Citation
Comparative Effectiveness of Managed Care on Quality of Care for Medicaid Adults with Disabilities

Material Information

Title:
Comparative Effectiveness of Managed Care on Quality of Care for Medicaid Adults with Disabilities
Creator:
Wegman, Martin P
Publisher:
University of Florida
Publication Date:
Language:
English

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Epidemiology
Committee Chair:
STRILEY,CATHERINE L
Committee Co-Chair:
SHENKMAN,ELIZABETH ANN
Committee Members:
SHORR,RONALD I
MULLER,KEITH E

Subjects

Subjects / Keywords:
medicaid

Notes

General Note:
Medicaid is the main source of health care for individuals who have chronic medical conditions severe enough to inhibit their gainful employment. Yet the complexity and longevity of health service needs for these individuals, often delivered through a fragmented system, contribute to expensive and low quality health care. In response, managed care delivery models have become a popular mechanism to increase integration and coordination across the spectrum of care while containing or ensuring greater predictability of costs. However, little is known about managed care's effects on quality and utilization of care for this subpopulation. There is concern that savings may be achieved by reducing quality of or access to care rather than through encouraging more appropriate care utilization- a result which would further jeopardize the well-being of an already vulnerable group. Thus, the objective of this dissertation is to assess the effects of comprehensive managed care on the quality and utilization of health care for adults in Medicaid managed care (MMC) who qualify due to disability, relative to more traditional fee-for-service (FFS) or primary-care case management (PCCM) health service delivery models. This goal will be achieved by leveraging a natural experiment in Texas where mandatory MMC enrollment was legislated in 2007 for a subset of Texas' 254 counties. The proposed studies utilize longitudinal quasi-experimental designs to provide ecologically valid information with immediate relevance to policy decision-makers. Through this work, we will increase the evidence base for Medicaid managed care, which currently serves more than 3.3 million enrollees nationally.

Record Information

Source Institution:
UFRGP
Rights Management:
All applicable rights reserved by the source institution and holding location.
Embargo Date:
8/31/2018

Downloads

This item has the following downloads:


Full Text

PAGE 1

COMPARATIVE EFFECT IVENESS OF MANAGED CARE ON QUALITY OF CARE FOR MEDICAID ADULTS WITH DISABILITIES By MARTIN P. WEGMAN A DISSERTATION PRESENTE D TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2017

PAGE 2

2017 Martin P Wegman

PAGE 3

3 ACKNOWLEDGEMENTS I am tremendously grateful for all who have played a role in my professional and personal development over the years To begin, I acknowledge the many funding organizations which have supported my graduate research and my professional education These entities include the National Institutes of Health ( specifically the National Institute of Mental Health and the National Center on Advancing Translational Sciences ), the Doris Duke Charitable Foundation, the Gold Humanism Foundation the Gilead Foundation, the American Medical Association Foundation and the American Medical Student Association With their funding these organizations have allowed me to pursue my passion for learning a n d have provided me tremendous latit ude to ponder and explore questions about how to make the wo rld a better place for marginalized members of our society I am grateful to the programs and support staff at the University of Florida Clinical and Translational Science Institute, the Department of Epidemiolog y, the Department of Health Outcomes and Policy and the College of Medicine. Specifically, I owe thanks to Patricia Sacks who has worked meticulously and tirelessly to help me meet dissertation deadlines. I am fortunate for the University of Florida MD PhD program which embrac ed my less traditional interests in population health and for the support of its tireless coordinator Skip Harris. I have learned volumes from my mentors. Nancy Hardt, MD and Anthony Delisle, PhD encouraged my continued engagement in social justice a nd humanism They also repeatedly developed fulfilling professional opportunities for me within their work from guest le cturing and teaching to invited presentations at national conferences I'm grateful to Frederick Altice, MD, MA who afforded me a tremendous opportunity to move

PAGE 4

4 overse a s and immerse myself in a new and very challenging health service delivery context. I never would have thought that such environments would so closely approximate th ose experienced by so many in Medicaid. I appreciate Dr. Altice's ready inclusion of me within his team of brilliant and dedicated researchers and his unwavering support of my work I am thankful for some of my first opportunities conducting formal research, both at the Laboratory for Laser En ergetics with William Donaldson, PhD and at the University of Rochester Department of Biomedical Engineering with Richard Waugh, PhD. These experiences solidified my desire to work in a stimulating academic style environment. High school marked the first time I remember having an intense passion for learning David Dye played a significant role in challenging me with mathematics beyond the limits of my knowledge. I responded by spending my free time completing problem sets in the most advanced math books I c ould locate John Maxwell encouraged my budding interest in computer science, challenging me to self study for the AB exam in sophomore year and inviting me to participate on our top ranked programming team. Several additional outstanding teachers deser ve mention, including William Sadler and Brad Bovenzi. I have been fortunate to work with several outstanding clinical mentors throughout my training These include Flavia Nobay, MD and Robert Hollander, MD who have been exemplars of professionalism and humanism. I also thank Dr. Hollander a s well as Heather Harrell, MD for introducing me to the fundamentals of medicine and clinical diagnosis.

PAGE 5

5 I owe gratitude to my PhD committee : Catherine Striley, PhD, MPE, MSW, Ronald Shorr, MD, MS, Keith Muller, PhD, Elizabeth Shenkman, PhD. These individuals have shown great patience as I have completed my dissertation work over the years. They have spent countless hours of their own time reviewing my work and advising me My greatest professional gratitude is owed to my primary mentor Dr. Shenkman for her advo ca c y on my behalf her unwavering support and her belie f in me even as I strayed from the norm. Finally, I offer most appreciation to my family. I am thankful to m y brother, Stephen, and sister, Katherine, for their love and support, especially as they have become adults themselves. I appreciate my daughter, Morgan, who reminds me to live each day fully and without regret. She has taught me how important empathy and love are in bringing about change. My earliest and most profound mentor was my father It is from him I developed a passion for using my hands and mind to develop solutions to any problem that might present itself. M y greatest supporter through much of my life was my mothe r -who in addition to providing me my first exposure to clinical medicine, gave birth to me, and helped me to thrive with wonderful meals, clean clothes, rewarding activities, and compassion. Both my parents created a n environment whe re I could pursue my intellectual passions without limit. I feel the deepest gratitude for my wife Courtney Courtney inspires me, encourag es me, and challeng es me to ensure a greater purpose to my work. From her I have learned about true generosity and humility.

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGEMENTS ................................ ................................ ............................... 3 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 11 LIST OF ACRO NYMS/TERMS ................................ ................................ ..................... 12 ABSTRACT ................................ ................................ ................................ ................... 14 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 16 Background and Significance ................................ ................................ ................. 16 Characteristics of Medicaid Enrollees with Disabilities ................................ ........... 17 Clinical and Health Characteristics ................................ ................................ ... 17 Demographic, Social and Economic Characteristics ................................ ........ 20 Conceptual Frameworks Describing Health Care Use and Health Outcomes ........ 21 Medicaid, the States and Enrollees with Disabilities ................................ ............... 24 Managed Care for Medicaid Enrollee s with Disabilities ................................ .......... 28 Medicaid Home and Community Based Service Alternatives ................................ 33 Managed Care and HCBS Together in Medicaid ................................ .................... 34 Existing Literature Evaluating Medicaid Managed Care and HCBS ........................ 34 Setting ................................ ................................ ................................ ..................... 37 2 QUALITY OF CARE FOR CHRONIC CONDITIONS AMONG DISABLED MEDICAID ENROLLEES: AN EVALU TATION OF A 1915(b) AND (c) WAIVER PROGRAM ................................ ................................ ................................ ............ 39 Introduction ................................ ................................ ................................ ............. 39 Methods ................................ ................................ ................................ .................. 42 Overview ................................ ................................ ................................ .......... 42 Population and Data Sources ................................ ................................ ........... 43 Outcome Variables ................................ ................................ ........................... 44 Predictor Variables ................................ ................................ ........................... 45 Model Specification ................................ ................................ .......................... 45 Results ................................ ................................ ................................ .................... 47 Discussion ................................ ................................ ................................ .............. 54 3 IMPACT OF STAR+PLUS ON BEHAVIORAL HEALTH CARE QUALITY .............. 57 Background and Significance ................................ ................................ ................. 57 Methods ................................ ................................ ................................ .................. 59

PAGE 7

7 Overview ................................ ................................ ................................ .......... 59 Sample and Data ................................ ................................ .............................. 61 Outcome Meas ures ................................ ................................ .......................... 62 Explanatory and Control Variables ................................ ................................ ... 63 Clustering Unit ................................ ................................ ................................ .. 64 Empirical Model and Analyses ................................ ................................ ......... 64 Results ................................ ................................ ................................ .................... 66 Discussion ................................ ................................ ................................ .............. 73 4 IMPACT OF STAR+PLUS ON RECEIPT OF PREVENTATIVE CARE .................. 78 Background and Significance ................................ ................................ ................. 78 Methods ................................ ................................ ................................ .................. 80 Over view ................................ ................................ ................................ .......... 80 Sample and Data ................................ ................................ .............................. 82 Outcome Measures ................................ ................................ .......................... 83 Explanatory and Control Variables ................................ ................................ ... 84 Empirical Model and Analyses ................................ ................................ ......... 85 Results ................................ ................................ ................................ .................... 87 Discussion ................................ ................................ ................................ .............. 93 5 CONCLUSION ................................ ................................ ................................ ........ 96 APPENDIX A TABLES FOR CHAPTER 1: INTRODUCTION ................................ ..................... 102 B TEXT DESCRIBING TREATMENT VARIABLE OPERATIONALIZATION AND BEHAVIOR ................................ ................................ ................................ ........... 114 C VIS STANDARD DIFFERENCE IN DIFFERENCE MODEL ................................ ........ 115 D FIGURES FOR CHAPTER 3: BEHAVIORAL HEALTH ................................ ........ 117 E TABLES FOR CHAPTER 3: BEHAVIORAL HEALTH CARE ................................ 125 F FIGURES FOR CHAPTER 4: PREVENTATIVE CARE ................................ ........ 1 41 G TABLES FOR CHAPTER 4: PREVENTATIVE CARE ................................ .......... 147 LIST OF REFERENCES ................................ ................................ ............................. 164 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 175

PAGE 8

8 LIST OF TABLES Table page 2 1 Characteristics of eligible enrollees in transition and comparison counties during the baseline and post baseline periods. ................................ .................. 47 2 2 Enrollee and county sample sizes in transition and comparison counties, by measure. ................................ ................................ ................................ ............ 49 2 3 Unadjusted measure adherence (raw proportions) in transition and comparison counties during the baseline and post baseline periods. ................. 51 2 4 Model estimated average post baseline compliance in transition and comparison counties, and their differences, by measure. ................................ ... 53 3 1 Demographic health and contextual characteristics for the study population ..... 66 3 2 Weighted, unadjusted measure performance ................................ ..................... 71 3 3 Predicted marginal means during 2010, by measure. ................................ ........ 72 4 1 Demographic health and contextual characteristics for the study population ..... 87 4 2 Weighted, unadjusted measure performance ................................ ..................... 92 4 3 Predicted marginal means during 2010, by measure. ................................ ........ 93 A 1 STAR+PLUS transition history ................................ ................................ ......... 102 A 2 Comparing Fee for Service (FFS) and P rimary Care Case Management (PCCM) on measure compliance at baseline and control variables ................. 104 A 3 Abbreviated description s of outcome measures ................................ ............... 105 A 4 Member eligibility counts FFS/PCCM ................................ ............................ 108 A 5 Member eligibility percentages FFS/PCCM ................................ ................... 109 A 6 Member eligibility counts STAR+PLUS ................................ ......................... 110 A 7 Member eligibility percentages STAR+PLUS ................................ ................ 111 A 8 Coefficients for final reduced models ................................ ................................ 112 E 1. Percentage of the sample qualifying for a given measure ................................ ... 125 E 2 AAM eligible enrollees: Characteristics of enrollees in transition and comparison counties during the baseline and post baseline periods ................ 125

