COMPARING OUTCOMES FOR DUAL ELIGIBLE BENEFICIARIES IN INTEGRATED CARE: FINAL REPORT September 2021 Office of the Assistant Secretary for Planning and Evaluation The Assistant Secretary for Planning and Evaluation (ASPE) advises the Secretary of the U.S. Department of Health and Human Services (HHS) on policy development in health, disability, human services, data, and science; and provides advice and analysis on economic policy. ASPE leads special initiatives; coordinates the Department's evaluation, research, and demonstration activities; and manages cross-Department planning activities such as strategic planning, legislative planning, and review of regulations. Integral to this role, ASPE conducts research and evaluation studies; develops policy analyses; and estimates the cost and benefits of policy alternatives under consideration by the Department or Congress. Office of Behavioral Health, Disability, and Aging Policy The Office of Behavioral Health, Disability, and Aging Policy (BHDAP) focuses on policies and programs that support the independence, productivity, health and well-being, and long-term care needs of people with disabilities, older adults, and people with mental and substance use disorders. NOTE: BHDAP was previously known as the Office of Disability, Aging, and Long-Term Care Policy (DALTCP). Only our office name has changed, not our mission, portfolio, or policy focus. This report was prepared under contract HHHSP2332016000211 between HHS's ASPE/BHDAP and Research Triangle Institute. For additional information about this subject, you can visit the BHDAP home page at https://aspe.hhs.gov/about/offices/bhdap or contact the ASPE Project Officer, at HHS/ASPE/BHDAP, Room 424E, H.H. Humphrey Building, 200 Independence Avenue, S.W., Washington, D.C. 20201; Jhamirah.Howard@hhs.gov. COMPARING OUTCOMES FOR DUAL ELIGIBLE BENEFICIARIES IN INTEGRATED CARE: Final Report Zhanlian Feng Joyce Wang Angela Gadaska Molly Knowles Susan Haber Melvin J. Ingber Valentina Grouverman RTI International September 2021 Prepared for Office of Behavioral Health, Disability, and Aging Policy Office of the Assistant Secretary for Planning and Evaluation U.S. Department of Health and Human Services Contract #HHSP2332016000211 The opinions and views expressed in this report are those of the authors. They do not reflect the views of the Department of Health and Human Services, the contractor or any other funding organization. This report was completed and submitted on September2020. CONTENTS Section Page Executive Summary ..........ccscscscccscccsseeessneeseeeceeeeceeessenessueeseecseeeeseesseesesenseseseseusesnessnsesseesesenesenees 1 ES.1 Background 20... ecesccssccsecssecsscesncsseeesecsscessccesessecsseeseseeessseessrenssereceessnsevacesaseesoness l ES.2 Method... eee eesesececceeceesenseeeeceeceecesenaeeneesaeeacecaeeseseneeasecenaeseseseneseeesatseseneneneneeseeees 1 ES.3 Key FindingS ..........cescssccsccsscsssccssesscsesccseceesccsnesessscesssesessscecasseassnesenesensscaseneseseseness 2 ES.4 Discussion and Conclusion ...........cceeesceccesceeceeeeeseneseceseneseeseeeseneneneneeseneneeeseeeseeseees 3 SECTION 1 Background...........ccecccsscessseesseeesecesseesneeseeeseeceseeseseeessesesaueeseeseseeseeessseessseneseneesnatens 5 1.1 Dual Eligible Beneficiaries 0.00... ceeesssessescssseereeesccescssecssesnsccescsscsssecsnseeneseeees 5 1.2 Integrated Care Models.............ccccssccsssesecessecessscesstecsnesecceeseseseeessteceeeseenersnessnaeeeanees 5 1.3. Challenges with Determining Outcomes Across Integrated Care Models .............. 7 L.A QYJOCHVES oe eee ecseccteeseesscesessteceeesacensccatssecssesesecaesseesnensseseeseasesserssesenecnesaeens 8 SECTION 2 Methods ........c.ceecesscssescsseeseeseeseeseececeseeaceseeseesaeceesonseeeneneceensaesensuesneeaeeeescenseeeenees 9 2.1 Data SOUICES ..... ce cceccesesessncesecesceseecessessceseecaceeasessenseseseseneesnsesssssesenesensesssssesoneseaseees 9 2.2 --_- Study Population... ccc cesscsssccssccssnessseeesasecsecesnecsnecesacesesesenecsseceeeeeenessnenssaeeeenees 9 2.3 Study MeASULES ........ cs eescsscccsccsscesccssccssccsscessesscessccssecssesscscscesseseeseessecsseresseeseeseasonees 9 2.4 Statistical AMAlySes 0.0.0... eecescssseseccscesesesscessesoseessessessessseenssecessesesseeseeseseenaeoass 10 SECTION 3 Results ........cccssesccsssseesceseeseesecseeseteaseseeacescesteseeenessnsessnseeceensneseneeseseeaceenseeenenesens 12 3.1 Descriptive Analysis Results .............cccsescsssssssscesscssecssccsscesecssscsucesscseesesesesecsseeens 12 3.2 Multivariate Analysis Results ............:.ccsccsscccsresscsessrecssscsesesssessseeessseceatecssesensesaes 14 SECTION 4 Discussion............ccccccccsssccsstecssecceecessnceseeecsseeseeeceaeesecessuessseeeeseeessueesnescsessnseseseeeenees 20 4.1 Summary of Key Findings .............ccsscssccssssssscssesecesecssccescesscssesenecesscesseneseresenecaes 20 4.2 _-_ Interpretations of Key Findings and Implications ..................cscsssscssecesseeesereeseeeees 22 4.3. Usability of MA Encounter Data for Research and Policy..............cscsssssscsseereeens 23 4.4 Limitations and Potential Areas for Future Research ................cesssscensesecereeeneeens 24 SECTION 5 Conclusion. ..........ccccecccsssesstecseeccecesseesseeecsseeseescnsesescessessseeeeseeessuecsassensesenseseneesenees 25 References 0.0... ccescscsescscescesscseeeecsesesccacessseesesesscesccscesesesenessuensesscenesesenesescsesacsnesensnesensenssenenesens R-1 Appendices A Methodology .........cccccsccscccecsssccssccsccscssssssecenscssecsessescesesacesaesesesssesaeeseesesesneceesees A-1 B Additional Descriptive Results .0..........cecsscssescsccsscesseessceesceseceeesseesscessessesenesenceaes B-1 C Full Regression Model Results ............ccccccsscssecccsesesseresereceneessnessnesesenessnseseecseeeees C-1 Exhibits Number ES-1. 10. 11. Page Multivariate regression associations between integrated care plan enrollment and service utilization and mortality among dual eligible beneficiaries in 2015, compared to a regular MA plan ..........ceececccsstecsecesceseeeecsessseseserecsueesereseneeseseesees 3 Characteristics of study population, by plan type...........csscscscssesesesescsesstscsesesesseseeess 13 Logistic regression results predicting inpatient hospitalization in 2015........... eee 15 Association between integrated care plan enrollment and any inpatient hospitalization among dual eligible beneficiaries in 2015, compared to a regular MA plat .........cccccccsssccssecsscecseccssesssseessuecsneccececseesssesesaesesenessnecsnecessesenesseanesanees 15 Logistic regression results predicting any ED Visit in 2015 ..............cesssesseceeseeeseeesees 16 Association between integrated care plan enrollment and any ED visit among dual eligible beneficiaries in 2015, compared to a regular MA plam................eeeeseeees 16 Logistic regression results predicting any institutional use in 2015.0... eeeeeeeee 17 Association between integrated care plan enrollment and any institutional use among dual eligible beneficiaries in 2015, compared to a regular MA plan............... 17 Logistic regression results predicting any HCBS use in 2015.0. eeeesseseeseeterenees 18 Association between integrated care plan enrollment and any HCBS use among dual eligible beneficiaries in 2015, compared to a regular MA plan............... 18 Logistic regression results predicting mortality 1n 2015.0... eeseereseeesetersrenetenees 19 Association between integrated care plan enrollment and mortality in 2015, compared to a regular MA plan ............csccsccsscseccsecceseescssccseecsaserscesssseesssccsaseessseeseness 19 Acknowledgements We would like to thank Dr. Edith G. Walsh for providing helpful comments on early drafts of this report. Acronyms The following acronyms are mentioned in this report and/or appendices. AIDS CMS D-SNP ED ESRD FAI FFS FIDE-SNP GAO HCBS HCC HIV HMO IDR LTSS MA MACPAC MedPAC MFFS MLTSS MMP MSC+ MSHO NF OR OREC Acquired Immunodeficiency Syndrome Centers for Medicare & Medicaid Services Dual Eligible Special Needs Plan Emergency Department End-Stage Renal Disease Financial Alignment Initiative Fee-For-Service Fully Integrated Dual Eligible Special Needs Plan U.S. Government Accountability Office Home and Community-Based Services Hierarchical Condition Category Human Immunodeficiency Virus Health Maintenance Organization Integrated Data Repository Long-Term Services and Supports Medicare Advantage Medicaid and CHIP Payment and Access Commission Medicare Payment Advisory Commission Managed Fee-For-Service Managed Long-Term Services and Supports Medicare-Medicaid Plan Minnesota Senior Care Plus Minnesota Senior Health Option Nursing Facility Odds Ratio Original Reason for Entitlement Code PACE Program of All-inclusive Care for the Elderly POS Point of Service SD Standard Deviation SNP Special Needs Plan EXECUTIVE SUMMARY ES.1 Background Dual eligible beneficiaries are an important subset of the Medicare and Medicaid populations because they have a high prevalence of chronic conditions and disabilities, substantial care needs, and high health care and long-term services and supports (LTSS) utilization and costs. The enrollment of dual eligible beneficiaries in managed care has grown significantly with the introduction of Medicare Advantage (MA) Dual Eligible Special Needs Plans (D-SNPs) that specifically target this population and of state-developed Medicaid managed long-term services and supports (MLTSS) plans or comprehensive Medicaid managed care plans that include LTSS. Integrated care models have the potential to coordinate the administration, financing, and delivery of primary, acute, and behavioral health care, as well as LTSS across the Medicare and Medicaid programs, providing significant opportunities to improve care delivery and experience of care for dual eligible beneficiaries. Examples of integrated care models include the Program of All-Inclusive Care for the Elderly (PACE), Fully Integrated Dual Eligible Special Needs Plans (FIDE-SNPs), and D-SNPs, which have varying degrees of benefit integration and administrative alignment. For policymakers, the ability to compare the quality of care and outcomes across the different models and determine their effectiveness is hindered by the lack of timely and accurate utilization data submitted by the managed care plans, referred to as encounter data. In 2019, the Centers for Medicare & Medicaid Services released the MA encounter data for 2015, the first year for which the nationwide Medicare encounter data on service use were considered to be reasonably complete and useable for research purposes. In this study, we used Medicare encounter data from 2015 to analyze and compare selected measures of service utilization and outcomes for dual eligible beneficiaries enrolled in three types of integrated care models--D-SNPs, FIDE-SNPs, or PACE--relative to their counterparts enrolled in regular, non-integrated MA plans. Our analysis did not include beneficiaries in plans under the Financial Alignment Initiative demonstrations; their service use and outcomes are being evaluated separately and are beyond the scope of this study. ES.2. Methods Our study population included full-benefit dual eligible beneficiaries who were consistently enrolled in either a regular, non-integrated MA plan or one of three specific types of integrated care MA plan--D-SNPs, FIDE-SNPs, or PACE--for all months they were enrolled in Medicare and alive in 2015. These four plan types were mutually exclusive. We created five dichotomous outcome measures pertaining to service use and mortality. All outcome measures were based on 2015 data. Any inpatient hospitalization: Whether a beneficiary had at least one inpatient hospital stay during the year. Any emergency department (ED) visit: Whether a beneficiary had at least one outpatient ED visit during the year that did not result in an inpatient admission. Any institutional use: Whether a beneficiary had any institutional use during the year (regardless of home and community-based services [HCBS] use). Institutional use includes Medicaid-covered stays in a nursing facility (NF), intermediate care facility, or inpatient psychiatric hospital. HCBS use: Whether a beneficiary had HCBS use (without institutional use). HCBS use includes services through waivers and state plans. Mortality: Whether a beneficiary died during the year. We conducted descriptive statistical analyses to compare dual eligible beneficiaries enrolled in D-SNPs, FIDE-SNPs, PACE, and regular, non-integrated MA plans. Then, we used multivariate logistic regression models to estimate the independent association of enrollment in the different plan types with each of the outcome measures. We controlled for demographic characteristics and an indicator for each state to account for variations in state policies and other state-specific factors that were not measured but could influence the outcome. We also used Hierarchical Condition Categories (HCCs) from 2014 risk adjustment data to control for beneficiary comorbidities. Depending on the outcome measure, we also applied additional specific model criteria. ES.3 Key Findings = Descriptive analyses show considerable differences in the demographic and health profiles of dual eligible beneficiaries across MA plan types in 2015: Beneficiaries in PACE were the oldest, on average, while those in D-SNPs were the youngest. A greater proportion of beneficiaries in D-SNPs were originally or currently eligible for Medicare due to disability, than those in any other plan types. Beneficiaries in PACE had the greatest number of comorbidities as measured by HCCs, followed by those in regular MA plans, then those in FIDE-SNPs, and finally those in D-SNPs. This same pattern holds when comparing their risk scores measured by the HCC system. Beneficiaries in PACE had the highest mortality rate, while those in D-SNPs had the lowest mortality rate. « After controlling for demographics and disease burden, multivariate analysis results (Exhibit ES-1) indicate that in 2015, compared to beneficiaries in a regular MA plan: * Beneficiaries in a D-SNP or PACE were less likely to be hospitalized, and those in FIDE-SNPs were more likely to be hospitalized. * Beneficiaries in a D-SNP or FIDE-SNP were more likely to visit the ED, while those in PACE were less likely to visit the ED. * Beneficiaries ina D-SNP, FIDE-SNP or PACE were much less likely to be institutionalized. * Beneficiaries ina D-SNP or FIDE-SNP were more likely to use HCBS. ¢ Beneficiaries ina D-SNP or FIDE-SNP were less likely to die, while those in PACE were no more likely to die. Exhibit ES-1. Multivariate regression associations between integrated care plan enrollment and service utilization and mortality among dual eligible beneficiaries in 2015, compared to a regular MA plan FIDE-SNP Any inpatient hospitalization Any ED visit Any institutional use -t a _t HCBS use +t +f n/a Mortality al =i) _ - indicates Jower odds of an outcome associated with an integrated plan type, compared to a regular MA plan. + indicates higher odds of an outcome associated with an integrated plan type, compared to a regular MA plan. Legend: = Favorable association, statistically significant (p < 0.05) = Unfavorable association, statistically significant (p < 0.05) n/a_| =Not applicable (PACE excluded from regression model of HCBS use) - = Statistically not significant (p > 0.05) ES.4 Discussion and Conclusion Our findings indicate that after controlling for observed case-mix differences in terms of demographic characteristics and health conditions measured by a comprehensive set of HCCs, full-benefit dual eligible beneficiaries enrolled in any of the three integrated care models (D- SNPs, FIDE-SNPs, or PACE) were significantly less likely to be institutionalized than those in regular, non-integrated MA plans. Beneficiaries in FIDE-SNPs or D-SNPs are also more likely to use HCBS than those in regular MA plans. In general, less use of institutional care and more of HCBS are preferred by beneficiaries and are also intended policy goals. However, our finding of greater odds of any ED visits among beneficiaries in D-SNPs or FIDE-SNPs and of inpatient hospitalizations among beneficiaries in FIDE-SNPs, compared to those in regular MA plans, may suggest unmet care needs despite the HCBS they have received. For beneficiaries in D- SNPs (many of whom are younger adults with disabilities), although they were institutionalized or hospitalized least frequently among all the MA plan types, they had the greatest odds of ED use. This may also indicate unmet needs among D-SNP enrollees at home and in the community, leading to more frequent use of ED services. The PACE program, known for its focus on HCBS provision and full integration of a range of medical services and LTSS, stands out from our analysis as a consistently "high performer." We found that full-benefit dual eligible beneficiaries in PACE are significantly less likely to be hospitalized, to visit the ED, or be institutionalized, while their mortality risk is not significantly higher, compared to regular MA enrollees. PACE is designed to enroll people who have frailty levels qualifying for NF care, but who are treated at home as long as possible. It is also noteworthy that beneficiaries in FIDE-SNPs or D-SNPs had significantly lower mortality risk than those in regular MA plans, after controlling for demographic characteristics and risk factors as measured by the HCCs. For beneficiaries in D-SNPs, their risk-adjusted low mortality risk might be attributable in part to unmeasured health characteristics of this population that were related to their relatively younger age but were not captured in the HCCs. As the population of full-benefit dual eligible beneficiaries enrolled in MA plans continues to grow in years to come, it becomes increasingly more important to understand their service utilization patterns and outcomes across different types of MA plans with varying degrees of coordination and integration of Medicare and Medicaid services. With the advent of nationwide MA encounter data from 2015 and onward that has become reasonably reliable and useable, researchers and policymakers can