standardized using the setting-specific labor share and the area wage index. Labor shares were set at 76.1 for HHAs, 68.8 for SNFs, 70.9 for IRFs, and 66.5 for LTCHs. Finally, we capped routine, therapy, and non-therapy ancillary wage-adjusted costs at the 99.5th percentile for stays within each setting, separately by whether a facility is a hospital-based or free-standing facility. PREDICTING THE COST OF STAYS USING PATIENT CHARACTERISTICS Under a PAC PPS, the payment for a stay would be based on the stay's predicted cost. Patient and stay characteristics are used to predict the actual cost of the stay. In the below list, two sets of characteristics-primary reason to treat and the patient's severity of illness-were taken from the hospital claim when there was a hospital stay within 30 days of the admission date for the PAC stay and were proxied from PAC claims for stays without a preceding hospitalization. For home health, we used the information from the prior hospital stay to create predictors for both 30- day periods that were in the 60-day episodes that was preceded by the hospital stay. We used measures of the following to predict the cost of stays: ◼ patient age and disability status 6 UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE ◼ primary reason to treat (Medicare Severity Diagnosis-Related Group, aggregated into "reason to treat" groups) ◼ patient comorbidities (observed in a prior hospital stay and the PAC stay) ◼ the number of body systems involved with the patient's comorbidities for patients in institutional settings ◼ days spent in the intensive and coronary care units during the prior hospital stay ◼ the patient's severity of illness using the All Patient Refined Diagnosis-Related Groups ◼ beneficiary's risk score based on patient diagnoses for the prior year ◼ impairments and treatments (including bowel incontinence, urinary incontinence, impaired vision, severe wounds or pressure ulcers, use of certain high-cost service items, and difficulty swallowing) ◼ proxies for patient's frailty ◼ patient's cognitive status ◼ patient functional status Most risk adjustors are based on administrative data other than the patient assessment. We used claims information from PAC stays and the preceding hospital stays, demographic information from the Medicare enrollment files, and beneficiary risk scores. Information on diagnoses and the primary reason for treatment was collected from prior hospital stay claims and from PAC stay claims for patients admitted from the community. Comorbidities data were collected from hospital stay claims where available and from the PAC stays claims. Indicators of ventilator care and severe wound care needs were obtained from the PAC stay claims. We used claims-based diagnoses and procedure codes for measures of frailty, cognitive function, and select PAC service use. We used codes from the International Classification of Diseases, 10th revision (ICD-10) in the PAC claims to indicate bowel and urinary incontinence, severe wound care, and the presence of ventilator care.7 We calculated a JEN Frailty Index for each stay using ICD-10 codes and included the 13 components of that index as predictors. 8 As proxies for impaired cognitive function, we used ICD-10 codes to identify patients in a coma or with dementia or Alzheimer's' disease. As indicators of serious mental illness, we used ICD-10 codes to identify patients with schizophrenia, bipolar disorder, and/or severe depression. We used ICD-10 codes for dysphagia as a proxy for swallowing difficulties in the post-acute setting. UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE 7 Our model of costs also includes indicators of total functional score based on six questions from the Continuity Assessment Record and Evaluation Item Set that are now reported on patient assessments. The total functional score is the sum of zero to five scores from the following six measures (with the relevant question number in brackets): ◼ Self care [GG0130 E], shower/bathe self: the ability to bathe self in shower or tub, including washing, rinsing, and drying self. Does not include transferring in/out of tub/shower. ◼ Mobility [GG0170 B], sit to lying: the ability to move from sitting on side of bed to lying flat on the bed ◼ Mobility [GG0170 I], walk 10 feet: once standing, the ability to walk at least 10 feet in a room, corridor, or similar space ◼ Self care [GG0130 C], toileting hygiene: the ability to maintain perineal hygiene, adjust clothes before and after using the toilet, commode, bedpan, or urinal. If managing an ostomy, include wiping the opening but not managing equipment ◼ Mobility [GG0170 A], roll left and right: the ability to roll from lying on back to left and right side, and return to lying on back ◼ Mobility [GG0170 D], sit to stand: the ability to safely come to a standing position from sitting in a chair or on the side of the bed For many stays, at least one activity was not measured. If the patient refused to perform the activity or it was not attempted due to environmental limitations, the activity was excluded from the total functional score and the score was reweighted to account for fewer responses. Stays with three or more such activities were eliminated from the analysis. Activities that were not attempted-either because the patient didn't perform the activity prior to the current illness or because of the patient's medical condition or safety concerns-were included as a zero, indicating full dependence on a helper to accomplish the activity. The following indicators of total functional score are included in the primary prediction models of costs with highest function as a reference category: ◼ 0 to less than 6 points (lowest function) ◼ 6 to less than 12 points ◼ 12 to less than 18 points ◼ 18 to less than 24 points 8 UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE ◼ 24 to 30 points (highest function) More detailed definitions of all of the predictors are reported in Appendix A, table A.2. We avoided including in the model indicators of service use that might be manipulated by providers (such as the amount of rehabilitation therapy, the number of therapy disciplines, or the use of oxygen without a link to a respiratory diagnosis), but we included indicators for ventilator care, tracheostomy care, and continuous positive airflow pressure because the costs of those services are significant and use is much less likely to be influenced by payment policy. The measure for continuous positive airflow pressure captures use only within institutional settings, since home health claims do not provide the procedure codes needed to identify its use in home health. As in our earlier work, we included in the payment model an indicator for care provided by HHAs. HHAs do not incur the same kinds or levels of costs as institutional providers, so we adjust for this with an indicator in the model for home health. (Details are explained below.) Inclusion of this indicator ensures that costs for home health cases are predicted correctly on average. In addition, two indicators-whether secondary diagnoses involved five or more body systems and continuous airflow pressure-are only measured for those in institutional settings. Severely ill patients include those with a severity of illness level 4 (the sickest), calculated using the all-patient refined-diagnosis related groups, and exclude patients treated in home health agencies. Costs were predicted using generalized linear models with a log link (Poisson regression models). Compared with ordinary least squares regression, the Poisson regression gives less emphasis to infrequent but exceptionally high-cost stays. In addition, Poisson models can more easily handle dependent variables with zero values (such as institutional stays with no NTA costs) than linear models with a logged dependent variable. Our approach uses two regression models to predict each stay's actual costs: one for routine plus therapy costs and another for NTA costs. The routine plus therapy cost model is based on stays from HHA and institutional PAC settings. The NTA model is based on stays from only the institutional PAC settings because these services are not part of the home health benefit so HHAs do not incur costs for them. The two models use the patient and stay characteristics as predictors, except for inclusion of an indicator of a home health stay as mentioned above. We combined the cost estimates generated by the models (including zero predicted NTA costs for HHA stays) to obtain total predicted costs. One method of evaluating the results is obtained by comparing total actual costs (including zero NTA costs for HHA stays) with the total predicted costs. UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE 9 COMPARING PAYMENTS AND COSTS To compare what estimated payments would be under PAC PPS with the costs and payments of stays, all costs and payments were standardized for variation in wages across geographic areas. This ensures that the comparison of costs, actual payments, and modelled payments takes place on an equal basis. Since estimated payments under the new systems were based on standardized costs, they did not need to be further adjusted for wage differences. Actual payments include relevant adjustments for rural location, teaching status, low-income share, outliers, and the amounts paid by the beneficiary (any coinsurance and deductibles). For SNFs, the total "actual" payments were simulated for the 2019 stays by staff at Acumen LLC based on the Patient Driven Payment Model-the payment system implemented in FY 2020. For HHAs, total payments were simulated for the 30-day periods in 2019 by staff at Abt Associates based on the Patient Driven Groupings Model (PDGM)-the episode length and payment system implemented in 2020. LTCH payments reflected what would have been paid under fully implemented dual-rate structure: LTCH rates for qualifying stays and the lower of the inpatient hospital PPS rate or 100 percent of the cost of the case. The PAC PPS payments combine an initial payment that is set to be proportional to total predicted costs that includes an outlier policy for low utilization stays and a high loss episode. Total PAC PPS dollars paid out were set equal to total actual payments (i.e., payment levels are set to be budget neutral across all PAC settings). To implement a low-utilization payment policy that works across all PAC settings, we defined a short stay outlier (SSO) for institutional stays that parallels the low-utilization payment adjustment (LUPA) definition used in home health. SSOs are defined as institutional stays in the bottom decile of length of stay within each setting (six or fewer days for SNFs and IRFs and seven or fewer days for LTCHs). LUPA cases were assigned based on whether the 30-day episode qualified as a LUPA episode under the PDGM rules.9 The high-loss outlier policy was implemented with separate pools and fixed- loss amounts for home health episodes and institutional stays, with each pool equal to 5 percent of payments. The combined outlier payment is calculated in the following steps: 1. For these SSO and LUPA cases, expected costs are set equal to 1.2 times the setting-specific average cost per day or per visit for the first day to reflect higher initial costs and set equal to average cost per day or per visit for subsequent days. For other stays, expected costs are set equal to the model prediction. Payments (before the addition of a high-loss outlier) are then set 10 UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE proportional to the expected cost, imposing the condition that the average implied PAC PPS payment equals the average of current payments. 2. In our primary model with total function score as a predictor, PAC PPS payments for non-SSO and non-LUPA stays are reduced by 5 percent to establish an outlier pool, which is then used to pay 80 percent of losses above $1,044 for HHAs and above $11,134 for institutional settings. In our model without function as a predictor, the pool is used to pay 80 percent of losses above $1,058 for HHAs and $11,726 for institutional providers. Evaluating the Design of the PAC PPS To evaluate the potential accuracy of a PAC PPS and estimate its impact on payments, we examined the accuracy of the payment models in aggregate (across all stays) and their effects on many patient groups. We created these groups to report the results of the PPS design, but the underlying prediction models remain the same across all groups. These groups "stress test" the models by looking at how well they perform for different clinical conditions and various definitions of medically complex patients. The following subsections detail the patient groups that we use in evaluating the models. CLINICAL CONDITION Measures of clinical condition were generally based on information (diagnoses and procedure codes) from claims for the preceding hospital stay. When there was not a prior acute hospital stay within 30 days (such as the two-thirds of home health care stays that are admitted from the community), we used claims for the PAC stay.10 For these stays, the Medicare Severity Diagnosis-Related Group assignment was simulated using diagnostic information from the PAC claim. For two clinical conditions, ventilator care and severe wound care, we based measures on information from the PAC claim instead of from a prior acute hospital stay claim to focus on the adequacy of payments for those with the condition observed during the PAC stay. Except for stays for patients with serious mental illness, the clinical condition groups are mutually exclusive, with stays first assigned to ventilator care, then severe wound care; all other stays are assigned to a major diagnosis category (MDC) based on the Medicare Severity Diagnosis-Related Group. We report on the following clinical conditions: ◼ ventilator care ◼ severe wound care UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE 11 ◼ stroke ◼ other neurology medical-medical stays assigned to MDC 1, excluding stroke ◼ other neurology surgical-surgical stays assigned to MDC 1, excluding stroke ◼ orthopedic medical-medical