PAGE 9

9 E 3 FUH eligible enrollees: Characteristics of enrollees in transition and comparison counties during the baseline and post baseline periods ................ 127 E 4 IET eligible enrollees: Characteristics of enrollees in transition and comparison counties during the baseline and post baseline periods ................ 130 E 5 Summary of sample size, by group and by measure ................................ ........ 133 E 6 Backward step wise regression results for AMM Acute ................................ .... 133 E 7 Backward step w ise regression results for AMM Cont ................................ ..... 133 E 8 Backward step wise regression results for FUH 7 ................................ ............ 134 E 9 Backward step wise regression results for FUH 30 ................................ .......... 134 E 10 Backward step wise regression results for IET EGMT ................................ ..... 135 E 11 Backward step wise regression results for IET INIT ................................ ......... 135 E 12 Full model coefficients for AMM Acute ................................ ............................. 136 E 13 Full model coefficients for AMM Cont ................................ ............................... 136 E 14 Full model coefficients for FUH 7 ................................ ................................ ...... 137 E 15 Final model coefficients for 30 ................................ ................................ .......... 138 E 16 Full model coefficients for IET EGMT ................................ ............................... 138 E 17 Full model coefficients for IET INIT ................................ ................................ ... 139 E 18 Final model coefficients for AMM Acute ................................ ........................... 139 E 19 Final model coefficients for AMM Cont ................................ ............................. 140 E 20 Final model coefficients for FUH 7 ................................ ................................ .... 140 E 21 Final model coefficients for FUH 30 ................................ ................................ .. 140 E 22 Final model coefficients for IET EGMT ................................ ............................. 140 E 23 Final model coefficients for IET INIT ................................ ................................ 140 G 1 Percentage of the sample qualifying for a given measure ................................ 147 G 2 AAP eligible enrollees: Characteristics of enrollees in transition and comparison counties during the baseline and post basel ine periods ................ 147

PAGE 10

10 G 3 BCS eligible enrollees: Characteristics of enrollees in transition and comparison counties during the baseline and post baseline periods ................ 150 G 4 CCS eligible enrollees: Characteristics of enrollees in transition and comparison counties during the baseline and post baseline periods ................ 153 G 5 COL eligible enrollees: Characteristics of enrollees in transition and comparison counties during the baseline and post baseline periods ................ 155 G 6 Summary of sample size, by group and by measure ................................ ........ 158 G 7 Backward step wise regression results for AAP ................................ ............... 159 G 8 Backward step wise regression results for BCS ................................ ............... 159 G 9 Backward step wise regression results for CCS ................................ ............... 160 G 10 Backward step wise regression results for COL ................................ ............... 160 G 11 Full model coeff icients for AAP ................................ ................................ ......... 161 G 12 Full model coefficients for BCS ................................ ................................ ......... 161 G 1 3 Full model coefficients for CCS ................................ ................................ ........ 162 G 14 Full model coefficients for COL ................................ ................................ ......... 162 G 15 Final model coefficients for AAP ................................ ................................ ....... 163 G 16 Final model coefficients for BCS ................................ ................................ ....... 163 G 17 Final model coefficients for CCS ................................ ................................ ...... 163 G 18 Final model coefficients for COL ................................ ................................ ....... 163

PAGE 11

11 LIST OF FIGURES Figure page 1 1 Institute of Medicine's Social Risk Factors of health care use, outcomes and cost model. ................................ ................................ ................................ ......... 22 1 2 Andersen's Health Service Utilization Framework. ................................ ............. 23 1 3 Map of Texas counties, with color indicating the date of STAR+PLUS implementation ................................ ................................ ................................ ... 37 1 4 STAR+PLUS expansion by transition group ................................ ....................... 38 D 1 Number of enrollees that qualify for a given measure, by year and by group ... 117 D 2 Sample spatial distribution ................................ ................................ ................ 117 D 3 Weighted, unadjusted measure performance over time, by group ................... 118 D 4 Regression assumption diagnostics for AMM Acute (final and full models) ...... 119 D 5 Regression assumption diagnostics for AMM CONT (final and full models) ..... 120 D 6 Regression assumption diagnostics for FUH 7 (final and full models) .............. 121 D 7 Re gression assumption diagnostics for FUH 30 (final and full models) ............ 122 D 8 Regression assumption diagnostics for IET EGMT (final and full models) ....... 123 D 9 Regression assumption diagnostics for IET INIT (final and full models) ........... 124 F 1 Number of enrollees that qualify for a given measure, by year and by group ... 141 F 2 Sample spatial distribution ................................ ................................ ................ 141 F 3 Weighted, unadjusted measure performance over time, by group ................... 142 F 4 Regression assumption diagnostics for AAP (full and final models) ................. 143 F 5 Regression assumption diagnostics for BCS (full and final models) ................. 144 F 6 Regression assumption diagnostics for CCS (full and final models) ................. 145 F 7 Regression assumption diagnostics for COL (full and final models) ................. 146

PAGE 12

12 LIST OF ACRONYMS/TERMS AMI Acute myocardial infarction AWD Adults with disabilities CABG Coronary artery bypass graft CHIP COPD Chronic obstructive pulmonary disease CRG (3M) Clinical risk group DD Difference in difference ED Emergency department FFS Fee for service HCBS Home based and community based services HEDIS The healthcare effectiveness data and information set HMO Health m aintenanc e o rganization ICD 9 CM The international classification of diseases, ninth revision, clinical modification ICHP Institute for child health policy ICM Integrated care management LTSS Long term services and supports MCO Managed care organization MMC Medicaid managed care MRSA Medicaid rural service area NCQA National committee for quality assurance PACE Program of all inclusive care PCCM Primary care case management

PAGE 13

13 PCI Percutaneous coronary intervention SAS Statistical analysis s ystem SSI Supplemental security income US United States

PAGE 14

14 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy COMPARATIVE EFFECT IVENESS OF MANAGED CARE ON QUALITY OF CARE FOR MEDICAID ADULTS WITH DISABILITIES By Martin P Wegman August 2017 Chair: Catherine Striley Cochair: Elizabeth Shenkman Major: Epidemiology Medicaid is the main source of health care for individuals who have chronic medical conditions severe enough to inhibit their gainful employment. Yet the complexity and longevity of health service needs for these individuals, often delivered through a fragmented system, contribu te to expensive and low quality health care. In response, managed care delivery models have become a popular mechanism to increase integration and coordination across the spectrum of care while containing or ensuring greater predictability of costs. However, little is known about effects on quality of care for this subpopulation. There is concern that savings may be achieved by reducing quality of or access to care rather than through encouraging more a ppropriate care utilization a result which would further jeopar dize the well being of an already vulnerable group. Thus, the objective of this dissertation is to assess the effects of comprehensive managed care on the quality of health care for adults in Medicaid managed care (MMC) who qualify due to disability relat ive to more traditional fee for service (FFS) or primary care case management (PCCM) health service delivery models. This goal will be achieved by leveraging a natural experiment in Texas where

PAGE 15

15 mandatory MMC enrollment was legislated in 2007 for a subset o The proposed studies utilize longitudinal quasi experimental designs to provide ecologically valid information with immediate relevance to policy decision makers. Through this work, we will increase the evi dence base for Medicaid man aged care, which currently serves more than 3.3 million enrollees nationally.

PAGE 16

16 CHAPTER 1 INTRODUCTION Background and Significance Despite a recent slowing of growth, United States (US) health care spending has reached unsustainable levels, accou nting for 17.1% of gross domestic product in 2013. 1 Paradoxically, such spending has led to only fair performance on health outcome and health care quality indicators in comparison to other developed nations. 2, 3 This has fueled intense focus on understanding the sources of increased spending 4 and relatively poor health outcomes 1 3,5 and has prompted development of new models of health care delivery and financing. 6, 7 Considerable attention, especially by governmental leaders and tax payers, has be en focused on refining health care programs financed by federal or state government. By number insured, the largest among these is Medicaid 1 a jointly funded state federal health care entitlement program serving primarily low income families, children, re lated caretakers, pregnant women, adults aged 65 and older, and individuals with disabilities. 8 Similar to other health care programs, expenditures in Medicaid are highly concentrated among a relatively small proportion of enrollees 2 For example, althoug h only 15% of Medicaid enrollees qualify due to disability 3 these individuals account for 1 Programs financed directly by the US government also include Medicare, Veterans Health Administration and Indian Health Serv ice, among others. The focus on Medicaid is driven by pragmatic and ethical motivations: ( 1) a favorable (i.e. unsaturated) research niche ( 2) stakeholder research needs ( 3) desire to generate evidence which could benefit traditionally underserved and marginalized population, and ( 4) data availability and expertise owing to my fellowship within the Institute for Child Health Policy, the External Quality Review Organization for Texas Medicaid. 2 The term enrollee is used throughout in a manner synonymous with beneficiary or recipient 3 To aid in readability, hereafter we use term s such as to represent enrollees who qualify for Medicaid due to a chronic medical impairment that inhibits engagement in any substantial gainful activity This is the disability criteria used to determine eligibility for

PAGE 17

17 close to half of all health care spending for the program. Correspondingly, disabled Medicaid adults cost an average of $20,000 per year compared to just $3,000 per Medicaid child. 9 Unfortunately, this increased spending is associated with poorer health care quality and lagging health outcomes when compared to the general population or to relatively healthier persons receiving Medicaid services. In the following sect ions, I describe the diverse population with disabilities served by the Medicaid program, discuss factors contributing to this increased cost, lower quality care and poorer outcomes, and describe key strategies to improve the value of health care for this important population which encompasses more than 9 million adults nationwide. 9 S ubsequently, I present my strategy to evaluate a large implementation of one of these strategies. Characteristics of Medicaid Enrollees with Disabilities Clinical and Health C haracteristics Although often grouped as a single population in discussing reform, individuals with disabilities served by Medicaid have marked variation in the types of disability they experience. Broadly classified, 42% qualify due to a physical conditio n, 26% qualify due to mood or psychotic disorders and 20% qualify due to intellectual disorders. 9 For age adults with spinal cord and traumatic brain ndividuals] with other serious, chronic illnesses and disorders such as diabetes and cardiac and pulmonary 10 This heterogeneity in turn leads to substantial variation in acute and long supplemental security income (SSI). In Texas, the focus of this analysis, SSI r ecipients are automatically eligible for Medicaid.

PAGE 18

18 term needs, forming the first set of challenges in providi ng efficient and patient centered care for this population. In addition to this variation in disability type and care needs, Medicaid enrollees with disabilities experience a very high burden of illness relative to the general population. This high illnes s burden proceeds from disability eligibility standards requiring that enrollees have chronic medical conditions severe enough to prevent substantial gainful activity for at least 12 months. Oftentimes, this illness burden takes the form of multiple co occ urring chronic health conditions. For example, i t is estimated that over 60% of Medicaid beneficiaries who qualify due to disability and over 95% of the costliest 5% receiving Medicaid have 2 or more chronic conditions. 11 Both this multi morbidity and poorer health status further complicate care delivery and disease management and lead to high levels of health care utilization. 12 One of the most prevalent and expensive forms of disease multi morbidity in Medicaid is co occurring physical and mental ill ness. Mental illness and/or substance abuse disorders co occur in approximately two third s of Medicaid disabled enrollees experiencing the most common physical health conditions (hypertension, diabetes, coronary heart disease, congestive heart failure and chronic obstructive pulmonary disease or asthma). 13 Among the highest cost 5% receiving Medicaid, 3 of the top 5 most frequent disease dyads and triads include psychiatric illness. 11 Among the general US adult population, the co occurrence of physical and mental health conditions is associated with substantially increased health care costs and reduced health care quality compared to those with a physical condition alone. For example, analysis of national health care claims revealed that monthly medical exp enditures for chronic conditions were increased $560 for individuals with co

PAGE 19

19 occurring depression and $710 for those with co occurring anxiety. 14 Adults with co morbid conditions are less likely to receive preventive care and their mental illness is more l ikely to go undiagnosed and untreated, compared to the general population and those with a physical condition alone. 15 17 A study of Medicaid adults with diabetes demonstrated an almost 20% decreased odds of receiving quality outpatient diabetes care and 32% increased odds of preventable hospitalizations for those also suffering from mental illness 18 This disease co occurr ence also has profound impacts on length and quality of life. As compared to adults with a single condition, mental and physical health comorbidity is associated with lost productivity, increased functional impairment and decreased health related quality o f life. 19 22 Most astounding, even after adjustment for physical health conditions, mental illness is associated with a 2 to 4 fold excess risk of premature death. 23 This is equivalent to a life expectancy 25 years shorter than the general population. 24 Further complicating the clinical morbidity burden experienced by this population, emerging evidence suggests that enrollees with disabilities served by Medicaid also have impairments in cognitive functioning domains implicated in poor disease management a nd limited patient engagement. For example, phone based assessments with a sample of Texas Medicaid enrollees with disabilities who have behavioral health conditions 25 revealed deficits in working memory, attention, verbal fluency and ability to encode nov el information nearly indistinguishable from reference populations with dementia. Moreover, cognitive functioning is highly related to health literacy, 26 which has been situated as a critical determinant of health and efficient health care utilization. 27 B y type of insurance, including the uninsured, individuals served by