begin to use these data to help address important policy questions surrounding the coordination and integration of care for the dual eligible population. Results from our exploratory analysis of the 2015 MA encounter data show promising early evidence in support of the effectiveness of several types of MA integrated care models relative to non-integrated MA plans, including PACE, FIDE-SNPs, and D-SNPs, in reducing the use of Medicaid-covered institutional care while increasing the use of HCBS, which is an important intended policy goal. This favorable finding, however, was not always accompanied by reductions in the utilization of more costly hospital care--and indeed, we found increases in ED use by beneficiaries in FIDE-SNPs or D-SNPs and increases in inpatient hospitalization among beneficiaries in FIDE-SNPs, compared to those in regular, non-integrated MA plans. These findings may suggest that there exist unmet care needs among some beneficiaries in FIDE- SNPs and D-SNPs despite their greater use of HCBS. Our analysis did not find any adverse association of enrollment in any of the three integrated care models with mortality; enrollment in a FIDE-SNP or D-SNP could even be protective. Additional research, enhanced with more rigorous design and improved quality of the MA encounter data, is needed to validate our findings and to inform ongoing policy discussions in this area. SECTION 1 BACKGROUND 1.1 Dual Eligible Beneficiaries In 2019, 12.2 million individuals were eligible for both Medicare and Medicaid (CMS, 2020a). The majority of dual eligible beneficiaries were aged 65 or older, and 39% were people with disabilities under 65 (CMS, 2020a). Dual eligible beneficiaries receive coverage for their acute care medical services (e.g., hospital, physician, prescription drugs, and post-acute care) through Medicare. Medicaid provides financial assistance for their Medicare premiums and cost- sharing, as well as coverage for services not included in Medicare, such as long-term services and supports (LTSS) or behavioral health services. ! Dual eligible beneficiaries are an important subset of the Medicare and Medicaid populations because they have a high prevalence of chronic conditions and disabilities, substantial care needs, and high health care and LTSS utilization (Walsh et al., 2010). Dual eligible beneficiaries are among the highest cost enrollees in each program. In 2013, they accounted for 15% of Medicaid enrollees but 32% of total Medicaid expenditures. Similarly, they made up 20% of the Medicare population but accounted for 34% of total Medicare expenditures (MedPAC, 2018). Historically, dual eligible beneficiaries received their Medicare and Medicaid services mostly through fee-for-service (FFS) arrangements. However, their enrollment in managed care has grown significantly with the introduction of Medicare Advantage (MA) Dual Eligible Special Needs Plans (D-SNPs) that specifically target this population (Verdier et al., 2016) and state-developed Medicaid managed long-term services and supports (MLTSS) plans. Between 2012 and 2018, enrollment of dual eligible beneficiaries in Medicare managed care grew from 23% to 40% (MedPAC, 2020). In 2018, 33% of dual eligible beneficiaries were enrolled in either an MLTSS plan or a comprehensive Medicaid managed care plan that may have included LTSS (CMS, 2018). 1.2 Integrated Care Models Person-centered care delivery models that offer the full range of services in an integrated care system for dual eligible beneficiaries have been shown to help address the fragmentation of care associated with the lack of coordination of Medicare and Medicaid benefits, financing, and incentives (Anderson, Feng, & Long, 2016). Integrated care models have the potential to coordinate the administration, financing, and delivery of primary, acute, and behavioral health 1 Different types of dual eligible beneficiaries receive different levels of Medicaid assistance. Full-benefit dual eligible beneficiaries receive the full range of Medicaid benefits offered in a given state along with Medicaid coverage of Medicare premiums and cost-sharing for Medicare services. Partial-benefit dual eligible beneficiaries only qualify for Medicaid assistance with Medicare premiums and may also pay the cost-sharing for Medicare services. care, as well as LTSS, across the Medicare and Medicaid programs, providing significant opportunities to improve care delivery and experience of care for dual eligible beneficiaries. Examples of integrated care models include the Program of All-Inclusive Care for the Elderly (PACE); Fully Integrated Dual Eligible Special Needs Plans (FIDE-SNPs); D-SNPs, which have varying degrees of benefit integration and administrative alignment; and the capitated and managed fee-for-service (MFFS) models under the Financial Alignment Initiative (FAI) demonstrations. As of 2018, over 800,000 dual eligible beneficiaries were receiving their Medicare and Medicaid services through one of these integrated care models (Medicare- Medicaid Coordination Office, 2018). PACE is a provider-based model that serves people aged 55 or older who are eligible for state-determined nursing facility (NF) level of care but are able to live in the community with supports at the time of enrollment. PACE provides coordinated acute care, chronic care, and LTSS, with the goal of keeping enrollees in the community. The two primary model components of PACE are: (1) an adult day health center where enrollees receive medical care and social services, and (2) an interdisciplinary team comprised of medical care providers, social workers, nutritionists, therapists, personal care attendants, and drivers. Payment is capitated for both Medicare and Medicaid on a per-member per-month basis, providing an incentive to invest in medical care to improve or maintain health and reduce LTSS needs, and in LTSS to support health and reduce medical care needs. Total enrollment in PACE as of August 2020 was 49,357 beneficiaries in 137 programs in 31 states (Integrated Care Resource Center, 2020). D-SNPs are a special type of MA plan that only serve dual eligible beneficiaries. They were first authorized in the Medicare Modernization Act of 2003 with the purpose of providing a coordinated Medicare and Medicaid benefit package and offering a higher level of integration than regular MA plans or traditional FFS Medicare. The Medicare Improvements for Patients and Providers Act of 2008--as amended by the Affordable Care Act--required all D-SNPs to have contracts with the Medicaid agencies in the states in which they operate. Although D-SNPs are required to coordinate the delivery of both Medicare and Medicaid services, the majority of these plans do not provide significant levels of care integration or administrative alignment. As of August 2020, 42 states and the District of Columbia had D-SNPs enrolling more than three million dual eligible beneficiaries (over 20% of the dual eligible population) (CMS, 2020a). FIDE-SNPs are a specific type of D-SNP that focus on achieving a high degree of integration of Medicare and Medicaid services while contracting separately with the Centers for Medicare & Medicaid Services (CMS) for the Medicare-covered benefits and with states for the Medicaid-covered benefits. Authorized by the Affordable Care Act in 2010, FIDE-SNPs must "provide dually-eligible beneficiaries access to Medicare and Medicaid benefits under a single managed care organization."" In particular, FIDE-SNP contracts with states must include LTSS, and some eligible FIDE-SNPs receive additional frailty-adjusted payments. As of August 2020, 2 Section 1853(a)(1)(B)(iv) of the Social Security Act and 42 CFR §422.2. 56 FIDE-SNPs were operating in ten states (Arizona, Idaho, Massachusetts, Minnesota, New Jersey, New York, Pennsylvania, Tennessee, Virginia, and Wisconsin), with a total national enrollment of 292,725 beneficiaries (CMS, 2020b). CMS established the FAI in 2011 to allow states to test integrated care and financing models for dual eligible beneficiaries. CMS made two financial alignment models available to states: (1) a capitated model in which health plans coordinate the full range of health care services, and (2) a MFFS model in which states are eligible to benefit financially from savings resulting from initiatives that improve quality and reduce costs (Chepaitis, 2015). On April 24, 2019, CMS announced that states have three new opportunities available to test integrated care models for dual eligible beneficiaries, including the capitated financial model, the MFFS model, or a state-developed model (CMS, 2019). Our analysis did not include beneficiaries under the FAI; their service use and outcomes are being evaluated separately and are beyond the scope of this study. 1.3. Challenges with Determining Outcomes Across Integrated Care Models For policymakers, the ability to compare across programs and determine their effectiveness is key when considering which programs should be further supported and expanded. While integrated care models provide the opportunity to improve care for dual eligible beneficiaries through coordination of care, several challenges exist when trying to compare the quality of care and outcomes across the different models. These models vary in program design and populations targeted. PACE is a provider- based model for individuals aged 55 or older who qualify for NF level of care, an inherently frail population. D-SNPs are managed care organizations that target dual eligible beneficiaries, and the level of integration and coverage of Medicaid services vary by plan. FIDE-SNPs have more stringent integration requirements than D-SNPs and require a single managed care organization to coordinate both Medicare and Medicaid services and benefits, including LTSS. While LTSS is covered by all FIDE-SNPs, D-SNPs may choose to instead contract with separate MLTSS programs. States also vary in their availability of home and community-based services (HCBS) programs, access to LTSS benefits, and types and levels of Medicaid services, which all may affect the care patterns of dual eligible beneficiaries. For example, states that offer fewer HCBS programs may have higher rates of NF admissions, regardless of the presence or effectiveness of integrated care programs. States also vary in their eligibility criteria to access LTSS with some states requiring stricter criteria such as higher levels of functional impairment among Medicaid beneficiaries to qualify for LTSS. These factors broadly impact the needs of the dual eligible population, and how much their care could be improved by integrated care programs. As a result of both the different composition of dual eligible beneficiaries across states and the varying levels of coverage, it is difficult to compare outcomes of individuals enrolled in programs across states. Analysis of the patterns of service use and outcomes for beneficiaries in integrated care plans is dependent upon data submitted by the managed care plans, referred to as encounter data. The lack of timely, accurate, and integrated Medicare and Medicaid encounter data is a major barrier towards providing a complete picture of the entire spectrum of services provided dual eligible beneficiaries. Starting in 2012, all MA plans are required by CMS to provide Medicare encounter data. In 2019, CMS released the MA encounter data for 2015, the first year for which the nationwide Medicare encounter data on service use were made available for research use. However, the reporting of Medicaid FFS and encounter data is uneven across the states. 1.4 Objectives In this study, we used Medicare encounter data from 2015 to analyze and compare selected measures of service utilization and outcomes for dual eligible beneficiaries enrolled in three types of integrated care models (Special Needs Plans [SNPs], FIDE-SNPs, and PACE) relative to their counterparts enrolled in regular, non-integrated MA plans. Thus, our analysis reflects the features of these integrated care models as of 2015, which might have evolved since then and differed somewhat from their current designs. Our analysis did not include beneficiaries who enrolled in Medicare-Medicaid Plans (MMPs) in 2015 under the FAI demonstrations; their service use and outcomes are being evaluated separately (CMS, 2020c) and beyond the scope of this study. SECTION 2 METHODS 2.1 Data Sources We used 2015 data from the CMS Integrated Data Repository (IDR) to identify our study population, including information on Medicare eligibility and enrollment, demographic characteristics, institutional or HCBS use, and mortality. To define measures of service utilization, including inpatient hospitalizations and outpatient emergency department (ED) visits, we used Medicare encounter data from 2015--the first year the encounter data were considered to be reasonably complete and useable for research purposes (Mulcahy et al., 2019). We applied a 4-year runout period through December 31, 2019, which ensured data completeness. In addition, we used 2014 Medicare risk adjustment data to obtain risk scores, Hierarchical Condition Categories (HCCs), and prior long-term institutional use. 2.2 Study Population We included full-benefit dual eligible beneficiaries who were consistently enrolled in either a regular, non-integrated MA plan or one of three specific types of integrated care MA plan--D-SNPs, FIDE-SNPs, or PACE--for all months they were enrolled in Medicare and alive in 2015.4 These four plan types were mutually exclusive. We excluded beneficiaries who ever switched between different types of integrated care plans or between integrated and non- integrated plans during the year. We further excluded beneficiaries enrolled in an MMP under the FAI demonstrations. 2.3 Study Measures We created five dichotomous outcome measures pertaining to service use and mortality. All outcome measures were based on 2015 data. Any inpatient hospitalization. Using Medicare encounter data, we assessed whether a beneficiary had at least one inpatient hospital stay during the year. Any ED visit. Using Medicare encounter data, we determined whether a beneficiary had at least one outpatient ED visit during the year that did not result in an inpatient admission. Any institutional use or HCBS use. We started with a monthly IDR indicator of whether a beneficiary was institutionalized, not institutionalized, or used HCBS. Institutional use includes Medicaid-covered stays in an NF, intermediate care facility, or inpatient psychiatric hospital. 3 We could only use 2014 risk adjustment data per our Data Use Agreement with CMS. 4 Compared to the other integrated care models, the FIDE-SNPs were more concentrated among only a few states. As of December 2015, 36 FIDE SNPs were operating in seven states (Arizona, California, Idaho, Massachusetts, Minnesota, New York, and Wisconsin) with 65% of total FIDE-SNP enrollment in Massachusetts and Minnesota (CMS, 2015). Both Massachusetts and Minnesota limited their FIDE-SNP programs to dual eligible beneficiaries aged 65 or older. Please see Exhibit B-2 in the appendix for the state distribution of FIDE-SNP beneficiaries in our study population. HCBS use includes services through waivers and state plans. We then created two separate, dichotomous indicators categorizing beneficiaries as those with any institutional use in at least one month (a small percentage of whom also used HCBS in at least one month) and those with HCBS use in at least one month but no institutional use in any month. It should be noted that the eligibility for HCBS varies across states, with waivers covering specific geographic areas and different subpopulations. This adds variability to the use of HCBS. We tried to account for this variability by using control variables capturing state effects. Mortality. Using the date of death from Medicare enrollment data in the IDR, we determined whether a beneficiary died during the year. 2.4 Statistical Analyses We conducted descriptive statistical analyses to compare dual eligible beneficiaries enrolled in D-SNPs, FIDE-SNPs, PACE, and regular, non-integrated MA plans. We present descriptive statistics on the outcome measures and on beneficiary characteristics such as age, sex, race/ethnicity, original and current reason for Medicare eligibility, risk scores, and HCCs. We used multivariate logistic regression models to examine the independent associations between enrollment in the different plan types and each of the dichotomous outcome measures in 2015, including any inpatient hospitalization, any ED visit, any institutional use (regardless of HCBS use), any HCBS use (without institutional use), and mortality. In all these models, we controlled for demographic characteristics and an indicator for each state to account for variations in state policies and other state-specific factors that were not measured but could influence the outcome. In addition, we included an indicator for beneficiaries with End-Stage Renal Disease (ESRD) dialysis status for at least one month in 2015, and an interaction term between an indicator for beneficiaries who originally became eligible for Medicare because of disability and another indicator for being aged 65 or older in 2015. Our study sample used for multivariate analysis was limited to beneficiaries with 2014 risk adjustment data, which we used to obtain HCC information; we controlled for HCCs in all models. In the models predicting any inpatient hospitalization, any ED visit, and mortality, we also controlled for prior long-term institutional use in 2014. In all models except the mortality model, we further controlled for exposure time (i.e., proportion of months observed during the year, which was directly related to and highly correlated with death). In the mortality model, we excluded beneficiaries from three states (California, Oregon, and Utah) due to data irregularities. These states had mortality rates of less than 1%, which is far lower than expected for the study population. In the model predicting HCBS use, we excluded beneficiaries enrolled in PACE, because PACE is a program designed to enroll people who can be served at home while qualified for NF care and it is not part of state HCBS waiver programs. 