stays assigned to MDC 8 ◼ orthopedic surgical-surgical stays assigned to MDC 8 ◼ respiratory medical-medical stays assigned to MDC 4 ◼ respiratory surgical-surgical stays assigned to MDC 4 ◼ cardiovascular medical-medical stays assigned to MDC 5 ◼ cardiovascular surgical-surgical stays assigned to MDC 5 ◼ infection medical-medical stays assigned to MDC 18 ◼ infection surgical-surgical stays assigned to MDC 18 ◼ hematology medical-medical stays assigned to MDC 16 or 17 ◼ hematology surgical-surgical stays assigned to MDC 16 or 17 ◼ rehabilitation medical-medical stays assigned diagnosis-related groups 945 or 946 ◼ skin medical-medical stays assigned to MDC 9 ◼ skin surgical-surgical stays assigned to MDC 9 ◼ serious mental illness-includes stays for beneficiaries with schizophrenia, bipolar disorder, or severe depression, identified using the hierarchical condition category indicators 57 or 58 in the PAC or preceding hospital stay. This group is not mutually exclusive with the other clinical groups; a stay can be assigned to another clinical group and to the serious mental illness group. ◼ kidney and urinary tract medical-medical stays assigned to MDC 11 ◼ liver medical-medical stays assigned to MDC 7 ◼ digestive medical-medical stays assigned to MDC 6 ◼ endocrine medical-medical stays assigned to MDC 10 ◼ mental illness medical-medical stays assigned to MDC 19 ◼ alcohol and drug use medical-medical stays assigned to MDC 20 ◼ HIV medical-medical stays assigned to MDC 25 12 UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE ◼ other medical-medical stays not otherwise grouped (including eye and ear, reproductive, and other factors influencing health) ◼ other surgical-surgical stays not otherwise grouped (including liver, gastrointestinal, or endocrine) A small number of cases that could not be assigned as medical or surgical were dropped from the analysis. In addition, we report groups with the following clinical conditions (these groups are not mutually exclusive and may overlap with other conditions): ◼ cancer-stays with cancer as primary reason for treatment ◼ transplant-stays with transplant as primary reason for treatment ◼ kidney and urinary-stays assigned to MDC 11 ◼ gastrointestinal or hepatobiliary-stays with primary reason for treatment as GI, liver, or pancreatic (MS-DRG in ranges 326–358; 368–395; 405–425; 432–446) ◼ vision impairment in the PAC or preceding hospital stay ◼ urinary incontinence in the PAC ◼ trauma-stays with a PAC Clinical Classifications Software Refined (CCSR) trauma code, an IRF rehabilitation impairment trauma code, or an MS-DRG for the prior hospitalization indicating trauma MEDICAL COMPLEXITY AND IMPAIRMENT To further evaluate stays, we examine groups of medically complex patients or those with impairments who meet the following conditions: ◼ health conditions affecting multiple body systems: patients in institutional settings with secondary diagnoses involving five or more body systems ◼ chronically critically ill: patients who spent eight or more days in an intensive care or coronary care unit during the preceding hospital stay or are on a ventilator in the PAC setting ◼ severity of illness level 4: institutional PAC patients assigned to the highest-severity group using the All Patient Refined Diagnosis-Related Group, based on diagnostic information from the preceding hospital stay or proxied for patients admitted without a hospital stay UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE 13 ◼ impaired cognition: patients who were in a coma or had dementia or Alzheimer's disease ◼ patient frailty: patients in approximately the top and bottom quartile of the JEN Frailty Index ◼ total function score: stays in the bottom quartile, middle two quartiles, and top quartile in a 30- point total functional score (defined above) defined from full year of stays OTHER STAY AND PATIENT CHARACTERISTICS We also examined the following patient groups: ◼ therapy use: For home health episodes, the groups are defined by the number of therapy visits: zero, one to four, five to nine, and ten or more; for institutional PAC stays, the groups are the four quartiles of per-diem therapy costs. ◼ disabled based on original reason for entitlement ◼ fully dual-eligible for Medicare and Medicaid, partially dual-eligible, or received the Low- Income Subsidy under part D ◼ beneficiaries with end-stage renal disease ◼ age 85 or older We also examined groups defined by the following stay characteristics: ◼ short stays: For institutional stays, patients with stays in the shortest decile for their setting (that is, less than or equal to six days for SNFs and IRFs, less than or equal to seven days for long-term care hospitals); for home health, 30-day episodes subject to the current low- utilization payment adjustment. ◼ community admissions: patients admitted from the community (with no hospital stay within the 30 days preceding the PAC stay, identified by the lack of a matching hospital claim) ◼ patients with a prior hospitalization within the 30 days preceding the PAC stay identified by a matching hospital claim Outcomes for short stays are examined by setting, while those for community admissions and those with a recent hospitalization are examined separately for home health and for institutional stays. PROVIDER CHARACTERISTICS We also examine payment accuracy by provider characteristics: 14 UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE ◼ facility type: hospital-based, freestanding facilities ◼ ownership: nonprofit, for-profit, and government facilities ◼ low-volume provider: bottom decile of provider size within setting in full-year file ◼ low-income share for IRFs: quintiles of provider share among IRFs in full-year file ◼ IRF teaching facilities ◼ provider shares of duals/LIS: quintiles of provider share of duals or LIS patient stays within setting among April–September stays ◼ geographic location: frontier, metro, rural micropolitan, rural adjacent, rural nonadjacent, and urban or rural core-based statistical areas ◼ provider share of race/ethnicity groups: top decile within setting of proportion of stays with white non-Hispanic patients, black non-Hispanic patients, and patients of other race/ethnicities. Race/ethnicity is assigned using the RTI measure of race and ethnicity, which takes the beneficiary's name into account to assign persons to Hispanic and Asian categories. Cutoffs for bottom decile is based on facilities with at least 25 stays in the full year file. In addition, we report the CMS region where the provider is located: ◼ region 1: CT, ME, MA, NH, RI, VT ◼ region 2: NJ, NY ◼ region 3: DC, DE, MD, PA, VA, WV ◼ region 4: AL, GA, FL, KY, MS, NC, SC, TN ◼ region 5: IL, IN, MI, MN, OH, WI ◼ region 6: AR, LA, OK, NM, TX ◼ region 7: IA, KS, MO, NE ◼ region 8: CO, MT, ND, SD, UT, WY ◼ region 9: AZ, CA, HI, NV ◼ region 10: AK, ID, OR, WA UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE 15 Findings In this section, we report on the regression models underlying the PAC PPS and provide estimates of the implied payment model's accuracy and expected impacts on payments for key subgroups. Results are presented for payment models with and without total functional score in the predictive model and an assumption that the systems would be implemented immediately and be budget neutral. In addition, we present impacts for two other scenarios. In the first, the PAC PPS with function as a predictor is implemented immediately with a 5 percent reduction in pooled PAC payments. In the second, this PAC PPS with reduced payments is implemented over three years. We report the estimated impacts for the first of the three years. Findings for the PAC PPS: Immediate Implementation of a Budget-Neutral Policy In Appendix A, table A.3, we report the coefficients and standard errors from the Poisson regression models that underlie our primary simulation of the PAC PPS system. This model includes indicators of functional score as predictors. Coefficients are reported separately for models of routine plus therapy costs and models of NTA costs. The exponentiated coefficients provide multipliers for predicted costs associated with a one unit increase in the predictor. The models of routine-plus-therapy costs are based on stays from both institutional and home health settings; the model of NTA costs is based on stays from institutional settings. The standard errors are clustered to account for the similarity of stays from the same provider. The prediction for each institutional stay is the sum of the predicted costs from the routine-plus-therapy and NTA cost models; the prediction for home health episodes is the predicted cost from the routine plus the therapy cost model. Altogether, the model explains 54.3 percent of the variation in total costs across all settings. The relatively high share of variance explained stems largely from including the home health setting indicator in the model. The coefficient on the home health indicator in the routine-plus-therapy model implies that, all else equal, home health costs for routine and therapy are 15 percent of those of institutional settings (see the exponentiated coefficient). As expected, we find a strong relationship between indicators of patient total functional score and both combined routine and therapy costs and non-therapy ancillary costs per stay after controlling for other stay and patient characteristics. (See the section of Appendix A, table A.3 labeled "Functional score.") For example, those with function scores less than six are estimated to have 49 percent higher 16 UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE routine and ancillary costs and 25 percent higher non-therapy ancillary costs than those in the reference group with the highest functional scores. We reject the hypothesis that the coefficients on the function measures equal zero (p=0.0 in each model) and, as shown in Appendix A, table A.4, when we estimate the model excluding the four measures of function score, the R2 statistic falls to 53.2. The pattern of R2 statistics obtained by estimating the payment model within each setting shows greater predictive power within LTCHs and IRFs than in HHAs and SNFs. The within-setting payment models explain 20 and 21 percent of the variation in LTCHs and IRFs, compared with 4 percent and 7 percent of the variation in SNFs and HHAs (data not shown). That the overall R2 statistic is much higher than the within-setting R2 statistics suggests that much of the predictive power comes from predicting the variation across settings and is consistent with the high R2 resulting from including a control for home health in the routine/therapy regression model. Average costs, predicted costs, current payments, and PAC PPS payments for the initial PAC PPS approach are reported in Appendix A, table A.5. The PAC PPS payments are based on the predictive model including total functional score as a predictor and assume immediate implementation of a budget-neutral system. The overall payment-to-cost ratio is 1.14, that is, average PAC PPS payments are 14 percent higher than average costs. This high level of profitability matches the overall level of profitability of actual payments and is the result of assuming budget neutrality. As noted earlier, budget neutrality ensures that average PAC PPS payments are set equal to average actual 2019 payments. In general, the ratios of PAC PPS payments to costs (i.e., profitability) are relatively even across the various patient and stay reporting groups. An interesting exception is cases with no or few therapy costs in home health (shown in Appendix A, table A.5, in the rows labeled "HHA, no therapy" and "HHA, 1–4 visits"), which would be considerably more profitable than home health stays with higher therapy costs under the modeled system. The payment-to-cost ratio for nontherapy home health cases is 1.85 as compared with ratios between 0.79 for home health stays with ten or more therapy visits. 11 Because the PAC PPS design does not consider the amount of therapy in establishing payments (because it is under the control of providers), it is not surprising that the design does not accurately predict the variation in costs with therapy as accurately as for other types of stays.12 This would tend to be true for any measure of service provision, which, with a few exceptions, are not included among predictors by design. The PAC PPS tends to shift payments toward SNFs and away from IRFs and LTCHs, as can be seen in the section "Provider characteristics" of table A.5 in Appendix A. The ratio of PAC PPS payments to UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE 17 current payments is 1.07 for SNFs, 0.83 for IRFs, and 0.94 for LTCHs. These ratios are close to those found using 2017 data and reported in Wissoker and Garrett (2019). Home health payments would be reduced slightly less than in the earlier work. Other notable differences in profitability and payments are observed for providers. We highlight a few cases that stand out. The findings show that chronically clinically ill LTCH stays are more profitable than LTCH stays overall, with a payment-to-cost ratio of 0.99 as compared with a ratio of 0.93 for all LTCH stays. In this sample, PAC PPS payments to hospital-based facilities would be lower than current payments (with a PAC PPS to current payment ratio of 0.97) and continue to be less profitable than freestanding facilities (with a payment-to-cost ratio of 0.92 compared with 1.17 for freestanding facilities). In Appendix A, table A.6, we show that if function is not included as a predictor in the payment model, profitability under the PAC PPS would be substantially below average for patients with low functional ability and above average for patients with high functional ability. This can be seen in the section labeled "Frailty, cognitive function, mental illness and functional score. Patients with function in the lowest quartile have an average payment-to-cost ratio of 1.06, while patients with function in the highest quartile have an average payment-to-cost ratio of 1.31. By contrast, table A.5 in Appendix A shows that with functional