PAGE 20

20 Medicaid have the lowest levels of health literacy, with for example, 60% being unable to correctly specify the time a medication is be taken after reading a prescription label. 28 These observations of diminished health literacy and relatively poor performance on executive functioning tasks suggest that cognitive barriers to engaging in appropriate health care use and disease management may be multifaceted. Demographic, S ocial and E conomic C haracteristics Persons with disabilities receiving Medicaid services are also diverse in terms of demographic characteristics Although representative state level Medicaid program enrollment data is not often readily available, enrollment in the Supplemental Security Income (SSI) program provides insights, in that SSI is the most common route to Medicaid eligibility for the disabled and confers automatic Medicaid eligibility in most states. Of the approximately 5 million adults aged 18 64 receivi ng SSI nationally, 57% are women. 29 Among the non institutionalized subset of this SSI population, 31% are African American and 14% are Hispanic 30 In addition, the living situation of adults receiving SSI (and thus qualifying for Medicaid due to disabili ty) reflects a greater degree of social and economic instability than the general population. For example, only 20% of these individuals are married, compared to a 60% rate among all US adults. In addition, 45% of adult SSI recipients have less than a high school education. 30 Economic circumstances reflect income eligibility requirements, whereby about 70% of households with a member receiving SSI have annual income below $10,000, 13% live in public housing, and 40% receive food stamps. 30 As discussed in mo re detail below, these social and economic characteristics are primary contributors to the inequitable illness burden and poor health care quality

PAGE 21

21 experienced by this population. Furthermore, these factors describe a local context where disease management and engagement in care are unlikely to be satisfactory Conceptual Frameworks Describing Health C are U se and Health O utcomes While clinical disease burden and related deficits appear to be driving health and health care disparities for the Medicaid disabled, it is important to situate these individual factors in the larger social and structural ecology that both accentuates and fundamentally determines these disparities T heory posits that social and economic stratification, along with the very structuring of political, social and economic institutions in society, lead to the inequitable access to and distribution of resources (e.g. financial, social, power, knowledge). abilit ies to attenuate and adapt to disease risks, and ultimately, promote health. 31 33 In many ways, this theory suggests that efforts to address proximal factors like risk behaviors will ultimatel y be ineffective, because prior disease paths will be replaced by new disease mechanisms, for which disadvantaged groups will still lack resources to deploy. For example, developed countries have seen unequal infectious disease burdens transition to unequa l non communicable disease burdens over the previous century, with similar marginalized groups disproportionately affected. Medicaid, Supplemental Security Income (SSI) for those individuals with disabilities, food vouchers and subsidized housing represent several important, yet insufficient, social programs which seek to diminish the imbalance in resources of those with limited income and wealth. The research described in this proposal does not seek to directly address these fundamental causes or improve t he relative position of disadvantaged groups; rather it contributes to efforts to improve the absolute value of care for a vulnerable population.

PAGE 22

22 \ Figure 1 1. Institute of Medicine's Social Risk Factors of health care use, outcomes and cost model. One recent conceptual framework developed by the Institute of Medicine (see Figure 1 1 ) provides more detail on the social factors implicated in health care use and outcomes, and is helpful in understanding differences not only bet ween groups of widely varying social position but also within a defined group. 34 For example, among Medicaid enrollees with disabilities with similar health conditions, differences in education and occupation portend different abilities to access and engag e with health care and achieve improved outcomes, despite enrollees having similar income and health status. In this same group, measures of race and ethnicity may capture well documented differences in access, patient provider trust and provision of care, or may serve as a proxy for cultural differences in beliefs and preferences for care. Socially and biologically derived gender specific differences in outcomes, and in accessing and receiving care would also be expected. Enrollees are also likely to experience different

PAGE 23

23 physical environments and communities, which can impact their ability to access health care or the saliency of and capacity in managing disease. Figure 1 2 Andersen's Health Service Utilization Framework. described originally by Andersen and shown in Figure 1 2. 34 The remaining care, including the acquisition of health insurance, and local or regional availability to and patterns of and prices for health care. For example, findings over the past decade, many of which were driven by the Dartmouth Atlas project, have documented high levels of between region variation and within region consistency in utilization and prices, suggesting the role of suppl y side driven utilization in health care, which is by definition

PAGE 24

24 independent of clinical need or individual enrollee characteristics. 35 context could also be envisaged to distinguish between enrollees living in the community and those who reside in an institutional nursing or assisted living facility, since the health care provided in these settings would be expected to differ. According to the Andersen model, the final and perhaps most proximal component driving health care use, as I have discussed first, is that of clinical need, which can be complicated by functional limitations (physical and mental), symptomology, and clinical multi morbidity. Medicaid the S tates and Enrollees with D isabilities Formed by the Social Security Amendme nts of 1965, Medicaid is the primary social program in the US providing health care to those with low income and limited resources. Social Security Amendments of 1972 replaced existing state administered welfare with the federally administered SSI program and largely extended Medicaid to all SSI recipients while greatly expanding income and disability related eligibility requirements. Since this time, additional amendments have generally extended eligibility to include additional disability segments. 36 S tate and the federal governments share responsibility for the Medicaid program in each respective state, with the federal government setting eligibility, services and financing guidelines and the state administering the program. For example, all participat ing states must offer the required inpatient, outpatient, laboratory and home health services. In addition, federal guidelines require that Medicaid beneficiaries have on diagnosis, type of illness or condition; and reimburse services which are

PAGE 25

25 37 Within these limits, states have flexibility in choosing to participate in Medicaid, in providing any number of optional health care services, in e xpanding eligibility to amount and duration of services. 37, 38 Furthermore, states can petition the federal Secretary of Health and Human Services for a variety of wai vers to many of these federal requirements. The first of these, enacted prior to Medicaid in 1962, is provided under section 1115 of the Social Security Act, and allows discretion for broad changes in eligibility, benefits, cost sharing, and provider payme nts. 39 The Omnibus Budget Reconciliation Act of 1981 provided an additional mechanism, with the 1915(b) waiver allowing states to require beneficiaries to enroll in managed care (discussed in greater detail below), and permitting these programs to offer different benefits targeted to groups living in selected geographic areas. Section 1915(c) waivers allow states to provide long term services (also discussed in greater detail below) in home and community settings as opposed to traditional institutional se ttings to targeted groups. Additional, related waivers in sections 1915, 1937 and 1945 may also be considered by states desiring flexibility in administering their programs. In addition to the general legal and administrative structure of Medicaid describe d above, there are several salient characteristics that provide additional useful context. First, in a manner similar to those with commercial insurance, Medicaid beneficiaries receive their care from many types of providers, including those in private sma ll group office based practices, as well as those in large groups, hospitals, academic medical centers and safety net health centers (such as federally qualified health centers). 40

PAGE 26

26 Health care professionals are typically paid through contracts with the sta te or with organizations contracted by the state to provide a defined schedule of services. Due to general shifts in organizational consolidation as well as targeted efforts among small group providers to limit Medicaid rolls, increasingly, care for Medica id enrollees is concentrated among a smaller proportion of all providers, mainly the large group and community health center organizations. 41 Second, access to health care services is substantially better than the uninsured receive and in some cases may e ven rival that received by those with employer sponsored insurance. 42 However, data also suggest several gaps, including limited specialist availability, particularly for critically needed mental health and substance abuse services. 43 Limited access to the se services is particularly ominous for enrollees with co occurring physical and behavioral conditions, as unmanaged behavioral health issues contribute to worsening physical health conditions. Similarly, adults served by Medicaid have very limited access to dental care 44 ; and new patient health care appointments are relatively restricted compared to commercially insured patients. 45 The reasons for limited provider participation in Medicaid compared to employer sponsored insurance are not entirely clear, al though low payments rates, on average 40% less than Medicare, 46 are suggested as a partial explanation. 47 In addition, provider shortages may be more likely in areas where Medicaid enrollees are concentrated. Limited after hours care and transportation bar riers represent additional obstacles to access faced by many adults receiving Medicaid services. Third, and similar to other programs in US health care Medicaid programs provide less than optimal care due to failures of care delivery, care coordination, overtreatment, pricing failures, administrative complexities and fraud and abuse. 4

PAGE 27

27 of employer sponsored insurance and Medicare, which routinely outpace inflation. 48 A lthough comparisons between payers on quality of care are limited, Medicaid enrollees are more likely to have less inclusive care, and have poorer health outcomes. 49, 50 Compared to those with employer sponsored insurance, Medicaid enrollees benefits are ty pically more limited and are often provided in an even more fragmented manner. This latter aspect is reinforced by service carve outs, whereby certain services (e.g. mental health) are provided by organizations separate from the other health care providers Fourth, Medicaid enrollees are subject to more frequent (i.e. often monthly) and burdensome application processes than most who have employer coverage, which leads to high rates of disenrollment or frequent coverage gaps. Those who qualify for Medicaid d ue to disability are the exception to this pattern, in that enrollment is longer term, resulting in more stability for this population. Fifth, for Medicaid enrollees with disabilities limiting gainful activity, long term services and supports (LTSS) form a n important and expensive subset of services provided by Medicaid. LTSS are used by 16% of Medicaid enrollees qualifying due to disability, with average annual costs of more than $60,000 per institutionalized adult and $30,000 for adults receiving these se rvices in a community setting. LTSS encompass the personal and medical care enrollees may need over extended periods of time delivered by both paid and unpaid (e.g. family and friends) providers. 51 This care often includes assistance with activities of dai ly living, instrumental activities of daily living and other services such as nursing facility care, adult daycare programs, home health aide services, personal care services, transportation, and supported employment. In light of many disabled enrollees us ing LTSS, coordination between the long term services and

PAGE 28

28 acute medical care is often cited as a barrier to improving quality and value. 52 Unfortunately, very few states have developed comprehensive systems where these aspects of care are coordinated. 53 Sixth, despite many commonalities across states in the characteristics described above, Medicaid programs also demonstrate substantial variation between states due political, economic, and demographic factors that vary across regions and over time. 54 Since states contribute more than 50% of their respective Medicaid budget s Medicaid budgeting is a key consideration in state fiscal decisions. This is further complicated by the fact that the need for Medicaid, similar to other welfare programs, is inver sely related to state prosperity. That is, increased expenditures and enrollment often coincide with decreased tax revenue. In addition, those served by Medicaid are almost exclusively individuals with limited political or social capital, such that outside public opinion, interest groups, political ideology and political culture dominate the discussion on policy changes and implementation. 54 Managed Care for Medicaid Enrollees with D isabilities The health disparities and high health care costs experienced by Medicaid enrollees with disabilities underscore the importance of focusing on this population; these results also suggest the failings of traditional health care delivery in adapting to the complexity, intensity and longevity of care required to meet th needs. 55 In response, and motivated to restrain or ensure greater predictability in state Medicaid budgets, policy makers have experimented with variations in Medicaid service delivery and financing for enrollees with disabilities. In creasingly, managed care health service delivery, and home and community based service alternatives to care have

PAGE 29

29 become central approaches 4 to address these concerns. 10, 11,56 In subsequent sections, I provide more context for these health care delivery in novations. Managed care draws its roots from efforts in the early 1900s to create more predictable revenue streams for providers and restrain costs for consumers and companies. Selected groups of individuals or companies arranged regular pre payments (prem iums) to provider groups or hospitals in return for no added cost or reduced price access to a schedule of medical services (outpatient, inpatient or both). Over time these arrangements became more like the independent practice association model of present where a third party, the health maintenance organization, or more generally the managed care organization (MCO), contracts with physicians or physician groups in private fee for service (FFS) practices. The 1960s marked the first time that the majority o f all medical care was paid for by a third party (including Medicare, Medicaid, most private insurance, and the Veterans Administration). This economic arrangement began to exacerbate the already existing moral hazard that typifies health insurance, leadin g to further increases in utilization and costs. The high prevalence of third party FFS payment, combined with the advances in technology and rising consumer expectations for health care are thought to be the cause of the inflationary health care costs fi rst occurring in the late 1960 -inflationary costs which have largely persisted (absent the Great Recession) to present day. 39 The concern over accelerating health care costs during the 1970s led to development of managed care as it is currently known, including varied formulations 4 Other common approaches include those, which, similar to managed care, seek to shift financial risk to the provider level (in order to drive accountability): global budgets, shared savings/risk, bundled payments and pay for performance programs, in payer specific or all payer formulations. Additional approaches include medical homes models of care delivery (which may be integrated with abovementioned financing models).