10 The current reason for Medicare eligibility, count of HCCs, and risk scores are presented in descriptive tables only. Additional methodological details on the data sources, study sample, and variables are included in Appendix A. 11 SECTION 3 RESULTS In this section we first summarize descriptive results comparing the characteristics of the beneficiaries in D-SNPs, FIDE-SNPs, PACE, and regular, non-integrated MA plans. We then present multivariate analysis results on the associations of enrollment in each of the integrated care plan types with the outcome measures, compared to enrollment in a regular MA plan. 3.1 Descriptive Analysis Results Select characteristics of the beneficiaries by plan type are shown in Exhibit 1. Descriptive statistics on HCCs by plan type and distribution of the study population across states by plan type are included in Appendix B. Not all characteristics or beneficiaries shown in Exhibit 1 were included in multivariate analysis, due to missing information or other sample restrictions. Descriptive statistics for the sample included in regression models are also available in Appendix B. Across the different plan types, the characteristics of, and service use by, beneficiaries in the different plan types varied. In 2015, beneficiaries in PACE had the highest unadjusted inpatient hospitalization rate (21.77%) and mortality rate (11.20%), while those in D-SNPs had the lowest unadjusted hospitalization (17.74%) and mortality (2.72%) rates. The opposite is true when examining ED visits: 24.82% of beneficiaries in PACE had an ED visit, compared to 36.71% of beneficiaries in D-SNPs. Using the indicators for institutional and HCBS status derived from IDR data, a greater proportion of beneficiaries in regular, non-integrated MA plans were institutionalized for some part of 2015 (24.62%), compared to any of the integrated care plan types. HCBS use was most common among beneficiaries in FIDE-SNPs (33.56%) but less so among those in D-SNPs (14.6%) or regular, non-integrated MA plans (16.24%). In terms of demographic characteristics, beneficiaries in D-SNPs were the youngest on average (mean age = 65.12 years), while those in PACE were the oldest (mean age = 78.82 years). Accordingly, the percentage of beneficiaries aged 85 or older was lowest in D-SNPs (7.41%) and highest in PACE (32.95%). A greater percentage of beneficiaries in PACE (71.17%) and FIDE-SNPs (68.50%) were female than those in D-SNPs (62.54%) and in regular, non-integrated MA plans (66.06%). A greater proportion of beneficiaries in D-SNPs were racial/ethnic minorities and were originally or currently eligible for Medicare benefits because of disability, compared to those in any other plan type. PACE beneficiaries had the highest average count of HCCs per beneficiary (3.84), followed by those in regular MA (2.87), then those in FIDE-SNPs (2.60), and finally those in D- SNPs (2.16). PACE beneficiaries also had the highest prevalence of most of the individual HCCs, compared to those in other plan types (see Appendix B). This same pattern holds when comparing average 2014 community risk scores. 12 Exhibit 1. Characteristics of study population, by plan type Characteristic Regular MA D-SNP FIDE- SNP PACE TOTAL N (all beneficiaries, 2015) 435,968 779,411 95,637 26,884 1,337,900 Outcome measures, 2015: Any inpatient 20.09 17.74 21.45 21.77 18.85 hospitalization, % Any ED visit, % 30.72 36.71 31.24 24.82 34.13 Institutionalized in at 24.62 2.19 17.60 6.41 10.69 least 1 month, % HCBS use in at least 1 16.24 14.60 33.56 f 16.57 month but not institutionalized in any month, % Died during year, % 9.45 2.72 8.68 11.20 5.51 Age, mean (SD) 72.13 (14.99) | 65.12 (15.01) | 76.81 (10.98) | 78.82 (10.11) | 68.51 (15.27) Age, grouped: < 65, % 23.64 38.69 6.44 8.49 30.87 65-74, % 29.78 34.01 35.93 26.99 32.63 75-84, % 24.71 19.90 33.41 31.57 22.63 85+, % 21.87 7.41 24.22 32.95 13.83 Female, % 66.06 62.54 68.50 71.17 64.29 Race/ethnicity: White, non-Hispanic, % 60.98 45.70 61.21 59.37 52.06 Black, non-Hispanic, % 18.19 24.91 12.00 24.55 21.79 Hispanic, % 11.46 15.18 9.56 7.57 13,41 Asian, % 5.77 9.58 11.05 5.51 8.36 Other, % 3.60 4.63 6.18 3.00 4.37 Original reason for Medicare eligibility: Old age and 62.70 49.14 75.07 67.86 55.79 survivors, % Disability, % 36.85 50.44 24.80 31.50 43.79 ESRD, % 0.14 0.15 0.06 0.20 0.14 Both disability and 0.31 0.28 0.08 0.43 0.28 ESRD, % Current reason for Medicare eligibility: Aged without ESRD, % 75.60 60.97 92.94 89.66 68.60 Aged with ESRD, % 0.93 0.48 0.65 1.95 0.67 Disabled without 23.00 38.11 6.27 7.87 30.30 ESRD, % Disabled with ESRD, % 0.39 0.34 0.11 0.43 0.34 13 Exhibit 2. (continued) Characteristic Regular MA D-SNP FIDE- SNP PACE TOTAL ESRD only, % 0.08 0.10 0.02 0.09 0.09 ESRD dialysis status for at 1.43 0.92 0.80 2.55 1.11 least 1 month in 2015, % N (beneficiaries with 393,404 687,819 89,949 25,665 1,196,837 2014 risk scores and HCCs) Community risk score, 1.66 (1.29) 1.25 (0.99) 1.59 (1.15) 2.15 (1.27) 1.43 (1.14) mean (SD) Long-term institutional 21.10 1.10 14.23 7.51 8.67 status for at least 1 month in 2014, % Count of HCCs, mean 2.87 (2.64) 2.16 (2.16) 2.60 (2.44) 3.84 (2.72) 2.46 (2.39) (SD) ¥ Percentage of beneficiaries in PACE with any HCBS is not reported because HCBS delivered by PACE are not under the various Medicaid waiver programs. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). 3.2. Multivariate Analysis Results In this section we present key results from multivariate logistic regression analysis of each outcome. We examined the independent association between enrollment in each type of integrated care plan, compared to enrollment in a regular, non-integrated MA plan, and a given outcome after controlling for all the covariates included in each model. We report the odds ratios (ORs) and 95% confidence intervals for each of the three integrated care plan types, the main predictor variable of interest in this study. Please see Appendix C for the full model results. What are the associations between different integrated care plans and inpatient hospitalizations? The logistic regression model results predicting any inpatient hospitalization are displayed in Exhibit 2 and Exhibit 3. PACE beneficiaries were significantly less likely to be hospitalized than those in regular MA (OR = 0.689; p < 0.001). Beneficiaries in D-SNPs were slightly less likely to be hospitalized compared to those in regular MA (OR = 0.970; p < 0.001). Beneficiaries in FIDE-SNPs were more likely to be hospitalized than those in regular MA (OR = 1.241; p< 0.001). 14 Exhibit 2. Logistic regression results predicting inpatient hospitalization in 2015 Plan Type . 0 (Reference = Regular MA) Odds Ratio 95% Confidence Interval D-SNP 0.970 el 0.958 0.981 FIDE-SNP 1.241 eK 1.207 1.277 PACE 0.689 ee 0.667 0.713 */ee eee - Significantly different from regular MA plan based on a p-value cutoff of 0.05/0.01/0.001 NOTES: In addition to MA plan types, the full regression model also controlled for beneficiary demographic characteristics, current ESRD dialysis status, an interaction term between being originally eligible for Medicare because of disability and currently being aged 65 or older, prior long-term institutional use, exposure time (proportion of months observed during the year), HCCs, and an indicator for each state. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). Exhibit 3. Association between integrated care plan enrollment and any inpatient hospitalization among dual eligible beneficiaries in 2015, compared to a regular MA plan D-SNP oe FIDE-SNP - PACE /--e-_ 0.75 1.00 1.25 Odds ratio and 95% Cl NOTES: In addition to MA plan types, the full regression model also controlled for beneficiary demographic characteristics, current ESRD dialysis status, an interaction term between being originally eligible for Medicare because of disability and currently being aged 65 or older, prior long-term institutional use, exposure time (proportion of months observed during the year), HCCs, and an indicator for each state. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). What are the associations between different integrated care plans and ED visits? As shown in Exhibit 4 and Exhibit 5, beneficiaries in D-SNPs and FIDE-SNPs were more likely to visit the ED at least once than beneficiaries in regular MA (OR = 1.160; p < 0.001 and OR = 1.141; p < 0.001, respectively). The opposite is true for beneficiaries in PACE; those in PACE were less likely to visit the ED (OR = 0.523; p < 0.001). 15 Exhibit 4. Logistic regression results predicting any ED visit in 2015 Plan Type , a (Reference = Regular MA) Odds Ratio 95% Confidence Interval D-SNP 1.160 *K 1.149 1.172 FIDE-SNP 1.141 eK 1.113 1.170 PACE 0.523 eK 0.507 0.539 */**/*** = Significantly different from regular MA plan based on a p-value cutoff of 0.05/0.01/0.001 NOTES: In addition to MA plan types, the full regression model also controlled for beneficiary demographic characteristics, current ESRD dialysis status, an interaction term between being originally eligible for Medicare because of disability and currently being aged 65 or older, prior long-term institutional use, exposure time (proportion of months observed during the year), HCCs, and an indicator for each state. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). Exhibit 5. Association between integrated care plan enrollment and any ED visit among dual eligible beneficiaries in 2015, compared to a regular MA plan D-SNP bo FIDE-SNP t-e4 PACE --e- 0.50 1.00 1.15 Odds ratio and 95% Cl NOTES: In addition to MA plan types, the full regression model also controlled for beneficiary demographic characteristics, current ESRD dialysis status, an interaction term between being originally eligible for Medicare because of disability and currently being aged 65 or older, prior long-term institutional use, exposure time (proportion of months observed during the year), HCCs, and an indicator for each state. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). What are the associations between different integrated care plans and institutional and HCBS use? We separately examined the association of integrated care plan enrollment with institutional use and with HCBS use as defined in the IDR. Institutional use includes Medicaid- covered stays in a NF, intermediate care facility, or inpatient psychiatric hospital. HCBS use includes services through Medicaid waivers and state plans. Regression results on institutional use are displayed in Exhibit 6 and Exhibit 7, and results from the HCBS model are presented in Exhibit 8 and Exhibit 9. Beneficiaries in D-SNPs are less likely to be institutionalized (OR = 0.127; p < 0.001) and more likely to use HCBS (OR = 1.046; p < 0.001), compared to those in regular MA. This same pattern holds when examining beneficiaries in FIDE-SNPs and their 16 institutional use (OR = 0.320; p < 0.001) and HCBS use (OR = 4.223; p < 0.001). Those in PACE are much less likely to be institutionalized (OR = 0.062; p < 0.001). PACE beneficiaries were excluded from the HCBS model because home-based care is the default treatment pattern for the program. Exhibit 6. Logistic regression results predicting any institutional use in 2015 Plan Type . 0 (Reference = Regular MA) Odds Ratio 95% Confidence Interval D-SNP 0.127 *eK 0.124 0.129 FIDE-SNP 0.320 eK 0.308 0.332 PACE 0.062 eK 0.058 0.065 */pe*/*** - Significantly different from regular MA plan based on a p-value cutoff of 0.05/0.01/0.001 NOTES: In addition to MA plan types, the full regression model also controlled for beneficiary demographic characteristics, current ESRD dialysis status, an interaction term between being originally eligible for Medicare because of disability and currently being aged 65 or older, exposure time (proportion of months observed during the year), HCCs, and an indicator for each state. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). Exhibit 7. Association between integrated care plan enrollment and any institutional use among dual eligible beneficiaries in 2015, compared to a regular MA plan D-SNP tel FIDE-SNP He PACE Fe 0.06 0.25 1.00 Odds ratio and 95% Cl NOTES: In addition to MA plan types, the full regression model also controlled for beneficiary demographic characteristics, current ESRD dialysis status, an interaction term between being originally eligible for Medicare because of disability and currently being aged 65 or older, exposure time (proportion of months observed during the year), HCCs, and an indicator for each state. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). 17 Exhibit 8. Logistic regression results predicting any HCBS use in 2015 Plan Type . 0 (Reference = Regular MA) Odds Ratio 95% Confidence Interval D-SNP 1.046 el 1.033 1.060 FIDE-SNP 4.223 eK 4.102 4.347 */**/*** = Sionificantly different from regular MA plan based on a p-value cutoff of 0.05/0.01/0.001 NOTES: In addition to MA plan types, the full regression model also controlled for beneficiary demographic characteristics, current ESRD dialysis status, an interaction term between being originally eligible for Medicare because of disability and currently being aged 65 or older, exposure time (proportion of months observed during the year), HCCs, and an indicator for each state. The model excluded beneficiaries in PACE. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). Exhibit 9. Association between integrated care plan enrollment and any HCBS use among dual eligible beneficiaries in 2015, compared to a regular MA plan D-SNP | fe FIDE-SNP Hed Odds ratio and 95% Cl NOTES: In addition to MA plan types, the full regression model also controlled for beneficiary demographic characteristics, current ESRD dialysis status, an interaction term between being originally eligible for Medicare because of disability and currently being aged 65 or older, exposure time (proportion of months observed during the year), HCCs, and an indicator for each state. The model excluded beneficiaries in PACE. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). What are the associations between the different integrated care plans and mortality? As displayed in Exhibit 10 and Exhibit 11, beneficiaries in D-SNPs and FIDE-SNPs were significantly less likely to die in 2015 than beneficiaries in regular MA (OR = 0.578; p < 0.001 and OR = 0.694; p < 0.001, respectively). There was not a statistically significant difference in mortality between beneficiaries in PACE and those in regular MA (OR = 0.958; p = 0.062). 18 Exhibit 10. Logistic regression results predicting mortality in 2015 Plan Type : 9 (Reference = Regular MA) Odds Ratio 95% Confidence Interval D-SNP 0.578 *eK 0.565 0.591 FIDE-SNP 0.694 *K 0.663 0.728 PACE 0.958 0.917 1.002 */ee eee - Significantly different from regular MA plan based on a p-value cutoff of 0.05/0.01/0.001 NOTES: In addition to MA plan types, the full regression model also controlled for beneficiary demographic characteristics, current ESRD dialysis status, an interaction term between being originally eligible for Medicare because of disability and currently being aged 65 or older, prior long-term institutional use, HCCs, and an indicator for each state. The model excluded beneficiaries from California, Oregon, and Utah. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). Exhibit 11. Association between integrated care plan enrollment and mortality in 2015, compared to a regular MA plan D-SNP /-e- FIDE-SNP }_e-__-_ PACE __» __j 0.55 0.75 1.00 Odds ratio and 95% Cl NOTES: In addition to MA plan types, the full regression model also controlled for beneficiary demographic characteristics, current ESRD dialysis status, an interaction term between being originally eligible for Medicare because of disability and currently being aged 65 or older, prior long-term institutional use, HCCs, and an indicator for each state. The model excluded beneficiaries from California, Oregon, and Utah. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). 19 SECTION 4 DISCUSSION In this section we summarize and discuss the major findings of this analysis--the first of its kind--to compare utilization and health outcomes across integrated care models using nationwide MA encounter data from 2015, the first year considered to have reasonably complete and useable encounter data for research purposes (Mulcahy et al., 2019). Where possible, we compare our findings to the existing literature. However, the previous literature on this topic is scant, because most of the existing studies have compared service use and outcomes between traditional FFS Medicare and MA beneficiaries, and we did not identify any studies that have compared various integrated care models with regular, non-integrated MA plans using national data. 4.1 Summary of Key Findings ¢ There are considerable differences in the health profile of full-benefit dual eligible beneficiaries across MA plan types. Beneficiaries in PACE programs had the greatest number of comorbidities as measured by HCCs, followed by those in regular MA, then those in FIDE-SNPs, and finally those in D-SNPs. This same pattern holds when comparing their risk scores. Consistent with our findings, other studies have also identified high rates of chronic conditions among PACE enrollees. For example, the average PACE enrollee has multiple acute and chronic medical conditions, such as heart or respiratory disease or diabetes (Hirth, Baskins, & Dever-Bumba, 2009) and PACE participants were more likely to be diagnosed with Alzheimer's disease or other forms of dementia compared to HCBS participants (42% and 29%, respectively) (Beauchamp et al., 2008). The limited existing literature comparing chronic conditions in D-SNPs to other types of MA plans reports different findings from ours. In contrast to our findings, a Government Accountability Office (GAO) study found a higher prevalence of some chronic conditions among D-SNP and Medicare FFS beneficiaries compared to dual eligible beneficiaries enrolled in other MA plans. For example, the study found 15% of D-SNP enrollees were diagnosed with a chronic or disabling mental illness, such as major depressive disorder or schizophrenia, compared to 10% of dual eligible beneficiaries enrolled in other regular MA plans (GAO, 2012). Our study, using HCCs that group mental illness diagnoses differently (combining major depressive, bipolar, and paranoid disorders together and including schizophrenia separately), found less variation in prevalence across MA types. * After controlling for demographics and disease burden, beneficiaries in D-SNPs or PACE were less likely to be hospitalized, and those in FIDE-SNPs were more likely to be hospitalized, compared to those in regular MA plans. 