score in the payment model, patients with function in the bottom and top quartiles have quite similar levels of profitability. These results are in line with those found in Garrett, Wissoker, and Skopec (2021) based on 2017 stays and functional score data that led to the decision to include function in the preferred payment model. Distribution of Impacts on Payments Next, we report the distribution of impacts on payments for stays and for providers. Table A.7 in Appendix A reports the distribution of the ratio of PAC PPS to current payments by stay reporting group. The columns indicate the size of the expected change in payments. Overall, we see that although the new system would be budget neutral, 19 percent of stays would be paid at least 25 percent less and 33 percent would be paid at least 25 percent more. These results show there is substantial variation in impacts across stays. For example, in table A.5 in Appendix A we see that payments increase by one percent for home health episodes without therapy visits. In table A.7, we see that 16 percent of these episodes would be paid at least 25 percent less and 40 percent paid 25 percent more. 18 UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE In Appendix A, table A.8 reports the distribution of the ratio of the percentage change in payments under a PAC PPS by provider groups. The percentage change in payments for each provider is calculated as the ratio of total PAC PPS payments to total current payments. The distributions are reported for the provider reporting groups used in earlier tables. Only providers with at least 20 stays are included in the reported distributions. Overall, as seen in the "all providers" row, 21 percent of providers with at least 20 stays have a decrease in payments of at least 10 percent, while 33 percent of providers have an increase of at least 10 percent. These patterns vary dramatically by setting. For HHAs, where impacts are set to be close to one, the distribution is roughly symmetric, with 12 percent having at least a 10 percent increase and 12 percent having at least a 10 percent decrease in payments. Among IRFs, the finding is quite unbalanced with 78 percent of IRFs having a decrease of at least 10 percent and less than 1 percent with an increase of at least 10 percent. Among SNFs and LTCHs, the distribution is a bit more balanced: Among SNFs, 51 percent have an increase of at least 10 percent as compared with 22 percent with a decrease of 10 percent; among LTCHs, 34 percent have a decrease of 10 percent as compared with 9 percent with an increase of at least 10 percent. In Appendix A, table A.9 we describe how the changes in payments from implementation of the PAC PPS are estimated to vary with the relative current profitability of facilities. Relative profitability of a provider is the provider's profitability divided by the average profitability in the setting. The table reports the counts of facilities for combinations of ranges of impacts and relative profitability. Facilities with below average current relative profitability tend to get an increase in payments with the PAC PPS while those with above average profitability tend to get a decrease in payments. Findings: Immediate Implementation of PAC PPS Payments with a 5 Percent Reduction Overall, a 5 percent reduction in overall payments leads to a reduction in the payment to cost ratio from 1.14 to 1.08 (see Appendix A, table A.10). The reduction is comparable in each of the four settings. Home health payment-to-cost ratios fall from 1.15 to 1.09, the SNF ratio falls from 1.22 to 1.15, the IRF ratio from 0.83 to 0.78, and the LTCH ratio from 0.94 to 0.90. The payment cut falls nearly evenly across settings, with a 5 percent reduction in payments for home health, a 5.0 to 5.1 percent reduction for SNFs and IRFs and a 4.1 percent reduction for LTCHs. The variation in the reduction across settings results from differential effects of the payment cut on outlier payments across the institutional settings. More details and findings by group of providers are reported in table A.10 in Appendix A. UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE 19 Findings: Phased-in Implementation of PAC PPS Payments with a 5 Percent Reduction The simulation of the first year of a three-year implementation of the PAC PPS with a 5 percent reduction works as expected. We assume that payments in the first year of transition are a blend of one-third of the PAC PPS payments and two-thirds of current PPS payments. As can be seen in Appendix A, table A.11, with a three-year transition, payments differ from actual payments by one-third as much as with the immediate implementation and a 5 percent reduction shown in table A.10. As a result, the payment-to-cost ratios also change from those observed with actual payments by one third as much as with an immediate implementation. For example, for all stays the payment-to-cost ratio in this simulation is 1.12-this is simply one third of the change from the payment-to-cost ratio based on actual payments (1.14) and the ratio with an immediate implementation of the 5 percent reduction (1.08). Conclusion In this report, we have provided additional methodological detail and data analyses used in the MedPAC report to Congress on a unified payment system for post-acute care. The implications of these findings for the design of a unified payment system, as well as likely impacts of moving from the current setting-specific prospective payment system to a unified payment system, are discussed in MedPAC's forthcoming 2023 report to Congress (MedPAC, forthcoming). 20 UPDATED SIMULATION OF A PROSPECTIVE PAYMENT SYSTEM FOR POST -ACUTE CARE Appendix A. Payment Models and Impacts TABLE A.1 Disposition of Full Year Stays for Post-Acute Files, Payment Year 2019 Inpatient Skilled nursing rehabilitation Long-term care Disposition Home health facility facility hospital All Included in annual sample 8,425,034 1,729,668 379,845 80,731 10,615,278 Two stays with same start date 1,209 489 218 27 1,943 Health maintenance organization/MA coverage 308,094 93,944 14,203 4,096 420,337 Long length of stay 0 718 230 11 959 No ratio of costs to charges 0 3,360 2,654 14 6,028 Common Medicare Environment record missing 114 0 0 0 114 No cost report 0 82,910 6,149 4,966 94,025 No charges reported 0 6,472 127 1,232 7,831 No MEDPAR record for LTCH stays 0 0 0 2,570 2,570 MSDRG group not assigned 183 107 5 4 299 No provider of service record 724 0 0 0 724 No payment or zero length of stay 0 119,830 4,395 885 125,110 No risk score 48 6 10 2 66 No routine costs 0 13,974 0 0 13,974 No wage index 0 100 0 0 100 Records missing due to file error 37 0 0 0 37 Facility located in US Territory 8,728 0 489 0 9,217 Patient is a minor 0 2 0 0 2 Data from the wrong year 0 51 29 0 80 Total 8,744,171 2,051,631 408,354 94,538 11,298,694 Sources: 2018–2019 Medicare post-acute care claims and assessments, Medicare 2019 risk score file, and Medicare cost reports for 2019. APPENDIX A 21 Notes: Skilled nursing facility, inpatient rehabilitation facility, and long-term care hospital claims are for stays beginning between October 2018 and September 2019; the home health claims are for 60-day episodes that ended between January and December 2019. MA=Medicare Advantage. TABLE A.2 Description and Sources of Model Predictors Model Predictors Characteristic Stay predictor Source ProAge Age at start of PAC stay, restricted to between 50 and 95: Age CMS-HCC risk score file minus 50, (Age minus 50)2, and indicator for age less than 50 Cognitive function Dementia with and without comorbidities (HCC51 and HCC52) Based on diagnoses from prior hospital stay and coma and current PAC stay; measures other than coma assigned using PACE/ESRD HCCs definitions Mental health Schizophrenia (HCC57) and major depressive disorder, bipolar Based on diagnoses from prior hospital stay disorder, paranoid disorder (HCC58) and current PAC stay; measures other than coma assigned using PACE/ESRD HCCs definitions Frailty Components of JEN Frailty Index included are minor Based on diagnoses from prior hospital stay ambulatory limitations; severe ambulatory limitations; cognitive and current PAC stay; calculated using ICD- developmental disability; chronic mental illness; dementia; 10-based JEN program from Westat. sensory disorders; self-care impairment; syncope; cancer; chronic medical disease; pneumonia; renal disorders; other systemic disorders (e.g., septicemia) Primary reason for treatment MSDRGs were assigned to broad categoriesa From prior hospital stay MSDRG if available; used PAC stay to proxy MSDRG if no prior stay found. Groupings exclude current ventilator cases Ventilator care Patient was on a ventilator during PAC stay PAC diagnosis Patient comorbidities Comorbidities Prior hospital stay and PAC stay secondary diagnoses combined to 22 groups of CMS- HCC PACE/ESRD categories. Respirator dependence is measured only in PAC stay. 22 APPENDIX A Model Predictors Characteristic Stay predictor Source Treatments and impairments Indicators of bowel incontinence, continuous positive airflow PAC diagnoses; vision from PAC and prior pressure in institutional setting, urinary incontinence, vision hospital stay diagnoses impairment, difficulty swallowing, with tracheostomy Risk score Risk score and squared risk score 2019 CMS-HCC risk score Total number of ICU and CCU days Total number of ICU and CCU days (capped at 15) From prior hospital stay claim Severity level APRDRG severity levels 1–4. Indicators for levels 2, 3, and 4 Stay assigned to APRDRG severity of illness levels 1–4 using claim from prior hospital stay (or proxied if no prior hospital stay within 30 days was found) Severe wound Includes nonhealing surgical wound, wound for a patient who is PAC diagnoses morbidly obese, fistula, osteomyelitis, or patient with a stage III, stage IV, or an unstageable pressure wound Number of body systems ≥ 5 Secondary diagnoses include five or more body systems and Count of comorbidities from prior hospital stay is in institutional setting stay and PAC stay Disabled Original reason for entitlement is disabled Medicare enrollment file Function score Indicators of total functional score between [0,6), [6,12), [12,18), Composite of measures at admission of six [18,24); assessment items: ability to perform toileting hygiene, bathe/wash, roll left/right, walk 10 feet, transfer from sit to lying, and transfer from sit to stand. Assessments with refusals and not attempted due to environmental limitations are excluded with the score increased to maintain the scale of 30 points. Home health agency patient Patient treated by a home health agency Home health claim Notes: PAC = post-acute care. CMS HCC = Centers for Medicare & Medicaid Services Hierarchical Condition Category. PACE = Programs of All-Inclusive Care for the Elderly. ESRD = end-stage renal disease. MSDRG = Medicare Severity Diagnosis-Related Groups. ICU = intensive care unit. CCU = coronary care unit. APRDRG = All Patients Refined Diagnosis- Related Groups. Frailty indicators are the 13 components of the JEN Frailty index. Comorbidity groups are alcohol or drug disease; cancer; cardiac and vascular; complications of device or graft; dementia; eye disorders; gastrointestinal and liver; head and spine; hematologic and immunologic disease; HIV/AIDS; mental illness; metabolic endocrine; neurological, excluding stroke; obesity; orthopedic; renal; respirator dependence; respiratory; septicemia and other systemic infection; skin disorders; stroke; and transplant. APPENDIX A 23 a Broad groups for primary reason for treatment are stroke; neurological surgical; neurological medical; respiratory with tracheostomy or ventilator care; respiratory surgical; respiratory medical; chronic obstructive pulmonary disease; cardiovascular surgical; cardiac medical; orthopedic spinal; orthopedic surgical; orthopedic medical; skin surgical; skin medical; endocrine and metabolic surgical; endocrine and metabolic medical; kidney and urinary surgical; kidney and urinary medical; infections surgical; infections medical (except septicemia); infections including septicemia; transplant; gastrointestinal surgical; gastrointestinal medical; liver and pancreas medical; liver and pancreas surgical; hematology (except cancer) surgical; hematology (except cancer) medical; cancer surgical; cancer medical; trauma, injury, and burns surgical; trauma, injury, and burns medical; mental medical; alcohol and drug abuse; HIV; male reproductive medical; female reproductive medical; other surgery; and other medical. TABLE A.3 Models of Costs per Stay Including Total Functional Score, based on April–September 2019 PAC Stays with Function Data Routine and Therapy Costs per Stay Nontherapy Ancillary Costs per Stay Cluster Cluster robust robust standard t- standard t- Predictor Coefficient error statistic Exp (coef) Coefficient error statistic Exp (coef) Age minus 50 (age restricted to 50–95) Age minus 50 0.002 0.0003 5.46 1.002 -0.002 0.0009 -2.63 0.998 Age minus 50 squared -0.000014 0.00001 -2.37 1.000 -0.000189 0.00002 -11.78 1.000 Age less than 50 0.006 0.0062 0.94 1.006 0.084 0.0161 5.22 1.088 Cognitive function Coma -0.051 0.0091 -5.59 0.950 -0.040 0.0199 -2.00 0.961 Dementia with and without complications (HCC51 and HCC52) -0.067 0.0105 -6.36 0.935 0.030 0.0334 0.90 1.031 Schizophrenia (HCC57) 0.076 0.0082 9.20 1.079 0.035 0.0201 1.74 1.036 Major depressive, bipolar, and paranoid disorders (HCC58) -0.006 0.0088 -0.67 0.994 -0.015 0.0269 -0.57 0.985 Frailty (JEN Frailty Index components) Minor ambulatory limitations 0.045 0.0038 11.65 1.046 -0.095 0.0280 -3.38 0.910 Severe ambulatory limitations 0.094 0.0019 49.70 1.099 -0.047 0.0063 -7.37 0.954 Cognitive developmental disorder -0.006 0.0079 -0.82 0.994 -0.077 0.0224 -3.44 0.926 Chronic mental illness 0.011 0.0018 6.03 1.011 0.037 0.0068 5.42 1.037 Dementia 0.047 0.0050 9.39 1.048 -0.045 0.0113 -4.03 0.956 Sensory disorders 0.021 0.0033 6.15 1.021 -0.047 0.0077 -6.02 0.954 Self-care impairment 0.026 0.0017 15.88 1.027 0.048 0.0060 7.96 1.049 Syncope 0.039 0.0022 17.54 1.040 0.022 0.0076 2.90 1.022 Cancer -0.059 0.0049 -12.08 0.943 -0.086 0.0150 -5.78 0.917 Chronic medical disease 0.005 0.0019 2.79 1.005 0.081 0.0062 13.06 1.084 24 APPENDIX A Routine and Therapy Costs per Stay Nontherapy Ancillary Costs per Stay Cluster Cluster robust robust standard t- standard t- Predictor Coefficient error statistic Exp (coef) Coefficient error statistic Exp (coef) Pneumonia 0.014 0.0027 5.25 1.014 0.104 0.0079 13.08 1.109 Renal disorders -0.004 0.0040 -1.00 0.996 -0.017 0.0110 -1.50 0.984 Systemic disorders (e.g., septicemia) 0.032 0.0016 19.93 1.033 0.093 0.0043 21.61 1.098 Primary reason for treatmenta Stroke 0.158 0.0055 28.89 1.171 0.011 0.0108 1.03 1.011 Neurological surgical 0.163 0.0070 23.15 1.177 0.069 0.0157 4.38 1.071 Neurological medical -0.001 0.0036 -0.25 0.999 -0.059 0.0098 -6.08 0.942 Respiratory with trach/vent 0.058 0.0089 6.48 1.059 0.172 0.0210 8.19 1.188 Respiratory surgical -0.074 0.0108 -6.82 0.929 0.054 0.0280 1.95 1.056 Respiratory medical -0.107 0.0038 -27.93 0.898 -0.096 0.0095 -10.18 0.908 COPD -0.070 0.0051 -13.72 0.932 0.052 0.0135 3.87 1.054 Cardiovascular surgical -0.072 0.0043 -16.66 0.930 -0.088 0.0102 -8.63 0.916 Cardiac medical -0.100 0.0035 -28.91 0.905 -0.155 0.0097 -16.04 0.856 Orthopedic spinal 0.022 0.0066 3.29 1.022 -0.055 0.0130 -4.23 0.947 Orthopedic medical 0.016 0.0032 4.98 1.016 -0.007 0.0072 -0.99 0.993 Skin surgical 0.024 0.0108 2.26 1.025 0.233 0.0346 6.73 1.262 Skin medical -0.062 0.0046 -13.46 0.940 -0.071 0.0137 -5.15 0.932 Endocrine and metabolic surgical 0.058 0.0089 6.48 1.059 0.264 0.0223 11.87 1.302 Endocrine and metabolic medical -0.074 0.0108 -6.82 0.929 -0.162 0.0109 -14.82 0.850 Kidney and urinary surgical -0.107 0.0038 -27.93 0.898 -0.138 0.0262 -5.26 0.871 Kidney and urinary medical -0.070 0.0051 -13.72 0.932 -0.279 0.0086 -32.32 0.757 Infections surgical -0.072 0.0043 -16.66 0.930 0.137 0.0142 9.71 1.147 Infections medical, except septicemia -0.100 0.0035 -28.91 0.905 0.174 0.0304 5.72 1.190 Infections septicemia 0.022 0.0066 3.29 1.022 -0.196 0.0121 -16.17 0.822 Transplant 0.016 0.0032 4.98 1.016 0.425 0.0732 5.80 1.529 GI surgical 0.024 0.0108 2.26 1.025 -0.037 0.0199 -1.84 0.964 GI medical -0.062 0.0046 -13.46 0.940 -0.194 0.0113 -17.17 0.824 Liver and pancreas medical 0.058 0.0089 6.48 1.059 -0.133 0.0310 -4.30 0.875 Liver and pancreas surgical -0.074 0.0108 -6.82 0.929 -0.177 0.0178 -9.98 0.838 Hematology, except cancer surgical -0.107 0.0038 -27.93 0.898 -0.285 0.0742 -3.84 0.752 Hematology, except cancer medical -0.070 0.0051 -13.72 0.932 -0.214 0.0182 -11.75 0.807 Cancer surgical -0.072 0.0043 -16.66 0.930 -0.090 0.0530 -1.69 0.914 Cancer medical -0.169 0.0127 -13.27 0.844 -0.197 0.0368 -5.35 0.821 APPENDIX A 25 Routine and Therapy Costs per Stay Nontherapy Ancillary Costs per Stay Cluster Cluster robust robust standard t- standard t- Predictor Coefficient error statistic Exp (coef) Coefficient error statistic Exp (coef) Trauma, injury, and burns surgical 0.084 0.0084 9.94 1.087 0.152 0.0246 6.19 1.165 Trauma, injury, and burns medical -0.046 0.0070 -6.65 0.955 -0.117 0.0178 -6.58 0.890 Mental medical -0.042 0.0083 -5.03 0.959 -0.133 0.0383 -3.47 0.875 Alcohol and drug abuse -0.177 0.0154 -11.44 0.838 -0.612 0.0315 -19.41 0.542 HIV -0.016 0.0314 -0.51 0.984 0.017 0.0609 0.28 1.017 Male reproductive medical -0.113 0.0145 -7.81 0.893 -0.166 0.0428 -3.88 0.847 Female reproductive medical -0.162 0.0209 -7.75 0.850 -0.269 0.0544 -4.95 0.764 Other surgery -0.020 0.0072 -2.80 0.980 0.074 0.0198 3.75 1.077 Other medical -0.035 0.0045 -7.83 0.965 -0.014 0.0133 -1.08 0.986 Ventilator in post-acute care 0.635 0.0164 38.63 1.887 1.424 0.0309 46.03 4.154 Comorbidities Alcohol or drug disease -0.019 0.0042 -4.52 0.981 -0.088 0.0114 -7.72 0.916 Cancer -0.006 0.0048 -1.29 0.994 -0.034 0.0150 -2.29 0.966 Cardiac and vascular 0.015 0.0015 9.51 1.015 0.116 0.0050 22.97 1.123 Complications of device or graft 0.023 0.0039 6.02 1.023 0.199 0.0106 18.77 1.221 Dementia 0.005 0.0094 0.56 1.005 -0.075 0.0315 -2.39 0.927 Eye disorders 0.000 0.0118 -0.03 1.000 -0.069 0.0380 -1.82 0.933 GI and liver 0.020 0.0028 7.32 1.020 0.133 0.0091 14.51 1.142 Head and spine 0.076 0.0049 15.61 1.079 0.072 0.0112 6.46 1.075 Hematologic and immunologic disease 0.015 0.0025 5.94 1.015 0.058 0.0083 6.93 1.059 HIV/AIDS 0.027 0.0135 2.01 1.027 0.334 0.0352 9.47 1.396 Mental illness -0.005 0.0090 -0.56 0.995 -0.117 0.0276 -4.24 0.890 Metabolic endocrine 0.025 0.0016 15.54 1.025 0.211 0.0055 38.32 1.236 Neurological, excluding stroke 0.033 0.0017 19.30 1.034 0.088 0.0055 16.03 1.092 Obesity 0.035 0.0026 13.46 1.036 0.091 0.0083 10.98 1.095 Orthopedic 0.052 0.0035 14.96 1.054 0.020 0.0089 2.28 1.021 Renal 0.001 0.0041 0.25 1.001 0.077 0.0108 7.14 1.080 Respirator dependence 0.190 0.0141 13.51 1.209 0.285 0.0263 10.83 1.330 Respiratory -0.005 0.0017 -2.97 0.995 0.188 0.0058 32.36 1.207 Septicemia and other systemic infection 0.027 0.0043 6.35 1.028 0.095 0.0114 8.35 1.100 Skin disorders 0.056 0.0026 21.28 1.057 0.130 0.0076 17.07 1.138 Stroke 0.024 0.0027 8.97 1.024 0.021 0.0072 2.87 1.021 Transplant 0.064 0.0102 6.20 1.066 0.135 0.0297 4.54 1.145 26 APPENDIX A Routine and Therapy Costs per Stay Nontherapy Ancillary Costs per Stay Cluster Cluster robust robust standard t- standard t- Predictor Coefficient error statistic Exp (coef) Coefficient error statistic Exp (coef) Functional score 0 to 5.999 0.399 0.0060 66.58 1.491 0.223 0.0286 7.78 1.249 6 to 11.999 0.382 0.0056 68.16 1.465 0.130 0.0253 5.15 1.139 12 to 17.999 0.230 0.0053 43.03 1.258 -0.044 0.0251 -1.75 0.957 18 to 23.999 0.113 0.0046 24.46 1.119 -0.136 0.0232 -5.83 0.873 Treatments and impairments Bowel incontinence 0.159 0.0106 15.02 1.173 0.197 0.0295 6.67 1.218 Urinary incontinence 0.093 0.0052 17.84 1.097 0.130 0.0113 11.47 1.139 Vision impairment -0.009 0.0048 -1.93 0.991 0.052 0.0160 3.25 1.053 Continuous positive airflow pressure 0.443 0.0151 29.35 1.558 0.999 0.0265 37.64 2.715 Swallowing 0.082 0.0029 28.05 1.085 -0.058 0.0111 -5.22 0.943 Tracheostomy -0.086 0.0220 -3.90 0.918 -0.023 0.0346 -0.66 0.977 Risk score -0.016 0.0008 -19.98 0.984 0.039 0.0023 16.87 1.040 Risk score squared 0.000 0.0001 7.19 1.000 -0.002 0.0002 -11.82 0.998 Total number of ICU and CCU days (capped) 0.002 0.0004 4.69 1.002 0.008 0.0011 7.83 1.008 Severity level Two 0.024 0.0019 12.53 1.024 0.098 0.0069 14.14 1.102 Three -0.001 0.0026 -0.20 0.999 0.118 0.0079 14.90 1.125 Four -0.015 0.0034 -4.40 0.985 0.174 0.0101 17.25 1.190 Wound care Pressure ulcer, stage III 0.133 0.0071 18.81 1.142 0.184 0.0177 10.40 1.202 Pressure ulcer, stage IV 0.202 0.0105 19.26 1.224 0.404 0.0183 22.13 1.498 Pressure ulcer, unstageable 0.051 0.0071 7.21 1.053 0.126 0.0182 6.93 1.135 Wound with morbid obesity 0.051 0.0092 5.55 1.052 0.041 0.0198 2.06 1.042 Fistula 0.236 0.0178 13.26 1.266 0.557 0.0325 17.12 1.745 Nonhealing surgical wound 0.178 0.0081 21.93 1.195 0.455 0.0236 19.30 1.576 Number of body systems ≥ 5 -0.018 0.0025 -7.11 0.983 0.029 0.0074 3.96 1.030 Disabled -0.017 0.0019 -8.56 0.984 -0.020 0.0059 -3.35 0.980 Home health agency patient -1.867 0.0060 -312.29 0.155 Constant 8.954 0.0102 876.22 7740.247 6.949 0.0455 152.72 1042.317 APPENDIX A 27 Routine and Therapy Costs per Stay Nontherapy Ancillary Costs per Stay Cluster Cluster robust robust standard t- standard t- Predictor Coefficient error statistic Exp (coef) Coefficient error statistic Exp (coef) Sample size 3,692,064 1,053,039 Combined Routine+Therapy and NTA model R2 0.543 Sources: 2019 Medicare acute hospital and post-acute care claims and assessments, Medicare 2019 risk score file, and Medicare cost reports for 2019. Notes: PAC = post-acute care. COPD = chronic obstructive pulmonary disease. GI = gastrointestinal. ICU = intensive care unit. CCU = coronary care unit. Models estimated using Poisson regression. Standard errors are clustered by provider. Data are all stays that began between April and September 2019 and had the CARE function variables on a matched assessment. Routine and therapy costs model is based on data from home health and institutional settings; the model of nontherapy ancillary costs is based only on data from institutional settings. a Orthopedic surgery is the omitted group. TABLE A.4 Models of Costs per Stay Excluding Total Functional Score, based on April–September 2019 PAC Stays with Function Data Routine and Therapy Costs per Stay Nontherapy Ancillary Costs per Stay Cluster Cluster robust robust standard t- standard t- Predictor Coefficient error statistic Exp (coef) Coefficient error statistic Exp (coef) Age minus 50 (age restricted to 50–95) Age minus 50 0.003 0.0004 7.20 1.003 -0.002 0.0009 -1.94 0.998 Age minus 50 squared -0.000002 0.00001 -0.32 1.000 -0.00018 0.00002 -11.11 1.000 Age less than 50 0.006 0.0062 0.98 1.006 0.082 0.0161 5.10 1.086 Cognitive function Coma -0.065 0.0091 -7.15 0.937 -0.052 0.0201 -2.58 0.949 Dementia with and without complications (HCC51 and HCC52) -0.029 0.0105 -2.77 0.971 0.062 0.0334 1.86 1.064 Schizophrenia (HCC57) 0.065 0.0083 7.80 1.067 0.037 0.0204 1.82 1.038 Major depressive, bipolar, and paranoid disorders (HCC58) -0.015 0.0089 -1.64 0.986 -0.018 0.0272 -0.66 0.982 28 APPENDIX A Routine and Therapy Costs per Stay Nontherapy Ancillary Costs per Stay Cluster Cluster robust robust standard t- standard t- Predictor Coefficient error statistic Exp (coef) Coefficient error statistic Exp (coef) Frailty (JEN Frailty Index components) Minor ambulatory limitations 0.052 0.0039 13.49 1.054 -0.094 0.0283 -3.33 0.910 Severe ambulatory limitations 0.118 0.0019 60.77 1.126 -0.021 0.0060 -3.55 0.979 Cognitive developmental disorder 0.026 0.0079 3.27 1.026 -0.042 0.0220 -1.90 0.959 Chronic mental illness 0.009 0.0018 4.80 1.009 0.033 0.0069 4.84 1.034 Dementia 0.047 0.0050 9.28 1.048 -0.047 0.0113 -4.15 0.954 Sensory disorders 0.014 0.0034 4.24 1.014 -0.053 0.0079 -6.67 0.949 Self-care impairment 0.030 0.0017 17.85 1.031 0.055 0.0059 9.21 1.056 Syncope 0.034 0.0023 14.82 1.034 0.015 0.0078 1.92 1.015 Cancer -0.062 0.0049 -12.65 0.940 -0.089 0.0150 -5.95 0.915 Chronic medical disease 0.000 0.0020 -0.09 1.000 0.074 0.0062 11.88 1.077 Pneumonia 0.016 0.0027 5.79 1.016 0.108 0.0081 13.30 1.114 Renal disorders -0.008 0.0041 -2.07 0.992 -0.021 0.0111 -1.87 0.979 Systemic disorders (e.g., septicemia) 0.043 0.0016 26.07 1.044 0.108 0.0045 23.82 1.114 Primary reason for treatmenta Stroke 0.113 0.0055 20.64 1.120 -0.033 0.0108 -3.08 0.967 Neurological surgical 0.119 0.0071 16.80 1.127 0.027 0.0157 1.71 1.027 Neurological medical -0.023 0.0036 -6.22 0.978 -0.091 0.0098 -9.31 0.913 Respiratory with trach/vent 0.028 0.0090 3.08 1.028 0.146 0.0211 6.90 1.157 Respiratory surgical -0.117 0.0109 -10.70 0.890 0.008 0.0280 0.28 1.008 Respiratory medical -0.138 0.0039 -35.76 0.871 -0.131 0.0092 -14.29 0.877 COPD -0.117 0.0051 -22.80 0.890 -0.012 0.0129 -0.92 0.988 Cardiovascular surgical -0.114 0.0044 -26.22 0.892 -0.137 0.0100 -13.78 0.872 Cardiac medical -0.138 0.0034 -40.27 0.871 -0.202 0.0088 -22.95 0.817 Orthopedic spinal 0.015 0.0067 2.16 1.015 -0.070 0.0132 -5.33 0.932 Orthopedic medical 0.004 0.0032 1.31 1.004 -0.017 0.0073 -2.40 0.983 Skin surgical -0.005 0.0109 -0.45 0.995 0.219 0.0347 6.29 1.244 Skin medical -0.075 0.0045 -16.58 0.927 -0.094 0.0135 -6.98 0.910 Endocrine and metabolic surgical -0.016 0.0092 -1.74 0.984 0.205 0.0218 9.42 1.228 Endocrine and metabolic medical -0.123 0.0044 -28.04 0.885 -0.205 0.0113 -18.22 0.814 Kidney and urinary surgical -0.148 0.0083 -17.87 0.863 -0.171 0.0256 -6.66 0.843 Kidney and urinary medical -0.138 0.0038 -35.92 0.871 -0.300 0.0085 -35.20 0.740 Infections surgical -0.009 0.0068 -1.25 0.9 92 0.109 0.0143 7.62 1.115 APPENDIX A 29 Routine and Therapy Costs per Stay Nontherapy Ancillary Costs per Stay Cluster Cluster robust robust standard t- standard t- Predictor Coefficient error statistic Exp (coef) Coefficient error statistic Exp (coef) Infections medical, except septicemia -0.099 0.0104 -9.54 0.906 0.137 0.0307 4.47 1.147 Infections septicemia -0.181 0.0054 -33.78 0.835 -0.228 0.0123 -18.52 0.796 Transplant 0.121 0.0347 3.49 1.129 0.359 0.0747 4.80 1.431 GI surgical -0.116 0.0059 -19.61 0.890 -0.081 0.0194 -4.16 0.922 GI medical -0.168 0.0042 -40.29 0.845 -0.227 0.0109 -20.87 0.797 Liver and pancreas medical -0.170 0.0107 -15.93 0.843 -0.177 0.0314 -5.63 0.838 Liver and pancreas surgical -0.244 0.0071 -34.32 0.784 -0.222 0.0170 -13.02 0.801 Hematology, except cancer surgical -0.215 0.0380 -5.66 0.806 -0.325 0.0751 -4.33 0.723 Hematology, except cancer medical -0.164 0.0074 -22.28 0.848 -0.247 0.0185 -13.35 0.781 Cancer surgical -0.105 0.0224 -4.69 0.900 -0.121 0.0533 -2.26 0.886 Cancer medical -0.197 0.0127 -15.48 0.821 -0.224 0.0369 -6.09 0.799 Trauma, injury, and burns surgical 0.076 0.0086 8.86 1.079 0.141 0.0246 5.74 1.151 Trauma, injury, and burns medical -0.081 0.0071 -11.47 0.922 -0.157 0.0173 -9.09 0.854 Mental medical -0.092 0.0083 -11.06 0.912 -0.181 0.0390 -4.65 0.834 Alcohol and drug abuse -0.236 0.0158 -14.92 0.790 -0.672 0.0315 -21.32 0.511 HIV -0.058 0.0316 -1.83 0.944 -0.011 0.0605 -0.18 0.989 Male reproductive medical -0.134 0.0146 -9.19 0.875 -0.191 0.0430 -4.44 0.826 Female reproductive medical -0.171 0.0209 -8.22 0.843 -0.267 0.0540 -4.94 0.766 Other surgery -0.051 0.0072 -7.04 0.950 0.044 0.0198 2.24 1.045 Other medical -0.064 0.0045 -14.14 0.938 -0.046 0.0133 -3.42 0.955 Ventilator in post-acute care 0.635 0.0167 38.08 1.886 1.444 0.0314 45.92 4.237 Comorbidities Alcohol or drug disease -0.036 0.0043 -8.30 0.965 -0.104 0.0114 -9.11 0.901 Cancer -0.009 0.0048 -1.80 0.991 -0.039 0.0149 -2.64 0.961 Cardiac and vascular 0.017 0.0016 10.64 1.017 0.119 0.0050 23.61 1.126 Complications of device or graft 0.030 0.0038 7.80 1.030 0.209 0.0105 19.88 1.232 Dementia -0.004 0.0094 -0.41 0.996 -0.076 0.0313 -2.42 0.927 Eye disorders -0.011 0.0119 -0.96 0.989 -0.084 0.0385 -2.17 0.920 GI and liver 0.030 0.0028 10.74 1.030 0.146 0.0090 16.22 1.157 Head and spine 0.109 0.0049 22.29 1.116 0.104 0.0115 9.03 1.110 Hematologic and immunologic disease 0.013 0.0025 5.14 1.013 0.055 0.0083 6.59 1.056 HIV/AIDS 0.017 0.0136 1.23 1.017 0.324 0.0354 9.15 1.382 Mental illness 0.004 