PAGE 30

30 (e.g. preferred provider organizations). Subsequently, the mid 1980s to early 1990s were characteriz ed by rapid growth in managed health care and consolidation of the industry. Mirroring private insurance trends, state policy makers began to implement managed care in Medicaid, initially through 1915(b) waivers enacted through the Omnibus Budget Reconcili ation Act of 1981. The mid 1990s were marked by backlash from the general public over utilization management (described in further detail below), leading to decreasing private managed care enrollment. During the same time, policymakers generally expanded m anaged care in their state Medicaid programs. 39 Managed care was almost exclusively first offered to or mandated for relatively healthier Medicaid children and their parents, rather than those with disabilities or the elderly, due to concerns over disrupti ng ongoing disease management, developing and maintaining appropriate provider networks and managing greater and more complex financial risk. However, beginning in the late 1990s, managed care has increasingly been made available to Medicaid adults with di sabilities. As of 2012, managed care was available to disabled Medicaid enrollees in over two thirds of US counties. 5 7, 58 Recently California, Texas, Florida, New York and Illinois for example, have passed legislation requiring millions of disabled benefic iaries to enroll in Medicaid managed care. 59 In managed care, MCOs leverage several tools to promote appropriate utilization of health care services. First, MCOs often modify the payment system such that providers and the insurance companies are better aligned. In Medicaid and many commercial implementations, this is often accomplish by contracting with independent primary care providers or provider groups and paying a recurri ng fixed amount (known

PAGE 31

31 5 for delivering a schedule of medical services. 57 Additional financial incentives are also sometimes used to encourage more appropriate referrals to speciali sts, who often remain in FFS payment systems. Second, MCOs use their purchasing power to negotiate lower rates with providers and provider groups, and exclude providers unwilling to accept discounted rates, or who may be deemed to not be utilizing health s ervices efficiently. Third, MCOs employ utilization management, including pre authorization for elective procedures or hospitalizations, and actively monitor inpatient stays (to reduce length of stay). Another example of a prior authorization involves requ iring referrals from a primary care physician in order for specialist services to be covered. MCOs may also place increased restrictions on the types of medications or medical testing covered. Utilization review involves MCOs monitoring service use furnish ed by pr oviders and incentivizing or d e incentivizing below or above average use, respectively. Fourth, MCOs use case management, which involves coordination of health care and social services for patients with chronic conditions. Oftentimes, these case ma nagement activities proceed from initial and recurring enrollee medical and service needs assessments. Additionally, MCOs may employ technologies and processes to increase inter provider and provider patient communication, mail, phone or in person benefic iary education, care transition supports, and additional targeted process improvement and quality incentives. 8,59 While cost sharing (including co payments, co insurance and deductibles) are key tools in employer sponsored and Medicare managed care, federa l law restricts cost sharing to only nominal levels for Medicaid managed care enrollees. 5 In this arrangement, providers ar (wholly or partially) for additional costs/provision of services beyond those additionally predicted.

PAGE 32

32 Much of the description thus far has centered on comprehensive risk based Medicaid managed care service delivery. In these managed care models, prospective payments ar e typically a uniform amount risk adjusted for a particular case mix (i.e. illness burden) cared for by the MCO. To provide services, 6 MCOs in turn contract with a network of providers, arranging various financial and administrative relationships, includin g payments to providers that may be bundled by care episode or remitted per unit of service. The Centers for Medicare & Medicaid Services, which administers Medicaid at a federal level, also considers primary care case management (PCCM) and limited benefit plans as managed care programs. In PCCM, providers receive standard FFS payments as well as a monthly case management fee in return for providing care management and coordination. In limited benefit plans, coverage is provided for a limited set of service s in exchange for a capitation payment. 39 These payment and service delivery models are, to varying degrees, different from the traditional FFS health care financing system, where the state only directly reimburses health care providers or their entities an amount for each unit of service they provide. When implemented in a third party payment manner, FFS incentivizes unnecessary health care use because providers financially benefit from all (nec essary and unnecessary) service delivery and patients have no direct financial burden. In such a system, uncoordinated care and worsening health outcomes are often financially rewarding for providers because this requires duplicated or increased interventi on and services. In comparison, under set prospective payments, it is thought that MCOs are 6 States vary on which benefits they include or exclude from their managed care programs. States often carve out or exclude certain Me dicaid services from the set of benefits that a comprehensive risk based managed care plan is responsible for providing enrollees

PAGE 33

33 incentivized to provide less costly (i.e. more proactive/preventive) care to delay or decrease the onset of higher cost utilization which would be incurred by the M CO at a later date. 60 Thus, the goal is to promote comprehensive, high quality, primary care centered services better matched to enrollee needs. 60 At the same time, services not decreasing subsequent health care utilization (e.g. services that solely impro ve quality of life) may be less likely to be provided. In addition, the short time horizon for capitation may promote temporary intervention in lieu of proactive or preventive care, which has benefits on a longer time scale. Medicaid Home and Community Ba sed Service A lternatives Home based and community based service (HCBS) alternatives to institutional care represent a second health service innovation increasingly employed in Medicaid programs. Since the early 1980s when the 1915(c) waiver option became available, HCBS use has grown rapidly. In 2011, more than 3.2 million Medicaid beneficiaries received HCBS, accounting for almost half of Medicaid expenditures on long term services and supports (LTSS). 58 Dukett and Guy 61 trace the rise in Medicaid HCBS to several factors. Key among these was increased awareness of the disproportionate costs of institutional based care and the tendency for Medicaid to favor institutionalization in eligibility and coverage tion, it was recognized that a large proportion of individuals in institutional settings were capable of living at home or in community residences, and many actually had a preference for doing so and experience unsatisfactory quality of life as a result. 61 This increased awareness was further punctuated by multiple court cases favoring deinstitutionalization. Given this background, HCBS are emphasized as cost effective, patient oriented

PAGE 34

34 options for enrollees to receive LTSS outside of the traditionally required hospital or institutional setting. 62 HCBS commonly encompass 7 specific service categories: case management (coordination of care and promoting access to medical and social services), homemaker (i.e. cooking, cleaning, laundry, some maintenance and organization), home health aide (covering certain nursing or nursing aid services for disease management), personal care (covering activities of daily living, such as toileting, eating, dressing and bathing), adult day health care (e.g. act ivities and stimulation paired with skilled health services, provided during the day outside of the home), habilitation (such as occupational therapy, physical therapy and speech and language therapy), and respite care. M anaged Care and HCBS T ogether in M edicaid Growing out of efforts to coordinate long term and acute services, as well as a desire to further stabilize growing costs, state policy makers have developed Medicaid programs which combine traditional managed care for acute and outpatient services with managed LTSS, including HCBS. One of the earliest examples of these programs, which still operates today, was developed exclusively for the elderly dual eligible (i.e. Medicare and Medicaid eligible) population: T he Program of All inclusive Care for the Elderly (PACE). Arizona was the first state to develop a program which also included disabled (non elderly) Medicaid enrollees, beginning in 1987. 63 Since this time, through combined 1915(b)/(c) waivers or 1115 demonst ration waivers, the number of integrated managed care and HCBS programs serving enrollees with disabilities has continued to grow, from 8 states in 2004 to 18 in 2014, with projections for continued expansion. 64 Existing Literature Evaluating Medicaid Man aged C are and HCBS Despite this growth in comprehensive managed care, including managed LTSS

PAGE 35

35 delivered through HCBS, little is known about the quality of care in these programs. As managed care implementation in Medicaid focused initially on relatively he althy children and young adults, much of the existing evidence on Medicaid managed care describes populations with substantially different health states, care utilization and health needs than those qualifying due to disability The literature focused on e valuating managed care for the elderly, namely evaluations of PACE, might also not serve as an appropriate reference, since the chronic condition distribution of enrollees with disabilities is different than for other Medicaid high use groups. For example, enrollees with disabilities are more likely to have psychiatric illness, substance abuse, and developmental disability diagnoses compared with aged Medicaid enrollees. 11 Among the limited studies that exist for the Medicaid disabled population, the large majority have investigated changes due to voluntary managed care enrollment. And so while these studies found expenditure reductions between 9% and 37%, selection bias is likely to diminish the validity of such findings. 9 The very limited literature on the effects of mandatory managed care for disabled adults has been mixed. Burns 66 found that beneficiaries in mandatory managed care were more likely to report waiting to se e a provider, difficulty obtaining specialty care and not receiving a flu shot. Yet these same beneficiaries were also more likely to have a usual source of care compared to FFS enrollees. Average total per beneficiary Medicaid expenditures did not differ by care delivery model. 59 We caution that a large number of main comparisons were made with the analytical models in these studies increasing the likelihood that some of these findings may have been due to chance alone. The studies were also limited in tha t all mandatory managed care was modeled

PAGE 36

36 similarly. This assumption is less than ideal considering the large program heterogeneity across the country. Harman et al 67 where per member per month costs for Adults with disabilities ( AWD ) and Temporary Assistance to Needy Families beneficiaries were examined in a difference in difference framework for the 2 counties relative to matched controls. Cost reductions were not seen for the total population, but enrollees with at least a 3 month enrollment prior to and following the transition did realize cost savings. Using the National Health Interview Survey, Coughlin, Lo ng and Graves 62 explored access to care, flu shot receipt and use of primary care providers, specialty physicians and the emergency department for HMO). The main finding revealed increases in the usual source of care reported by Medicaid adults with disabilities in managed care counties. It is important to note however, that the program status for each enrollee was unknown and based on the county proxy for which there was considerable discordance. 62 The paucity of information on the effects of managed care for disabled adults 58 also reflects the inconsistent and heterogeneous reporting of quality and cost information 59, 60 which occurs at a national level. Recognizing this, the congressionally authorized Medicaid and ( CHIP ) Payment and that assess the impact of current programs and new service delivery inno vations on 57 A dditionally, the Department of Health and Human Services Office of the Inspector General identified concerns with inconsistent monitoring of the quality of

PAGE 37

37 and the nature of these programs puts beneficiaries at risk for receiving inadequate 68 With this context, we next describe our experimental setting S etting In the past decade, the state of Texas legislatively mandated the transition of Medicaid service delivery from FFS and PCCM models to comprehensive managed care for adults receiving SSI for their disabilities. The program implemented is called STAR+PLUS. 8 The program delivers acute and long term services and supports through 1 system, employs service coordinators who coordinate supports and develop individual plans of care; and emphasizes home and community based services. Figure 1 3. Map of Texas counties, with color indicating the date of STAR+PLUS implementation

PAGE 38

38 Figure 1 4. STAR+PLUS expansion by transition group Figure s 1 3 and 1 4 depict the history of implementation of the STAR+PLUS throughout the state of Texas over the past 2 decades. STAR+PLUS was first adopted in Harris County in 1998. Nine years later, in January and February of 2007, STAR+PLUS expanded to 28 additional count ies. In February and September of 2011, and in March of 2012 STAR+PLUS was expanded to an additional 13, 21 and 27 counties, respectively. 62 remaining 164 counties, comprising the Medicaid Rural Serv ice Areas (MRSAs). This phased implementation by county provides an opportune natural experiment that can be analyzed with high levels of ecological and internal validity.

PAGE 39

39 CHAPTER 2 QUALITY OF CARE FOR CHRONIC CONDITIONS AMONG DISABLED MEDICAID ENROLLEES: AN EVALUTATION OF A 1915(b) AND (c) WAIVER PROGRAM Introduction Home and community based service (HCBS) alternatives to institutional care have been emphasized as cost effective, patient oriented approaches that allow Medicaid enrollees to receive long term services and supports (LTSS) in their homes and communities. 58 In 2011, more than 3.2 million Medicaid beneficiaries received HCBS, accounting for almost half of Medicaid expenditures on LTSS. 58 There has been rapid growth in the use of managed care to provide LTSS through 1915 (b)/(c) managed care/HCBS waivers or 1115 demonstration waivers, increasing from 8 state Medicaid programs in 2004 to 18 programs in 2014. 64 However, little is known about the quality of care delivered through th ese programs. The Department of Health and Human Services (DHHS) Office of the Inspector General identified concerns with inconsistent monitoring of the quality of vulnerable, and t he nature of these programs puts beneficiaries at risk for receiving 68 Although there is considerable research on the effect of Medicaid managed care delivery for children, their parents and other low income beneficiaries, there is a pauc ity of information on the effects of managed care and HCBS waiver programs for adults with disabilities (AWD), which may be due to the relatively recent Reprinted with permission from Wegman MP, Herndon JB, Muller KE, et al. Quality of care for chronic conditions among disabl ed Medicaid enrollees: An evaluation of a 1915(b) and (c) waiver program. Med Care 2015;53(7):599 606.