20 We were not able to identify other studies that compared hospitalization rates across integrated care models with regular MA plans. However, other studies found that D-SNP or FIDE-SNP enrollees had lower inpatient utilization compared to the Medicare FFS dual eligible population. For example, a descriptive analysis of D-SNP beneficiaries determined they averaged 2,821 inpatient days per 1,000 enrollees per year compared to 3,327 inpatient days per 1,000 enrollees per year for FFS dual eligible beneficiaries (Lewin Group, 2011). Another study found that preventable hospitalization rates among D-SNP enrollees were 14% lower and risk- adjusted hospital readmission rates were 25% lower than in Medicare FFS (Avalere, 2012). Anderson, Long & Feng (2020) found a significantly lower rate of inpatient hospital stays among enrollees in the Minnesota Senior Health Option (MSHO), a FIDE-SNP model, compared to enrollees in the Minnesota Senior Care Plus (MSC+), a Medicaid-only managed care plan with Medicare FFS. The literature also indicates that PACE enrollees are less likely to be hospitalized and spend fewer days in the hospital compared to control groups (Beauchamp et al., 2008; MedPAC, 2012; Segelman et al., 2014). ¢ Based on multivariate analyses, beneficiaries in D-SNPs or FIDE-SNPs were more likely to visit the ED at least once, while those in PACE were less likely to visit the ED, compared to those in regular MA plans. Compared to our analyses, the findings related to ED use in the literature are mixed. We were not able to identify any studies that compared beneficiaries enrolled in integrated models to beneficiaries enrolled in regular MA plans. But among studies that analyzed ED use among integrated care programs, one study differed from our analysis and found that FIDE-SNP beneficiaries were 6% less likely to have an outpatient ED visit compared to dual eligible beneficiaries in Medicaid managed care (Anderson et al., 2016). Another study found that after adjusting for demographic characteristics and certain disease conditions, D-SNP enrollees had a 9% lower ED visit rate compared to Medicare FFS dual eligible beneficiaries (Murugan, Drozd, & Dietz, 2012). Consistent with our analysis, another study conducted a descriptive analysis and found ED use by D-SNP enrollees (919 ED visits per 1,000 enrollees per year) and by FIDE- SNP enrollees (917 ED visits per 1,000 enrollees per year) was higher compared to dual eligible beneficiaries in FFS (844 ED visits per 1,000 enrollees per year) (Lewin Group, 2011). ¢ After risk-adjustment, beneficiaries in D-SNPs, FIDE-SNPs or PACE were much less likely to be institutionalized thanthose in regular MA plans. Previous studies of FIDE-SNP enrollees varied in NF utilization outcomes. Although we were not able to identify studies that compared integrated care programs with regular MA plans, one study of FIDE-SNP enrollees also determined that enrollment was associated with a 16% lower risk of long-stay NF admission after risk adjustment compared to the Medicare FFS dual eligible population (JEN Associates, 2013). However, unlike our analysis, another study (Anderson et al., 2020) found no significant difference in long-term NF use between enrollees in the MSHO, a FIDE-SNP model, and enrollees in the MSC+, a Medicaid-only managed care plan with Medicare FFS, after risk adjustment. 21 The literature also showed mixed findings on NF use among PACE enrollees. Unlike our analysis, one multivariate analysis (Beauchamp et al., 2008) and one descriptive study (Nadash, 2004) found that NF use was higher among PACE enrollees compared to HCBS participants and participants in Medicaid MLTSS plans. Conversely, other studies were consistent with our analyses and found that NF use was lower in PACE enrollees when compared to PACE eligible or HCBS waiver dual eligible beneficiaries after risk adjustment (MedPAC, 2012; Segelman et al., 2015). * Beneficiaries in FIDE-SNPs and D-SNPs were more likely to receive HCBS compared to those in regular MA plans. We did not identify any studies that compared HCBS utilization of dual eligible beneficiaries enrolled in integrated care programs with dual eligible beneficiaries enrolled in regular MA plans. However, one study of dual eligible beneficiaries in Minnesota found that enrollees in MSHO (a FIDE-SNP model) had greater use of primary care and HCBS than enrollees in MSC+ (a less integrated Medicaid-only managed care plan) after risk adjustment (Anderson et al., 2020). ¢ Beneficiaries in D-SNPs or FIDE-SNPs were less likely to die than those enrolled in regular MA. There was no evidence that those in PACE were more or less likely to die, compared to those in regular MA plans. There is limited literature that examines similar mortality comparisons. One multivariate analysis that compared FIDE-SNP enrollees with FFS dual eligible beneficiaries determined that FIDE-SNP enrollees had a 17% lower risk of death compared to FFS beneficiaries (JEN Associates, 2013). Overall, studies of PACE enrollees found lower mortality rates compared to HCBS waiver enrollees and FFS dual eligible beneficiaries (Chatterji et al., 1998; Ghosh, Schmitz, & Brown, 2015; JEN Associates, 2015; Wieland et al., 2010). 4.2 Interpretations of Key Findings and Implications Our findings indicate that after controlling for observed case-mix differences in terms of demographic characteristics and health conditions measured by a comprehensive set of HCCs, full-benefit dual eligible beneficiaries enrolled in any of the three integrated care models (D-SNPs, FIDE-SNPs, or PACE) were significantly less likely to be institutionalized than their counterparts in regular, non-integrated MA plans. Beneficiaries in FIDE-SNPs or D-SNPs are also more likely to use HCBS than those in regular MA plans. In general, less use of institutional care and more of HCBS are preferred by beneficiaries and are also intended federal policy goals (e.g., federal initiatives to support state efforts to rebalance LTSS such as Money Follows the Person program or the Balancing Incentive Program) (Musumeci & Reaves, 2014; CMS, n.d.). However, our finding of greater odds of ED visits among beneficiaries in D-SNPs or FIDE-SNPs and of inpatient hospitalizations among beneficiaries in FIDE-SNPs, compared to those in regular MA plans, may suggest unmet care needs of beneficiaries despite the HCBS they have 22 received. Based on descriptive data, beneficiaries in D-SNPs (many of whom are younger adults with disabilities) were institutionalized or hospitalized least frequently among all the MA plan types, but they visited the ED most frequently. This may also indicate unmet needs among D- SNP enrollees at home and in the community, leading to more frequent use of ED services. The PACE program, well known for its focus on HCBS provision and full integration of a range of medical services and LTSS, stands out from our analysis as a consistently "high performer." We found that full-benefit dual eligible beneficiaries in PACE are significantly less likely to be hospitalized, to visit the ED, or be institutionalized, while their mortality risk is not greater despite their higher frailty levels, compared to regular MA enrollees. It is also noteworthy that beneficiaries in FIDE-SNPs or D-SNPs had significantly lower mortality risk than those in regular MA plans, after controlling for demographic characteristics and risk factors as measured by the HCCs. For beneficiaries in D-SNPs, their risk-adjusted low mortality risk might be attributable in part to unmeasured health characteristics of this population that were related to their relatively younger age but were not captured in the HCCs. Although we applied an extensive list of risk adjustment characteristics in the model to account for case-mix differences across plan types, there are always potentially unobserved factors that could account for some degree of estimated differences. For example, the D-SNP population is considerably younger and has a lower disease burden than other plan populations. We have adjusted for these differences. However, if severity of the diseases in the young population is less than that of older populations with the same conditions, we cannot measure that directly. 4.3 Usability of MA Encounter Data for Research and Policy For years, the lack of reliable MA encounter data has been a major barrier for researchers, policymakers and other stakeholders to track health service utilization and outcomes for dual eligible beneficiaries in managed care plans in general and in various integrated care models in particular (Brennan, 2018; Creighton, Duddy-Tenbrunsel, & Michel, 2019). The recent release by CMS of the Research Identifiable File MA encounter data made this analysis possible. Our findings on inpatient hospitalizations and outpatient ED visits were based on MA encounter data for 2015, the first year for which the encounter data were considered to be reasonably complete and of acceptable quality, in line with data validation findings by others (Mulcahy et al., 2019). Using the beneficiary and MA contract or plan identification information on the 2015 encounter data, we were able to identify and classify beneficiaries into the three integrated care plan types of interest versus those in regular, non-integrated MA plans, and to link with their hospital inpatient and outpatient encounter data for comparison. As far as utilization of major health care services is concerned, such as hospital inpatient stays and ED visits, we consider the 2015 encounter data to be reasonably reliable for this analysis. Given the newness of these data and the scarcity of published studies using these data, we consider our study to be an exploratory analysis. 23 4.4 Limitations and Potential Areas for Future Research In addition to potential issues about the quality of MA encounter data, we note several limitations of this study. First, although we controlled for beneficiary demographic information and a comprehensive set of HCCs as risk factors in our multivariate regression models, it is possible that unmeasured disease severity or frailty factors, together with the lack of functional impairment measures, could drive the residual differences in the observed outcomes and therefore potentially bias our estimated effects of integrated care plan types on each outcome. Second, we identified a potential issue with the mortality data for our study population from three states (California, Oregon, and Utah), where the mortality rate in 2015 was unusually low relative to the national average. We were unable to ascertain whether the data were erroneous and opted to exclude beneficiaries in the three states from the mortality model. We note, however, all the other outcome measures appeared to be reasonable for beneficiaries in those three states. There could be reporting errors in the IDR data and this warrants further investigation. Lastly, in this study we conducted a population based analysis that included the entire population of full-benefit dual eligible beneficiaries in 2015 who were in one of the three MA integrated care models or in a regular MA plan and met all other study inclusion criteria. This approach is appropriate for an exploratory analysis to compare beneficiary outcomes across the various MA plan types. Future research could be enhanced by selecting a comparison group of beneficiaries in regular, non-integrated MA plans who have similar characteristics and risk profiles to those in a given type of integrated care model and incorporating this comparison group in multivariate analysis. The comparison group selection should also take into consideration the fact that D-SNPs, FIDE-SNPs, and PACE programs are more concentrated in some states than others. Depending on sample sizes, the comparison group could be selected within states or among states with similar penetration of integrated care programs. CMS and many states have prioritized improving care and reducing costs of care for dual eligible beneficiaries by supporting integrated care models. The recent proliferation of non- integrated care MA options, such as D-SNP "look-alike" plans, has come under state and federal scrutiny (CMS, 2019). Future research of outcomes among dual eligible beneficiaries enrolled in integrated care programs may provide policymakers additional support to address such non- integrated MA options that target dual eligible beneficiaries. 24 SECTION 5 CONCLUSION As the population of full-benefit dual eligible beneficiaries enrolled in MA plans continues to grow, it becomes increasingly important to understand their service utilization patterns and outcomes across different types of MA plans with varying degrees of coordination and integration of Medicare and Medicaid services. With the advent of nationwide MA encounter data from 2015 and onward that has become reasonably reliable and useable, researchers and policymakers can begin to use these data to help address important policy questions surrounding the coordination and integration of care for the dual eligible population. Results from our exploratory analysis of the 2015 MA encounter data show promising early evidence in support of the effectiveness of several types of MA integrated care models, including PACE, FIDE-SNPs, and D-SNPs, in reducing the use of Medicaid-covered institutional care while increasing the use of HCBS waiver services, which is an important intended policy goal. This favorable finding, however, was not always accompanied by reductions in the utilization of more costly hospital care--and indeed, we found increases in ED use by beneficiaries in FIDE- SNPs or D-SNPs and increases in inpatient hospitalization among beneficiaries in FIDE-SNPs, compared to their counterparts in regular, non-integrated MA plans. These findings may suggest that there exist unmet care needs among some beneficiaries in FIDE-SNPs and D-SNPs despite their greater use of HCBS waiver services. Our analysis did not find any adverse association of enrollment in any of the three integrated care models with mortality; enrollment in a FIDE-SNP or D-SNP could even be protective. Additional research, enhanced with more rigorous design and improved quality of the MA encounter data, is needed to validate our findings and to inform ongoing policy discussions in this area. 25 REFERENCES Anderson, W., Feng, Z., & Long, S. (2016). Minnesota Managed Care Longitudinal Data Analysis. RTI International. Retrieved July 29 from https://aspe.hhs.gov/reports/minnesota- managed-care-longitudinal-data-analysis-0. Anderson, W.L., Long, S.K., & Feng, Z. (2020). Effects of integrating care for Medicare- Medicaid dually eligible seniors in Minnesota. J Aging Soc Policy, 32(1), 31-54. doi.org/10.1080/08959420.2018.1485396. Avalere. (2012). Dual Eligible Population Analysis for SCAN Health Plan: Hospitalizations and Readmissions. Retrieved from https://www.nchc.org/wp- content/uploads/2016/04/20120401_scan-dual-eligibles-analysis_ final for-posting.pdf. Beauchamp, J., Cheh, V., Schmitz, R., Kemper, P., & Hall, J. (2008). The Effect of the Program of All-Inclusive Care for the Elderly (PACE) on Quality. Mathematica Policy Research. https://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and- Reports/Reports/downloads/beauchamp_2008.pdf. Chatterji, P., Bustein, N.R., Kidder, D., & White, A. (1998). Evaluation of the Program of All- Inclusive Care for the Elderly (PACE) demonstration: The impact of PACE on participant outcomes. Retrieved from https://innovation.cms.gov/files/migrated-medicare-demonstration- x/evalpace-outcomeimpact.pdf. Chepaitis, A., Greene, A.M., Hoover, S., Khatutsky, G., Lyda-McDonald, B., Ormond, C., Walsh, E., Zheng, N., Booth, M., Fralich, J., Justice, D., & Weiss, A. (2015). Report on Early Implementation of the Demonstrations under the Financial Alignment Initiative. https://www.cms.gov/Medicare-Medicaid-Coordination/Medicare-and-Medicaid- Coordination/Medicare-Medicaid-Coordination- Office/FinancialAlignmentInitiative/Downloads/MultistateIssueBriefFAL pdf. Centers for Medicare & Medicaid Services (CMS). (2015). SVP Comprehensive Report-- December 2015. https://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics- Trends-and-Reports/MCRAdvPartDEnrolData/Special-Needs-Plan-SNP-Data-Items/SNP- Comprehensive-Report-2015-12. Centers for Medicare & Medicaid Services (CMS). (2018). Medicaid Managed Care Enrollment Report. Retrieved from https://www.medicaid.gov/medicaid/managed-care/downloads/2018- medicaid-managed-care-enrollment-report.pdf. Centers for Medicare & Medicaid Services (CMS). (2019). State Medicaid Director Letter: Three New Opportunities to Test Innovative Models of Integrated Care for Individuals Dually Eligible for Medicaid and Medicare. Retrieved from https://www.medicaid.gov/federal-policy-guidance/downloads/smd19002.pdf. Centers for Medicare & Medicaid Services (CMS). (2020a). Medicare-Medicaid Coordination Office: FY 2019. Report to Congress. Retrieved from https://www.cms.gov/Medicare- Medicaid-Coordination/Medicare-and-Medicaid-Coordination/Medicare-Medicaid- Coordination-Office/Downloads/FY -2018-Report-to-Congress.pdf. Centers for Medicare & Medicaid Services (CMS). (2020b). SVP Comprehensive Report-- August 2020. Retrieved from https://www.cms.gov/research-statistics-data-and- systemsstatistics-trends-and-reportsmcradvpartdenroldataspecial-needs/snp-comprehensive- report-2020-08. Centers for Medicare & Medicaid Services (CMS). (2020c). Financial Alignment Initiative: Evaluations. https://www.cms.gov/Medicare-Medicaid-Coordination/Medicare-and- Medicaid-Coordination/Medicare-Medicaid-Coordination- Office/FinancialAlignmentInitiative/Evaluations. Centers for Medicare & Medicaid Services (CMS). (n.d.) Balancing Long-Term Services and Supports. https://www.medicaid.gov/medicaid/long-term-services-supports/balancing-long- term-services-supports/index.html. Creighton, S., Duddy-Tenbrunsel, R., & Michel, J. (2019). The Promise and Pitfalls of Medicare Advantage Encounter Data. https://www.healthaffairs.org/do/10.1377/hblog20190221.696651/full/. Government Accountability Office (GAO). (2012). Medicare Special Needs Plans: CMS should improve Information available about Dual-Eligible Plans' Performance. http://www.gao.gov/assets/650/64829 1 .pdf. Ghosh, A., Schmitz, R., & Brown, R. (2015). Effect of PACE on Costs, Nursing Home Admissions, and Mortality: 2006-2011. https://aspe.hhs.gov/reports/effect-pace-costs- nursing-home-admissions-mortality-2006-2011-0. Hirth, V., Baskins, J., & Dever-Bumba, M. (2009). Program of All-Inclusive Care (PACE): Past, present, and future. Journal of the American Medical Directors Association, 10(3), 155-160. doi.org/10.1016/j.jamda.2008.12.002. Integrated Care Resource Center. (2020). Program of All-Inclusive Care for the Elderly (PACE) Total Enrollment by State and by Organization. JEN Associates. (2013). Massachusetts Senior Care Option 2005-2010 Impact on Enrollees: Nursing Home Entry Utilization. http://www.mass.gov/eohhs/docs/masshealth/sco/sco- evaluation-nf-entry-rate-2004-through-2010-enrollment-cohorts.pdf. JEN Associates. (2015). Massachusetts PACE Evaluation Nursing Facility Residency and Mortality Summary Report. http://www.mass.gov/eohhs/docs/masshealth/pace/ma-pace- evaluation-nf-residency-and-mortality-summary-report-11-23-15.pdf. R-2 Kaiser Family Foundation. (2014). Medicaid Beneficiaries Who Need Home and Community- Based Services: Supporting Independent Living and Community Integration. https://www.kff.org/wp-content/uploads/2014/03/8568-medicaid-beneficiaries-who-need- home-and-community-based-servcies.pdf. Medicaid.gov. (2019). 