0.0091 0.41 1.004 -0.114 0.0278 -4.10 0.892 30 APPENDIX A Routine and Therapy Costs per Stay Nontherapy Ancillary Costs per Stay Cluster Cluster robust robust standard t- standard t- Predictor Coefficient error statistic Exp (coef) Coefficient error statistic Exp (coef) Metabolic endocrine 0.028 0.0016 17.64 1.029 0.216 0.0056 38.27 1.241 Neurological, excluding stroke 0.041 0.0017 23.90 1.042 0.095 0.0055 17.29 1.099 Obesity 0.058 0.0026 21.98 1.059 0.118 0.0077 15.20 1.125 Orthopedic 0.067 0.0035 19.10 1.069 0.034 0.0088 3.83 1.034 Renal 0.003 0.0041 0.74 1.003 0.078 0.0109 7.15 1.081 Respirator dependence 0.206 0.0143 14.41 1.228 0.306 0.0270 11.32 1.358 Respiratory -0.012 0.0017 -6.79 0.988 0.180 0.0060 30.23 1.197 Septicemia and other systemic infection 0.035 0.0044 8.07 1.036 0.105 0.0118 8.89 1.111 Skin disorders 0.085 0.0026 32.38 1.088 0.163 0.0081 20.14 1.177 Stroke 0.038 0.0027 14.05 1.039 0.036 0.0075 4.78 1.036 Transplant 0.050 0.0103 4.84 1.051 0.119 0.0300 3.96 1.126 Functional score 0 to 5.999 0.399 0.0060 66.58 1.491 0.223 0.0286 7.78 1.249 6 to 11.999 0.382 0.0056 68.16 1.465 0.130 0.0253 5.15 1.139 12 to 17.999 0.230 0.0053 43.03 1.258 -0.044 0.0251 -1.75 0.957 18 to 23.999 0.113 0.0046 24.46 1.119 -0.136 0.0232 -5.83 0.873 Treatments and impairments Bowel incontinence 0.185 0.0107 17.38 1.204 0.219 0.0307 7.14 1.245 Urinary incontinence 0.105 0.0052 20.17 1.111 0.137 0.0114 12.03 1.147 Vision impairment -0.001 0.0048 -0.14 0.999 0.060 0.0163 3.67 1.062 Continuous positive airflow pressure 0.462 0.0153 30.14 1.587 1.022 0.0272 37.64 2.779 Swallowing 0.113 0.0029 38.86 1.120 -0.023 0.0101 -2.29 0.977 Tracheostomy -0.082 0.0222 -3.69 0.922 -0.015 0.0346 -0.44 0.985 Risk score -0.011 0.0008 -13.83 0.989 0.042 0.0024 17.97 1.043 Risk score squared 0.0003 0.0001 4.58 1.000 -0.002 0.0002 -12.33 0.998 Total number of ICU and CCU days (capped) 0.003 0.0005 6.25 1.003 0.009 0.0011 8.31 1.009 Severity level Two 0.028 0.0019 14.58 1.028 0.104 0.0070 14.75 1.109 Three 0.007 0.0027 2.77 1.007 0.132 0.0082 16.23 1.142 Four 0.001 0.0035 0.33 1.001 0.202 0.0102 19.77 1.224 Wound care APPENDIX A 31 Routine and Therapy Costs per Stay Nontherapy Ancillary Costs per Stay Cluster Cluster robust robust standard t- standard t- Predictor Coefficient error statistic Exp (coef) Coefficient error statistic Exp (coef) Pressure ulcer, stage III 0.148 0.0071 20.89 1.160 0.193 0.0177 10.90 1.213 Pressure ulcer, stage IV 0.222 0.0104 21.31 1.248 0.420 0.0190 22.14 1.522 Pressure ulcer, unstageable 0.068 0.0070 9.73 1.071 0.141 0.0180 7.82 1.151 Wound with morbid obesity 0.045 0.0092 4.82 1.046 0.033 0.0200 1.67 1.034 Fistula 0.236 0.0180 13.12 1.266 0.562 0.0330 17.02 1.754 Nonhealing surgical wound 0.170 0.0082 20.78 1.186 0.459 0.0238 19.28 1.582 Number of body systems ≥ five -0.015 0.0025 -5.87 0.985 0.040 0.0078 5.05 1.040 Disabled -0.010 0.0019 -5.04 0.990 -0.012 0.0061 -2.02 0.988 Home health agency patient -1.923 0.0059 -323.80 0.146 Constant 9.185 0.0087 1059.43 9749.064 6.951 0.0331 210.13 1043.959 Sample size 3,692,064 1,053,039 Combined Routine+Therapy and NTA model R2 0.532 Sources: 2019 Medicare acute hospital and post-acute care claims, Medicare 2019 risk score file, and Medicare cost reports for 2019. Notes: PAC = post-acute care. COPD = chronic obstructive pulmonary disease. GI = gastrointestinal. ICU = intensive care unit. CCU = coronary care unit. Models estimated using Poisson regression. Standard errors are clustered by provider. Data are all stays that began between April and September 2019 and had the CARE function variables on a matched assessment. Routine and therapy costs model is based on data from home health and institutional settings; the model of nontherapy ancillary costs is based only on data from institutional settings. Functional score is measured on a 30-point scale that increases with functionality. a Orthopedic surgery is the omitted group. 32 APPENDIX A TABLE A.5 Comparison of Actual Costs, Predicted Costs, Actual Payments, and Payments (Including Outliers) under a PAC PPS for PAC Stays from April–September 2019, with Function in the Model, 5 Percent Outlier Pool and Short-Stay Outlier Payments Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predicted 2019 under payment to actual cost cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) All 5,496 5,496 6,266 6,259 1.14 1.00 3,692,064 71.5 22.8 4.8 1.0 Clinical group Ventilator 62,681 62,681 71,415 75,178 1.20 1.05 8,451 0.0 0.6 0.8 98.6 Severe wounds 7,655 7,470 8,773 8,791 1.15 1.00 157,912 70.7 19.8 4.4 5.0 Stroke 11,473 11,453 12,641 12,978 1.13 1.03 88,788 43.7 31.2 24.6 0.4 Other neurology medical 4,251 4,239 4,949 4,831 1.14 0.98 344,129 81.2 13.2 5.5 0.1 Other neurology surgical 11,185 11,202 12,123 12,556 1.12 1.04 32,188 45.4 26.1 27.6 1.0 Orthopedic medical 3,703 3,704 4,163 4,251 1.15 1.02 523,541 85.1 11.6 3.2 0.1 Orthopedic surgical 7,672 7,701 8,407 8,773 1.14 1.04 363,782 55.1 35.4 9.4 0.2 Respiratory medical 5,185 5,242 5,976 5,900 1.14 0.99 290,637 69.7 26.6 2.7 1.0 Respiratory surgical 5,546 5,627 6,318 6,257 1.13 0.99 11,808 67.0 25.5 5.7 1.8 Cardiovascular medical 3,884 3,881 4,390 4,397 1.13 1.00 477,884 79.8 17.9 1.9 0.4 Cardiovascular surgical 6,220 6,281 7,010 7,002 1.13 1.00 123,119 62.4 27.4 9.2 1.0 Infection medical 7,414 7,448 8,887 8,401 1.13 0.95 184,382 52.5 41.7 3.9 2.0 Infection surgical 9,696 9,902 11,142 11,259 1.16 1.01 39,087 47.6 41.1 7.3 4.0 Hematology medical 4,601 4,607 5,308 5,174 1.12 0.97 37,193 71.7 25.6 2.4 0.3 Hematology surgical 5,924 6,021 6,718 6,697 1.13 1.00 3,225 59.9 31.9 7.7 0.6 Rehabilitation medical 9,295 8,868 9,623 10,071 1.08 1.05 2,179 45.6 32.3 22.1 0.0 Skin medical 3,417 3,386 4,105 3,894 1.14 0.95 115,348 85.5 13.1 1.1 0.3 Skin surgical 6,168 6,285 7,089 7,171 1.16 1.01 10,213 66.9 28.4 2.9 1.8 Kidney and urine medical 5,463 5,438 6,290 6,197 1.13 0.99 220,018 66.9 30.5 2.3 0.3 Liver medical 4,989 4,980 5,721 5,496 1.10 0.96 30,063 65.3 31.3 2.8 0.5 Digestive medical 5,096 5,100 5,906 5,742 1.13 0.97 126,434 68.1 29.2 2.3 0.4 Endocrine medical 4,384 4,358 5,159 4,956 1.13 0.96 123,093 76.1 21.7 1.9 0.3 Mental illness medical 4,002 3,997 4,752 4,590 1.15 0.97 56,973 78.7 20.6 0.5 0.2 Alcohol and drug use 8,167 8,192 9,805 9,183 1.12 0.94 3,889 25.7 67.7 6.1 0.6 medical HIV medical 7,982 7,758 10,837 8,939 1.12 0.82 1,220 56.4 37.5 3.2 2.9 Other medical 3,364 3,367 3,930 3,828 1.14 0.97 181,258 85.6 10.7 3.5 0.1 APPENDIX A 33 Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predicted 2019 under payment to actual cost cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) Other surgical 7,076 7,050 8,047 7,998 1.13 0.99 135,250 60.1 31.4 6.5 2.0 Other clinical conditions Cancer 4,083 4,083 4,708 4,521 1.11 0.96 13,852 75.1 21.0 3.5 0.4 Transplant 5,961 5,961 6,004 6,655 1.12 1.11 3,830 79.3 4.9 14.2 1.7 Kidney/urinary 5,618 5,624 6,497 6,405 1.14 0.99 249,759 66.3 30.6 2.5 0.5 GI or hepatobiliary 5,341 5,341 6,098 6,015 1.13 0.99 223,312 66.8 29.3 3.1 0.9 Vision impairment 6,139 6,139 7,081 7,008 1.14 0.99 83,938 68.2 24.2 6.7 1.0 Urinary incontinence 6,342 6,342 7,043 7,278 1.15 1.03 99,207 72.7 11.8 14.7 0.7 Trauma 6,118 5,798 6,975 6,704 1.10 0.96 160,667 71.2 19.1 8.8 0.9 Frailty, cognitive function, mental illness, and functional score Least frail 2,032 2,059 2,424 2,352 1.16 0.97 581,928 95.6 3.8 0.6 0.0 Most frail 9,179 9,171 10,380 10,445 1.14 1.01 1,245,183 47.3 41.9 8.6 2.2 Cognitively impaired 6,618 6,615 7,748 7,576 1.14 0.98 747,570 63.7 31.8 3.5 1.0 Serious mental illness 7,550 7,552 8,875 8,656 1.15 0.98 386,130 54.7 41.5 3.0 0.8 Function, 0–25th 9,742 9,636 11,283 10,981 1.13 0.97 853,908 51.5 39.2 6.5 2.9 percentile Function, 25–75th 5,001 5,040 5,634 5,749 1.15 1.02 1,889,877 72.5 21.4 5.7 0.4 percentile Function, 75–100th 2,658 2,674 3,009 3,022 1.14 1.00 948,279 87.5 10.7 1.5 0.4 percentile Severely ill (SOI level 4) 17,346 17,394 19,825 19,685 1.13 0.99 238,693 0.0 75.9 14.0 10.0 Multiple body systems 17,064 17,064 19,762 19,449 1.14 0.98 353,374 0.0 78.2 14.0 7.8 Chronically critically ill 12,806 12,631 14,443 14,533 1.13 1.01 149,404 43.9 37.2 9.0 10.0 Highest acuity 16,123 15,860 18,188 18,321 1.14 1.01 82,410 36.5 38.6 9.7 15.2 Other stay and patient characteristics HHA, no therapy visits 872 1,615 1,600 1,609 1.85 1.01 637,749 100.0 0.0 0.0 0.0 HHA, 1–4 therapy visits 987 1,679 1,695 1,729 1.75 1.02 583,154 100.0 0.0 0.0 0.0 HHA, 5–9 therapy visits 1,884 1,688 2,167 2,033 1.08 0.94 862,746 100.0 0.0 0.0 0.0 HHA,10+ therapy visits 3,045 1,770 2,524 2,393 0.79 0.95 555,376 100.0 0.0 0.0 0.0 34 APPENDIX A Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predicted 2019 under payment to actual cost cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) I-PAC, therapy costs per 15,227 15,943 19,465 18,139 1.19 0.93 263,871 0.0 90.7 0.4 8.8 day in 0–25th percentile I-PAC, therapy costs per 13,911 14,412 16,140 16,755 1.20 1.04 266,970 0.0 97.9 0.3 1.8 day in 25–50th percentile I-PAC, therapy costs per 14,140 14,522 14,369 16,657 1.18 1.16 266,716 0.0 96.2 2.2 1.7 day in 50–75th percentile I-PAC, therapy costs per 16,982 15,320 17,918 16,782 0.99 0.94 255,482 0.0 32.7 66.2 1.1 day in > 75th percentile HHA community 1,535 1,676 1,819 1,902 1.24 1.05 1,761,523 100.0 0.0 0.0 0.0 admitted HHA stays with prior 1,988 1,705 2,367 2,013 1.01 0.85 877,502 100.0 0.0 0.0 0.0 hospital stay I-PAC community 15,467 14,078 18,029 16,309 1.05 0.90 85,802 0.0 67.2 30.2 2.5 admitted I-PAC stays with prior 15,006 15,130 16,861 17,152 1.14 1.02 967,237 0.0 81.0 15.6 3.4 hospital stay Disabled 5,702 5,702 6,709 6,503 1.14 0.97 883,582 71.1 22.8 4.6 1.6 Dual eligible or LIS 6,016 5,968 7,254 6,847 1.14 0.94 1,255,694 69.0 26.3 3.3 1.4 ESRD 6,670 6,589 7,983 7,466 1.12 0.94 171,844 65.3 26.7 5.6 2.4 Very old (85+) 5,187 5,179 5,829 5,911 1.14 1.01 1,138,287 72.6 23.9 3.2 0.4 SNF shortest 10th 2,311 14,178 2,488 4,046 1.75 1.63 85,909 0.0 100.0 0.0 0.0 percentile IRF shortest 10th 6,639 14,667 10,940 4,657 0.70 0.43 17,935 0.0 0.0 100.0 0.0 percentile LTCH shortest 10th 7,898 26,056 8,223 4,783 0.61 0.58 3,597 0.0 0.0 0.0 100.0 percentile IRF short stay outlier (<=3 3,499 15,551 3,584 2,433 0.70 0.68 5,282 0.0 0.0 100.0 0.0 days) HHA LUPA 353 1,664 332 483 1.37 1.46 230,005 100.0 0.0 0.0 100.0 Provider setting HHA 1,685 1,685 2,001 1,939 1.15 0.97 2,639,025 100.0 0.0 0.0 0.0 SNF 13,179 14,301 14,957 16,014 1.22 1.07 840,922 0.0 100.0 0.0 0.0 IRF 18,393 15,621 21,344 17,621 0.96 0.83 176,755 0.0 0.0 100.0 0.0 APPENDIX A 35 Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predicted 2019 under payment to actual cost cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) LTCH 42,647 29,838 42,564 39,834 0.93 0.94 35,362 0.0 0.0 0.0 100.0 LTCH chronically 47,699 36,461 50,303 47,458 0.99 0.94 22,124 0.0 0.0 0.0 100.0 critically ill by law Provider characteristics Hospital-based 7,507 5,890 7,089 6,872 0.92 0.97 372,747 69.1 9.8 21.2 0.0 Freestanding 5,270 5,451 6,174 6,190 1.17 1.00 3,319,317 71.7 24.2 2.9 1.1 Nonprofit 6,036 5,734 6,146 6,563 1.09 1.07 904,608 69.0 23.7 6.8 0.5 For-profit 5,160 5,294 6,177 6,012 1.17 0.97 2,665,098 73.3 21.7 3.9 1.1 Government 8,811 8,118 9,108 9,383 1.06 1.03 122,358 50.6 39.9 9.2 0.4 Low-volume provider, 15,659 12,618 16,187 15,427 0.99 0.95 16,581 22.0 58.9 14.3 4.8 bottom decile IRF low-income share 16,966 15,288 20,656 17,060 1.01 0.83 41,505 0.0 0.0 100.0 0.0 0–20th percentile IRF low-income share 20– 17,236 15,399 20,851 17,166 1.00 0.82 40,196 0.0 0.0 100.0 0.0 40th percentile IRF low-income share 40– 18,480 15,711 21,301 17,653 0.96 0.83 37,279 0.0 0.0 100.0 0.0 60th percentile IRF low-income share 60– 18,972 15,843 21,561 17,914 0.94 0.83 31,816 0.0 0.0 100.0 0.0 80th percentile IRF low-income share 21,698 16,148 23,140 18,929 0.87 0.82 24,139 0.0 0.0 100.0 0.0 80th+ percentile Teaching (IRF only) 21,743 16,512 23,612 19,345 0.89 0.82 18,558 0.0 0.0 100.0 0.0 Dual/LIS share 0–20th 5,645 5,702 5,804 6,489 1.15 1.12 952,563 68.7 27.0 3.6 0.7 percentile in setting Dual/LIS share 20–40th 4,545 4,681 5,159 5,300 1.17 1.03 1,134,141 77.2 18.9 3.3 0.6 percentile in setting Dual/LIS share 40–60th 4,942 5,019 5,817 5,706 1.15 0.98 811,907 75.3 19.3 4.6 0.8 percentile in setting Dual/LIS share 60–80th 6,507 6,404 7,782 7,282 1.12 0.94 498,054 65.5 25.1 8.0 1.3 percentile in setting Dual/LIS share 80 –100th 8,478 7,735 10,689 8,989 1.06 0.84 295,399 57.8 29.7 9.5 3.1 percentile in setting 36 APPENDIX A Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predicted 2019 under payment to actual cost cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) White Non-Hispanic 5,324 4,906 5,651 5,715 1.07 1.01 239,036 75.9 17.6 5.2 1.3 share, top decile in setting Black Non-Hispanic share, 9,634 9,208 11,305 10,635 1.10 0.94 155,293 46.2 42.4 9.2 2.2 top decile in setting Other race/ethnicity 8,954 8,745 11,987 9,960 1.11 0.83 190,893 49.7 40.1 7.8 2.4 share, top decile in setting Geographic location Frontier 5,700 4,958 6,028 5,956 1.04 0.99 9,317 72.9 27.1 0.0 0.0 Metro 5,441 5,495 6,288 6,242 1.15 0.99 3,191,331 71.7 22.1 5.2 1.1 Rural micropolitan 5,915 5,554 6,183 6,417 1.08 1.04 326,658 69.8 26.5 3.3 0.5 Rural adjacent 5,905 5,546 6,186 6,452 1.09 1.04 106,236 69.0 30.6 0.4 0.0 Rural nonadjacent 5,411 5,158 5,761 5,974 1.10 1.04 67,812 72.0 26.8 1.1 0.0 Urban CBSA based 5,441 5,494 6,287 6,241 1.15 0.99 3,195,018 71.7 22.0 5.2 1.1 Rural CBSA based 5,847 5,503 6,131 6,368 1.09 1.04 496,901 69.9 27.5 2.3 0.3 Regions 1: CT, MA, M, NH, RI, VT 4,417 5,052 5,535 5,628 1.27 1.02 243,902 72.5 23.6 3.2 0.6 2: NY, NJ 6,538 6,396 8,317 7,308 1.12 0.88 276,819 64.0 32.0 3.7 0.4 3: DE, DC, MD, PA, VA, 5,561 5,649 6,025 6,407 1.15 1.06 403,044 70.1 24.0 5.3 0.6 WV 4: AL, FL, GA, KY, MS, NC, 5,111 5,075 5,504 5,799 1.13 1.05 908,759 75.1 19.7 4.4 0.9 SC, TN 5: IL, IN, MI, MN, OH, WI 5,700 5,830 6,261 6,597 1.16 1.05 607,492 68.7 26.5 4.0 0.8 6: AR, LA, NM, OK, TX 6,021 5,589 6,681 6,453 1.07 0.97 472,780 71.8 17.3 8.5 2.4 7: IA, KS, MO, NE 6,494 6,481 6,861 7,406 1.14 1.08 154,319 63.4 29.5 6.0 1.2 8: CO, MT, ND, SD, UT, 6,312 5,731 6,321 6,674 1.06 1.06 79,673 69.0 25.6 4.7 