PAGE 40

40 expansion of these programs or inconsistent and heterogeneous reporting of quality information. 57 60 O f the existing research, Burns found that Medicaid AWD in mandatory managed care were more likely to wait to see a provider, report difficulty obtaining specialty care, and less likely to receive a flu shot compared with fee for service (FFS) enrollees; ye t they were also more likely to report having a usual source of care.59 Coughlin, Long and Graves also found that Medicaid managed care was positively associated with having a usual source of preventive care among AWD compared with FFS.62 Neither study f ocused specifically on AWD enrolled in HCBS waiver programs. More than one third of AWD in Medicaid have three or more chronic conditions, and the chronic condition profile of AWD is different than for other Medicaid high use groups. For example, AWD are more likely to have psychiatric illness, substance abuse and developmental disability diagnoses compared with aged Medicaid enrollees.70 The high rates of chronic disease comorbidity and differential chronic condition profile highlights the importance of understanding the impacts of HCBS on the quality of chronic disease care among Medicaid AWD.65,70 Ensuring the delivery of recommended care contributes to improved disease management and the ability to remain in a home setting. The purpose of our study w as to examine the effects of a large acute care and HCBS program delivered through managed care organizations (MCOs) in Texas Medicaid, the STAR+PLUS program, on the quality of chronic disease care for AWD. In 2012, Texas Medicaid enrollees accounted for one half of all enrollees nationally in Medicaid managed care LTSS programs. 71 A primary focus of STAR+PLUS is to improve the quality of care for enrollees with disabilities through coordinated and

PAGE 41

41 comprehensive care. The program delivers acute and LTSS t hrough a single system; employs service coordinators who develop individual care plans and assist enrollees in receiving needed care; and emphasizes HCBS alternatives to institutional care. 8 Additionally, in a fully capitated health care delivery model li ke STAR+PLUS, evidence suggests that care which has the potential to reduce future visits, such as medications for chronic condition management, is provided more frequently compared with fee for service payment. 72 Given this evidence and the key program objectives to provide more integrated and coordinated care compared to the pre existing FFS and PCCM Medicaid program components, we hypothesized that the quality of care for chronic conditions would improve after ST AR+PLUS enrollment and relative to a comparison group that remained enrolled in FFS or PCCM. To test this, we performed a series of longitudinal mixed model analyses with a comparison group. Our study offers several contributions to the literature. First, it focuses specifically on Medicaid AWD <65 years in contrast to the more frequently studied Medicare Medicaid dual eligible populations. 73 76 Second, most research on HCBS has examined health care expenditures or access to care rather than specifi c quality indicators. 5,17 Third, we study a program with mandatory enrollment, overcoming the limitation of potential selection bias in prior research on waiver programs with voluntary enrollment. Thus, our study extends existing knowledge by examining th e effects of HCBS delivered through an integrated managed care program on the quality of care provided for common chronic conditions among Medicaid AWD.

PAGE 42

42 Methods Overview Currently, there are 13 STAR+PLUS service areas (SAs) that encompass all 254 counties in Texas. SAs are contiguous counties grouped together to organize health care delivery for Texas Medicaid. STAR+PLUS was phased in over time by SA, and some individual SAs transitioned asynchronously (i.e., subsets of counties within a SA transitioned i n different years). STAR+PLUS was piloted in 1998 in the Harris SA (initially comprised of Harris County). The program expanded to 40 additional counties in January and February of 2007, comprising subsets of 4 SAs. The remainder of the counties in thes e 4 SAs, plus six additional SAs, transitioned in 2011 and 2012. On September 1, 2014, STAR+PLUS completed statewide expansion (see Figure 2 1 and Appendix A for a listing of county transitions). 18 Texas selected SAs for initial implementation based on t he presence of a strong health care infrastructure in order to increase the likelihood of successful program implementation. The phased implementation a llowed us to compare the quality of care enrollees received for a range of chronic conditions before an d after their transition to STAR+PLUS and relative to enrollees who were phased in later. We used Texas Medicaid administrative data from January 2006 December 2010 to estimate the treatment effect of the STAR+PLUS program on chronic care quality, focusin g on the 2007 program expansions to allow for sufficient post transition data for analysis. This time frame includes a baseline year (2006), a transition year (2007), and three post transition for enrollees in the treatment counties. FFS and PCCM enrollees in counties that did not switch to STAR+PLUS during the study period served as the comparison group. We did not further distinguish between FFS and PCCM

PAGE 43

43 enrollees in our analyses based on existing research indicating few differences in access to care among adult Medicaid enrollees in general and those with disabilities in particular. 19 22 In addition, post analysis comparison of the control variables and baseline measure compliance between FFS and PCCM revealed only small differences (see Appendi Population and Data Sources The study population included individuals 21 64 years old who were enrolled during the study time period and qualified for Supplemental Security Income (SSI ) and Medicaid due to disability. Dual Medicare Medicaid eligible were excluded because Medicare data were not available for those enrollees. Individuals <21 years old were excluded because STAR+PLUS enrollment was voluntary for this group. We excluded enrollees in the 1998 Harris SA pilot because it was not possible to generate separate program effect estimates to compare mature versus newly implemented STAR+PLUS with only a single observational unit. Person level administrative enrollment and claims/e ncounter data provided by sex, race/ethnicity, county, service area, monthly enrollment, and delivery model (STAR+PLUS or FFS/PCCM). Enrollment records were linked to claim s data that included International Classification of Diseases (ICD 9 CM) diagnosis codes, Current Procedural Terminology (CPT) codes, and National Drug Codes. These data were supplemented with county, Zip Code Tabulation Area (ZCTA), and census tract lev el data from the Area Resource File, U.S. Census Bureau, and U.S. Department of Commerce to capture geographic contextual factors. The sample size varied for each

PAGE 44

44 outcome according to the eligibility inclusion criteria for each quality measure described b elow. Outcome Variables We used chronic care quality indicators from the National Committee for Quality ): (1) Use of Appropriate Medication for People with Asthma, (2) Medication Management for People with Asthma at the 75% level, (3) Pharmacotherapy for COPD Exacerbation, (4) Cholesterol Management for People with Cardiovascular Conditions, (5) Persistence of Beta Blocker Treatment after a Heart Attack, and (6) Comprehensive Diabe tes Care. These measures were selected because they reflect quality of care for the Agency for 23 and were suggested by the Centers for Medicare & Medicaid Services (CMS) as initial health care quali ty core measures for adults in Medicaid. Strong performance on these indicators is linked to improved health outcomes. 24 We used NCQA certified software (Inovalon, Quality Spectrum Insight v15.2011), applying 2012 HEDIS technical specifications to determine person level compliance for each measure in each study year (see Appendix C, for measure definitions). 25 For COPD exacerbation pharmacotherapy, we calculated the average of the compliance rates for appropriate corticosteroid and bronchodilator di spensing. For diabetes care, we constructed a person level composite measure used in prior research that averages the compliance rates of the subcomponents: annual hemoglobin A1c testing, eye exam, LDL cholesterol screening, and nephropathy screening. 26 F or the remaining measures, person level dichotomous indicators of compliance were created.

PAGE 45

45 Predictor Variables For each member year, we computed our main predictor variable as the number of STAR+PLUS enrollment months (0 12) (see Appendix D, for more de tail on the main level and contextual variables. 59,77 First, for each outcome, person level measure compliance in 2006 was used to control for baseline differences between the two studied groups. We constructed a variable to indicate if >5 enrollees were residing at the same address to identify group living arrangements. We also used the 3M Clinical Risk Groups (CRGs), which uses ICD 9 CM diagnosis codes from health care encounte rs for individuals enrolled >6 months to assign enrollees to the following hierarchically defined health status categories: healthy, significant acute conditions (e.g., chest pains), minor chronic conditions (e.g., migraine), moderate chronic conditions (e .g., asthma and diabetes), or major chronic conditions (e.g., cystic fibrosis and cancer). 78 Less than 1% of individuals lacked sufficient enrollment history for classification and were excluded. Additional individual characteristics included age at base line (in years), gender, and race/ethnicity (non Hispanic white, non Hispanic black, Hispanic other). Contextual geographic census tract (or ZCTA if census tract was unavailable) and county level median household income. Dummy variables for each year were controlled for secular trends. The full model included two three and four way predictor interactions and predictor interactions with time to test for time varyin g treatment effects. Model Specification Although SAs represent administrative clustering and the approximate unit at which STAR+PLUS was implemented, there were too few units to have sufficient

PAGE 46

46 statistical power. In addition, some SAs were transitioned i n phases. Accordingly, we selected an intermediate clustering unit (county), which provides sufficient power and accounts for potential non independence among enrollees. The county reasonably represents the context within which enrollees seek health care and captures administrative similarity of the care environment, such as available health care resources that impact health service use. We employed a two stage, multilevel approach using general linear models. First, we computed person level models sep arately for each year and outcome. Using individually adjusted averages and assuming unstructured correlations over time, we then built the second tier model for each outcome, spanning the post baseline years (2007 2010). With these 6 full models, fixed order backwards selection of predictor variables with = 0.05 was used to arrive at the reduced models. Finding general agreement in terms retained between the models, we established one final reduced model form for consistency. R 2 statistics were then estimated. 79 Finally, we computed the least squares mean predicted compliance rates for the STAR+PLUS and FFS/PCCM counties using group specific covariate distributions. The difference between these estimates reflects the STAR+PLUS effect. Our modeling approach is a mathematical generalization of a difference in difference (DD) approach. The standard DD model compares differences between 2 groups at 2 time points to isolate and test the presence of an effect, assumed to be a deviat ion from the baseline difference. In a design with more than 2 time points (e.g., in our study with 4 post baseline measurements), DD requires aggregation of the post period measurements as a single time point or computation of 4 separate DD models for ea ch outcome. Our strategy models all post period measurements simultaneously,

PAGE 47

47 while accounting for the covariance of repeated measurements, thus increasing power relative to the standard DD approaches. This generalized approach also relaxes the DD paralle l trend assumption requiring similar rates of growth between groups by allowing for differing slopes (see Appendix E, for more technical detail). Analyses were performed using SAS 9.3 (SAS Institute Inc., Cary, NC, USA). Results Table 2 1 provides summary demographic and health status information for Texas Medicaid AWD meeting the inclusion criteria, stratified by time period and delivery system (STAR+PLUS counties or FFS/PCCM counties). Eighty percent of enrollees were assigned to the most severe health status category. Two thirds were the sample was Hispanic or non Hispanic black. Over half lived in census tracts designated as impoverishe d or extremely impoverished. Table 2 1. Characteristics of eligible enrollees in transition and comparison counties during the baseline and post baseline periods. Transition counties (STAR+PLUS) Comparison counties (FFS/PCCM) Baseline (n=8,068) Average Post (n=9,571) Baseline (n=21,746) Average Post (n=16,714) Age (mean; std) 52.0 (9.8) 52.1 (9.6) 52.8 (9.5) 52.6 (9.6) 21 29 4.1% 3.8% 3.5% 3.9% 30 39 7.7% 7.5% 6.8% 6.9% 40 49 19.9% 20.7% 18.3% 18.5% 50 59 42.7% 43.2% 43.0% 43.1% 60 64 25.6% 24.8% 28.4% 27.6% Female (n;%) 5304 (65.7%) 6350 (66.3%) 14281 (65.7%) 10836 (64.8%) Race/Ethnicity (n;%) White, non Hispanic 2460 (30.5%) 2876 (30.1%) 7811 (35.9%) 6536 (39.1%) Black, non Hispanic 975 (12.1%) 1225 (12.8%) 4392 (20.2%) 3028 (18.1%)

PAGE 48

48 Table 2 1. Continued. Transition counties (STAR+PLUS) Comparison counties (FFS/PCCM) Baseline (n=8,068) Average Post (n=9,571) Baseline (n=8,068) Average Post (n=9,571) Hispanic 4159 (51.5%) 4834 (50.5%) 7492 (34.5%) 5684 (34.0%) Other 474 (5.9%) 637 (6.7%) 2051 (9.4%) 1466 (8.8%) Health status* (n;%) Healthy 216 (2.7%) 319 (3.3%) 543 (2.5%) 375 (2.2%) Significant Acute 45 (0.6%) 58 (0.6%) 90 (0.4%) 70 (0.4%) Minor Chronic 69 (0.9%) 90 (0.9%) 168 (0.8%) 142 (0.8%) Moderate Chronic 1197 (14.8%) 1329 (13.9%) 2833 (13.0%) 2160 (12.9%) Major Chronic 6541 (81.1%) 7776 (81.2%) 18112 (83.3%) 13967 (83.6%) Census tract poverty (mean; std) 23.1% (0.120) 23.1% (0.121) 26.1% (0.136) 26.0% (0.132) 0.0% 4.9% 3.5% 3.7% 2.6% 1.9% 5.0% 9.9% 8.8% 9.3% 7.4% 7.3% 10.0% 19.9% 33.4% 32.3% 27.9% 29.0% Poverty Area (20.0% 39.9%) 45.3% 45.7% 46.0% 45.9% Extreme poverty area (> 40.0%) 9.1% 9.1% 16.1% 15.9% County level median income (mean; std) $39,660 ($16,531) $39,977 ($16,780) $36,495 ($14,430) $35,930 ($12,927) Facility residence (n;%) 1751 (21.7%) 1876 (19.6%) 4849 (22.3%) 3677 (22.0%) Years eligible for study, 2006 2010 (mean; std) 3.90 (0.38) 3.63 (0.75) All descriptive statistics were stable over the study period. However, enrollees residing in STAR+PLUS counties were more likely to be Hispanic and less likely to be non Hispanic black or non Hispanic white compared with those in FFS/PCCM counties. STAR+ PLUS enrollees also resided in areas with slightly lower levels of poverty and higher median household income. Distributions of age, gender and health status were similar.