2017 Managed Care Enrollment by Program and Population (Duals). Retrieved July 29 from https://data.medicaid.gov/Enrollment/2017-Managed-Care- Enrollment-by-Program-and-Popula/388z-vsy3. Medicare Payment Advisory Commission (MedPAC). (2012). Report to Congress: Chapter 3, Care Coordination Programs for Dual-Eligible Beneficiaries. Retrieved from http://medpac.gov/docs/default-source/reports/junl2_ch03.pdf?sfvrsn=0. Medicare Payment Advisory Commission (MedPAC). (2020). A Data Book: Health Care Spending and the Medicare Program. Retrieved from http://www.medpac.gov/docs/default- source/data-book/july2020 databook_entirereport_sec.pdf?sfvrsn=0. Medicare Payment Advisory Commission (MedPAC), & Medicaid and CHIP Payment and Access Commission (MACPAC), (2018). Beneficiaries Dually Eligible for Medicare and Medicaid Data Book. Mulcahy, A.W., Sorbero, M.E., Mahmud, A., Wilks, A., Gildner, J., Hornsby, A., & Pignotti, A. (2019). Measuring Health Care Utilization in Medicare Advantage Encounter Data: Methods, Estimates, and Considerations for Research. RAND Corporation. https://www.rand.org/pubs/research_reports/RR2681.html. Murugan, V., Drozd, E., & Dietz, K. (2012). A Comparison of the Mercy Care Plan Population to Nationwide Dual-Eligible Medicare Beneficiaries. Avalere. http://avalere.com/research/docs/20120627_Avalere_ Mercy Care White Paper.pdf. Nadash, P. (2004). Two models of managed long-term care: Comparing PACE with a Medicaid- only plan. Gerontologist, 44(5), 644-654. doi.org/10.1093/geront/44.5.644. Segelman, M., Szydlowski, J., Kinosian, B., McNabney, M., Raziano, D.B., Eng, C., van Reenen, C., & Temkin-Greener, H. (2014, Feb). Hospitalizations in the Program of All- Inclusive Care for the elderly. J Am Geriatr Soc, 62(2), 320-324. doi.org/10.1111/4gs.12637. Segelman, M., Cai, X., van Reenen, C., & Temkin-Greener, H. (2017). Transitioning from community-based to institutional long-term care: Comparing 1915(c) waiver and PACE enrollees. Gerontologist, 57(2), 300-308. The Lewin Group. (2011). 2010 SNP Alliance Profile and Advanced Practice Report. http://www.nhpg.org/media/12019/201 1lewinprofilefinal.pdf. R-3 Verdier, J., Kruse, A., Sweetland Lester, R., Philip, A-M., & Chelminsky, D. (2016). State Contracting with Medicare Advantage Dual Eligible Special Needs Plans: Issues and Options. Retrieved July 29 from http://www.chcs.org/media/ICRC_DSNP_Issues_Options.pdf. Walsh, E.G., Freiman, M.P., Haber, S., Bragg, A., Ouslander, J., & Wiener, J.M. (2010). Cost Drivers for Dually Eligible Beneficiaries: Potentially Avoidable Hospitalizations from Long- Term and Post-Acute Care Settings. Wieland, D., Boland, R., Baskins, J., & Kinosian, B. (2010). Five-year survival in a Program of All-inclusive Care for elderly compared with alternative institutional and home- and community-based care. J Gerontol A Biol Sci Med Sci, 65(7), 721-726. doi.org/10.1093/gerona/g1q040. R4 APPENDIX A METHODOLOGY After describing key data sources and critical components of our analytic file construction in detail, we summarize the variables used in our analyses in Exhibit A-1. Then we describe our study population and samples used for descriptive and multivariate analyses. A.1 Data sources The Centers for Medicare & Medicaid Services (CMS) IDR was used for all analyses. All data were accessed between January and August of 2020. Key tables, or views, are described below: ¢ V2 MDCR_BENE FCT: 2015 indicators of eligibility, demographic characteristics, and institutional/HCBS outcomes: - Note: This table consolidates information from multiple source tables. The IDR has also been transitioning into restructured BENE_FCT_TRANS tables. ¢ V2 MDCR_CNTRCT_PBP NUM: 2015 indicators of MA enrollment plan information, including specific type of integrated care plan. * V2 MDCR_BENE RISK SCRE: 2014 risk adjustment data on risk scores and long term institutional status. ¢ V2_MDCR_BENE_RISK PTC_F_ SCRE: Hierarchical condition categories (HCCs). * V2 _MDCR_CLM: 2015 encounter data claims header information for utilization measures. * V2 MDCR_ CLM LINE: 2015 encounter data claims line information for utilization measures. * V2 MDCR_BENE: 2015 indicator of mortality outcome. A.2 Analytic file construction Full-benefit dual eligibility. Beneficiaries were considered full-benefit dual eligible if they met full-benefit criteria for all months they were enrolled in Medicare and alive in 2015. Full-benefit status was indicated by BENE DUAL _STUS_CD = 02 (Qualified Medicare Beneficiaries plus full Medicaid), 04 (Specified Low-Income Medicare Beneficiaries plus full Medicaid), or 08 (other full-benefit duals). Beneficiaries also had to be Part A and Part B eligible for all months (BENE PTA STUS CD='Y' and BENE PTB STUS CD='Y'). MA plan information. Beneficiaries were considered MA enrollees for a given month if CNTRCT_PBP_PTAB_SK > 0. After restricting our sample to beneficiaries who were enrolled in MA for all months, we examined more detailed MA plan information in the table CNTRCT_PBP_ NUM. We used CNTRCT_SPCL_ PLAN IND CD =3 to indicate monthly enrollment in a D-SNP plan, and CNTRCT_SPCL_PLAN_IND_CD = 9 to indicate monthly enrollment in a FIDE-SNP plan. We used CNTRCT_PBP_TYPE CD = 20 to indicate PACE enrollment. Remaining beneficiary months were considered enrollment in a regular non- integrated MA plan. We then created four mutually exclusive categories at the beneficiary-year level by excluding beneficiaries who switched between types of integrated care plans, or between integrated and non-integrated care, within the year. We then used CNTRCT_PBP_TYPE_CD = 48 (MMP HMO) or = 49 (MMP HMOPOS) to indicate MMP enrollment. Encounter data outcomes. We identified unique beneficiary claims from 2015 using the 5-part key described in Exhibit A-1. For all claims and claim lines, we restricted observations to those marked final action (CLM_FINL_ACTN_IND='Y' and CLM_ LINE FINL_ACTN_IND='Y'). In addition, we used the institutional admission date variable CLM_ACTV_CARE_ FROM _DT from inpatient claims to assess potential overlap with ED claims. We excluded ED claims where for the same beneficiary, their ED claim through date (CLM_THRU_DT) overlapped with an inpatient admission date. Thus, our measure of ED visits excluded those that resulted in an inpatient admission. For all 2015 hospital inpatient and outpatient encounter data, we applied a 4-year runout period, through 12/31/2019 to ensure data completeness. While our study was only authorized to analyze encounters with service dates in 2015, we accessed the IDR encounter data in 2020, allowing us to use a longer runout period. For inpatient claims, we found that although the vast majority of claims were submitted within 2 years, a notable quantity were not submitted until 3 years later, even continuing into the 4th year. We reviewed data from two types of inpatient claim codes (values indicated by CLM_TYPE CD and description from CLM_TYPE CD_DESC): ¢ 4011 =011X Medicare Part C ENC Hospital Inpatient (Including Medicare Part A). ¢ 4041 =041X Medicare Part C ENC Religious Non-medical Health Care Institutions-- Hospital Inpatient. We did not find any claims for CLM_TYPE_CD = 4041. We also reviewed the following outpatient claim code types: * 4012 =012X Medicare Part C ENC Hospital Inpatient (Medicare Part B only). ¢ 4013 =013X Medicare Part C ENC Hospital Outpatient. * 4014=014X Medicare Part C ENC Hospital Laboratory Services Provided to Non- patients. ¢ 4085 = 085X Medicare Part C ENC Special Facility CAH Critical Access Hospital. For ED claims, we also restricted data to ED revenue center codes, where CLM_LINE_REV_CTR_CD = 045x or 0981 (0450, 0451, 0452, 0456, 0459, 0981). Note that we checked for values of 0453, 0454, 0455, 0457, 0458 but did not find those to be populated. Institutional/HCBS use. The monthly indicator, BENE DUAL INSTNL_STUS_IND_ SW, categorizes beneficiary months as 1 = Institutionalized, 2 = Not institutionalized, 3 = HCBS, and 9 = Unknown. Institutional use includes Medicaid-covered stays in an NF, intermediate care facility, or inpatient psychiatric hospital for the entire span of eligibility for a given month. HCBS use includes services delivered under a Section 1115 demonstration, under a 1915(c) or (d) waiver, under a state plan amendment under 1915(i), or through enrollment in a Medicaid managed care organization with a contract under Section 1903(m) or under Section 1932 of the Social Security Act. After examining this indicator for all months during the year, we categorized beneficiaries as having institutional use in at least 1 month (a small percentage of whom also used HCBS in at least 1 month), which constituted our institutional use outcome measure, and HCBS use in at least 1 month but no institutional use in any month, which constituted our HCBS use outcome measure. In addition, we were able to use the monthly variables BENE_LT_INSTNL_(MONTH)_RCNCLD_IND from the risk adjustment data to identify whether a beneficiary had any long-term institutional use in 2014, which we used as a covariate indicator for prior long-term institutional use for select multivariate models. Note this is a more restricted definition of institutional use than our outcome measure, as it is focused on long-term institutional use only. Mortality outcome. We used BENE_DEATH_DT from the V2. MDCR_BENE table to determine whether a beneficiary died in 2015. We found that several states (California, Oregon, and Utah) had death rates of less than 1% for our study population, which is far lower than expected. Thus, we excluded these states from the mortality analysis. Original reason for entitlement code (OREC). Although V2. MDCR_BENE FCT has an indicator for (OREC), we found this variable to have a high rate of missingness. Instead, we defined OREC using the variable BENE _MDCR_ENTLMT_RSN_CD from the IDR table V2_MDCR_BENE MDCR_ENTLMT RSN. A-3 Exhibit A-1. Selected variables and data source IDR table Variable Description Full-benefit dual eligibility (2015) V2_MDCR_BENE FCT BENE DUAL STUS_CD Monthly dual status code to indicate full-benefit dual eligibility V2_MDCR_BENE FCT BENE PTA STUS_CD Part A eligibility V2_MDCR_BENE FCT BENE PTB STUS CD Part B eligibility MA plan information (2015) V2_MDCR_BENE FCT CNTRCT_PBP_PTAB SK MA plan enrollment V2_MDCR_CNTRCT_PBP_NUM V2_MDCR _CNTRCT PBP_ NUM | CNTRCT_ SPCL_PLAN IND CD SNP indicator (D-SNP, FIDE-SNP) V2_MDCR_CNTRCT PBP_ NUM | CNTRCT_PBP TYPE CD Plan type indicator (PACE, MMP) Encounter data outcomes (2015) V2_MDCR_CLM V2_MDCR_CLM_LINE GEO_BENE SK CLM _DT_SGNTR_SK CLM_TYPE_CD CLM_NUM SK CLM_FROM DT 5-part key to identify unique beneficiary claims V2_MDCR_CLM V2_MDCR_CLM_LINE CLM_FINL_ACTN_IND CLM_LINE FINL_ACTN_ IND Final action claims header and line information V2_MDCR_CLM V2_MDCR_CLM_LINE CLM_TYPE_CD CLM_TYPE_CD DESC Indicates type of claim Used to identify inpatient and outpatient claims V2_MDCR_CLM_LINE CLM_LINE_REV_CTR CD Revenue center code V2_MDCR_CLM_LINE CLM_THRU_DT Claim through date V2_MDCR_CLM_DT_SGNTR CLM_ACTV_CARE FROM DT Date the beneficiary was admitted for an institutional claim Mortality outcome (2015) V2_MDCR_BENE | BENE_DEATH_ DT Death Institutional/HCBS outcome (2015) V2_MDCR_BENE FCT BENE DUAL INSTNL_STUS_ IND | Monthly indicator of _sw institutional or HCBS use Covariate used in multivariate model (2015 unless otherwise indicated) V2_MDCR_BENE_ FCT GEO_MDCD FIPS STATE CD State code V2_MDCR_BENE_FCT BENE_MDCR_ STUS_CD Current reason for Medicare entitlement V2_MDCR_BENE_FCT BENE_AGE CNT Continuous age (categorical used in multivariate model) V2_MDCR BENE FCT BENE_SEX_CD Sex Exhibit A-1 (continued) IDR table Variable Description V2_MDCR_BENE FCT BENE RACE CD Race V2_MDCR_BENE RISK PTC_F_ | BENE PTC_HCC X 2014 HCCs SCRE V2_MDCR_BENE_ RISK SCRE BENE_LT_INSTNL_ 2014 long-term institutional use (MONTH) RCNCLD_IND in any month, dichotomized to year in our analyses V2_MDCR_BENE ESRD DLYS | BENE_ESRD DLYS_TYPE CD ESRD dialysis in 2015 BENE_RNG BGN_DT BENE_RNG END DT Included in descriptive analysis only (2015 unless otherwise indicated) V2_MDCR_BENE RISK SCRE BENE_CMNTY_NUM 2014 community risk score V2_MDCR_BENE MDCR_ENTL | BENE MDCR ENTLMT_RSN CD _| Original reason for Medicare MT_RSN entitlement Source: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). Study population To identify the beneficiaries in our study population, we started by selecting beneficiaries with at least 1 month of full-benefit dual eligibility in 2015 (N = 8,431,292). We then restricted our population to beneficiaries who had Medicare Part A and B, were full-benefit dual eligible and who were consistently enrolled in a non-integrated MA plan or specific type of integrated care plan for all available months (N = 1,539,821). The vast majority of beneficiaries we excluded were not enrolled in an MA plan for all months or did not have full- benefit dual eligibility in all months. A small fraction of beneficiaries were excluded because they switched between integrated care and non-integrated care, or among integrated care plan types. After excluding beneficiaries enrolled in a MMP, we finalized the overall study sample (N = 1,337,900) used for the descriptive analyses. The study sample used for the multivariate analyses differed. First, since the HCCs were used as covariates for all models, the sample was restricted to beneficiaries with 2014 risk adjustment data (N = 1,196,829). Then we excluded a small number of other beneficiaries with missing covariates, leaving the final sample for the hospitalization, ED, and institutional models (N = 1,196,141). For the HCBS model, we excluded beneficiaries in PACE (N = 1,170,480). For the mortality model, we excluded beneficiaries in California, Utah, and Oregon (N = 936,833). APPENDIX B ADDITIONAL DESCRIPTIVE RESULTS In Appendix B, we provide additional descriptive results on various study populations. In Exhibit 1, we presented the full study population for our major descriptive statistics, along with a subset of measures from the 2014 risk adjustment data. In Exhibit B-1, we restricted the population to those with 2014 risk adjustment data, and present beneficiary HCCs. In Exhibit B- 2, we present the distribution by state for all plan types, for the full study population, and no restrictions applied. Finally, Exhibit B-3 presents the smaller population used for our multivariate analyses. It is restricted to everyone with 2014 risk adjustment data, as well as no other missing covariates. Exhibit B-1. Percentage of beneficiaries in each plan type with individual HCCs HCC ee D-SNP |FIDE-SNP| PACE | TOTAL N (beneficiaries with 2014 risk scores and HCCs) 393,402 687,813 89,949 25,665 | 1,196,829 HCC1: HIV/AIDS, % 0.67 1.22 0.35 0.26 0.95 HCC2: Septicemia, Sepsis, Systemic Inflammatory Response Syndrome/Shock, % 4.22 2.13 3.21 4.55 2.95 HCC6: Opportunistic Infection, % 0.32 0.31 0.29 0.38 0.31 HCC8: Metastatic Cancer and Acute Leukemia, % 0.62 0.52 0.71 0.71 0.57 HCC9: Lung and Other Severe Cancers, % 0.94 0.78 1.16 1.13 0.87 HCC10: Lymphoma and Other Cancers, % 1.03 0.80 1.15 1.21 0.91 HCC11: Colorectal, Bladder, and Other Cancers, % 1.59 1.29 1.89 1.95 1.45 HCC12: Breast, Prostate, and Other Cancers and 415 371 481 5 43 3.97 Tumors, % HCC17: Diabetes with Acute Complications, % 0.68 0.58 0.63 1.13 0.63 HCC18: Diabetes with Chronic Complications, % 24.21 20.07 23.56 33.77 21.99 HCC19: Diabetes without Complication, % 15.17 15.71 16.51 12.13 15.52 HCC21: Protein-Calorie Malnutrition, % 5.04 1.95 2.95 5.91 3.13 HCC22: Morbid Obesity, % 9.58 10.36 6.99 11.66 9.88 HCC23: Other Significant Endocrine and Metabolic 3.91 3.25 3.86 7.10 3.60 Disorders, % HCC27: End-Stage Liver Disease, % 0.52 0.58 0.50 0.73 0.55 HCC28: Cirrhosis of Liver, % 0.70 0.82 0.80 0.97 0.79 HCC29: Chronic Hepatitis, % 1.08 2.05 1.08 1.30 1.64 HCC33: Intestinal Obstruction/Perforation, % 2.16 1.45 1.86 2.71 1.74 HCC34: Chronic Pancreatitis, % 0.36 0.39 0.30 0.38 0.37 HCC35: Inflammatory Bowel Disease, % 0.90 0.75 0.79 0.91 0.80 HCC39: Bone/Joint/Muscle Infections/Necrosis, % 1.55 1.08 1.15 2.07 1.26 HCC40: Rheumatoid Arthritis and Inflammatory 6.52 6.65 6.21 6.97 6.58 Connective Tissue Disease, % HCC46: Severe Hematological Disorders, % 0.55 0.44 0.50 0.56 0.48 HCC47: Disorders of Immunity, % 1.20 1.11 1.22 1.47 1.16 HCC48: Coagulation Defects and Other