0.6 WY 9: AZ, CA, HI, NV 4,859 4,993 6,542 5,641 1.16 0.86 452,636 75.9 19.3 3.9 0.8 10: AK, ID, OR, WA 5,420 5,373 6,132 6,185 1.14 1.01 92,640 72.1 25.2 2.3 0.4 Sources: 2019 Medicare acute hospital and post-acute care claims and assessments, Medicare 2019 risk score file, and Medicare cost reports for 2019. Notes: PAC = post-acute care. PPS = prospective payment system. HHA = home health agency. SNF = skilled nursing facility. IRF = inpatient rehabilitation facility. LTCH = long-term care hospital. SOI = severity of illness. I-PAC = institutional post-acute care. ESRD = end-stage renal disease. LUPA = low-utilization payment adjustment. CBSA = core-based statistical area. LIS = Low-income subsidy program for Part D enrollees. Data are all stays that began between April and September 2019 and had the CARE function variables on a APPENDIX A 37 matched assessment. Patients' level of frailty was determined using the JEN Frailty Index. Chronically critically ill stays include patients who spent eight or more days in an intensive care or coronary care unit during the preceding hospital stay or were on a ventilator in the PAC setting. LTCH chronically critically ill by law stays include LTCH patients who spent three or more days in an intensive care or coronary care unit during the preceding hospital stay or were on a ventilator in the LTCH. Severely ill stays include institutional-setting patients who were categorized as severity of illness level 4, usually during the immediately preceding hospital stay. Multiple body systems include institutional patients with secondary diagnoses involving five or more body systems. Highest-acuity patients were institutional patients categorized as severity of illness level 4, on dialysis, and who had severe wounds or a pressure ulcer. Race/ethnicity shares are based on the RTI race measure, with top decile based on shares in facilities within a setting with at least 25 stays. TABLE A.6 Comparison of Actual Costs, Predicted Costs, Actual Payments, and Payments (Including Outliers) under a PAC PPS for April–September 2019 PAC Stays, with 5 Percent Outlier Pool and Short-Stay Outlier Payments, Functional Score Omitted from the Model Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predicted 2019 under payment to actual cost cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) All 5,496 5,496 6,266 6,258 1.14 1.00 3,692,064 71.5 22.8 4.8 1.0 Clinical group Ventilator 62,681 62,681 71,415 75,280 1.20 1.05 8,451 0.0 0.6 0.8 98.6 Severe wounds 7,655 7,428 8,773 8,745 1.14 1.00 157,912 70.7 19.8 4.4 5.0 Stroke 11,473 11,457 12,641 13,003 1.13 1.03 88,788 43.7 31.2 24.6 0.4 Other neurology medical 4,251 4,240 4,949 4,831 1.14 0.98 344,129 81.2 13.2 5.5 0.1 Other neurology surgical 11,185 11,205 12,123 12,572 1.12 1.04 32,188 45.4 26.1 27.6 1.0 Orthopedic medical 3,703 3,703 4,163 4,246 1.15 1.02 523,541 85.1 11.6 3.2 0.1 Orthopedic surgical 7,672 7,700 8,407 8,757 1.14 1.04 363,782 55.1 35.4 9.4 0.2 Respiratory medical 5,185 5,243 5,976 5,906 1.14 0.99 290,637 69.7 26.6 2.7 1.0 Respiratory surgical 5,546 5,632 6,318 6,264 1.13 0.99 11,808 67.0 25.5 5.7 1.8 Cardiovascular medical 3,884 3,882 4,390 4,400 1.13 1.00 477,884 79.8 17.9 1.9 0.4 Cardiovascular surgical 6,220 6,286 7,010 7,003 1.13 1.00 123,119 62.4 27.4 9.2 1.0 Infection medical 7,414 7,451 8,887 8,407 1.13 0.95 184,382 52.5 41.7 3.9 2.0 Infection surgical 9,696 9,910 11,142 11,277 1.16 1.01 39,087 47.6 41.1 7.3 4.0 Hematology medical 4,601 4,609 5,308 5,179 1.13 0.98 37,193 71.7 25.6 2.4 0.3 Hematology surgical 5,924 6,024 6,718 6,711 1.13 1.00 3,225 59.9 31.9 7.7 0.6 Rehabilitation medical 9,295 9,091 9,623 10,287 1.11 1.07 2,179 45.6 32.3 22.1 0.0 Skin medical 3,417 3,405 4,105 3,911 1.14 0.95 115,348 85.5 13.1 1.1 0.3 Skin surgical 6,168 6,333 7,089 7,210 1.17 1.02 10,213 66.9 28.4 2.9 1.8 Kidney and urine medical 5,463 5,439 6,290 6,197 1.13 0.99 220,018 66.9 30.5 2.3 0.3 Liver medical 4,989 4,981 5,721 5,505 1.10 0.96 30,063 65.3 31.3 2.8 0.5 Digestive medical 5,096 5,104 5,906 5,749 1.13 0.97 126,434 68.1 29.2 2.3 0.4 38 APPENDIX A Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predicted 2019 under payment to actual cost cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) Endocrine medical 4,384 4,358 5,159 4,957 1.13 0.96 123,093 76.1 21.7 1.9 0.3 Mental illness medical 4,002 3,999 4,752 4,587 1.15 0.97 56,973 78.7 20.6 0.5 0.2 Alcohol and drug use 8,167 8,196 9,805 9,171 1.12 0.94 3,889 25.7 67.7 6.1 0.6 medical HIV medical 7,982 7,802 10,837 8,951 1.12 0.83 1,220 56.4 37.5 3.2 2.9 Other medical 3,364 3,368 3,930 3,828 1.14 0.97 181,258 85.6 10.7 3.5 0.1 Other surgical 7,076 7,055 8,047 8,008 1.13 1.00 135,250 60.1 31.4 6.5 2.0 Other clinical conditions Cancer 4,083 4,083 4,708 4,528 1.11 0.96 13,852 75.1 21.0 3.5 0.4 Transplant 5,961 5,961 6,004 6,672 1.12 1.11 3,830 79.3 4.9 14.2 1.7 Kidney/urinary 5,618 5,625 6,497 6,405 1.14 0.99 249,759 66.3 30.6 2.5 0.5 GI or hepatobiliary 5,341 5,341 6,098 6,019 1.13 0.99 223,312 66.8 29.3 3.1 0.9 Vision impairment 6,139 6,139 7,081 7,007 1.14 0.99 83,938 68.2 24.2 6.7 1.0 Urinary incontinence 6,342 6,342 7,043 7,276 1.15 1.03 99,207 72.7 11.8 14.7 0.7 Trauma 6,118 5,762 6,975 6,668 1.09 0.96 160,667 71.2 19.1 8.8 0.9 Frailty, cognitive function, mental illness, and functional score Least frail 2,032 2,059 2,424 2,351 1.16 0.97 581,928 95.6 3.8 0.6 0.0 Most frail 9,179 9,167 10,380 10,445 1.14 1.01 1,245,183 47.3 41.9 8.6 2.2 Cognitively impaired 6,618 6,614 7,748 7,572 1.14 0.98 747,570 63.7 31.8 3.5 1.0 Serious mental illness 7,550 7,553 8,875 8,657 1.15 0.98 386,130 54.7 41.5 3.0 0.8 Function, 0–25th 9,742 8,920 11,283 10,306 1.06 0.91 853,908 51.5 39.2 6.5 2.9 percentile Function, 25–75th 5,001 5,126 5,634 5,829 1.17 1.03 1,889,877 72.5 21.4 5.7 0.4 percentile Function, 75–100th 2,658 3,148 3,009 3,469 1.31 1.15 948,279 87.5 10.7 1.5 0.4 percentile Severely ill (SOI level 4) 17,346 17,383 19,825 19,702 1.14 0.99 238,693 0.0 75.9 14.0 10.0 Multiple body systems 17,064 17,064 19,762 19,462 1.14 0.98 353,374 0.0 78.2 14.0 7.8 Chronically critically ill 12,806 12,603 14,443 14,523 1.13 1.01 149,404 43.9 37.2 9.0 10.0 Highest acuity 16,123 15,833 18,188 18,324 1.14 1.01 82,410 36.5 38.6 9.7 15.2 APPENDIX A 39 Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predicted 2019 under payment to actual cost cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) Other stay and patient characteristics HHA, no therapy visits 872 1,640 1,600 1,635 1.88 1.02 637,749 100.0 0.0 0.0 0.0 HHA, 1–4 therapy visits 987 1,681 1,695 1,729 1.75 1.02 583,154 100.0 0.0 0.0 0.0 HHA, 5–9 therapy visits 1,884 1,687 2,167 2,030 1.08 0.94 862,746 100.0 0.0 0.0 0.0 HHA,10+ therapy visits 3,045 1,739 2,524 2,363 0.78 0.94 555,376 100.0 0.0 0.0 0.0 I-PAC, therapy costs per 15,227 15,860 19,465 18,066 1.19 0.93 263,871 0.0 90.7 0.4 8.8 day in 0–25th percentile I-PAC, therapy costs per 13,911 14,416 16,140 16,765 1.21 1.04 266,970 0.0 97.9 0.3 1.8 day in 25–50th percentile I-PAC, therapy costs per 14,140 14,567 14,369 16,711 1.18 1.16 266,716 0.0 96.2 2.2 1.7 day in 50–75th percentile I-PAC, therapy costs per 16,982 15,356 17,918 16,796 0.99 0.94 255,482 0.0 32.7 66.2 1.1 day in > 75th percentile HHA community 1,535 1,667 1,819 1,893 1.23 1.04 1,761,523 100.0 0.0 0.0 0.0 admitted HHA stays with prior 1,988 1,722 2,367 2,028 1.02 0.86 877,502 100.0 0.0 0.0 0.0 hospital stay I-PAC community 15,467 13,951 18,029 16,174 1.05 0.90 85,802 0.0 67.2 30.2 2.5 admitted I-PAC stays with prior 15,006 15,141 16,861 17,166 1.14 1.02 967,237 0.0 81.0 15.6 3.4 hospital stay Disabled 5,702 5,702 6,709 6,500 1.14 0.97 883,582 71.1 22.8 4.6 1.6 Dual eligible or LIS 6,016 5,928 7,254 6,807 1.13 0.94 1,255,694 69.0 26.3 3.3 1.4 ESRD 5,227 5,272 5,757 5,975 1.14 1.04 171,844 65.3 26.7 5.6 2.4 Very old (85+) 6,670 6,587 7,983 7,468 1.12 0.94 1,138,287 72.6 23.9 3.2 0.4 SNF shortest 10th 2,311 14,099 2,488 4,047 1.75 1.63 85,909 0.0 100.0 0.0 0.0 percentile IRF shortest 10th 6,639 15,109 10,940 4,657 0.70 0.43 17,935 0.0 0.0 100.0 0.0 percentile LTCH shortest 10th 7,898 25,687 8,223 4,784 0.61 0.58 3,597 0.0 0.0 0.0 100.0 percentile 40 APPENDIX A Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predicted 2019 under payment to actual cost cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) IRF short stay outlier (<=3 3,499 15,397 3,584 2,434 0.70 0.68 5,282 0.0 0.0 100.0 0.0 days) HHA LUPA 353 1,677 332 483 1.37 1.46 230,005 100.0 0.0 0.0 100.0 Provider setting HHA 1,685 1,685 2,001 1,938 1.15 0.97 2,639,025 100.0 0.0 0.0 0.0 SNF 13,179 14,299 14,957 16,024 1.22 1.07 840,922 0.0 100.0 0.0 0.0 IRF 18,393 15,650 21,344 17,616 0.96 0.83 176,755 0.0 0.0 100.0 0.0 LTCH 42,647 29,737 42,564 39,662 0.93 0.93 35,362 0.0 0.0 0.0 100.0 LTCH chronically 47,699 22,124 0.0 0.0 0.0 100.0 critically ill by law 36,347 50,303 47,303 0.99 0.94 Provider characteristics Hospital-based 7,507 5,954 7,089 6,918 0.92 0.98 372,747 69.1 9.8 21.2 0.0 Freestanding 5,270 5,444 6,174 6,184 1.17 1.00 3,319,317 71.7 24.2 2.9 1.1 Nonprofit 6,036 5,785 6,146 6,612 1.10 1.08 904,608 69.0 23.7 6.8 0.5 For-profit 5,160 5,276 6,177 5,994 1.16 0.97 2,665,098 73.3 21.7 3.9 1.1 Government 8,811 8,141 9,108 9,402 1.07 1.03 122,358 50.6 39.9 9.2 0.4 Low-volume provider, 15,659 12,627 16,187 15,415 0.98 0.95 16,581 22.0 58.9 14.3 4.8 bottom decile IRF low-income share 16,966 15,312 20,656 17,058 1.01 0.83 41,505 0.0 0.0 100.0 0.0 0–20th percentile IRF low-income share 20– 17,236 15,421 20,851 17,152 1.00 0.82 40,196 0.0 0.0 100.0 0.0 40th percentile IRF low-income share 40– 18,480 15,680 21,301 17,580 0.95 0.83 37,279 0.0 0.0 100.0 0.0 60th percentile IRF low-income share 60– 18,972 15,901 21,561 17,932 0.95 0.83 31,816 0.0 0.0 100.0 0.0 80th percentile IRF low-income share 21,698 16,244 23,140 18,994 0.88 0.82 24,139 0.0 0.0 100.0 0.0 80th+ percentile Teaching (IRF only) 21,743 16,534 23,612 19,329 0.89 0.82 18,558 0.0 0.0 100.0 0.0 Dual/LIS share 0–20th 5,645 5,748 5,804 6,534 1.16 1.13 952,563 68.7 27.0 3.6 0.7 percentile in setting Dual/LIS share 20–40th 4,545 4,687 5,159 5,305 1.17 1.03 1,134,141 77.2 18.9 3.3 0.6 percentile in setting APPENDIX A 41 Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predicted 2019 under payment to actual cost cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) Dual/LIS share 40–60th 4,942 5,010 5,817 5,697 1.15 0.98 811,907 75.3 19.3 4.6 0.8 percentile in setting Dual/LIS share 60–80th 6,507 6,367 7,782 7,245 1.11 0.93 498,054 65.5 25.1 8.0 1.3 percentile in setting Dual/LIS share 80 –100th 8,478 7,653 10,689 8,904 1.05 0.83 295,399 57.8 29.7 9.5 3.1 percentile in setting White Non-Hispanic 5,324 4,942 5,651 5,748 1.08 1.02 239,036 75.9 17.6 5.2 1.3 share, top decile in setting Black Non-Hispanic share, 9,634 9,156 11,305 10,586 1.10 0.94 155,293 46.2 42.4 9.2 2.2 top decile in setting Other race/ethnicity 8,954 8,609 11,987 9,824 1.10 0.82 190,893 49.7 40.1 7.8 2.4 share, top decile in setting Geographic location Frontier 5,700 4,998 6,028 5,992 1.05 0.99 9,317 72.9 27.1 0.0 0.0 Metro 5,441 5,494 6,288 6,241 1.15 0.99 3,191,331 71.7 22.1 5.2 1.1 Rural micropolitan 5,915 5,564 6,183 6,428 1.09 1.04 326,658 69.8 26.5 3.3 0.5 Rural adjacent 5,905 5,542 6,186 6,450 1.09 1.04 106,236 69.0 30.6 0.4 0.0 Rural nonadjacent 5,411 5,153 5,761 5,967 1.10 1.04 67,812 72.0 26.8 1.1 0.0 Urban CBSA based 5,441 5,493 6,287 6,240 1.15 0.99 3,195,018 71.7 22.0 5.2 1.1 Rural CBSA based 5,847 5,508 6,131 6,374 1.09 1.04 496,901 69.9 27.5 2.3 0.3 Regions 1: CT, MA, M, NH, RI, VT 4,417 5,164 5,535 5,738 1.30 1.04 243,902 72.5 23.6 3.2 0.6 2: NY, NJ 6,538 6,290 8,317 7,206 1.10 0.87 276,819 64.0 32.0 3.7 0.4 3: DE, DC, MD, PA, VA, 5,561 5,685 6,025 6,444 1.16 1.07 403,044 70.1 24.0 5.3 0.6 WV 4: AL, FL, GA, KY, MS, NC, 5,111 5,047 5,504 5,773 1.13 1.05 908,759 75.1 19.7 4.4 0.9 SC, TN 5: IL, IN, MI, MN, OH, WI 5,700 5,868 6,261 6,636 1.16 1.06 607,492 68.7 26.5 4.0 0.8 6: AR, LA, NM, OK, TX 6,021 5,562 6,681 6,420 1.07 0.96 472,780 71.8 17.3 8.5 2.4 7: IA, KS, MO, NE 6,494 6,551 6,861 7,473 1.15 1.09 154,319 63.4 29.5 6.0 1.2 8: CO, MT, ND, SD, UT, 6,312 5,825 6,321 6,764 1.07 1.07 79,673 69.0 25.6 4.7 0.6 WY 9: AZ, CA, HI, NV 4,859 4,953 6,542 5,599 1.15 0.86 452,636 75.9 19.3 3.9 0.8 42 APPENDIX A Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predicted 2019 under payment to actual cost cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) 10: AK, ID, OR, WA 5,420 5,393 6,132 6,204 1.14 1.01 92,640 72.1 25.2 2.3 0.4 Sources: 2019 Medicare acute hospital and post-acute care claims, Medicare 2019 risk score file, and Medicare cost reports for 2019. Notes: PAC = post-acute care. PPS = prospective payment system. HHA = home health agency. SNF = skilled nursing facility. IRF = inpatient rehabilitation facility. LTCH = long-term care hospital. SOI = severity of illness. I-PAC = institutional post-acute care. ESRD = end-stage renal disease. LUPA = low-utilization payment adjustment. CBSA = core-based statistical area. LIS = Low-income subsidy program for Part D enrollees. Data are all stays that began between April and September 2019 and had the CARE function variables on a matched assessment. Patients' level of frailty was determined using the JEN Frailty Index. Chronically critically ill stays include patients who spent eight or more days in an intensive care or coronary care unit during the preceding hospital stay or were on a ventilator in the PAC setting. LTCH chronically critically ill by law stays include LTCH patients who spent three or more days in an intensive care or coronary care unit during the preceding hospital stay or were on a ventilator in the LTCH. Severely ill stays include institutional-setting patients who were categorized as severity of illness level 4, usually during the immediately preceding hospital stay. Multiple body systems include institutional patients with secondary diagnoses involving five or more body systems. Highest-acuity patients were institutional patients categorized as severity of illness level 4, on dialysis, and who had severe wounds or a pressure ulcer. Race/ethnicity shares are based on the RTI race measure, with top