PAGE 49

49 Table 2 2. Enrollee and county sample sizes in transition and comparison counties, by measure. Measure Number of enrollees Transition Counties Comparison Counties Baseline Post Baseline* Baseline Post Baseline* Use of Appropriate Medication for People with Asthma 429 291 739 491 352 508 Medication Management for People with Asthma 363 322 599 393 274 386 Pharmacotherapy for COPD Exacerbation 280 308 513 1260 1,034 1,508 Cholesterol Management for People with Cardiovascular Conditions 1420 1,186 1,777 3173 2,989 3,788 Persistence of Beta Blocker Treatment after a Heart Attack 69 64 87 251 149 194 Comprehensive Diabetes Care 7293 7,827 9,709 20,168 13,036 17,045 Use of Appropriate Medication for People with Asthma 26 24 26 86 70 86 Medication Management for People with Asthma 26 24 26 79 62 79 Pharmacotherapy for COPD Exacerbation 28 25 28 133 121 133 Cholesterol Management for People with Cardiovascular Conditions 28 28 154 126 154 Persistence of Beta Blocker Treatment after a Heart Attack 11 8 11 46 41 46 Comprehensive Diabetes Care 28 28 207 177 207 In the Post Baseline period, data from four years were available; therefore, the range of enrollees/counties included in the analytic sample across these four years is given

PAGE 50

50 Table 2 2 describes the sample size range (enrollees and counties) by delivery system and outcome measure (see Appendix F, for full observation patterns by group and measure). The enrollee sample was largest for the diabetes care measure, reflecting relatively high prevalence. In cont rast, the smallest enrollee sample was for Persistence of Beta Blocker Treatment after a Heart Attack. Lower proportions of FFS/PCCM counties (225 in total) are represented in each measure compared to STAR+PLUS counties (28 in total), reflecting the small er populations of many rural FFS/PCCM counties. The sample sizes remained relatively stable over time for each measure. Table 2 3 provides unadjusted measure adherence rates by delivery system and time period. Measure adherence was similar betw een groups at baseline. However, baseline adherence rates varied widely between measures. For example, approximately 80% of enrollees with persistent asthma were dispensed at least asthma controller medication, whereas <50% of those enrollees remained on the medication for the majority of the baseline year. In 2006, members discharged with COPD exacerbation received sustained bronchodilator or systemic corticosteroid 55% of time. Members were given long term beta blocker therapy for approximately 50% of di scharg es after a heart attack

PAGE 51

51 Table 2 3. Unadjusted measure adherence (raw proportions) in transition and comparison counties during the baseline and post baseline periods. Transition counties (STAR+PLUS) Comparison counties (FFS/PCCM) HEDIS Measure Baseline Average Post 2007 2008 2009 2010 Baseline Average Post 2007 2008 2009 2010 Use of Appropriate Medication for People with Asthma 84.6% 82.9% 84.1% 84.9% 82.3% 81.1% 80.0% 77.8% 80.0% 77.8% 78.8% 75.0% Medication Management for People with Asthma 43.3% 42.9% 42.6% 42.8% 44.6% 41.6% 47.8% 47.0% 45.7% 47.8% 45.9% 48.8% Pharmacotherapy for COPD Exacerbation 55.5% 82.0% 73.2% 84.7% 82.3% 85.0% 56.6% 57.9% 55.5% 57.2% 58.5% 59.7% Cholesterol Management for People with Cardiovascular Conditions 74.6% 76.0% 61.6% 78.6% 79.1% 80.9% 76.1% 79.0% 76.8% 79.3% 79.9% 80.0% Persistence of Beta Blocker Treatment after a Heart Attack 50.7% 74.2% 54.7% 75.0% 82.8% 79.5% 52.6% 44.8% 47.9% 44.4% 43.6% 42.6% Comprehensive Diabetes Care 71.7% 73.1% 63.4% 75.0% 75.0% 77.4% 74.5% 75.8% 75.6% 76.1% 74.9% 76.7%

PAGE 52

52 Two of the 6 measures demonstrated sizable pre post improvements in STAR+PLUS, but not in FFS/PCCM. Beta blocker treatment after a heart attack increased from 51% in the baseline year to 74% in the post transition period for STAR+PLUS enrollees; while rat es for FFS/PCCM enrollees decreased from 53% to 45%. Similarly, the percentage of STAR+PLUS enrollees who received pharmacotherapy following a hospitalization or emergency department visit for COPD increased from 56% to 82%, while rates for FFS/PCCM enroll ees remained stable at 58%. Little change between the pre post periods was observed for the remaining measures (diabetes, asthma, and cholesterol management). Table 2 4 provides fit statistics and model based estimates of the predicted means between group s for the final reduced models. R 2 statistics ranged from 0.166 to 0.517 indicating fair to good model fit. Our model did not converge for the diabetes outcome, and so while the point estimates for this measure are reliable, the standard errors and model fit are undetermined. Also note that the effect of treatment was stable across time (i.e., there was no interaction between time and the STAR+PLUS program variable). Correspondingly, all model based estimates reflect the predicted county level average ef fects of STAR+PLUS implementation on measure adherence across the 4 post implementation years.

PAGE 53

53 Table 2 4. Model estimated average post baseline compliance in transition and comparison counties, and their differences, by measure. HEDIS Measure Transition County Proportion Comparison County Proportion Mean Difference (Transition Comparison) (99.167% CI) p value 0.05/6 = 0.00833) Final model R 2 (p value) Respiratory Condition Management Use of Appropriate Medication for People with Asthma 0.819 0.801 0.018 ( 0.128, 0.164) .724 0.278 (0.0006) Medication Management for People with Asthma 0.509 0.494 0.015 ( 0.132, 0.161) .785 0.1663 (0.0023) Pharmacotherapy for COPD Exacerbation 0.677 0.393 0.285 (0.216, 0.354) < .001 0.3582 (<.0001) Cardiovascular Condition Management Cholesterol Management for People with Cardiovascular Conditions 0.763 0.744 0.020 ( 0.239, 0.200) .697 0.517 (0.0023) Persistence of Beta Blocker Treatment after a Heart Attack 0.814 0.495 0.320 (0.068, 0.572) .001 0.283 (0.011) Diabetes Care Comprehensive Diabetes Care 0.618 0.638 0.020 ( 1.000, 1.000) .707 --The model based estimates, which were adjusted for control variables, administrative clustering, and correlation between outcomes over time, align closely with the unadjusted results. In STAR+PLUS counties, 28.5% (95% CI: 21.6%, 35.4%) more enrollees rece ived appropriate medication following COPD exacerbation compared to FFS/PCCM counties. In addition, receipt of beta blocker following heart

PAGE 54

54 attack discharge demonstrated improvement of 32.0% (95% CI; 6.8%, 57.2%) in the STAR+PLUS counties relative to the FFS/PCCM counties. We did not find statistically significant differences for the remaining measures (see Appendix G, for final model coefficients and standard errors). Discussion Measuring the impact of Medicaid managed care HCBS waiver programs is criti cal given rapid expansion of these models nationally. In this study, we examined the effect of the Texas STAR+PLUS HCBS waiver program on the quality of chronic disease care for Medicaid AWD. Our results demonstrate large and sustained improvements in ca re following both heart attack and COPD exacerbation. However, differences were not observed in the quality of ambulatory care for diabetics or asthmatics, or for cholesterol screening for those with cardiovascular conditions. Further research is necess ary to identify the pathways through which the observed improvements were achieved and the reasons why improvements were not seen in all the measures. However, it is worth noting that the 2 measures for which we found significant improvements, Persistence of Beta Blockers after a Heart Attack and Pharmacotherapy for COPD Exacerbation, focus on care processes linked to an acute event. Thus, it may be that managed care quality improvement protocols were more readily implemented in the context of an acute ev ent. For example, prior authorization is sought during inpatient admissions concomitant with an acute event, thereby providing care coordinators a near real time opportunity to influence care. In contrast, difficulty with patient follow up for the remain ing measures may be a particularly important factor contributing to the lack of significant findings. The immediacy of costs

PAGE 55

55 of poor quality may also be an important motivator for these differences. For example, with the acutely linked care examined here the costs of poor quality care are realized quickly in re admissions and emergency department visits, while the consequences of inadequate ambulatory diabetes care or lipid screening occur over longer time spans. Our study has several strengths. First, the study was conducted in Texas, which has the second largest Medicaid program in the US. The population in STAR+PLUS is racially and ethnically diverse providing greater insight into the effects of a comprehensive HCBS program in a broad populati on. Second, this study is based on a natural experiment in which STAR+PLUS was mandatorily phased into different counties for Medicaid AWD, providing a high level of internal and ecological validity. As in all observational studies, there is possibility of residual confounding. However, bias introduced by non random phase in is largely attenuated through the study design, which included the baseline value of the outcome and modeled separate slopes for each study group. To threaten internal validity, an external influence would have (1) needed to mirror the implementation of STAR+PLUS, that is occurring only in the transitioned counties and during the same period under study in this analysis and (2) not been closely correlated to the repeatedly measured c ontextual and individual control variables. 80 Our study also has limitations that should be considered when interpreting the results. First, the NCQA certified software that we used to calculate the chronic care measures uses health care claims and encou nter data, the quality of which may be affected by coding practices. As part of ongoing quality of care evaluation for the Texas Medicaid program, we conduct encounter data validation of the administrative claims

PAGE 56

56 data against medical records following CMS external quality review protocols. 81 A random sample of >1,100 medical records for STAR+PLUS are reviewed annually by certified medical record coders and compared to claims/encounter data fields (e.g., ICD 9 CM codes, CPT codes, date of service, place o f service, and rendering provider) with >92% agreement, lending confidence in the data quality. Second, Medicaid managed care HCBS waiver programs implemented through MCOs differ throughout the U.S. Therefore, it is possible that the findings in our stud y are not generalizable to other Medicaid programs. Even so, the pattern of improved care linked to specific acute events, versus that delivered in routine care settings is seen frequently. 82 Further, information about the structure of the STAR+PLUS prog ram is available, and policymakers and health care providers can examine the extent to which the program design characteristics are similar to existing or proposed programs. 8,83 84 Given variability in HCBS programs, future research should examine specifi c types of HCBS received and their association with quality of care. It also would be prudent to explore potential heterogeneity in program implementation and see if this heterogeneity leads to differences in quality improvements between health plans. Fi nally, this study examined process of care measures; Future work should examine the extent to which these findings translate into improved health outcomes. In summary, in one large Medicaid managed care HCBS program, the quality of chronic disease care lin ked to acute events improved while that provided during routine encounters appeared unaffected. Additional research is needed to further evaluate and refine care for this vulnerable population.

PAGE 57

57 CHAPTER 3 IMPACT OF STAR+PLUS ON BEHAVIORAL HEALTH CARE Q UALITY Background and Significance Mental and substance use disorders (i.e. behavioral illnesses) are the leading contributors to disability worldwide. 24,8 5,86 These disorders portend decreased quality of life, decreased functioning, and dramatic rates of premature mortality. 87,88 Yet despite the profound individual and societal burdens, persons with behavioral illness are much less likely to receive behavioral health treatment or guideline recommended care. 90 92 In the US, Medicaid is the largest source of funding for behavioral health services 93 with an estimated 35% of non elderly enrollees having a chronic mental illness and 11.5% suffering from a substance use disorder. 94,53 Most of these individuals qualify for Medicaid by meeting eligibility requirements for disability, effectively concentra ting those with high levels of behavioral illness severity and persistence, and high prevalence of co occurring physical illness, in this group. 11,13,95 In this context, routine and timely behavioral health care and management, especially integrated with o ther needed health services, is critical. For example, in addition to acute behavioral health and medical illness needs, many Medicaid enrollees with disabilities regularly receive long term services and supports (LTSS), including assistance with activitie s of daily living, nursing facility care, adult daycare programs, home health aide services, personal care services, transportation, and supported employment. Correspondingly, lack of coordination of physical and mental health services as well as long term and acute services are often cited as key barriers to improving quality and value of care for this population. 52,90,96

PAGE 58

58 In response, and motivated to restrain or ensure greater predictability in state Medicaid budgets, policy makers have experimented with variations in Medicaid service delivery and financing for enrollees qualifying due to disability. Increasingly, states have implemented programs combining traditional comprehensive managed care for acute and outpatient services with managed LTSS, either pr ovided in an institutional setting or through home community based service (HCBS) alternatives. 10,56,70 For non elderly Medicaid enrollees with disabilities, this reform has occurred primarily through 1915 (b)/(c) managed care/HCBS waivers and 1115 demonst ration waivers, increasing in use from 8 states in 2004 to 18 states in 2014. 58 Although there exists considerable research on the effects of Medicaid managed care delivery for children, their parents and other relatively healthier low income beneficiaries information on managed care for enrollees with severe and persistent physical and behavioral health conditions (i.e. those qualifying due to disability) is limited. In particular, there is an absence of research on care for behavioral health conditions, particularly when examining services delivered through managed care systems that encompass managed LTSS, including HCBS. 97 The purpose of this study was to examine the effects of a large acute care and HCBS program delivered through managed care organizati ons (MCOs) in Texas Medicaid, the STAR+PLUS program, on the quality of behavioral health care for individuals qualifying for Medicaid due to disability. A primary focus of STAR+PLUS is to improve care by delivering acute services and LTSS through a single system; employing service coordinators who develop individual care plans and assist enrollees in receiving needed care; and emphasizing HCBS alternatives to institutional care. 98 At