Specified 5.26 3.72 458 6.96 436 Hematological Disorders, % HCCS54: Drug/Alcohol Psychosis, % 1.14 1.05 1.01 1.71 1.09 B-1 Exhibit B-1 (continued) Regular HCC MA D-SNP |FIDE-SNP| PACE | TOTAL HCCS55: Drug/Alcohol Dependence, % 4.28 4.82 3.24 5.51 4.53 HCC57: Schizophrenia, % 4.30 6.01 3.73 4.91 5.25 HCCS58: Major Depressive, Bipolar, and Paranoid 18.24 15.62 17.40 27.20 16.86 Disorders, % HCC70: Quadriplegia, % 0.94 0.30 0.62 0.52 0.54 HCC71: Paraplegia, % 0.57 0.38 0.49 0.56 0.45 HCC72: Spinal Cord Disorders/Injuries, % 1.04 0.84 0.97 1.46 0.93 HCC73: Amyotrophic Lateral Sclerosis and Other 0.09 0.04 0.09 0.11 0.06 Motor Neuron Disease, % HCC74: Cerebral Palsy, % 0.79 1.05 0.60 0.53 0.92 HCC75: Myasthenia Gravis/Myoneural Disorders and Guillain-Barre Syndrome/Inflammatory and 1.12 0.98 0.98 1.50 1.04 Toxic Neuropathy, % HCC76: Muscular Dystrophy, % 0.12 0.12 0.10 0.11 0.12 HCC77: Multiple Sclerosis, % 1.17 0.68 0.82 1.17 0.86 HCC78: Parkinson's and Huntington's Diseases, % 3.36 1.08 2.95 5.25 2.06 HCC79: Seizure Disorders and Convulsions, % 7.05 6.33 5.40 8.73 6.55 HCC80: Coma, Brain Compression/Anoxic 0.43 0.23 0.36 0.44 0.31 Damage, % HCC82 : Respirator Dependence/Tracheostomy 051 031 0.46 0.46 0.39 Status, % HCC83: Respiratory Arrest, % 0.05 0.03 0.05 0.10 0.04 HCC84: Cardio-Respiratory Failure and Shock, % 4.50 2.75 3.76 7.82 3.51 HCC85: Congestive Heart Failure, % 20.62 12.31 18.60 29.73 15.88 HCC86: Acute Myocardial Infarction, % 1.46 0.88 1.27 1.77 1.12 HCC87 : Unstable Angina and Other Acute Ischemic 2.24 1.96 2.34 2.85 2.10 Heart Disease, % HCC88: Angina Pectoris, % 4.44 4.15 3.97 7.33 4.30 HCC96: Specified Heart Arrhythmias, % 15.10 7.89 15.30 21.20 11.10 HCC99: Cerebral Hemorrhage, % 0.96 0.41 0.79 1.39 0.64 HCC100: Ischemic or Unspecified Stroke, % 6.33 3.29 5.23 8.65 4,55 HCC103: Hemiplegia/Hemiparesis, % 4.83 2.43 4.30 10.73 3.54 HCC104: Monoplegia, Other Paralytic 031 0.21 0.27 0.83 0.26 Syndromes, % HCC106: Atherosclerosis of the Extremities with 1.06 0.49 0.71 1.60 0.71 Ulceration or Gangrene, % HCC107: Vascular Disease with Complications, % 2.63 1.83 2.72 3.95 2.20 HCC108: Vascular Disease, % 29.17 18.18 24.96 39.90 22.77 HCC110: Cystic Fibrosis, % 0.02 0.02 0.01 0.01 0.02 HCC111: Chronic Obstructive Pulmonary 22.17 19.07 19.65 | 29.03 | 20.35 Disease, % HCC1 12: Fibrosis of Lung and Other Chronic Lung 0.74 0.70 0.93 1.05 0.74 Disorders, % HCCl1 14: Aspiration and Specified Bacterial 191 0.76 1.63 1.89 1.23 Pneumonias, % HCC115: Pneumococcal Pneumonia, Empyema, 0.42 0.28 0.36 0.55 0.34 Lung Abscess, % HCC122: Proliferative Diabetic Retinopathy and 1.72 143 1.66 2.85 1.57 Vitreous Hemorrhage, % HCC124: Exudative Macular Degeneration, % 1.43 0.66 1.86 2.34 1.04 Exhibit B-1 (continued) Regular HCC MA D-SNP |FIDE-SNP| PACE | TOTAL HCC134: Dialysis Status, % 1.18 0.71 0.63 2.13 0.89 HCC135: Acute Renal Failure, % 6.66 3.70 6.34 8.32 4.97 HCC136: Chronic Kidney Disease, Stage 5, % 0.75 0.67 0.68 1.34 0.71 1%. : Chronic Kidney Disease, Severe (Stage 1.22 0.76 1.38 3.04 1.01 HCC157: Pressure Ulcer of Skin with Necrosis Through to Muscle, Tendon, or Bone, % 0.41 0.11 0.23 0.29 0.22 HCC158: Pressure Ulcer of Skin with Full Thickness 0.87 0.20 0.50 1.16 0.46 Skin Loss, % HCC161: Chronic Ulcer of Skin, Except Pressure, % 3.82 2.16 3.58 4.99 2.87 HCC162: Severe Skin Burn or Condition, % 0.02 0.02 0.02 0.03 0.02 HCC166: Severe Head Injury, % 0.03 0.01 0.01 0.03 0.02 HCC167: Major Head Injury, % 1.10 0.71 1.02 1.62 0.88 HCC169: Vertebral Fractures without Spinal Cord 1.96 0.91 1.99 3.14 1.38 Injury, % HCC170: Hip Fracture/Dislocation, % 2.41 0.73 2.03 2.98 1.43 HCC173: Traumatic Amputations and 0.58 0.38 0.54 0.71 0.47 Complications, % HCC176: Complications of Specified Implanted 2.30 1.55 1.90 267 1.85 Device or Graft, % HCC186: Major Organ Transplant or Replacement 0.16 0.17 0.15 0.12 0.16 Status, % HCC188: Artificial Openings for Feeding or 11 0.91 1.82 1.95 1.40 Elimination, % HCC189: Amputation Status, Lower 1.12 0.74 0.93 1.84 0.90 Limb/Amputation Complications, % Source: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). B-3 Exhibit B-2. Percentage of beneficiaries located in each state, by plan type State Regular MA (%) | D-SNP(%) | FIDE-SNP(%) | PACE(%) | TOTAL (%) N (total population) 435,968 779,411 95,637 26,884 1,337,900 Alaska 0.00 0.00 0.00 0.00 0.00 Alabama 0.51 1.61 0.00 0.47 1.12 Arkansas 0.88 0.74 0.00 0.50 0.73 Arizona 2.06 7.94 6.22 0.00 5.75 California 25.68 16.70 9.75 11.61 19.03 Colorado 1.74 0.77 0.00 7.79 1.17 Connecticut 1.58 0.40 0.00 0.00 0.75 District of Columbia 0.10 0.19 0.00 0.00 0.14 Delaware 0.13 0.04 0.00 0.38 0.08 Florida 8.36 9.02 0.01 2.93 8.04 Georgia 2.52 2.19 0.01 0.00 2.10 Hawaii 0.54 2.17 0.00 0.00 1.44 Iowa 0.90 0.01 0.00 0.77 0.32 Idaho 0.46 0.00 1.31 0.00 0.25 Tilinois 2.23 0.80 0.01 0.00 1.19 Indiana 2.06 0.05 0.00 0.00 0.70 Kansas 0.50 0.00 0.00 0.93 0.18 Kentucky 0.57 0.24 0.00 0.00 0.32 Louisiana 0.81 1.19 0.00 0.99 0.98 Massachusetts 1.22 0.02 33.16 10.95 3.00 Maryland 0.81 0.26 0.00 0.47 0.42 Maine 0.39 0.07 0.00 0.00 0.17 Michigan 1.88 1.01 0.00 3.88 1.28 Minnesota 0.62 0.00 36.20 0.00 2.79 Missouri 2.00 1.00 0.00 0.51 1.25 Mississippi 0.21 0.57 0.00 0.00 0.40 Montana 0.08 0.00 0.00 0.00 0.03 North Carolina 2.40 1.56 0.00 3.14 1.75 North Dakota 0.06 0.00 0.00 0.31 0.03 Nebraska 0.44 0.00 0.00 0.22 0.15 New Hampshire 0.04 0.00 0.00 0.00 0.01 New Jersey 1.23 0.96 0.00 2.33 1.01 New Mexico 1.05 0.81 0.00 1.29 0.84 Nevada 0.48 0.00 0.00 0.00 0.16 New York 10.69 16.15 11.19 15.95 14.01 Ohio 2.40 0.38 0.00 0.81 1.02 Oklahoma 1.27 0.00 0.00 0.46 0.42 Oregon 2.23 2.50 0.00 3.12 2.25 Pennsylvania 5.00 11.53 0.00 15.69 8.66 Rhode Island 1.23 0.00 0.00 0.88 0.42 South Carolina 3.12 2.41 0.00 1.21 2.45 South Dakota 0.07 0.00 0.00 0.00 0.02 Tennessee 1.07 5.89 0.00 0.99 3.80 Texas 2.69 5.91 0.00 3.44 4.39 Utah 0.45 0.84 0.00 0.00 0.64 Virginia 0.95 0.07 0.00 4.01 0.43 Vermont 0.06 0.00 0.00 0.00 0.02 Exhibit B-2 (continued) State Regular MA (%) | D-SNP(%) | FIDE- SNP (%) | PACE (%) | TOTAL (%) Washington 1.98 2.25 0.00 1.66 1.99 Wisconsin 1.71 1.69 2.12 2.09 1.74 West Virginia 0.54 0.01 0.00 0.00 0.18 Wyoming 0.02 0.00 0.00 0.23 0.01 Source: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). B-5 Exhibit B-3. Characteristics of study population included in multivariate regression models, by plan type Characteristic Reguat | p-SNP_ | FIDE-SNP | PACE | TOTAL N 393,248 687,315 89,917 25,661 1,196,141 Outcome measures, 2015: 21.06 18.24 21.99 22.16 19.53 Any inpatient admission, % 21.06 18.24 21.99 22.16 19.53 Any ED visit, % 31.47 37.34 31.56 25.18 34.72 Institutionalized in at least 1 27.03 2.40 18.54 6.53 11.80 month, % HCBS in at least one month but not 17.49 15.58 34.55 t 17.70 institutionalized in any month, % Died during year*, % 14.02 3.62 10.09 13.40 7.61 Age, mean (SD) 73.30 (14.93) | 65.81 (15.26) | 77.49 (10.81) | 79.45 (9.85) | 68.51 (15.27) Age, grouped: <65, % 22.56 38.64 6.18 7.48 30.25 65-74, % 26.04 30.63 32.99 25.23 29.19 75-84, % 27.19 22,37 35.14 32.87 25.14 85+, % 24.20 8.35 25.68 34.41 15.43 Female, % 66.74 62.78 68.92 71.66 64.73 Race/ethnicity: White, non-Hispanic, % 62.11 46.19 62.30 59.76 52.93 Black, non-Hispanic, % 18.46 25.11 11.82 24.39 21.91 Hispanic, % 10.84 14.95 9.43 741 13.02 Asian, % 5.44 9.41 10.68 5.50 8.12 Other, % 3.15 4.34 5.77 2.94 4.02 Original reason for Medicare eligibility: Old age and survivors, % 62.36 47.73 74.23 68.04 54.97 Disability, % 37.19 51.84 25.64 31.37 44.62 ESRD, % 0.12 0.13 0.05 0.16 0.12 Both disability and ESRD, % 0.33 0.30 0.08 0.43 0.29 Current reason for Medicare eligibility: Aged without ESRD, % 76.47 60.89 93.14 90.59 69.07 Aged with ESRD, % 1.01 0.52 0.68 1.97 0.72 Disabled without ESRD, % 22.06 38.16 6.06 6.94 29.78 Disabled with ESRD, % 0.4 0.35 0.11 0.42 0.35 ESRD only, % 0.07 0.09 0.02 0.07 0.07 ESRD dialysis status for at least one 1.51 0.95 0.82 2.53 1.16 month in 2015, % N (beneficiaries with 2014 risk scores) 393,248 687,315 89,917 25,661 1,196,141 Community risk score, mean (SD) 1.66 (1.29) | 1.25 (0.99) | 1.59(1.15) | 2.15 (1.27) | 1.43 (1.14) Long-term institutional status for at least 22.80 1.21 14.95 7.67 9.48 one month in 2014, % Individual HCCs: HCC1: HIV/AIDS, % 0.67 1.22 0.35 0.26 0.95 HCC2: Septicemia, Sepsis, Systemic 4.22 2.13 3.21 4.56 2.95 Inflammatory Response Syndrome/Shock, % HCC6: Opportunistic Infection, % 0.32 0.31 0.29 0.38 0.31 HCC8: Metastatic Cancer and Acute 0.62 0.52 0.71 0.71 0.57 Leukemia, % HCC9: Lung and Other Severe 0.94 0.78 1.16 1.13 0.87 Cancers, % B-6 Exhibit B-3 (continued) Characteristic ae D-SNP | FIDE-SNP | PACE TOTAL HCC10: Lymphoma and Other 1.03 0.80 1.15 1.21 0.91 Cancers, % HCC11: Colorectal, Bladder, and 1.59 1.29 1.89 1.95 1.45 Other Cancers, % HCC12: Breast, Prostate, and Other 4.15 3.71 4.81 5.43 3.97 Cancers and Tumors, % HCC17: Diabetes with Acute 0.68 0.57 0.63 1.13 0.63 Complications, % HCC18: Diabetes with Chronic 24.22 20.08 23.57 33.77 21.99 Complications, % HCC19: Diabetes without 15.17 15.71 16.51 12.13 15.52 Complication, % HCC21: Protein-Calorie 5.04 1.95 2.95 5.91 3.13 Malnutrition, % HCC22: Morbid Obesity, % 9.58 10.36 6.99 11.66 9.88 HCC23: Other Significant Endocrine 3.91 3.25 3.86 7.10 3.60 and Metabolic Disorders, % HCC27: End-Stage Liver Disease, % 0.52 0.58 0.50 0.73 0.55 HCC28: Cirrhosis of Liver, % 0.70 0.82 0.80 0.97 0.79 HCC29: Chronic Hepatitis, % 1.08 2.06 1.08 1.30 1.64 HCC33: Intestinal 2.16 1.45 1.86 2.71 1.74 Obstruction/Perforation, % HCC34: Chronic Pancreatitis, % 0.36 0.39 0.30 0.38 0.37 HCC35: Inflammatory Bowel 0.90 0.75 0.79 0.91 0.80 Disease, % HCC39: Bone/Joint/Muscle 1.55 1.08 1.15 2.07 1.26 Infections/Necrosis, % HCC40: Rheumatoid Arthritis and 6.52 6.65 6.21 6.97 6.58 Inflammatory Connective Tissue Disease, % HCC46: Severe Hematological 0.55 0.44 0.50 0.56 0.48 Disorders, % HCC47: Disorders of Immunity, % 1,20 1.11 1.22 1.47 1.16 HCC48: Coagulation Defects and 5.26 3.72 4.58 6.96 4.36 Other Specified Hematological Disorders, % HCCS54: Drug/Alcohol Psychosis, % 1.14 1.05 1.01 1.71 1.09 HCCS55: Drug/Alcohol 4.27 4.82 3.24 5.51 4.54 Dependence, % HCC57: Schizophrenia, % 4.30 6.01 3.73 4.90 5.25 HCCS58: Major Depressive, Bipolar, 18.23 15.62 17.40 27.20 16.86 and Paranoid Disorders, % HCC70: Quadriplegia, % 0.94 0.30 0.62 0.52 0.54 HCC71: Paraplegia, % 0.56 0.38 0.49 0.56 0.45 HCC72: Spinal Cord 1.04 0.84 0.97 1.46 0.93 Disorders/Injuries, % HCC73: Amyotrophic Lateral 0.09 0.04 0.09 0.11 0.06 Sclerosis and Other Motor Neuron Disease, % HCC74: Cerebral Palsy, % 0.79 1.05 0.60 0.53 0.92 B-7 Exhibit B-3 (continued) Characteristic ae D-SNP_ | FIDE-SNP | PACE TOTAL HCC7S5: Myasthenia 1,12 0.98 0.98 1.50 1.04 Gravis/Myoneural Disorders and Guillain-Barre Syndrome/Inflammatory and Toxic Neuropathy, % HCC76: Muscular Dystrophy, % 0.12 0.12 0.10 0.11 0.12 HCC77: Multiple Sclerosis, % 1.17 0.68 0.82 1.17 0.86 HCC78: Parkinson's and Huntington's 3.36 1.08 2.95 5.25 2.06 Diseases, % HCC79: Seizure Disorders and 7.05 6.33 5.40 8.73 6.55 Convulsions, % HCC80: Coma, Brain 0.43 0.23 0.36 0.44 0.31 Compression/Anoxic Damage, % HCC82: Respirator 0.51 0.31 0.46 0.46 0.39 Dependence/Tracheostomy Status, % HCC83: Respiratory Arrest, % 0.05 0.03 0.05 0.10 0.04 HCC84: Cardio-Respiratory Failure 4.50 2.75 3.76 7.83 3.51 and Shock, % HCC85: Congestive Heart Failure, % 20.62 12.31 18.60 29.73 15.89 HCC86: Acute Myocardial 1.46 0.88 1.27 1.77 1.12 Infarction, % HCC87: Unstable Angina and Other 2.24 1.96 2.34 2.85 2.10 Acute Ischemic Heart Disease, % HCC88: Angina Pectoris, % 4.44 4.15 3.97 7.33 4.30 HCC96: Specified Heart 15.10 7.89 15.30 21.20 11.10 Arrhythmias, % HCC99: Cerebral Hemorrhage, % 0.96 0.41 0.79 1.40 0.64 HCC100: Ischemic or Unspecified 6.33 3.29 5.23 8.65 4.55 Stroke, % HCC103: Hemiplegia/Hemiparesis, % 4.83 2.43 4.30 10.74 3.54 HCC104: Monoplegia, Other 0.31 0.21 0.27 0.83 0.26 Paralytic Syndromes, % HCC106: Atherosclerosis of the 1.06 0.49 0.71 1.60 0.71 Extremities with Ulceration or Gangrene, % HCC107: Vascular Disease with 2.63 1.83 2.71 3.95 2.20 Complications, % HCC108: Vascular Disease, % 29.18 18.19 24.97 39.90 22.78 HCC110: Cystic Fibrosis, % 0.02 0.02 0.01 0.01 0.02 HCC111: Chronic Obstructive 22.17 19.08 19.65 29.03 20.35 Pulmonary Disease, % HCC112: Fibrosis of Lung and Other 0.74 0.70 0.93 1.05 0.74 Chronic Lung Disorders, % HCC114: Aspiration and Specified 1.91 0.76 1.63 1.89 1.23 Bacterial Pneumonias, % HCC115: Pneumococcal Pneumonia, 0.42 0.28 0.36 0.55 0.34 Empyema, Lung Abscess, % HCC122: Proliferative Diabetic 1.72 1.43 1.65 2.85 1.57 Retinopathy and Vitreous Hemorrhage, % B-8 Exhibit B-3 (continued) Regular Characteristic MA D-SNP FIDE-SNP PACE TOTAL HCC124: Exudative Macular 1.43 0.66 1.86 2.34 1.04 Degeneration, % HCC134: Dialysis Status, % 1.18 0.71 0.63 2.13 0.89 HCC135: Acute Renal Failure, % 6.67 3.70 6.34 8.32 4.98 HCC136: Chronic Kidney Disease, 0.75 0.67 0.68 1.34 0.71 Stage 5, % HCC137: Chronic Kidney Disease, 1,22 0.76 1.37 3.04 1.01 Severe (Stage 4), % HCC157: Pressure Ulcer of Skin with 0.41 0.11 0.23 0.29 0.22 Necrosis Through to Muscle, Tendon, or Bone, % HCC158: Pressure Ulcer of Skin with 0.87 0.20 0.50 1.16 0.46 Full Thickness Skin Loss, % HCC161: Chronic Ulcer of Skin, 3.82 2.16 3.58 4.99 2.87 Except Pressure, % HCC162: Severe Skin Burn or 0.02 0.02 0.02 0.03 0.02 Condition, % HCC166: Severe Head Injury, % 0.03 0.01 0.01 0.03 0.02 HCC167: Major Head Injury, % 1.10 0.71 1.02 1.63 0.88 HCC169: Vertebral Fractures without 1.96 0.91 1.99 3.13 1.38 Spinal Cord Injury, % HCC170: Hip Fracture/Dislocation, % 2.41 0.73 2.03 2.98 1.43 HCC173: Traumatic Amputations and 0.58 0.38 0.54 0.71 0.47 Complications, % HCC176: Complications of Specified 2.30 1.55 1.90 2.67 1.85 Implanted Device or Graft, % HCC186: Major Organ Transplant or 0.16 0.17 0.15 0.12 0.16 Replacement Status, % HCC188: Artificial Openings for 2.11 0.91 1.82 1.95 1.40 Feeding or Elimination, % HCC189: Amputation Status, Lower 1.12 0.74 0.93 1.84 0.90 Limb/Amputation Complications, % Count of HCCs, mean (SD) 2.87 (2.64) | 2.16 (2.16) | 2.60(2.44) | 3.84(2.72) | 2.46 (2.39) * Mortality rate excludes beneficiaries from California, Oregon, and Utah. + Percentage of beneficiaries in PACE with any HCBS is not reported because HCBS delivered by PACE are not under the various Medicaid waiver programs. Source: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). B-9 APPENDIX C FULL REGRESSION MODEL RESULTS Exhibit C-1. Full logistic regression model results predicting any inpatient hospitalization in 2015 Parameter Odds Ratio 95% Confidence Interval Plan type (reference = regular MA) D-SNP 0.970 *** 0.958 0.981 FIDE-SNP 1.24] *** 1.207 1.277 PACE 0.689 *** 0.667 0.713 Age group (reference = 65-74) < 65 0.996 0.981 1.012 75-84 1.258 *** 1.241 1.275 85+ 1.506 *** 1.481 1,531 Male 0.997 0.987 1.008 Race (reference = White) Black 1.000 0.987 1.013 Hispanic 0.806 *** 0.793 0.820 Asian 0.627 *** 0.613 0.642 Other race/ethnicity 0.808 *** 0.786 0.830 Long-term institutional use in 2014 0.499 *** 0.490 0.509 Proportion of months with data available in 2015 0.292 *** 0.283 0.300 ESRD patient with dialysis status 5.188 *** 4.904 5.490 HCC 1: HIV/AIDS 1,239 *** 1,182 1.299 HCC 2: Septicemia, Sepsis, Systemic Inflammatory Response 1.179 *** 1.148 1.211 Syndrome/Shock HCC 6: Opportunistic Infections 1.385 *** 1.288 1.490 HCC 8: Metastatic Cancer and Acute Leukemia 1.444 *** 1.367 1.525 HCC 9: Lung and Other Severe Cancers 1.395 *** 1.334 1.458 HCC 10: Lymphoma and Other Cancers 1.269 *** 1.213 1.328 HCC 11: Colorectal, Bladder, and Other Cancers 1.176 *** 1.134 1.220 HCC 12: Breast, Prostate, and Other Cancers and Tumors 1.034 ** 1.010 1.058 HCC 17: Diabetes with Acute Complications 1.805 *** 1.714 1.900 HCC 18: Diabetes with Chronic Complications 1.333 *** 1.316 1.349 HCC 19: Diabetes without Complication 1.215 *** 1,199 1,232 HCC 21: Protein-Calorie Malnutrition 0.935 *** 0.911 0.960 HCC 22: Morbid Obesity 1,220 *** 1.202 1.240 HCC 23: Other Significant Endocrine and Metabolic Disorders 1.119 *** 1.093 1.146 HCC 27: End-Stage Liver Disease 1.608 *** 1.521 1.700 HCC 28: Cirrhosis of Liver 1.369 *** 1.305 1.435 HCC 29: Chronic Hepatitis 1.190 *** 1.148 1.234 HCC 33: Intestinal Obstruction/Perforation 1.249 *** 1.209 1.291 HCC 34: Chronic Pancreatitis 1.753 *** 1.643 1.872 HCC 35: Inflammatory Bowel Disease 1.327 *** 1.265 1.391 HCC 39: Bone/Joint/Muscle Infections/Necrosis 1.304 *** 1.254 1.356 HCC 40: Rheumatoid Arthritis and Inflammatory Connective Tissue 1.166 *** 1.145 1.187 Disease HCC 46: Severe Hematological Disorders 1.577 *** 1.487 1.673 C-1 Exhibit C-1 (continued) 95% Confidence Parameter Odds Ratio I nterval HCC 47: Disorders of Immunity 1.204 *** 1.156 1.253 HCC 48: Coagulation Defects and Other Specified Hematological 1.128 *** 1.104 1.152 Disorders HCC 54: Drug/Alcohol Psychosis 1.925 *** 1.851 2.001 HCC 55: Drug/Alcohol Dependence 1.386 *** 1.357 1.416 HCC 57: Schizophrenia 1.452 *** 1.422 1.483 HCC 58: Major Depressive, Bipolar, and Paranoid Disorders 1.124 *** 1.109 1.139 HCC 70: Quadriplegia 1.183 _*** 1.112 1.259 HCC 71: Paraplegia 1.499 *** 1.407 1.598 HCC 72: Spinal Cord Disorders/Injuries 1.155 *** 1.103 1.209 HCC 73: Amyotrophic Lateral Sclerosis and Other Motor Neuron 1.375 *** 1.159 1.631 Disease HCC 74: Cerebral Palsy 0.948 0.899 1,001 HCC 75: Myasthenia Gravis/Myoneural Disorders and Guillain-Barre 1.075 *** 1.030 1.122 Syndrome/Inflammatory and Toxic Neuropathy HCC 76: Muscular Dystrophy 1.105 0.966 1.263 HCC 77: Multiple Sclerosis 1.463 *** 1.395 1.534 HCC 78: Parkinson's and Huntington's Diseases 1.239 *** 1.201 1.278 HCC 79: Seizure Disorders and Convulsions 1.265 *** 1,242 1.288 HCC 80: Coma, Brain Compression/Anoxic Damage 0.953 0.881 1.031 HCC 82: Respirator Dependence/Tracheostomy Status 1.225 *** 1.145 1.310 HCC 83: Respiratory Arrest 1.251 * 1.026 1.526 HCC 84: Cardio-Respiratory Failure and Shock 1,364 *** 1,332 1.397 HCC 85: Congestive Heart Failure 1,350 *** 1.332 1.368 HCC 86: Acute Myocardial Infarction 1.242 *** 1.194 1.292 HCC 87: Unstable Angina and Other Acute Ischemic Heart Disease 1.323 *** 1.285 1.362 HCC 88: Angina Pectoris 1.120 *** 1.096 1.145 HCC 96: Specified Heart Arrhythmias 1.282 *** 1.263 1.301 HCC 99: Cerebral Hemorrhage 