decile based on shares in facilities within a setting with at least 25 stays. TABLE A.7 Estimated Distribution of the Changes in Payments under MedPAC's Model of a PAC PPS for PAC Stays Ratio of Model to Actual Payments Decrease in payment Increase in payment 10% to 5% to About 5% to 10% to Reporting category N > 25% 25% 10% 1% to 5% the same 1% to 5% 10% 25% > 25% All stays 3,692,064 0.19 0.20 0.05 0.03 0.02 0.03 0.04 0.11 0.33 Clinical group Ventilator 8,451 0.21 0.19 0.07 0.05 0.02 0.04 0.04 0.09 0.29 Severe wounds 157,912 0.14 0.17 0.06 0.04 0.02 0.04 0.06 0.14 0.33 Stroke 88,788 0.12 0.15 0.06 0.05 0.02 0.04 0.05 0.10 0.40 Other neurology medical 344,129 0.14 0.21 0.06 0.04 0.02 0.04 0.06 0.15 0.28 Other neurology surgical 32,188 0.11 0.15 0.06 0.05 0.02 0.05 0.05 0.11 0.38 Orthopedic medical 523,541 0.10 0.23 0.07 0.04 0.02 0.03 0.04 0.12 0.35 Orthopedic surgical 363,782 0.21 0.23 0.05 0.03 0.01 0.02 0.03 0.07 0.36 APPENDIX A 43 Ratio of Model to Actual Payments Decrease in payment Increase in payment 10% to 5% to About 5% to 10% to Reporting category N > 25% 25% 10% 1% to 5% the same 1% to 5% 10% 25% > 25% Respiratory medical 290,637 0.21 0.18 0.04 0.03 0.01 0.03 0.04 0.10 0.35 Respiratory surgical 11,808 0.28 0.19 0.03 0.02 0.01 0.02 0.03 0.08 0.34 Cardiovascular medical 477,884 0.20 0.19 0.04 0.03 0.01 0.03 0.04 0.12 0.33 Cardiovascular surgical 123,119 0.23 0.22 0.04 0.03 0.01 0.02 0.03 0.08 0.34 Infection medical 184,382 0.26 0.16 0.04 0.03 0.01 0.02 0.03 0.09 0.35 Infection surgical 39,087 0.17 0.17 0.05 0.04 0.02 0.03 0.04 0.08 0.39 Hematology medical 37,193 0.28 0.16 0.03 0.03 0.01 0.03 0.05 0.11 0.30 Hematology surgical 3,225 0.30 0.14 0.03 0.02 0.01 0.02 0.03 0.08 0.35 Rehabilitation medical 2,179 0.14 0.20 0.06 0.03 0.02 0.03 0.04 0.08 0.39 Skin medical 115,348 0.25 0.19 0.06 0.05 0.03 0.05 0.05 0.10 0.22 Skin surgical 10,213 0.25 0.17 0.05 0.03 0.02 0.03 0.04 0.09 0.33 Kidney and urine medical 220,018 0.21 0.18 0.04 0.03 0.01 0.03 0.04 0.10 0.35 Liver medical 30,063 0.33 0.13 0.03 0.02 0.01 0.03 0.05 0.10 0.30 Digestive medical 126,434 0.26 0.17 0.04 0.03 0.01 0.03 0.04 0.10 0.33 Endocrine medical 123,093 0.25 0.18 0.05 0.04 0.02 0.04 0.05 0.11 0.27 Mental illness medical 56,973 0.11 0.18 0.06 0.04 0.02 0.03 0.04 0.10 0.43 Alcohol and drug use 3,889 0.27 0.10 0.03 0.02 0.01 0.02 0.04 0.09 0.42 medical HIV medical 1,220 0.22 0.20 0.04 0.04 0.01 0.02 0.05 0.09 0.33 Other medical 181,258 0.17 0.22 0.06 0.04 0.02 0.04 0.05 0.11 0.29 Other surgical 135,250 0.22 0.18 0.04 0.03 0.01 0.03 0.04 0.08 0.37 Other clinical conditions Cancer 13,852 0.33 0.15 0.03 0.03 0.01 0.03 0.05 0.11 0.25 Transplant 3,830 0.05 0.17 0.08 0.06 0.03 0.05 0.07 0.11 0.38 Kidney/urinary 249,759 0.21 0.18 0.04 0.03 0.01 0.03 0.04 0.10 0.36 GI or hepatobiliary 223,312 0.26 0.17 0.04 0.03 0.01 0.03 0.04 0.10 0.34 Vision impairment 83,938 0.18 0.19 0.05 0.04 0.02 0.03 0.04 0.11 0.35 44 APPENDIX A Ratio of Model to Actual Payments Decrease in payment Increase in payment 10% to 5% to About 5% to 10% to Reporting category N > 25% 25% 10% 1% to 5% the same 1% to 5% 10% 25% > 25% Urinary incontinence 99,207 0.08 0.17 0.07 0.05 0.02 0.04 0.05 0.11 0.40 Trauma 160,667 0.22 0.20 0.06 0.05 0.02 0.04 0.05 0.10 0.26 Frailty, cognitive function, mental illness, and functional score Least frail 581,928 0.18 0.23 0.05 0.04 0.02 0.03 0.05 0.12 0.27 Most frail 1,245,183 0.20 0.17 0.05 0.03 0.01 0.03 0.04 0.09 0.39 Cognitively impaired 747,570 0.18 0.18 0.05 0.04 0.02 0.03 0.04 0.11 0.35 Serious mental illness 386,130 0.19 0.17 0.05 0.03 0.02 0.03 0.04 0.10 0.38 Function, 0–25th 853,908 0.14 0.18 0.06 0.04 0.02 0.03 0.04 0.10 0.39 percentile Function, 25–75th 1,889,877 0.17 0.21 0.05 0.04 0.02 0.03 0.04 0.11 0.33 percentile Function, 75–100th 948,279 0.26 0.19 0.04 0.03 0.01 0.03 0.04 0.10 0.29 percentile Severely ill (SOI level 4) 238,693 0.23 0.12 0.03 0.03 0.01 0.02 0.03 0.08 0.46 Multiple body systems 353,374 0.23 0.12 0.03 0.03 0.01 0.02 0.03 0.08 0.45 Chronically critically ill 149,404 0.20 0.17 0.05 0.03 0.01 0.03 0.04 0.09 0.38 Highest acuity 82,410 0.21 0.16 0.05 0.03 0.01 0.03 0.04 0.09 0.38 Other stay and patient characteristics HHA, no therapy visits 637,749 0.16 0.15 0.04 0.03 0.01 0.03 0.06 0.12 0.40 HHA, 1–4 therapy visits 583,154 0.16 0.15 0.04 0.03 0.01 0.03 0.05 0.15 0.40 HHA, 5–9 therapy visits 862,746 0.21 0.28 0.06 0.04 0.02 0.03 0.04 0.11 0.21 HHA,10+ therapy visits 555,376 0.17 0.31 0.09 0.06 0.03 0.05 0.05 0.11 0.14 I-PAC, therapy costs per 263,871 0.26 0.09 0.03 0.02 0.01 0.02 0.03 0.07 0.48 day in 0–25th percentile I-PAC, therapy costs per 266,970 0.20 0.09 0.03 0.02 0.01 0.02 0.03 0.08 0.51 day, 25–50th percentile APPENDIX A 45 Ratio of Model to Actual Payments Decrease in payment Increase in payment 10% to 5% to About 5% to 10% to Reporting category N > 25% 25% 10% 1% to 5% the same 1% to 5% 10% 25% > 25% I-PAC, therapy costs per 266,716 0.14 0.09 0.03 0.02 0.01 0.02 0.03 0.08 0.58 day, 50–75th percentile I-PAC, therapy costs per 255,482 0.25 0.21 0.06 0.04 0.02 0.03 0.03 0.07 0.29 day, > 75th percentile HHA community admitted 1,761,523 0.10 0.18 0.05 0.04 0.02 0.04 0.05 0.15 0.36 HHA stays with prior 877,502 0.33 0.32 0.06 0.04 0.01 0.03 0.04 0.06 0.12 hospital stay I-PAC community admitted 85,802 0.29 0.17 0.04 0.03 0.01 0.03 0.03 0.07 0.34 I-PAC stays with prior 967,237 0.20 0.12 0.04 0.03 0.01 0.02 0.03 0.08 0.48 hospital stay Disabled 883,582 0.21 0.19 0.05 0.03 0.02 0.03 0.04 0.11 0.33 Dual eligible or LIS 1,255,694 0.20 0.18 0.05 0.03 0.02 0.03 0.04 0.11 0.34 ESRD 171,844 0.27 0.17 0.04 0.03 0.01 0.03 0.04 0.10 0.30 Very old (85+) 1,138,287 0.16 0.20 0.05 0.04 0.02 0.03 0.04 0.11 0.35 SNF shortest 10th 85,909 0.00 0.01 0.00 0.00 0.00 0.01 0.01 0.05 0.91 percentile IRF shortest 10th 17,935 0.91 0.03 0.00 0.00 0.00 0.00 0.00 0.01 0.04 percentile LTCH shortest 10th 3,597 0.77 0.14 0.03 0.02 0.01 0.01 0.01 0.02 0.01 percentile IRF short stay outlier (<=3 5,282 0.74 0.08 0.00 0.00 0.00 0.00 0.01 0.05 0.12 days) HHA LUPA 230,005 0.00 0.00 0.00 0.00 0.00 0.01 0.09 0.07 0.82 Sources: 2019 Medicare acute hospital and post-acute care claims and assessments, Medicare 2019 risk score file, and Medicare cost reports for 2019. Notes: PAC = post-acute care. PPS = prospective payment system. HHA = home health agency. SNF = skilled nursing facility. IRF = inpatient rehabilitation facility. LTCH = long-term care hospital. SOI = severity of illness. I-PAC = institutional post-acute care. ESRD = end-stage renal disease. LUPA = low-utilization payment adjustment. LIS = Low-income subsidy program for Part D enrollees. Data are all stays that began between April and September 2019 and had the CARE function variables on a matched assessment. Patients' level of frailty was determined using the JEN Frailty Index. Chronically critically ill stays include patients who spent eight or more days in an intensive care or coronary care unit during the preceding hospital stay or were on a ventilator in the PAC setting. LTCH chronically critically ill by law stays include LTCH patients who spent three or more days in an intensive care or coronary care unit during the preceding hospital stay or were on a ventilator in the LTCH. Severely ill stays include institutional-setting patients who were categorized as severity of illness level 4, usually during the immediately preceding hospital stay. Multiple body systems include institutional patients with secondary diagnoses involving five or more body systems. Highest-acuity patients were institutional patients categorized as severity of illness level 4, on dialysis, and who had severe wounds or a pressure ulcer. 46 APPENDIX A TABLE A.8 Estimated Distribution of the Changes in Payments under MedPAC's Model of a PAC PPS for Providers with 20 or More Stays Ratio of Model to Actual Payments Decrease in payment Increase in payment 10% to 5% to About 5% to 10% to Reporting category N > 25% 25% 10% 1% to 5% the same 1% to 5% 10% 25% > 25% Provider setting All providers 19,979 0.05 0.16 0.12 0.12 0.05 0.09 0.08 0.15 0.18 HHA 7,881 0.00 0.12 0.20 0.21 0.09 0.14 0.11 0.10 0.02 SNF 10,766 0.09 0.13 0.06 0.06 0.03 0.06 0.07 0.20 0.31 IRF 1,010 0.10 0.68 0.13 0.06 0.01 0.01 0.01 0.00 0.00 LTCH 322 0.04 0.30 0.17 0.13 0.07 0.11 0.09 0.07 0.02 Provider characteristics Hospital-based 1,768 0.05 0.30 0.14 0.10 0.04 0.06 0.06 0.08 0.18 Freestanding 18,211 0.05 0.14 0.12 0.12 0.05 0.09 0.08 0.16 0.18 Nonprofit 4,036 0.02 0.17 0.12 0.08 0.03 0.05 0.06 0.15 0.31 For-profit 14,733 0.06 0.15 0.13 0.13 0.06 0.10 0.09 0.15 0.14 Government 1,210 0.05 0.19 0.10 0.09 0.03 0.08 0.10 0.16 0.20 Low-volume provider, 89 0.10 0.45 0.17 0.07 0.00 0.12 0.02 0.04 0.02 bottom decile IRF low-income share 203 0.09 0.67 0.13 0.07 0.00 0.02 0.01 0.00 0.00 0–20th percentile IRF low-income share 20– 197 0.12 0.70 0.10 0.04 0.01 0.02 0.01 0.01 0.00 40th percentile IRF low-income share 40– 210 0.06 0.70 0.14 0.06 0.01 0.00 0.00 0.00 0.00 60th percentile IRF low-income share 60– 209 0.09 0.67 0.17 0.06 0.00 0.00 0.00 0.00 0.00 80th percentile IRF low-income share 178 0.16 0.65 0.08 0.04 0.02 0.01 0.01 0.01 0.01 80th+ percentile Teaching (IRF only) 90 0.09 0.74 0.10 0.03 0.02 0.01 0.00 0.00 0.00 Dual/LIS share 0–20th 4,099 0.01 0.12 0.13 0.11 0.05 0.07 0.07 0.13 0.32 percentile in setting APPENDIX A 47 Ratio of Model to Actual Payments Decrease in payment Increase in payment 10% to 5% to About 5% to 10% to Reporting category N > 25% 25% 10% 1% to 5% the same 1% to 5% 10% 25% > 25% Dual/LIS share 20–40th 4,454 0.02 0.14 0.14 0.14 0.05 0.08 0.07 0.15 0.21 percentile in setting Dual/LIS share 40–60th 4,190 0.04 0.16 0.14 0.14 0.06 0.09 0.08 0.15 0.15 percentile in setting Dual/LIS share 60–80th 4,016 0.06 0.18 0.11 0.12 0.05 0.10 0.09 0.16 0.11 percentile in setting Dual/LIS share 80th+ 3,220 0.17 0.18 0.09 0.08 0.04 0.10 0.12 0.15 0.07 percentile in setting White Non-Hispanic stays, 1,761 0.02 0.19 0.14 0.12 0.05 0.08 0.07 0.13 0.20 top decile by setting Black Non-Hispanic stays, 1,757 0.10 0.16 0.09 0.08 0.05 0.09 0.11 0.20 0.13 top decile by setting Other race/ethnicity stays, 1,761 0.02 0.19 0.14 0.12 0.05 0.08 0.07 0.13 0.20 top decile by setting Geographic location Frontier 120 0.03 0.22 0.14 0.13 0.05 0.10 0.08 0.14 0.11 Metro 15,997 0.06 0.16 0.12 0.12 0.05 0.09 0.08 0.14 0.17 Rural micropolitan 2,358 0.03 0.15 0.12 0.12 0.04 0.07 0.08 0.17 0.20 Rural adjacent 992 0.02 0.14 0.12 0.14 0.04 0.11 0.09 0.16 0.17 Rural nonadjacent 631 0.03 0.18 0.12 0.11 0.05 0.10 0.07 0.16 0.18 Urban CBSA based 16,021 0.06 0.16 0.12 0.12 0.05 0.09 0.08 0.14 0.17 Rural CBSA based 3,956 0.03 0.15 0.12 0.12 0.04 0.09 0.08 0.17 0.19 Regions 1: CT, MA, M, NH, RI, VT 990 0.04 0.21 0.15 0.09 0.03 0.06 0.06 0.18 0.19 2: NY, NJ 1,024 0.24 0.25 0.10 0.06 0.02 0.04 0.04 0.09 0.14 3: DE, DC, MD, PA, VA, WV 1,847 0.01 0.14 0.14 0.11 0.03 0.07 0.06 0.13 0.30 4: AL, FL, GA, KY, MS, NC, 3,844 0.01 0.13 0.11 0.13 0.06 0.08 0.08 0.18 0.21 SC, TN 5: IL, IN, MI, MN, OH, WI 3,978 0.03 0.15 0.11 0.10 0.04 0.09 0.10 0.18 0.20 6: AR, LA, NM, OK, TX 3,396 0.03 0.15 0.13 0.15 0.07 0.12 0.09 0.13 0.12 48 APPENDIX A Ratio of Model to Actual Payments Decrease in payment Increase in payment 10% to 5% to About 5% to 10% to Reporting category N > 25% 25% 10% 1% to 5% the same 1% to 5% 10% 25% > 25% 7: IA, KS, MO, NE 1,191 0.01 0.17 0.11 0.11 0.04 0.07 0.06 0.14 0.28 8: CO, MT, ND, SD, UT, WY 582 0.02 0.19 0.14 0.12 0.04 0.09 0.07 0.14 0.19 9: AZ, CA, HI, NV 2,619 0.17 0.15 0.13 0.12 0.06 0.11 0.10 0.11 0.05 10: AK, ID, OR, WA 508 0.03 0.17 0.14 0.17 0.06 0.09 0.08 0.14 0.12 Sources: 2019 Medicare acute hospital and post-acute care claims and assessments, Medicare 2019 risk score file, and Medicare cost reports for 2019. Notes: PAC = post-acute care. PPS = prospective payment system. HHA = home health agency. SNF = skilled nursing facility. IRF = inpatient rehabilitation facility. LTCH = long-term care hospital. CBSA = core-based statistical area. Underlying data are the 3.7 million stays that began between April and September 2019 and had the CARE function variables on a matched assessment. Table restricted to facilities with at least 20 stays. Deciles of facility race/ethnicity of patients based on shares with race/ethnicity groups defined Race/ethnicity shares are based RTI race measure, with top decile based on shares in facilities within a setting with at least 25 stays. APPENDIX A 49 TABLE A.9 Estimated Change in Payments under PAC PPS by Current Relative Medicare Profitability Decrease in Payments Increase in Payments About Provider 10% to 1% to the 1% to 10% to Relative profitability count < 25% 25% 10% same 10% 25% > 25% Below average <.75 2,846 0 47 223 82 611 702 1,181 .75 - .9 4,519 8 376 1,025 309 949 811 1,041 About average .9 - 1.1 6,906 137 1,191 2,083 396 1,135 985 979 Above average 1.1 - 1.25 2,876 208 762 857 130 400 283 236 >1.25 2,832 694 748 652 111 330 196 101 Provider counts 1047 3124 4840 1028 3425 2977 3538 Sources: 2019 Medicare acute hospital and post-acute care claims and assessments, Medicare 2019 risk score file, and Medicare cost reports for 2019. Notes: PAC = post-acute care. PPS = Prospective Payment System. Relative profitability is the ratio of the provider's profitability (the ratio of the provider's average payment under current policy to the average stay cost) to the setting's average profitability. Ratios below 1.0 indicate below-average profitability; ratios above 1.0 indicate above-average profitability. Only providers with at least 20 stays were included in the analysis (N=19,979). Data are 3.7 million stays that began between April and September 2019 and had the CARE function variables on a matched assessment. 