PAGE 59

59 the same time, STAR+PLUS employs a fully capitated health care delivery mo del, which can result in ambiguous effects on care (including behavioral health care) provision. For example, MCOs often employ mechanisms to restrict access to certain types of care, namely expensive or low value care, yet efficacious or high value servic es may be simultaneously limited. Conversely, MCOs typically implement programs to increase disease management and receipt of recommended care, in order to slow health declines and prevent future higher cost utilization. The importance of encouraging more timely and routine behavioral health care seems especially salient given the role of behavioral health on overall functioning, costs and care for co existing physical conditions. Thus, we hypothesized that the quality of behavioral health care would impro ve after STAR+PLUS enrollment and relative to a comparison group that remained enrolled in FFS or Primary Care Case Management (PCCM). Methods Overview Over the past two decades, the state of Texas has legislatively mandated the transition of Medicaid ser vice delivery from FFS and PCCM models to STAR+PLUS for adults receiving SSI for their disabilities. STAR+PLUS was first implemented in Harris County in 1998. Nine years later, in January and February of 2007, STAR+PLUS expanded to 28 additional counties. In February and September of 2011, and in March of 2012 STAR+PLUS was expanded to an additional 13, 21 and 27 counties, respectively. 62 164 counties, comprising the Medicaid Rural Service A reas (MRSAs). (Figure s 1 3 and 1 4 provides more detail on implementation of the STAR+PLUS on the timing and

PAGE 60

60 counties affected by the transition). This phased implementation by county provides an opportune natural experiment that can be analyzed with high levels of ecological and internal validity. The approach used in this study is similar to that described previously. 99 We estimate the intermediate term (4 year) changes in behavioral health care for the initial primary STAR+PLUS implementation: January/February 2007. To accomplish this, we used a generalization of the commonly employed difference in difference (DD) framew ork also known as a longitudinal time trend or pre post design with comparison group. More specifically, among eligible enrollees living in the counties which transition to STAR+PLUS, we estimated the change between the outcome measured during the post i mplementation period and the outcome measured in the pre implementation period, adjusted for potential confounding factors as described below. Then, we compared this difference to the corresponding adjusted change experienced by eligible enrollees living i n counties remaining in FFS or PCCM models during the same time period. To improve communication of the analysis plan and proposed group composition for the analysis we defined 6 STAR+PLUS implementation groups (A F) comprising all Texas counties. Figure 1 4 displays these groups in an implementation graphic; this type of figure is commonly reported in stepped wedge designs, which share many features of this study, except for the non randomized ordering of program implementation. (Appendix A provides speci fic detail on the counties included in each group).

PAGE 61

61 To examine the intermediate term effects of the January/February 2007 implementation, outcomes reflecting 2006 (pre) and 2010 (post) were utilized. A single post implementation year was selected to limit computational complexity of modeling a n imbalanced longitudinal cluster design; the latest year (2010) for which data was available was selected to ensure sufficient time for implementation of the program. The experimental group was comprised of group B, t he 28 counties in which STAR+PLUS was implemented in early 2007. The comparison group was selected from counties that were legislated to transition to STAR+PLUS after 2010, groups D, E and F. Note that group C (comprised of Dallas and Tarrant service areas ) was not used in assembling comparison counties because Texas implemented an Integrated Care Management (ICM) program in these regions from February 2008 May 2009. The ICM program is comparable in form to a non capitated version of STAR+PLUS. 69 Enrolle es in a non transitioned area were receiving care from either FFS or PCCM health care delivery models. We did not distinguish between FFS and PCCM in our analyses based on research indicating few differences among adult Medicaid enrollees and those with di sabilities in particular 62,100 102 In addition, comparison of the control variables and pre implementation quality of care between FFS and PCCM this study. Sample and Data Since only enrollees >21 years old were mandatorily transitioned to STAR+PLUS and Medicare data was unavailable for our sample, we restricted our analysis to adults 21 64 years of age who qualified for Medicaid due to disability, but who were not dua l eligible. This sample is further evaluated for eligibility for each outcome measure,

PAGE 62

62 according to HEDIS definitions, resulting in different samples for each outcome analysis. Person level administrative enrollment and encounter data provided by the Texa s race/ethnicity, county, monthly enrollment, and delivery model. Enrollment records were linked to professional, inpatient and pharmacy encounter data that included Internationa l Classification of Diseases (ICD 9 CM) diagnosis codes, Current Procedural Terminology codes, and National Drug Codes. These data were supplemented with county Zip Code Tabulation Area and census tract level data from the Area Resource File, US Census Bureau, and US Department of Commerce to capture geographic contextual factors. Outcome Measures Encounter data was assessed using National Committee for Quality Assurance (NCQA) certified software to calculate enrollee level HEDIS behavioral health care measures for the calendar years 2006 and 2010. We selected 3 measures, each with 2 submeasures: 7 and 30 day follow up after inpatient hospitalization for mental illness; initiation of and engagement in alcohol and other drug dependence treatment after a new episode of dependence; and short (12 weeks) and longer term (6 months) coverage of antidepressants for those with depression. These measures were selected because they reflect quality of care for the Agency for Health Care Research and 103 and were suggested by the Centers for Medicare & Medicaid Services as initial health care quality core measures for adults enrolled in Medicaid. 104 Strong performance on these indicators is linked to improved health outcomes. 105 The dich otomous indicator of compliance for each submeasure was

PAGE 63

63 modeled and reported separately to provide more nuanced information on the quality of care. Explanatory and Control Variables For each enrollee year, we computed the main predictor variable, time enr olled in STAR+PLUS in a given post transition year, thus capturing any instances of differential transition timing and scale up or partial Medicaid enrollment. Although the pre post design with comparison group provides some protection against selection bi as, we controlled for enrollee level and county level characteristics which could influence the outcome and associate with enrollment in STAR+PLUS, as model. 106,107 Age gender, and race/ethnicity (non Hispanic white, non Hispanic black, Hispanic and other) were assigned using demographic information contained in the enrollment files. Because we do not have a specific indicator for group living facility, we constructed a variable to indicate if 5 or more enrollees were residing at the same address. This variable allowed us to distinguish between those residing in an individual residence as opposed to a group living arrangement. Health status was measured using the 3M Clinical Risk Groups (CRGs), which uses ICD 9 CM diagnosis codes from health care encounters to assign individuals to hierarchically defined health status groups 108 To be classified, individuals must have been enrolled for at least 6 months. Individuals we re classified into the following health status categories: healthy, significant acute conditions (e.g., chest pain), minor chronic conditions (e.g., migraine), moderate chronic conditions (e.g., asthma and diabetes), or major chronic conditions (e.g., cyst ic fibrosis and cancer). Those who did not have a

PAGE 64

64 sufficient enrollment history to be classified into one of the preceding categories were treated as missing values. Contextual variables included the percentage of the population living in poverty, percent age of the population unemployed, and median household income, all assigned using the Zip Code Tabulation Area if valid census tract was unavailable). Median county income was also included. These data were obtained from the Area Health Resource File, U.S. Census Bureau and U.S. Department of Commerce. We also included a variable measuring density of Medicaid health care utilization (Medicaid inpatient discharges per Medicaid enrollee), assigned by county, from the Area Health Resource File. The full models also included 2 3 and 4 way predictor interactions, between race/ethnicity, gender and the main explanatory variable. Clustering Unit Although service areas represent administrative clustering and the approximate unit a t which STAR+PLUS was implemented, there are too few units to have sufficient statistical power. In addition, some service areas were transitioned in phases. Accordingly, we selected an intermediate clustering unit (county), which provides sufficient power and accounts for potential non independence among enrollees. The county reasonably represents the context within which enrollees seek health care and captures administrative similarity of the care environment, such as available health care resources that impact health service use. Empirical Model and Analyses We employed a longitudinal design widely used in analyzing clinical trials. This design was analyzed using a 2 stage, multilevel approach using general linear

PAGE 65

65 models. 78 First, we computed person leve l models separately for the post transition year (2010) and outcome, assuming compound symmetry correlation structure between participants nested by county. Using individually adjusted averages and weights derived from cluster sizes and intracluster correl ation, we then built the second tier model for each outcome. We applied standard collinearity approaches to reduce the predictor space when computationally necessary. 109 Fixed order backwards selection of predictor variables with = 0.05 was used to arrive at the reduced models. Assumption diagnostics were conducted for the final models; no major concerns with linearity, homogeneity or Gaussian distribution of residuals were noted We performed sensitivity analyses of our results, whereby influence points if detected were removed Predicted marginal means, including 95% confidence intervals were produced for the main explanatory variable. Analyses were conducted using R (Version 3.3.0 ; ) and Rstudio (Version 0.99.902) using lme4 ( Version 1.1.12) and lsmeans (Version 2.23) packages for the main analyses and model summarization. 110 113 It is important to note that this modeling approach is a mathematical generalization of a difference in difference (DD) approach, where the standard D D model compares differences between 2 groups at 2 timepoints to isolate and test the presence of an effect, assumed to be a deviation from the baseline difference. 114 Our generalized approach relaxes the DD parallel trend assumption requiring similar rate s of growth between groups by allowing for differing slopes. This approach also improves computational convergence for unbalanced cluster a characteristic of our data due to extremely unequal county sizes. See Appendix C for more detail on this novel mod eling approach for analyzing longitudinal clustered designs.

PAGE 66

66 Recent analyses of quality of care measures using data derived from the entire population reveal low levels of enrollee month missingness (<5%) for response or main individual level covariates. 99 For these data, list wise deletion was performed. Results The overall study sample was comprised of 9645 enrollees clustered across 214 Texas counties, with ~90% cases with complete data. The subgroup of enrollees eligible for the i nitiation and e ngagement in a lcohol/ d rug t reatment measures represented the largest sub group, with 6529 enrollees Table 3 1 provides a summary of demographic, health and contextual characteristics for the study population eligible for at least one of the studied measure s, a s well as by measure subgroup Table 3 1. Demographic health and contextual characteristics for the study population Combined Antidepressant Medication Management S ubgroup Follow up After Hospitalization for Mental Illness S ubgroup Initiation and Engagement of Alcohol and Other Drug Dependence Treatment S ubgroup Unique enrollees 9645 1291 3150 6529 Unique counties 214 146 173 205 Enrollees with complete data 8661 (89.8) 1196 (92.6) 2775 (88.1) 5893 (90.3) Counties with complete data, by year 209 (97.7) 140 (95.9) 165 (95.4) 201 (98) Counties with complete data in post and pre years 140 (95.9) 165 (95.4) 201 (98) Age [mean (SD)] 44.4 (11.3) 46.1 (11.6) 41 (11.5) 44.9 (10.9) Age categories [n (%)] 21 30 1333 (13.8) 156 (12.1) 664 (21.1) 778 (11.9)

PAGE 67

67 Table 3 1. Continued Combined Antidepressant Medication Management Subgroup Follow up After Hospitalization for Mental Illness Subgroup Initiation and Engagement of Alcohol and Other Drug Dependence Treatment Subgroup 30 39 1674 (17.4) 199 (15.4) 683 (21.7) 1106 (16.9) 40 49 3081 (31.9) 368 (28.5) 981 (31.1) 2168 (33.2) 50 59 2739 (28.4) 395 (30.6) 668 (21.2) 1931 (29.6) 60 64 818 (8.5) 173 (13.4) 154 (4.9) 546 (8.4) Female [n (%)] 4979 (51.6) 904 (70) 1854 (58.9) 2912 (44.6) Race/ethnicity [n (%)] Non Hispanic black 1818 (18.8) 169 (13.1) 561 (17.8) 1339 (20.5) Hispanic 3030 (31.4) 544 (42.1) 959 (30.4) 1973 (30.2) Other 548 (5.7) 87 (6.7) 182 (5.8) 337 (5.2) Non Hispanic white 4249 (44.1) 491 (38) 1448 (46) 2880 (44.1) Health status [n (%)] Healthy 278 (3.1) 69 (5.7) 3 (0.1) 211 (3.5) Significant acute 91 (1) 14 (1.1) 0 (0) 77 (1.3) Minor chronic 182 (2) 83 (6.8) 2 (0.1) 100 (1.7) Moderate chronic 2208 (24.9) 274 (22.5) 759 (26.6) 1524 (25.2) Major chronic 6124 (68.9) 779 (63.9) 2094 (73.3) 4144 (68.4) Health status not available 762 (7.9) 72 (5.6) 292 (9.3) 473 (7.2) County metro classification [n (%)] Large metro (>1M) 3176 (32.9) 461 (35.7) 1157 (36.7) 2097 (32.1) Medium metro (250K 1M) 2260 (23.4) 386 (29.9) 719 (22.8) 1431 (21.9) Small metro (<250k) 1969 (20.4) 215 (16.7) 622 (19.7) 1365 (20.9)