1.078 * 1.017 1.142 HCC 100: Ischemic or Unspecified Stroke 1.237 *** 1.210 1.264 HCC 103: Hemiplegia/Hemiparesis 1.121 *** 1.092 1.150 HCC 104: Monoplegia, Other Paralytic Syndromes 1.216 *** 1.118 1.323 HCC 106: Atherosclerosis of the Extremities with Ulceration or 1.402 *** 1.333 1.476 Gangrene HCC 107: Vascular Disease with Complications 1.247 *** 1.211 1.284 HCC 108: Vascular Disease 1.078 *** 1.065 1.091 HCC 110: Cystic Fibrosis 2.886 *** 2.196 3.793 HCC 111: Chronic Obstructive Pulmonary Disease 1.491 *** 1.474 1.508 HCC 112: Fibrosis of Lung and Other Chronic Lung Disorders 1.291 *** 1.227 1.359 HCC 114: Aspiration and Specified Bacterial Pneumonias 0.983 0.945 1.023 HCC 115: Pneumococcal Pneumonia, Empyema, Lung Abscess 1.114 ** 1.038 1.195 HCC 122: Proliferative Diabetic Retinopathy and Vitreous Hemorrhage | 1.178 *** 1.137 1.220 HCC 124: Exudative Macular Degeneration 1.063 ** 1.017 1.111 HCC 134: Dialysis Status 0.635 _*** 0.595 0.678 HCC 135: Acute Renal Failure 1.349 *** 1.321 1.377 HCC 136: Chronic Kidney Disease, Stage 5 1.070 * 1.016 1.128 HCC 137: Chronic Kidney Disease, Severe (Stage 4) 1.311 *** 1.257 1.367 HCC 157: Pressure Ulcer of Skin with Necrosis Through to Muscle, 1.336 *** 1.221 1.461 Tendon, or Bone Exhibit C-1 (continued) 95% Confidence Parameter Odds Ratio I nterval HCC 158: Pressure Ulcer of Skin with Full Thickness Skin Loss 1.074 * 1.009 1.143 HCC 161: Chronic Ulcer of Skin, Except Pressure 1.279 *** 1.247 1.312 HCC 162: Severe Skin Burn or Condition 0.900 0.671 1.208 HCC 166: Severe Head Injury 0.823 0.596 1.138 HCC 167: Major Head Injury 1.102 *** 1.049 1.158 HCC 169: Vertebral Fractures without Spinal Cord Injury 1.246 *** 1,201 1,293 HCC 170: Hip Fracture/Dislocation 1.032 0.995 1.071 HCC 173: Traumatic Amputations and Complications 1.049 0.986 1.117 HCC 176: Complications of Specified Implanted Device or Graft 1.300 *** 1.259 1.342 HCC 186: Major Organ Transplant or Replacement Status 1.467 *** 1.325 1.624 HCC 188: Artificial Openings for Feeding or Elimination 1.362 *** 1.312 1.413 HCC 189: Amputation Status, Lower Limb/Amputation Complications 1.312 *** 1.255 1,372 State (reference = California) Alaska 2.036 0.636 6.519 Alabama 1.776 *** 1.698 1.858 Arkansas 2.180 *** 2.072 2.293 Arizona 1.572 *** 1.533 1.611 Colorado 1.378 *** 1.317 1.443 Connecticut 1.851 *** 1.757 1.951 District of Columbia 1.495 *** 1.316 1.698 Delaware 1.447 *** 1,224 1.709 Florida 1.477 *** 1.445 1.510 Georgia 1.599 *** 1.545 1.655 Hawaii 1.785 *** 1.708 1.865 Iowa 1.999 *** 1.852 2.158 Idaho 1.183 *** 1.073 1.304 Illinois 2.124 *** 2.036 2.216 Indiana 2.050 *** 1.945 2.161 Kansas 2.304 *** 2.093 2.537 Kentucky 2.545 *** 2.363 2.742 Louisiana 1.810 *** 1.724 1.901 Massachusetts 1.344 *** 1.296 1.394 Maryland 1.148 *** 1.063 1.240 Maine 1.286 *** 1.142 1.447 Michigan 1.707 *** 1.636 1.782 Minnesota 1.278 *** 1,229 1,329 Missouri 1.892 *** 1.817 1.970 Mississippi 2.039 *** 1.902 2.185 Montana 1.213 0.899 1.638 North Carolina 1.600 *** 1.541 1.660 North Dakota 0.359 *** 0.232 0.555 Nebraska 1.979 *** 1.771 2.213 New Hampshire 2.854 *** 2.028 4.016 New Jersey 1.586 *** 1.511 1.665 New Mexico 1.582 *** 1.498 1.671 Nevada 1.587 *** 1.419 1.776 New York 1.711 *** 1.679 1.744 Ohio 2.064 *** 1.975 2.156 Oklahoma 2.204 *** 2.063 2.354 C-3 Exhibit C-1 (continued) Parameter Odds Ratio 95% Confidence Interval Oregon 1.458 *** 1.407 1.511 Pennsylvania 1.857 *** 1.818 1.897 Rhode Island 1.562 *** 1.457 1.675 South Carolina 1.611 *** 1.559 1.664 South Dakota 0.520 ** 0.348 0.777 Tennessee 1.852 *** 1.801 1.904 Texas 1.621 *** 1.579 1.664 Utah 1.602 *** 1.507 1.703 Virginia 1.898 *** 1.776 2.029 Vermont 1.415 0.997 2.010 Washington 1.605 *** 1.549 1.662 Wisconsin 1.681 *** 1.620 1.744 West Virginia 2.356 *** 2.139 2.596 Wyoming 2.168 *** 1.445 3.253 */**/*** - Significantly different from regular MA plan based on a p-value cutoff of 0.05/0.01/0.001 NOTES: The model also included an interaction term between an indicator for a beneficiary who originally became eligible for Medicare because of disability and another indicator for being aged 65 or older in 2015. This interaction term is not shown in the table because the OR for an interaction term is not directly interpretable. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). Exhibit C-2. Full logistic regression model results predicting any ED visit 95% Confidence Parameter Odds Ratio I nterval Plan type (reference = regular MA) D-SNP 1.160 *** 1.149 1.172 FIDE-SNP 1.141 *** 1.113 1.170 PACE 0.523 *** 0.507 0.539 Age group (reference = 65-74) < 65 1.572 *** 1.553 1.591 75-84 1.078 *** 1.065 1.090 85+ 1.198 *** 1.181 1,215 Male 0.820 *** 0.813 0.828 Race (reference = White) Black 1.149 *** 1.137 1.161 Hispanic 0.899 *** 0.888 0.911 Asian 0.522 *** 0.512 0.532 Other race/ethnicity 0.763 *** 0.746 0.780 Long-term institutional use in 2014 0.377 *** 0.370 0.384 Proportion of months with data available in 2015 2.323 *** 2.247 2.401 ESRD patient with dialysis status 2.377 *** 2.255 2.506 HCC 1: HIV/AIDS 1.179 *** 1.133 1.226 HCC 2: Septicemia, Sepsis, Systemic Inflammatory Response 1.094 *** 1.066 1.123 Syndrome/Shock HCC 6: Opportunistic Infections 1.171 *** 1.093 1,255 HCC 8: Metastatic Cancer and Acute Leukemia 1.235 *** 1.173 1.300 HCC 9: Lung and Other Severe Cancers 1.149 *** 1.102 1.199 HCC 10: Lymphoma and Other Cancers 1.143 *** 1.098 1.191 HCC 11: Colorectal, Bladder, and Other Cancers 1.078 *** 1.043 1.114 HCC 12: Breast, Prostate, and Other Cancers and Tumors 1.099 *** 1.077 1.121 HCC 17: Diabetes with Acute Complications 1.527 *** 1.453 1.604 HCC 18: Diabetes with Chronic Complications 1.254 *** 1.241 1.267 HCC 19: Diabetes without Complication 1.175 *** 1.162 1.188 HCC 21: Protein-Calorie Malnutrition 0.904 *** 0.881 0.926 HCC 22: Morbid Obesity 1.167 *** 1.151 1.182 HCC 23: Other Significant Endocrine and Metabolic Disorders 1.096 *** 1.072 1.119 HCC 27: End-Stage Liver Disease 1.212 *** 1.150 1.278 HCC 28: Cirrhosis of Liver 1.222 *** 1.170 1.276 HCC 29: Chronic Hepatitis 1.219 *** 1.183 1,257 HCC 33: Intestinal Obstruction/Perforation 1.229 *** 1,192 1,267 HCC 34: Chronic Pancreatitis 1.612 *** 1.512 1.718 HCC 35: Inflammatory Bowel Disease 1.263 *** 1.210 1.319 HCC 39: Bone/Joint/Muscle Infections/Necrosis 1.086 *** 1.047 1.127 HCC 40: Rheumatoid Arthritis and Inflammatory Connective Tissue 1.216 *** 1.197 1.235 Disease HCC 46: Severe Hematological Disorders 1.255 *** 1.186 1.327 HCC 47: Disorders of Immunity 1.054 ** 1.016 1.094 HCC 48: Coagulation Defects and Other Specified Hematological 1.119 *** 1.097 1.141 Disorders HCC 54: Drug/Alcohol Psychosis 1.793 *** 1.727 1.861 HCC 55: Drug/Alcohol Dependence 1.356 *** 1.331 1.382 HCC 57: Schizophrenia 1.160 *** 1.139 1.180 HCC 58: Major Depressive, Bipolar, and Paranoid Disorders 1.258 *** 1.244 1.272 HCC 70: Quadriplegia 0.883 *** 0.833 0.936 c-5 Exhibit C-2 (continued) 95% Confidence Parameter Odds Ratio I nterval HCC 71: Paraplegia 1.034 0.975 1.097 HCC 72: Spinal Cord Disorders/Injuries 1.174 *** 1.127 1.222 HCC 73: Amyotrophic Lateral Sclerosis and Other Motor Neuron 1.057 0.897 1,245 Disease HCC 74: Cerebral Palsy 0.814 *** 0.781 0.849 HCC 75: Myasthenia Gravis/Myoneural Disorders and Guillain-Barre 1.079 *** 1.039 1.121 Syndrome/Inflammatory and Toxic Neuropathy HCC 76: Muscular Dystrophy 0.996 0.891 1.114 HCC 77: Multiple Sclerosis 1.071 ** 1.027 1.118 HCC 78: Parkinson's and Huntington's Diseases 1.241 *** 1.206 1.277 HCC 79: Seizure Disorders and Convulsions 1.317 *** 1.296 1.338 HCC 80: Coma, Brain Compression/Anoxic Damage 0.951 0.885 1.022 HCC 82: Respirator Dependence/Tracheostomy Status 0.942 0.883 1,005 HCC 83: Respiratory Arrest 1.009 0.833 1,223 HCC 84: Cardio-Respiratory Failure and Shock 1.061 *** 1.037 1.086 HCC 85: Congestive Heart Failure 1.119 *** 1.105 1.133 HCC 86: Acute Myocardial Infarction 1,231 *** 1.186 1.278 HCC 87: Unstable Angina and Other Acute Ischemic Heart Disease 1.468 *** 1.428 1.508 HCC 88: Angina Pectoris 1.219 *** 1.195 1,243 HCC 96: Specified Heart Arrhythmias 1.222 *** 1.205 1.239 HCC 99: Cerebral Hemorrhage 1.097 *** 1.041 1.157 HCC 100: Ischemic or Unspecified Stroke 1,198 *** 1.173 1,222 HCC 103: Hemiplegia/Hemiparesis 0.994 0.971 1.018 HCC 104: Monoplegia, Other Paralytic Syndromes 0.987 0.914 1.065 HCC 106: Atherosclerosis of the Extremities with Ulceration or 1.012 0.963 1.064 Gangrene HCC 107: Vascular Disease with Complications 1.184 *** 1.152 1.217 HCC 108: Vascular Disease 1.031 *** 1.021 1.042 HCC 110: Cystic Fibrosis 1.198 0.920 1.559 HCC 111: Chronic Obstructive Pulmonary Disease 1.322 *** 1.308 1.336 HCC 112: Fibrosis of Lung and Other Chronic Lung Disorders 1.232 *** 1.178 1.289 HCC 114: Aspiration and Specified Bacterial Pneumonias 0.940 ** 0.904 0.977 HCC 115: Pneumococcal Pneumonia, Empyema, Lung Abscess 1.034 0.966 1,105 HCC 122: Proliferative Diabetic Retinopathy and Vitreous Hemorrhage | 1.056 *** 1.023 1.090 HCC 124: Exudative Macular Degeneration 1.113 *** 1.070 1.158 HCC 134: Dialysis Status 0.901 *** 0.848 0.957 HCC 135: Acute Renal Failure 1.152 *** 1.130 1.175 HCC 136: Chronic Kidney Disease, Stage 5 0.920 *** 0.877 0.966 HCC 137: Chronic Kidney Disease, Severe (Stage 4) 1.028 0.988 1.070 HCC 157: Pressure Ulcer of Skin with Necrosis Through to Muscle, 0.955 0.873 1.044 Tendon, or Bone HCC 158: Pressure Ulcer of Skin with Full Thickness Skin Loss 0.908 ** 0.853 0.966 HCC 161: Chronic Ulcer of Skin, Except Pressure 1.105 *** 1.079 1.132 HCC 162: Severe Skin Burn or Condition 0.921 0.710 1.196 HCC 166: Severe Head Injury 1.144 0.867 1.508 HCC 167: Major Head Injury 1.241 *** 1.189 1.297 HCC 169: Vertebral Fractures without Spinal Cord Injury 1.234 *** 1.193 1.277 HCC 170: Hip Fracture/Dislocation 1.043 * 1.007 1.079 HCC 173: Traumatic Amputations and Complications 1.043 0.983 1.106 C-6 Exhibit C-2 (continued) 95% Confidence Parameter Odds Ratio I nterval HCC 176: Complications of Specified Implanted Device or Graft 1.306 *** 1.268 1.346 HCC 186: Major Organ Transplant or Replacement Status 1.071 0.974 1.178 HCC 188: Artificial Openings for Feeding or Elimination 1.315 *** 1.269 1.362 HCC 189: Amputation Status, Lower Limb/Amputation Complications 1.023 0.981 1.067 State (ref = California) Alaska 1.952 0.715 5.330 Alabama 1.499 *** 1.445 1.555 Arkansas 1.797 *** 1.720 1.878 Arizona 1.314 *** 1.289 1.340 Colorado 1.437 *** 1.384 1.491 Connecticut 1.794 *** 1.714 1.877 District of Columbia 1.217 *** 1.098 1.349 Delaware 1.168 * 1.008 1.354 Florida 0.998 0.980 1.015 Georgia 1.622 *** 1.577 1.668 Hawaii 1.434 *** 1.383 1.486 Iowa 1.957 *** 1.830 2.093 Idaho 1.693 *** 1.569 1.826 Illinois 1.484 *** 1.430 1.539 Indiana 1.800 *** 1.719 1.886 Kansas 1.616 *** 1.478 1.767 Kentucky 2.032 *** 1.900 2.174 Louisiana 1.813 *** 1.743 1.886 Massachusetts 1.416 *** 1,373 1.460 Maryland 0.975 0.913 1.041 Maine 2.026 *** 1.849 2.220 Michigan 1.533 *** 1.479 1.589 Minnesota 1.113 *** 1.076 1.152 Missouri 1.634 *** 1.579 1.692 Mississippi 1.761 *** 1.661 1.866 Montana 1.033 0.803 1.329 North Carolina 1.543 *** 1.496 1,591 North Dakota 0.435 *** 0.304 0.624 Nebraska 1.700 *** 1.544 1.873 New Hampshire 1.618 ** 1.141 2.294 New Jersey 1.302 *** 1.251 1.355 New Mexico 1.369 *** 1.310 1.430 Nevada 1.124 * 1.017 1.242 New York 1.077 *** 1.060 1.093 Ohio 1.882 *** 1.811 1.955 Oklahoma 1.801 *** 1.699 1.909 Oregon 1.279 *** 1.244 1.315 Pennsylvania 1.480 *** 1.455 1.505 Rhode Island 1,493 *** 1.404 1.587 South Carolina 1.814 *** 1.768 1.862 South Dakota 0.581 ** 0.415 0.813 Tennessee 1.650 *** 1.614 1.688 Texas 1.314 *** 1.287 1.342 Utah 1.204 *** 1.148 1.263 C-7 Exhibit C-2 (continued) 95% Confidence Parameter Odds Ratio I nterval Virginia 1.700 *** 1.603 1.803 Vermont 1.620 *** 1.220 2.150 Washington 1.416 *** 1.376 1.457 Wisconsin 1.596 *** 1.548 1.646 West Virginia 2.233 *** 2.044 2.440 Wyoming 1.973 *** 1.370 2.840 */**/*** - Significantly different from regular MA plan based on a p-value cutoff of 0.05/0.01/0.001 NOTES: The model also included an interaction term between an indicator for a beneficiary who originally became eligible for Medicare because of disability and another indicator for being aged 65 or older in 2015. This interaction term is not shown in the table because the OR for an interaction term is not directly interpretable. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). C-8 Exhibit C-3. Full logistic regression model results predicting any institutional use Parameter Odds Ratio 2070 Configence nterval Plan type (reference = regular MA) D-SNP 0.127 *** 0.124 0.129 FIDE-SNP 0.320 *** 0.308 0.332 PACE 0.062 *** 0.058 0.065 Age group (reference = 65-74) <65 0.348 *** 0.337 0.359 75-84 2.344 *** 2.294 2.395 85+ 6.392 *** 6.245 6.542 Male 1.101 *** 1.083 1.120 Race/ethnicity (reference = White) Black 0.477 *** 0.467 0.488 Hispanic 0.237 *** 0.228 0.246 Asian 0.307 *** 0.295 0.320 Other race/ethnicity 0.407 *** 0.388 0.427 Proportion of months with data available in 2015 0.306 *** 0.295 0.317 ESRD patient with dialysis status 1.872 *** 1.719 2.039 HCC 1: HIV/AIDS 0.921 0.811 1.047 HCC 2: Septicemia, Sepsis, Systemic Inflammatory Response 1.317 *** 1.269 1.366 Syndrome/Shock HCC 6: Opportunistic Infections 0.781 *** 0.689 0.885 HCC 8: Metastatic Cancer and Acute Leukemia 0.860 *** 0.789 0.938 HCC 9: Lung and Other Severe Cancers 0.796 *** 0.742 0.855 HCC 10: Lymphoma and Other Cancers 0.859 *** 0.802 0.921 HCC 11: Colorectal, Bladder, and Other Cancers 0.717 *** 0.678 0.758 HCC 12: Breast, Prostate, and Other Cancers and Tumors 0.800 *** 0.772 0.829 HCC 17: Diabetes with Acute Complications 1.394 *** 1.285 1.512 HCC 18: Diabetes with Chronic Complications 0.976 * 0.958 0.995 HCC 19: Diabetes without Complication 0.969 ** 0.949 0.990 HCC 21: Protein-Calorie Malnutrition 1.900 *** 1.840 1.962 HCC 22: Morbid Obesity 0.989 0.962 1.017 HCC 23: Other Significant Endocrine and Metabolic Disorders 0.861 *** 0.830 0.894 HCC 27: End-Stage Liver Disease 1.270 *** 1.156 1.396 HCC 28: Cirrhosis of Liver 1.069 0.981 1.164 HCC 29: Chronic Hepatitis 0.836 *** 0.768 0.911 HCC 33: Intestinal Obstruction/Perforation 0.938 ** 0.894 0.983 HCC 34: Chronic Pancreatitis 0.798 *** 0.705 0.903 HCC 35: Inflammatory Bowel Disease 0.807 *** 0.746 0.872 HCC 39: Bone/Joint/Muscle Infections/Necrosis 1.130 *** 1.066 1.197 HCC 40: Rheumatoid Arthritis and Inflammatory Connective Tissue 0.787 *** 0.764 0.812 Disease HCC 46: Severe Hematological Disorders 1.004 0.915 1.101 HCC 47: Disorders of Immunity 0.893 ** 0.836 0.955 HCC 48: Coagulation Defects and Other Specified Hematological 1.036 * 1.004 1.069 Disorders HCC 54: Drug/Alcohol Psychosis 1.405 *** 1,321 1.495 HCC 55: Drug/Alcohol Dependence 0.810 *** 0.777 0.845 HCC 57: Schizophrenia 3.335 *** 3.221 3.454 HCC 58: Major Depressive, Bipolar, and Paranoid Disorders 1.893 *** 1.856 1.930 HCC 70: Quadriplegia 3.773 *** 3.493 4.076 HCC 71: Paraplegia 2.034 *** 1.864 2.220 c-9 Exhibit C-3 (continued) Parameter Odds Ratio 270 Confieence nterval HCC 72: Spinal Cord Disorders/Injuries 1.257 *** 1.176 1.344 HCC 73: Amyotrophic Lateral Sclerosis and Other Motor Neuron 1.663 *** 1.342 2.060 Disease HCC 74: Cerebral Palsy 2.003 *** 1.858 2.159 HCC 75: Myasthenia Gravis/Myoneural Disorders and Guillain-Barre | 0.954 0.892 1.019 Syndrome/Inflammatory and Toxic Neuropathy HCC 76: Muscular Dystrophy 1.520 *** 1.241 1.863 HCC 77: Multiple Sclerosis 2.822 *** 2.650 3.006 HCC 78: Parkinson's and Huntington's Diseases 2.419 *** 2.334 2.506 HCC 79: Seizure Disorders and Convulsions 1.762 *** 1.714 1.811 HCC 80: Coma, Brain Compression/Anoxic Damage 1.477 *** 1.334 1.634 HCC 82: Respirator Dependence/Tracheostomy Status 1.143 ** 1.039 1.257 HCC 83: Respiratory Arrest 0.699 * 0.524 0.934 HCC 84: Cardio-Respiratory Failure and Shock 1.035 0.999 1.072 HCC 85: Congestive Heart Failure 1.175 *** 1.152 1.198 HCC 86: Acute Myocardial Infarction 0.799 *** 0.755 0.844 HCC 87: Unstable Angina and Other Acute Ischemic Heart Disease 0.709 *** 0.676 0.743 HCC 88: Angina Pectoris 0.636 *** 0.613 0.659 HCC 96: Specified Heart Arrhythmias 1.092 *** 1.070 1.114 HCC 99: Cerebral Hemorrhage 1.400 *** 1,302 1.506 HCC 100: Ischemic or Unspecified Stroke 1.560 *** 1.516 1.606 HCC 103: Hemiplegia/Hemiparesis 1.979 *** 1.916 2.045 HCC 104: Monoplegia, Other Paralytic Syndromes 1.526 *** 1.358 1.715 HCC 106: Atherosclerosis of the Extremities with Ulceration or 2.490 *** 2.328 2.663 Gangrene HCC 107: Vascular Disease with Complications 1.473 *** 1.412 1.536 HCC 108: Vascular Disease 2.063 *** 2.030 2.098 HCC 110: Cystic Fibrosis 0.841 0.444 1.593 HCC 111: Chronic Obstructive Pulmonary Disease 0.859 *** 0.844 0.875 HCC 112: Fibrosis of Lung and Other Chronic Lung Disorders 0.798 *** 0.733 0.868 HCC 114: Aspiration and Specified Bacterial Pneumonias 1.402 *** 1.332 1.475 HCC 115: Pneumococcal Pneumonia, Empyema, Lung Abscess 1.045 0.941 1.160 HCC 122: Proliferative Diabetic Retinopathy and Vitreous 0.950 0.896 1.007 Hemorrhage HCC 124: Exudative Macular Degeneration 0.844 *** 0.800 0.890 HCC 134: Dialysis Status 0.670 *** 0.605 0.742 HCC 135: Acute Renal Failure 1.137 *** 1.105 1.170 HCC 136: Chronic Kidney Disease, Stage 5 1.090 * 1.006 1.182 HCC 137: Chronic Kidney Disease, Severe (Stage 4) 1.007 0.948 1.068 HCC 157: Pressure Ulcer of Skin with Necrosis Through to Muscle, 2.134 *** 1.913 2.381 Tendon, or Bone HCC 158: Pressure Ulcer of Skin with Full Thickness Skin Loss 2.214 *** 2.057 2.384 HCC 161: Chronic Ulcer of Skin, Except Pressure 1.422 *** 1.374 1.472 HCC 162: Severe Skin Burn or Condition 1.076 0.698 1.659 HCC 166: Severe Head Injury 2.026 *** 1.368 3.000 HCC 167: Major Head Injury 1.400 *** 1.310 1.496 HCC 169: Vertebral Fractures without Spinal Cord Injury 1.198 *** 1.143 1.256 HCC 170: Hip Fracture/Dislocation 1.803 *** 1.729 1.881 HCC 173: Traumatic Amputations and Complications 0.994 0.910 1.086 c-10 Exhibit C-3 (continued) Parameter Odds Ratio ove Confiaence nterval HCC 176: Complications of Specified Implanted Device or Graft 0.905 *** 0.863 0.950 HCC 186: Major Organ Transplant or Replacement Status 0.749 ** 0.603 0.931 HCC 188: Artificial Openings for Feeding or Elimination 1.684 *** 1.605 1.768 HCC 189: Amputation Status, Lower Limb/Amputation 1.301 *** 1.221 1.386 Complications State (reference = California) Alaska 1.162 0.111 12.129 Alabama 4.234 *** 3.897 4.600 Arkansas 2.680 *** 2.468 2.909 Arizona 2.001 *** 1.912 2.094 Colorado 5.078 *** 4.803 5.368 Connecticut 5.281 *** 4.959 5.623 District of Columbia 1.218 0.918 1.617 Delaware 15.447 *** 12.611 18.920 Florida 0.583 *** 0.557 0.611 Georgia 4.131 *** 3.927 4.345 Hawaii 4.071 *** 3.772 4.393 Iowa 2.446 *** 2.232 2.680 Idaho 1.882 *** 1.675 2.114 Illinois 5.030 *** 4.761 5.314 Indiana 3.861 *** 3.633 4.103 Kansas 4.451 *** 3.990 4.965 Kentucky 7.303 *** 6.636 8.038 Louisiana 6.156 *** 5.718 6.628 Massachusetts 3.122 *** 2.974 3.277 Maryland 17.185 _*** 15.776 18.720 Maine 2.113 *** 1.856 2.406 Michigan 4.230 *** 3.991 4.482 Minnesota 5.689 *** 5.419 5.973 Missouri 2.397 *** 2.256 2.548 Mississippi 2.112 *** 1.809 2.467 Montana 4.450 *** 3.439 5.759 North Carolina 3.647 *** 3.460 3.844 North Dakota 8.471 *** 6.495 11.048 Nebraska 2.087 *** 1.812 2.404 New Hampshire 8.369 *** 5.679 12,332 New Jersey 2.918 *** 2.718 3.133 New Mexico 1.766 *** 1.613 1.934 Nevada 0.695 *** 0.594 0.814 New York 3.307 *** 3.208 3.409 Ohio 5.355 *** 5.091 5.634 Oklahoma 1.577 *** 1.442 1.725 Oregon 1.298 *** 1.222 1.379 Pennsylvania 4.494 *** 4.342 4.650 Rhode Island 4.019 *** 3.741 4.318 South Carolina 1.114 ** 1.043 1.191 South Dakota 8.594 *** 6.580 11,225 Tennessee 2.281 *** 2.153 2.417 Texas 2.861 *** 2.738 2.991 C-11 Exhibit C-3 (continued) Parameter Odds Ratio 279 Configence nterval Utah 1.753 *** 1.551 1.982 Virginia 2.081 *** 1.903 2.274 Vermont 2.491 *** 1.742 3.564 Washington 2.287 *** 2.169 2.412 Wisconsin 4.281 *** 4.067 4.505 West Virginia 5.599 *** 5.057 6.199 Wyoming 2.074 * 1.180 3.644 */#*/*** = Significantly different from regular MA plan based on a p-value cutoff of 0.05/0.01/0.001 NOTES: The model also included an interaction term between an indicator for a beneficiary who originally became eligible for Medicare because of disability and another indicator for being aged 65 or older in 2015. This interaction term is not shown in the table because the OR for an interaction term is not directly interpretable. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). C-12 Exhibit C-4. Full logistic regression model results predicting any HCBS use 95% Confidence Parameter Odds Ratio I nterval Plan type (reference = regular MA) D-SNP 1.046 *** 1.033 1.060 FIDE-SNP 4.223 *** 4.102 4.347 Age group (reference = 65-74) < 65 1.651 *** 1.623 1.679 75-84 1.681 *** 1.655 1.707 85+ 2.380 *** 2.337 2.424 Male 0.956 *** 0.945 0.967 Race/ethnicity (reference = White) Black 1.152 *** 1.136 1.169 Hispanic 0.802 *** 0.787 0.818 Asian 1.177 *** 1.150 1.205 Other race/ethnicity 1.048 ** 1.019 1.077 Proportion of months with data available in 2015 1.376 *** 1.329 1.425 ESRD patient with dialysis status 1.254 *** 1.178 1,335 HCC 1: HIV/AIDS 1.434 *** 1.363 1.507 HCC 2: Septicemia, Sepsis, Systemic Inflammatory Response 0.966 * 0.937 0.997 Syndrome/Shock HCC 6: Opportunistic Infections 1.236 *** 1.138 1,342 HCC 8: Metastatic Cancer and Acute Leukemia 1.416 *** 1.332 1.505 HCC 9: Lung and Other Severe Cancers 1.314 *** 1.249 1.381 HCC 10: Lymphoma and Other Cancers 1.130 *** 1.074 1.189 HCC 11: Colorectal, Bladder, and Other Cancers 1.107 *** 1.062 1.153 HCC 12: Breast, Prostate, and Other Cancers and Tumors 1.061 *** 1.034 1.088 HCC 17: Diabetes with Acute Complications 1.076 * 1.011 1.145 HCC 18: Diabetes with Chronic Complications 1.264 *** 1.246 1.281 HCC 19: Diabetes without Complication 1.060 *** 1.044 1.075 HCC 21: Protein-Calorie Malnutrition 0.823 *** 0.799 0.849 HCC 22: Morbid Obesity 1,253 *** 1,232 1,275 HCC 23: Other Significant Endocrine and Metabolic Disorders 1.157 *** 1.126 1.188 HCC 27: End-Stage Liver Disease 1.078 * 1.008 1.154 HCC 28: Cirrhosis of Liver 0.994 0.937 1.054 HCC 29: Chronic Hepatitis 0.811 *** 0.775 0.848 HCC 33: Intestinal Obstruction/Perforation 1.082 *** 1.042 1.123 HCC 34: Chronic Pancreatitis 1.017 0.937 1.104 HCC 35: Inflammatory Bowel Disease 0.980 0.927 1.036 HCC 39: Bone/Joint/Muscle Infections/Necrosis 1.036 0.990 1.083 HCC 40: Rheumatoid Arthritis and Inflammatory Connective Tissue 1.149 *** 1.126 1.172 Disease HCC 46: Severe Hematological Disorders 1.009 0.941 1.081 HCC 47: Disorders of Immunity 1.140 *** 1.089 1.194 HCC 48: Coagulation Defects and Other Specified Hematological 1.050 *** 1.025 1.077 Disorders HCC 54: Drug/Alcohol Psychosis 0.784 *** 0.746 0.825 HCC 55: Drug/Alcohol Dependence 0.812 *** 0.790 0.835 HCC 57: Schizophrenia 0.771 *** 0.752 0.790 HCC 58: Major Depressive, Bipolar, and Paranoid Disorders 0.903 *** 0.890 0.916 HCC 70: Quadriplegia 1.663 *** 1.565 1.766 HCC 71: Paraplegia 2.767 *** 2.594 2.950 HCC 72: Spinal Cord Disorders/Injuries 1.720 *** 1.642 1.801 C-13 Exhibit C-4 (continued) Parameter Odds Ratio 2970 Konfigence nterval HCC 73: Amyotrophic Lateral Sclerosis and Other Motor Neuron 2.047 *** 1.720 2.435 Disease HCC 74: Cerebral Palsy 6.629 *** 6.345 6.926 HCC 75: Myasthenia Gravis/Myoneural Disorders and Guillain-Barre 1.184 *** 1.130 1.240 Syndrome/Inflammatory and Toxic Neuropathy HCC 76: Muscular Dystrophy 2.558 *** 2.265 2.890 HCC 77: Multiple Sclerosis 1.736 *** 1.656 1.820 HCC 78: Parkinson's and Huntington's Diseases 1.378 *** 1.334 1.424 HCC 79: Seizure Disorders and Convulsions 1.726 *** 1.694 1.759 HCC 80: Coma, Brain Compression/Anoxic Damage 1.025 0.944 1.114 HCC 82: Respirator Dependence/Tracheostomy Status 1.261 *** 1.171 1.359 HCC 83: Respiratory Arrest 1.584 *** 1.273 1.970 HCC 84: Cardio-Respiratory Failure and Shock 1.227 *** 1.194 1.262 HCC 85: Congestive Heart Failure 1.212 *** 1.193 1,231 HCC 86: Acute Myocardial Infarction 0.989 0.944 1.035 HCC 87: Unstable Angina and Other Acute Ischemic Heart Disease 1.111 *** 1.074 1.149 HCC 88: Angina Pectoris 1,122 *** 1.094 1.151 HCC 96: Specified Heart Arrhythmias 1.022 * 1.005 1.039 HCC 99: Cerebral Hemorrhage 0.921 ** 0.865 0.980 HCC 100: Ischemic or Unspecified Stroke 1,189 *** 1.161 1.217 HCC 103: Hemiplegia/Hemiparesis 1.586 *** 1.544 1.629 HCC 104: Monoplegia, Other Paralytic Syndromes 1.513 *** 1.384 1.653 HCC 106: Atherosclerosis of the Extremities with Ulceration or 1.149 *** 1.085 1.217 Gangrene HCC 107: Vascular Disease with Complications 1.084 *** 1.049 1.120 HCC 108: Vascular Disease 0.917 *** 0.905 0.929 HCC 110: Cystic Fibrosis 0.928 0.647 1.333 HCC 111: Chronic Obstructive Pulmonary Disease 1.088 *** 1.074 1.103 HCC 112: Fibrosis of Lung and Other Chronic Lung Disorders 1.121 *** 1.059 1.187 HCC 114: Aspiration and Specified Bacterial Pneumonias 0.825 *** 0.789 0.863 HCC 115: Pneumococcal Pneumonia, Empyema, Lung Abscess 0.959 0.883 1.042 HCC 122: Proliferative Diabetic Retinopathy and Vitreous 1.226 *** 1.178 1.276 Hemorrhage HCC 124: Exudative Macular Degeneration 1.189 *** 1.135 1.245 HCC 134: Dialysis Status 1.112 ** 1.034 1.197 HCC 135: Acute Renal Failure 1.162 *** 1.135 1.189 HCC 136: Chronic Kidney Disease, Stage 5 1.034 0.974 1.099 HCC 137: Chronic Kidney Disease, Severe (Stage 4) 1.219 *** 1.162 1.280 HCC 157: Pressure Ulcer of Skin with Necrosis Through to Muscle, 0.838 *** 0.759 0.926 Tendon, or Bone HCC 158: Pressure Ulcer of Skin with Full Thickness Skin Loss 0.795 *** 0.741 0.852 HCC 161: Chronic Ulcer of Skin, Except Pressure 1.213 *** 1.179 1.247 HCC 162: Severe Skin Burn or Condition 1.217 0.894 1.658 HCC 166: Severe Head Injury 1.044 0.751 1.451 HCC 167: Major Head Injury 1.273 *** 1.210 1.340 HCC 169: Vertebral Fractures without Spinal Cord Injury 1.172 *** 1.124 1.222 HCC 170: Hip Fracture/Dislocation 1.030 0.990 1.072 HCC 173: Traumatic Amputations and Complications 1.205 *** 1.126 1.290 HCC 176: Complications of Specified Implanted Device or Graft 1.165 *** 1.124 1.207 C-14 Exhibit C-4 (continued) Parameter Odds Ratio 2970 Ronigence nterval HCC 186: Major Organ Transplant or Replacement Status 0.879 * 0.777 0.995 HCC 188: Artificial Openings for Feeding or Elimination 1.036 0.994 1.080 HCC 189: Amputation Status, Lower Limb/Amputation Complications | 1.480 *** 1.410 1.554 State (reference = California) Alaska 3.558 0.806 15.711 Alabama 2.427 *** 2.281 2.582 Arkansas 3.558 *** 3.331 3.801 Arizona 4.791 *** 4.646 4.942 Colorado 14.125 *** 13.518 14.760 Connecticut 7.450 *** 7.057 7.865 District of Columbia 5.813 *** 5.128 6.591 Delaware 2.157 *** 1.723 2.700 Florida 4.164 *** 4.040 4.292 Georgia 5.490 *** 5.279 5.710 Hawaii 4.011 *** 3.815 4.216 Iowa 11.442 *** 10.617 12.331 Idaho 11.165 *** 10.288 12.115 Illinois 10.119 *** 9.685 10.572 Indiana 4.481 *** 4.206 4.775 Kansas 11.458 *** 10.398 12.627 Kentucky 2.721 *** 2.446 3.028 Louisiana 2.390 *** 2.232 2.559 Massachusetts 0.813 *** 0.775 0.852 Maryland 1.764 *** 1.594 1.953 Maine 1.169 0.946 1.445 Michigan 1.816 *** 1.696 1.943 Minnesota 6.175 *** 5.938 6.422 Missouri 6.542 *** 6.249 6.848 Mississippi 11.409 *** 10.685 12.182 Montana 6.568 *** 5.029 8.577 North Carolina 1.036 0.965 1.112 North Dakota 1.084 0.618 1.903 Nebraska 3.372 *** 2.911 3.907 New Hampshire 6.682 *** 4.599 9.709 New Jersey 1,558 *** 1.437 1.690 New Mexico 21.736 *** 20.721 22.801 Nevada 9.443 *** 8.501 10.488 New York 7.893 *** 7.694 8.097 Ohio 13.881 *** 13.288 14.499 Oklahoma 9.431 *** 8.810 10.094 Oregon 23.493 *** 22.710 24.304 Pennsylvania 6.287 *** 6.113 6.466 Rhode Island 3.781 *** 3.484 4.103 South Carolina 4.780 *** 4.598 4.970 South Dakota 5.689 *** 4.283 7.558 Tennessee 2.766 *** 2.660 2.876 Texas 4.034 *** 3.898 4.175 Utah 5.655 *** 5.304 6.029 Virginia 11.369 *** 10.601 12,192 C-15 Exhibit C-4 (continued) Parameter Odds Ratio 2070 Ronigence nterval Vermont 7.628 *** 5.461 10.654 Washington 13.084 *** 12.628 13.557 Wisconsin 10.317 *** 9.930 10.718 West Virginia 5.679 *** 5.101 6.323 Wyoming 15.508 *** 9.536 25.222 */**/*** - Significantly different from regular MA plan based on a p-value cutoff of 0.05/0.01/0.001 NOTES: The model also included an interaction term between an indicator for a beneficiary who originally became eligible for Medicare because of disability and another indicator for being aged 65 or older in 2015. This interaction term is not shown in the table because the OR for an interaction term is not directly interpretable. The model excluded beneficiaries in PACE. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). C-16 Exhibit C-5. Full logistic model results predicting mortality Parameter Odds Ratio 2970 Eonfigence nterval Plan type (reference = regular MA) D-SNP 0.578 *** 0.565 0.591 FIDE-SNP 0.694 *** 0.663 0.728 PACE 0.958 0.917 1.002 Age group (reference = 65-74) < 65 0.526 *** 0.509 0.545 75-84 1.652 *** 1.611 1.694 85+ 3.298 *** 3.210 3.389 Male 1.354 *** 1.329 1.379 Race/ethnicity (reference = White) Black 0.757 *** 0.740 0.775 Hispanic 0.530 *** 0.510 0.551 Asian 0.449 *** 0.425 0.473 Other race/ethnicity 0.623 *** 0.589 0.659 Long-term institutional use in 2014 2.310 *** 2.258 2.362 ESRD patient with dialysis status 2.336 *** 2.156 2.532 HCC 1: HIV/AIDS 1.241 *** 1.113 1.385 HCC 2: Septicemia, Sepsis, Systemic Inflammatory Response 0.946 ** 0.909 0.983 Syndrome/Shock HCC 6: Opportunistic Infections 1,117 0.993 1.256 HCC 8: Metastatic Cancer and Acute Leukemia 4,539 *** 4.245 4.853 HCC 9: Lung and Other Severe Cancers 2.287 *** 2.154 2.427 HCC 10: Lymphoma and Other Cancers 1.353 *** 1.260 1.453 HCC 11: Colorectal, Bladder, and Other Cancers 1.324 *** 1.253 1.399 HCC 12: Breast, Prostate, and Other Cancers and Tumors 1.026 0.987 1.067 HCC 17: Diabetes with Acute Complications 1.221 *** 1.118 1.332 HCC 18: Diabetes with Chronic Complications 1.045 *** 1.023 1.068 HCC 19: Diabetes without Complication 0.997 0.974 1.022 HCC 21: Protein-Calorie Malnutrition 1.360 *** 1.315 1.406 HCC 22: Morbid Obesity 0.859 *** 0.832 0.887 HCC 23: Other Significant Endocrine and Metabolic Disorders 1.049 * 1.009 1.091 HCC 27: End-Stage Liver Disease 2.397 *** 2.208 2.603 HCC 28: Cirrhosis of Liver 1.716 *** 1.587 1.856 HCC 29: Chronic Hepatitis 1.295 *** 1.197 1.401 HCC 33: Intestinal Obstruction/Perforation 0.915 *** 0.870 0.963 HCC 34: Chronic Pancreatitis 1.404 *** 1.256 1.569 HCC 35: Inflammatory Bowel Disease 0.929 0.853 1.012 HCC 39: Bone/Joint/Muscle Infections/Necrosis 0.984 0.923 1.048 HCC 40: Rheumatoid Arthritis and Inflammatory Connective Tissue 0.986 0.954 1.019 Disease HCC 46: Severe Hematological Disorders 1.235 *** 1.129 1.351 HCC 47: Disorders of Immunity 1.418 *** 1.334 1.508 HCC 48: Coagulation Defects and Other Specified Hematological 1.070 *** 1.035 1.106 Disorders HCC 54: Drug/Alcohol Psychosis 1.079 * 1.005 1.159 HCC 55: Drug/Alcohol Dependence 1.107 *** 1.058 1.158 HCC 57: Schizophrenia 0.854 *** 0.815 0.895 HCC 58: Major Depressive, Bipolar, and Paranoid Disorders 0.912 *** 0.891 0.933 HCC 70: Quadriplegia 1.474 *** 1,363 1.595 HCC 71: Paraplegia 1.026 0.914 1.151 C-17 Exhibit C-5 (continued) Parameter Odds Ratio 2070 Ronigence nterval HCC 72: Spinal Cord Disorders/Injuries 1.007 0.930 1.091 HCC 73: Amyotrophic Lateral Sclerosis and Other Motor Neuron 2.074 *** 1.656 2.598 Disease HCC 74: Cerebral Palsy 0.772 *** 0.681 0.875 HCC 75: Myasthenia Gravis/Myoneural Disorders and Guillain-Barre 1.004 0.934 1.078 Syndrome/Inflammatory and Toxic Neuropathy HCC 76: Muscular Dystrophy 1,395 ** 1.103 1.765 HCC 77: Multiple Sclerosis 1.069 0.977 1.169 HCC 78: Parkinson's and Huntington's Diseases 1.428 *** 1.373 1.485 HCC 79: Seizure Disorders and Convulsions 1.057 *** 1.024 1.092 HCC 80: Coma, Brain Compression/Anoxic Damage 0.893 0.793 1.006 HCC 82: Respirator Dependence/Tracheostomy Status 1.221 *** 1.106 1,349 HCC 83: Respiratory Arrest 1.114 0.832 1.493 HCC 84: Cardio-Respiratory Failure and Shock 1.427 *** 1.379 1.477 HCC 85: Congestive Heart Failure 1.353 *** 1.326 1.382 HCC 86: Acute Myocardial Infarction 1.128 *** 1.067 1.192 HCC 87: Unstable Angina and Other Acute Ischemic Heart Disease 0.977 0.930 1.026 HCC 88: Angina Pectoris 0.918 *** 0.883 0.954 HCC 96: Specified Heart Arrhythmias 1,253 *** 1.227 1.280 HCC 99: Cerebral Hemorrhage 1.007 0.928 1.093 HCC 100: Ischemic or Unspecified Stroke 1,107 *** 1.072 1.143 HCC 103: Hemiplegia/Hemiparesis 1.069 *** 1.028 1.110 HCC 104: Monoplegia, Other Paralytic Syndromes 1.027 0.895 1.179 HCC 106: Atherosclerosis of the Extremities with Ulceration or 1.403 *** 1.307 1.505 Gangrene HCC 107: Vascular Disease with Complications 1.045 0.998 1.094 HCC 108: Vascular Disease 1.113 *** 1.092 1.135 HCC 110: Cystic Fibrosis 1.789 * 1.047 3.059 HCC 111: Chronic Obstructive Pulmonary Disease 1.375 *** 1.349 1.403 HCC 112: Fibrosis of Lung and Other Chronic Lung Disorders 1.256 *** 1.153 1.370 HCC 114: Aspiration and Specified Bacterial Pneumonias 1.128 *** 1.073 1.186 HCC 115: Pneumococcal Pneumonia, Empyema, Lung Abscess 1.276 *** 1.153 1.412 HCC 122: Proliferative Diabetic Retinopathy and Vitreous 1.097 ** 1.029 1.169 Hemorrhage HCC 124: Exudative Macular Degeneration 0.954 0.899 1.013 HCC 134: Dialysis Status 1.193 *** 1.083 1.314 HCC 135: Acute Renal Failure 1.257 *** 1.220 1.295 HCC 136: Chronic Kidney Disease, Stage 5 1.359 *** 1.254 1.474 HCC 137: Chronic Kidney Disease, Severe (Stage 4) 1.610 *** 1.518 1.707 HCC 157: Pressure Ulcer of Skin with Necrosis Through to Muscle, 1.356 *** 1.213 1.516 Tendon, or Bone HCC 158: Pressure Ulcer of Skin with Full Thickness Skin Loss 1.218 *** 1,132 1.311 HCC 161: Chronic Ulcer of Skin, Except Pressure 1.285 *** 1.238 1.334 HCC 162: Severe Skin Burn or Condition 0.958 0.594 1.544 HCC 166: Severe Head Injury 0.806 0.495 1.313 HCC 167: Major Head Injury 0.990 0.916 1.071 HCC 169: Vertebral Fractures without Spinal Cord Injury 1.095 *** 1.038 1.154 HCC 170: Hip Fracture/Dislocation 0.989 0.944 1.036 HCC 173: Traumatic Amputations and Complications 1.135 ** 1.036 1.244 c-18 Exhibit C-5 (continued) Parameter Odds Ratio 2970 Konfigence nterval HCC 176: Complications of Specified Implanted Device or Graft 0.905 *** 0.859 0.953 HCC 186: Major Organ Transplant or Replacement Status 0.967 0.804 1.162 HCC 188: Artificial Openings for Feeding or Elimination 1.196 *** 1.136 1.260 HCC 189: Amputation Status, Lower Limb/Amputation Complications | 1.379 *** 1.289 1.474 State (reference = New York) Alaska 1.072 0.126 9.133 Alabama 1.098 * 1.005 1.201 Arkansas 1.482 *** 1.365 1.610 Arizona 1.028 0.983 1.075 Colorado 1.317 *** 1.242 1.397 Connecticut 0.916 * 0.854 0.983 District of Columbia 1.245 * 1.005 1,542 Delaware 0.949 0.776 1.160 Florida 0.812 *** 0.781 0.845 Georgia 1.211 *** 1.147 1.279 Hawaii 1.468 *** 1.359 1.586 Iowa 0.984 0.877 1.103 Idaho 1.560 *** 1.375 1.769 Illinois 1.134 *** 1.063 1.210 Indiana 1.254 *** 1.166 1.349 Kansas 1.417 *** 1.251 1.606 Kentucky 1.494 *** 1,345 1.660 Louisiana 1.342 *** 1,238 1.455 Massachusetts 1.044 0.990 1,100 Maryland 1.158 *** 1.064 1.260 Maine 1.523 *** 1.326 1.749 Michigan 1.540 *** 1.452 1.634 Minnesota 1.543 *** 1.460 1.630 Missouri 1.236 *** 1.158 1.319 Mississippi 1.451 *** 1.277 1.650 Montana 1.450 * 1.059 1.984 North Carolina 1.392 *** 1.319 1.468 North Dakota 1.203 0.911 1.590 Nebraska 1.007 0.845 1.199 New Hampshire 1.808 ** 1,252 2.611 New Jersey 0.894 ** 0.827 0.965 New Mexico 1.253 *** 1.149 1.366 Nevada 1.123 0.961 1.312 Ohio 0.191 *** 0.172 0.211 Oklahoma 1.163 ** 1.052 1.286 Pennsylvania 1.200 *** 1.161 1.241 Rhode Island 0.864 *** 0.796 0.938 South Carolina 1.160 *** 1.095 1.228 South Dakota 1.506 ** 1.146 1.980 Tennessee 1.506 *** 1.435 1.580 Texas 1.396 *** 1.339 1.456 Virginia 1.504 *** 1.377 1.642 Vermont 1.192 0.757 1.876 C-19 Exhibit C-5 (continued) Parameter Odds Ratio 2970 Ronigence nterval Washington 1.405 *** 1.334 1.481 Wisconsin 0.894 *** 0.843 0.948 West Virginia 1.476 *** 1.315 1.657 Wyoming 1.987 ** 1.235 3.198 */e* eee - Significantly different from regular MA plan based on a p-value cutoff of 0.05/0.01/0.001 NOTES: The model also included an interaction term between an indicator for a beneficiary who originally became eligible for Medicare because of disability and another indicator for being aged 65 or older in 2015. This interaction term is not shown in the table because the OR for an interaction term is not directly interpretable. The model excluded beneficiaries in California, Oregon, and Utah. SOURCE: RTI analysis of MA encounter data (2015), Medicare enrollment and eligibility data (2015), and Medicare risk adjustment data (2014). c-20