50 APPENDIX A TABLE A.10 Comparison of Actual Costs, Predicted Costs, Actual Payments, and PAC PPS Payments (Including Outliers) under PAC PPS for April– September 2019 Stays, with 5 Percent Outlier Pool and Short-Stay Outlier Payments: Five Percent Cut in Payments Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predicted 2019 under payment to actual cost cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) All 5,496 5,496 6,266 5,946 1.08 0.95 3,692,064 71.5 22.8 4.8 1.0 Provider setting HHA 1,685 1,685 2,001 1,842 1.09 0.92 2,639,025 100.0 0.0 0.0 0.0 SNF 13,179 14,301 14,957 15,200 1.15 1.02 840,922 0.0 100.0 0.0 0.0 IRF 18,393 15,621 21,344 16,732 0.91 0.78 176,755 0.0 0.0 100.0 0.0 LTCH 42,647 29,838 42,564 38,198 0.90 0.90 35,362 0.0 0.0 0.0 100.0 Provider characteristics Hospital-based 7,507 5,890 7,089 6,528 0.87 0.92 372,747 69.1 9.8 21.2 0.0 Freestanding 5,270 5,451 6,174 5,880 1.12 0.95 3,319,317 71.7 24.2 2.9 1.1 Nonprofit 6,036 5,734 6,146 6,236 1.03 1.01 904,608 69.0 23.7 6.8 0.5 For-profit 5,160 5,294 6,177 5,711 1.11 0.92 2,665,098 73.3 21.7 3.9 1.1 Government 8,811 8,118 9,108 8,913 1.01 0.98 122,358 50.6 39.9 9.2 0.4 Low-volume provider, 15,659 12,618 16,187 14,676 0.94 0.91 16,581 22.0 58.9 14.3 4.8 bottom decile Dual/LIS share 0–20th 5,645 5,702 5,804 6,165 1.09 1.06 952,563 68.7 27.0 3.6 0.7 percentile in setting Dual/LIS share 20–40th 4,545 4,681 5,159 5,035 1.11 0.98 1,134,141 77.2 18.9 3.3 0.6 percentile in setting Dual/LIS share 40–60th 4,942 5,019 5,817 5,421 1.10 0.93 811,907 75.3 19.3 4.6 0.8 percentile in setting Dual/LIS share 60–80th 6,507 6,404 7,782 6,917 1.06 0.89 498,054 65.5 25.1 8.0 1.3 percentile in setting Dual/LIS share 80–100th 8,478 7,735 10,689 8,540 1.01 0.80 295,399 57.8 29.7 9.5 3.1 percentile in setting White Non-Hispanic 5,324 4,906 5,651 5,432 1.02 0.96 239,036 75.9 17.6 5.2 1.3 share, top decile in setting Black Non-Hispanic share, 9,634 9,208 11,305 10,108 1.05 0.89 155,293 46.2 42.4 9.2 2.2 top decile in setting APPENDIX A 51 Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predicted 2019 under payment to actual cost cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) Other race/ethnicity 8,954 8,745 11,987 9,461 1.06 0.79 190,893 49.7 40.1 7.8 2.4 share, top decile in setting Geographic location Frontier 5,700 4,958 6,028 5,656 0.99 0.94 9,317 72.9 27.1 0.0 0.0 Metro 5,441 5,495 6,288 5,930 1.09 0.94 3,191,331 71.7 22.1 5.2 1.1 Rural micropolitan 5,915 5,554 6,183 6,093 1.03 0.99 326,658 69.8 26.5 3.3 0.5 Rural adjacent 5,905 5,546 6,186 6,124 1.04 0.99 106,236 69.0 30.6 0.4 0.0 Rural nonadjacent 5,411 5,158 5,761 5,670 1.05 0.98 67,812 72.0 26.8 1.1 0.0 Urban CBSA based 5,441 5,494 6,287 5,930 1.09 0.94 3,195,018 71.7 22.0 5.2 1.1 Rural CBSA based 5,847 5,503 6,131 6,046 1.03 0.99 496,901 69.9 27.5 2.3 0.3 Regions 1: CT, MA, M, NH, RI, VT 4,417 5,052 5,535 5,345 1.21 0.97 243,902 72.5 23.6 3.2 0.6 2: NY, NJ 6,538 6,396 8,317 6,940 1.06 0.83 276,819 64.0 32.0 3.7 0.4 3: DE, DC, MD, PA, VA, 5,561 5,649 6,025 6,086 1.09 1.01 403,044 70.1 24.0 5.3 0.6 WV 4: AL, FL, GA, KY, MS, NC, 5,111 5,075 5,504 5,510 1.08 1.00 908,759 75.1 19.7 4.4 0.9 SC, TN 5: IL, IN, MI, MN, OH, WI 5,700 5,830 6,261 6,267 1.10 1.00 607,492 68.7 26.5 4.0 0.8 6: AR, LA, NM, OK, TX 6,021 5,589 6,681 6,131 1.02 0.92 472,780 71.8 17.3 8.5 2.4 7: IA, KS, MO, NE 6,494 6,481 6,861 7,037 1.08 1.03 154,319 63.4 29.5 6.0 1.2 8: CO, MT, ND, SD, UT, 6,312 5,731 6,321 6,340 1.00 1.00 79,673 69.0 25.6 4.7 0.6 WY 9: AZ, CA, HI, NV 4,859 4,993 6,542 5,358 1.10 0.82 452,636 75.9 19.3 3.9 0.8 10: AK, ID, OR, WA 5,420 5,373 6,132 5,878 1.08 0.96 92,640 72.1 25.2 2.3 0.4 Sources: 2019 Medicare acute hospital and post-acute care claims and assessments, Medicare 2019 risk score file, and Medicare cost reports for 2019. Notes: PAC = post-acute care. PPS = prospective payment system. HHA = home health agency. SNF = skilled nursing facility. IRF = inpatient rehabilitation facility. LTCH = long-term care hospital. SOI = severity of illness. I-PAC = institutional post-acute care. ESRD = end-stage renal disease. LUPA = low-utilization payment adjustment. CBSA = core-based statistical area. LIS = Low-income subsidy program for Part D enrollees. Data are all stays that began between April and September 2019 and had the CARE function variables on a matched assessment. Race/ethnicity shares are based RTI race measure, with top decile based on shares in facilities within a setting with at least 25 stays. 52 APPENDIX A TABLE A.11 Comparison of Actual Costs, Predicted Costs, Actual Payments, and PAC PPS Payments (Including Outliers) under PAC PPS for April– September 2019 Stays, with 5 Percent Outlier Pool and Short-Stay Outlier Payments: First Year of a 3-Year Transition with a Five Percent Cut in Payments Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predict 2019 under payment to actual cost ed cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) All 5,496 5,496 6,266 6,159 1.12 0.98 3,692,064 71.5 22.8 4.8 1.0 Provider setting HHA 1,685 1,685 2,001 1,948 1.16 0.97 1,685 100.0 0.0 0.0 0.0 SNF 13,179 14,301 14,957 15,038 1.14 1.01 13,179 0.0 100.0 0.0 0.0 IRF 18,393 15,621 21,344 19,807 1.08 0.93 18,393 0.0 0.0 100.0 0.0 LTCH 42,647 29,838 42,564 41,108 0.96 0.97 42,647 0.0 0.0 0.0 100.0 Provider characteristics Hospital-based 7,507 5,890 7,089 6,902 0.92 0.97 372,747 69.1 9.8 21.2 0.0 Freestanding 5,270 5,451 6,174 6,076 1.15 0.98 3,319,317 71.7 24.2 2.9 1.1 Nonprofit 6,036 5,734 6,146 6,176 1.02 1.00 904,608 69.0 23.7 6.8 0.5 For-profit 5,160 5,294 6,177 6,021 1.17 0.97 2,665,098 73.3 21.7 3.9 1.1 Government 8,811 8,118 9,108 9,043 1.03 0.99 122,358 50.6 39.9 9.2 0.4 Low-volume provider, 15,659 12,618 16,187 15,683 1.00 0.97 16,581 22.0 58.9 14.3 4.8 bottom decile Dual/LIS share 0–20th 5,645 5,702 5,804 5,924 1.05 1.02 952,563 68.7 27.0 3.6 0.7 percentile in setting Dual/LIS share 20–40th 4,545 4,681 5,159 5,117 1.13 0.99 1,134,141 77.2 18.9 3.3 0.6 percentile in setting Dual/LIS share 40–60th 4,942 5,019 5,817 5,685 1.15 0.98 811,907 75.3 19.3 4.6 0.8 percentile in setting Dual/LIS share 60–80th 6,507 6,404 7,782 7,493 1.15 0.96 498,054 65.5 25.1 8.0 1.3 percentile in setting Dual/LIS share 80 – 100th percentile in 8,478 7,735 10,689 9,973 1.18 0.93 295,399 57.8 29.7 9.5 3.1 setting White Non-Hispanic share, top decile in 5,324 4,906 5,651 5,578 1.05 0.99 239,036 75.9 17.6 5.2 1.3 setting APPENDIX A 53 Ratio of Ratio of PAC PPS Distribution of Stays by Setting Actual Payment PAC PPS payment Actual Predict 2019 under payment to actual cost ed cost payment PAC PPS to actual 2019 HHA SNF IRF LTCH Reporting category ($) ($) ($) ($) cost payment Stay count (%) (%) (%) (%) Black Non-Hispanic share, top decile in 9,634 9,208 11,305 10,906 1.13 0.96 155,293 46.2 42.4 9.2 2.2 setting Other race/ethnicity share, top decile in 8,954 8,745 11,987 11,145 1.24 0.93 190,893 49.7 40.1 7.8 2.4 setting Geographic location Frontier 5,700 4,958 6,028 5,904 1.04 0.98 9,317 72.9 27.1 0.0 0.0 Metro 5,441 5,495 6,288 6,169 1.13 0.98 3,191,331 71.7 22.1 5.2 1.1 Rural micropolitan 5,915 5,554 6,183 6,153 1.04 1.00 326,658 69.8 26.5 3.3 0.5 Rural adjacent 5,905 5,546 6,186 6,165 1.04 1.00 106,236 69.0 30.6 0.4 0.0 Rural nonadjacent 5,411 5,158 5,761 5,731 1.06 0.99 67,812 72.0 26.8 1.1 0.0 Urban CBSA based 5,441 5,494 6,287 6,168 1.13 0.98 3,195,018 71.7 22.0 5.2 1.1 Rural CBSA based 5,847 5,503 6,131 6,103 1.04 1.00 496,901 69.9 27.5 2.3 0.3 Regions 1: CT, MA, M, NH, RI, VT 4,417 5,052 5,535 5,472 1.24 0.99 243,902 72.5 23.6 3.2 0.6 2: NY, NJ 6,538 6,396 8,317 7,858 1.20 0.94 276,819 64.0 32.0 3.7 0.4 3: DE, DC, MD, PA, VA, 5,561 5,649 6,025 6,046 1.09 1.00 403,044 70.1 24.0 5.3 0.6 WV 4: AL, FL, GA, KY, MS, 5,111 5,075 5,504 5,506 1.08 1.00 908,759 75.1 19.7 4.4 0.9 NC, SC, TN 5: IL, IN, MI, MN, OH, WI 5,700 5,830 6,261 6,263 1.10 1.00 607,492 68.7 26.5 4.0 0.8 6: AR, LA, NM, OK, TX 6,021 5,589 6,681 6,498 1.08 0.97 472,780 71.8 17.3 8.5 2.4 7: IA, KS, MO, NE 6,494 6,481 6,861 6,919 1.07 1.01 154,319 63.4 29.5 6.0 1.2 8: CO, MT, ND, SD, UT, 6,312 5,731 6,321 6,327 1.00 1.00 79,673 69.0 25.6 4.7 0.6 WY 9: AZ, CA, HI, NV 4,859 4,993 6,542 6,147 1.26 0.94 452,636 75.9 19.3 3.9 0.8 10: AK, ID, OR, WA 5,420 5,373 6,132 6,047 1.12 0.99 92,640 72.1 25.2 2.3 0.4 Sources: 2019 Medicare acute hospital and post-acute care claims and assessments, Medicare 2019 risk score file, and Medicare cost reports for 2019. Notes: PAC = post-acute care. PPS = prospective payment system. HHA = home health agency. SNF = skilled nursing facility. IRF = inpatient rehabilitation facility. LTCH = long-term care hospital. I-PAC = institutional post-acute care. CBSA = core-based statistical area. LIS = Low-income subsidy program for Part D enrollees. Data are all stays that began between April and September 2019 and had the CARE function variables on a matched assessment. Race/ethnicity shares are based on the RTI race measure, with top decile based on shares in facilities within a setting with at least 25 stays. 54 APPENDIX A Notes 1 In Wissoker and Garrett (2018b), we found that costs and thus profitability varies for home health care by position in a sequence of stays. It led commissioners to ask whether undesirable incentives related to transfers and multiple stays for the same patient that could arise in a stay-based payment system might be avoided with an episode-based payment system. Wissoker and Garrett (2019) modeled the trade-offs between a stay- and episode-based payment model and concluded that while the episode-based PPS could, on average, pay with accuracy comparable to a stay-based system, it would also provide strong incentives to shorten episodes of care, with possible implications for efficiency and patient care. 2 A small share of the IRF cases do not line up with the 2019 payment year. We include 3.4 percent of the IRF stays that were discharged under FY year 2020 rules, while excluding a similar share that began in FY 2018 but were discharged under FY 2019 rules. 3 Hospital-based facilities (i.e., those based in acute-care hospitals) account for 8 percent of home-health stays, 4 percent of SNF stays, and 44 percent of IRF stays. No LTCH stays are considered hospital based. 4 Stays or episodes are considered usable if they have fewer than three missing responses for the six relevant functional status items used in this analysis. Responses are considered missing if the patient refuses or if measurement was not attempted due to environmental imitations (e.g., lack of equipment, weather constraints). 5 Because the overhead share of the total cost of a stay was similar across settings (though the levels differed), we did not model fixed and variable costs separately. 6 We imputed medical social service cost per minute for 317 episodes as the median cost per minute for hospital- based and free-standing agencies. 7 Severe wound care includes care for patients with a nonhealing surgical wound; patients who are morbidly obese with a wound; patients with a fistula; patients with osteomyelitis; and/or patients with a stage III, stage IV, or an unstageable pressure wound. 8 The JEN Frailty Index is an algorithm developed by JEN Associates Inc., now part of Westat, to identify frail older adults who may be at risk of institutionalization. It is based on 13 grouped categories of diseases or signs found to be significantly related to need for long-term care services, either concurrently or in the future. The algorithm uses diagnoses codes from claims. 9 We follow the LUPA policy under the current PDGM prospective payment system. The LUPA threshold ranges from two to six visits within a 30-day payment period and depends on the patient's clinical grouping. 10 For home health stays, we include information from the prior hospital stay for both 30-day periods of the 60-day episode. 11 A parallel, though weaker, pattern can be seen for institutional stays, with relatively low profitability for stays with high per diem therapy costs. 12 Although payments in this analysis are simulated for the new payment systems for home health and SNFs, the use of therapy and other resources were determined under the old payment system. The findings might change when using data from 2020 and years after since incentives to provide more therapy are reduced under PDGM and PDPM. NOTES 55 References Garrett, Bowen, Doug Wissoker, and Laura Skopec. 2021. Evaluating Potential Proxies for Patient Functional Status in a Unified Post-Acute Care Payment System for Medicare. Washington, DC: Medicare Payment Advisory Commission. MedPAC (Medicare Payment Advisory Commission). Forthcoming. 2023 Report to the Congress: Medicare and the Health Care Delivery System. Washington, DC: Medicare Payment Advisory Commission. Wissoker, Doug. 2017. "Modeling the Transition to a Unified Post-Acute Care Prospective Payment System." Washington, DC: Medicare Payment Advisory Commission. Wissoker, Doug and Bowen Garrett. 2016. Designing a Unified Prospective Payment System for Postacute Care. Washington, DC: Medicare Payment Advisory Commission. ---. 2018a. Characteristics, Costs, and Payments for Stays within a Sequence of Post-Acute Care. Washington, DC: Medicare Payment Advisory Commission. ---. 2018b. Should There be a Spell-based Unit of Payment for Medicare Home Health. Washington, DC: Medicare Payment Advisory Commission. ---. 2019. Simulating an Episode-Based Unified Payment System for Post-Acute Care. Washington, DC: Medicare Payment Advisory Commission. 56 REFERENCES About the Authors Doug Wissoker is an economist and senior fellow in the Statistical Methods Group and the Labor, Human Services, and Population Center at the Urban Institute. Bowen Garrett an economist and senior fellow in the Health Policy Center at the Urban Institute. ABOUT THE AUTHORS 57 STATEMENT OF INDEPENDENCE The Urban Institute strives to meet the highest standards of integrity and quality in its research and analyses and in the evidence-based policy recommendations offered by its researchers and experts. We believe that operating consistent with the values of independence, rigor, and transparency is essential to maintaining those standards. As an organization, the Urban Institute does not take positions on issues, but it does empower and support its experts in sharing their own evidence-based views and policy recommendations that have been shaped by scholarship. Funders do not determine our research findings or the insights and recommendations of our experts. Urban scholars and experts are expected to be objective and follow the evidence wherever it may lead. 500 L'Enfant Plaza SW Washington, DC 20024 www.urban.org