PAGE 68

68 Table 3 1. Continued Combined Antidepressant Medication Management Subgroup Follow up After Hospitalization for Mental Illness Subgroup Initiation and Engagement of Alcohol and Other Drug Dependence Treatment Subgroup Adjacent to metro 1604 (16.6) 154 (11.9) 465 (14.8) 1190 (18.2) Nonadjacent to metro (>20k) 350 (3.6) 36 (2.8) 119 (3.8) 233 (3.6) Nonadjacent to metro (<20k) 286 (3) 39 (3) 68 (2.2) 213 (3.3) Census tract poverty [mean (SD)] 23.9 (12.8) 25.7 (13.5) 23.2 (12.8) 23.9 (12.6) Census tract poverty categories [n (%)] 0.0% 4.9% 274 (2.9) 33 (2.6) 97 (3.1) 170 (2.7) 5.0% 9.9% 844 (8.9) 97 (7.6) 324 (10.5) 550 (8.6) 10.0% 19.9% 3103 (32.7) 367 (28.8) 1024 (33.2) 2115 (33) Poverty area (20.0% 39.9%) 4121 (43.5) 579 (45.5) 1292 (41.9) 2834 (44.2) Extreme poverty area (> 40.0%) 1135 (12) 197 (15.5) 345 (11.2) 743 (11.6) Census tract not available 168 (1.7) 18 (1.4) 68 (2.2) 117 (1.8) Census tract unemployment level [n (%)] 0.0% 4.9% 1804 (19) 222 (17.4) 603 (19.6) 1213 (18.9) 5.0% 9.9% 4588 (48.4) 617 (48.5) 1460 (47.4) 3126 (48.8) 10.0% 19.9% 2931 (30.9) 421 (33.1) 969 (31.5) 1957 (30.5) > 20.0% 153 (1.6) 13 (1) 49 (1.6) 115 (1.8) Census tract not available 169 (1.8) 18 (1.4) 69 (2.2) 118 (1.8)

PAGE 69

69 Table 3 1. Continued Combined Antidepressant Medication Management Subgroup Follow up After Hospitalization for Mental Illness Subgroup Initiation and Engagement of Alcohol and Other Drug Dependence Treatment Subgroup Census tract household income [mean (SD)] 38215.6 (15467.6) 36589.6 (14893.3) 39124 (15789.4) 37957.5 (15218.1) County Medicaid discharge density [mean (SD)] 1 (0.6) 1.1 (0.6) 1 (0.6) 1 (0.6) Census tract determined from address [n (%)] 7669 (79.5) 1045 (80.9) 2473 (78.5) 5190 (79.5) Facility residence [n (%)] 2036 (21.4) 273 (21.4) 716 (23) 1323 (20.6) Months enrolled in Medicaid [n (%)] 12 7585 (78.6) 1109 (85.9) 2408 (76.4) 5159 (79) 9 11 1016 (10.5) 103 (8) 350 (11.1) 714 (10.9) 5 8 628 (6.5) 36 (2.8) 227 (7.2) 418 (6.4) 2 5 414 (4.3) 41 (3.2) 165 (5.2) 238 (3.6) 1 2 (0) 2 (0.2) 0 (0) 0 (0) STAR+PLUS transition cohort [n (%)] A 0 (0) 0 (0) 0 (0) 0 (0) B 3886 (40.3) 554 (42.9) 1420 (45.1) 2553 (39.1) C 0 (0) 0 (0) 0 (0) 0 (0) D 761 (7.9) 87 (6.7) 239 (7.6) 547 (8.4) E 1720 (17.8) 318 (24.6) 464 (14.7) 1114 (17.1) F 3278 (34) 332 (25.7) 1027 (32.6) 2315 (35.5) There are several important aspects to note. First, as expected, the study population has significant morbidity, with approximately two thirds belonging to the most severe health status category. Second, more than half of the study population lives in locations deemed poverty or extreme poverty areas, portending the challenging context

PAGE 70

70 in which these patients live. Third, nearly 20% of the study population lives at addresses with 5 or more enrollees, suggesting either a high degree of institutionalization or high levels of residential clustering in this population. Fourth, we observed high levels of continuous enrollment, with more than 85% of the sample enrolled in Medicaid for more than 9 months in the year. Almost the entire study population lives in metro politan or adjacent areas. The majority of the study population is non white or Hispanic In addition, many sample characteristics vary by the measure being evaluated, reflecting differences in the subpopulations being diagnosed with or presenting for the qualifying conditions (i.e. depression, mental illness hospitalization, and drug/alcohol hospitalization/emergency department visit). Those hospitalized for mental illness are younger, with more than 40% under 40 years of age Seventy percent of those qualifying for the antidepressant management measure are female, compared to only 45% of the subgroup presenting acutely with an alcohol/drug use disorder. Table E 1 demonstrates that only a small proportion (~20%) of the study population was eligible for multiple measures T he most common dyad or triad was having both a qualifying mental illness hospitalization and a substance abuse diagnosis during a hospit alization/emergency department visit (14%). Tables E2 4 present separate descriptive statistics for the transition and comparison groups for each measure. The large majority of characteristics are very similar between groups, except for race/ethnicity and metro politan classification. Enrollees from transition counties were almost twice as likely to be Hispanic ; much of this difference is attributable to a lesser proportion of white, non Hispanic enrollees. Enrollees in transition counties were more likely l iving in large metro politan areas (~80%) as compared to those from comparison

PAGE 71

71 counties who were living in medium, small and adjacent metro politan areas. Of note, the characteristics demonstrated stability over time. For patients hospitalized with mental il lness individuals were eligible for the measure multiple times per year, however, three quarters met eligibility criteria only once, and ~1 6 % qualified for the measure twice. The total sample size for each measure (over time) by group is presented in Figure D 1 and a summary of the number of observations per county is presented in Table E 6. Changes in sample size over time and the degree of clustering by county were similar be tween the groups. Figure D 2 provides a spatial description of the sample sizes by group. Table 3 2. Weighted, unadjusted measure performance Measure Type Measure Sub Group 2006 2007 2008 2009 2010 Antidepressant Medication Management 3 months Comparison 64.0 58.5 55.7 57.4 54.5 Transition 54.6 57.0 51.8 53.0 52.2 6 months Comparison 45.1 43.4 38.0 39.4 39.3 Transition 36.2 36.1 38.5 37.4 39.7 Follow up After Hospitalization for Mental Illness 7 day Comparison 18.2 18.9 18.4 20.3 20.5 Transition 20.7 11.2 14.9 22.7 27.3 30 day Comparison 41.1 40.6 36.8 44.8 45.6 Transition 43.7 19.0 27.4 49.4 51.8 Initiation and Engagement of Alcohol and Other Drug Dependence Treatment Initiation Comparison 46.4 46.6 36.8 34.4 32.8 Transition 45.4 44.8 38.0 34.8 35.5 Engagement Comparison 4.2 2.7 3.7 2.5 3.2 Transition 3.9 2.8 4.8 3.0 5.8 Weighted, u nadjusted outcome measure performance is illustrated in Table 3 2 and Figure D 3. Across all measures, compliance is rarely greater than 50%, except for the measure reflecting antidepressant medication receipt for the 3 month period following a new depression diagnosis. For some measures, compliance is ver y low,

PAGE 72

72 including rates of follow up within 7 days following hos pitalization for mental illness and intermediate term engagement in alcohol/drug treatment. During the baseline year, transition and comparison counties appeared to have similar rates of foll ow up after hospitalization for mental illness and ini ti ation/engagement of substance abuse treatment However, during the same baseline period, enrollees in transition counties were about 10% less likely than those in comparison counties to receive antide pressant mediations. There was notable variation in measure performance over the transition period. Table 3 3 provides the predicted marginal means during 2010 for each measure Of note, differences between comparison and transitioned counties were not sig nificant, except for engagement in substance abuse treatment following a hospitalization/emergency visit with substance abuse diagnosis. For this measure, t ransition counties we re twice as likely to engage eligible patients in substance abuse care -that includes both initiating substance abuse care, and ensuring receipt of two or more additional related services within the subsequent 30 days. Yet, expressed on an absolute scale, the difference due STAR+PLUS is small, at 3%. Sensitivity analyses involving the removal of influential points was not found to substantially impact the se results. Table 3 3. Predicted marginal means during 2010, by measure. Measure Measure Sub FFS/PCCM (Comparison) STAR+PLUS (Transition) Difference p value Antidepressant Medication Management 3 months 0.55 (0.52 0.58) 0.51 (0.46 0.55) 0.04 0.119 6 months 0.39 (0.35 0.42) 0.39 (0.34 0.44) 0.00 0.909

PAGE 73

73 Table 3 3. Continued Measure Measure Sub FFS/PCCM (Comparison) STAR+PLUS (Transition) Difference p value Follow up After Hospitalization for Mental Illness 7 day 0.22 (0.20 0.25) 0.25 (0.19 0.31) 0.02 0.491 30 day 0.47 (0.44 0.50) 0.52 (0.47 0.58) 0.05 0.097 Initiation and Engagement of Alcohol and Other Drug Dependence Treatment Initiation 0.32 (0.30 0.34) 0.34 (0.30 0.38) 0.03 0.241 Engagement 0.03 (0.03 0.04) 0.06 (0.05 0.07) 0.03 0.000 Discussion Measuring the impact of Medicaid managed care and HCBS on quality of behavioral health care is critical, given the high prevalence, cost and quality of life impacts associated with poorly managed behavioral health conditions. 115 In this study, we found that the Texas STAR+PLUS program did not affect rates of follow up after mental illness hospitalization or linkage to substance abuse treatment after a new substance abuse diagnosis. Additionally, STAR+PLUS was not associated with increased antidepressant medicat ion receipt within either 3 or 6 months following a new depressio n diagnosis. However, STAR+PLUS did significantly increase rates of more sustained engagement with substance abuse treatment, although the absolute increase was small (3%). To our knowledge, no studies have examined the impact of managed care on behavioral health care delivery for Medicaid patients with disabilities. Recently however a few studies have examined the role of various emerging payment reforms on

PAGE 74

74 behavioral health care. For example, Busch 116 found that implementation of accountable care organiza tions in Medicare was not associated with changes in follow up after mental health admissions, rates of depression diagnosis or mental health readmissions spending. Barry 117 found slight decreases in mental health care use, but not in mental health spendin g in an analysis of a global payment model in Massachusetts. Thus, similar to our analysis, global payment reforms have not been shown to have substantial impact on behavioral health care delivery. There are several explanations for the limited STAR+PLUS effect on the selected behavioral health measures. First, all of these measures essentially require engagement in outpatient care after an acute episode -an episode marking a period of considerable instability and morbidity in a patient's life. Although STAR+PLUS utilized care coordinators and other follow up supports, it is unclear whether such services were targeted towards acute behavioral health needs. Complex social and economic barriers often occurring in Medicaid populations could also have diminis hed the impact of any provided supports. Furthermore, the very conditions for which the enrollees need treatment mental illness or substance abuse disorder can lead to cognitive and behavioral impediments to engagement in care, especially in comparison to many physical health conditions. Thus, the low rates of measure performance may require considerable system redesign or additional investments to realize any substantial gains. In a related manner, there is a well known shortage of mental health provi ders, particularly those serving Medicaid, and this may inhibit the formation and growth of satisfactory managed care organization p rovider networks 118 Many managed care organizations have limited experience delivering behavioral health services, as these

PAGE 75

75 provisions have historically been carved out. Finally, it is likely that managed care organizations, similar to accountable care organizations in Medic a re, were focused on alternative health care priorities during initial implementation. 116 In addition t o examining the impact of STAR+PLUS, we found unsatisfactory performance for all behavioral health measures, regardless of study group. This poor performance was fairly stable over time. Even further, several rates were substantially below national average s published by NCQA. 119 For example, in 2010, 43% of Medicaid managed care patients nationally received follow up after hospitalization for a mental illness within 7 days, but only 22 25% in our sample received this level of care. Rates of initiation of su bstance abuse treatment in our sample were also lower than national rates (32 34% vs 43%), and rates of intermediate term engagement were roughly one third the national average (3 6% vs 14%). Receipt of antidepressant medications after a new depression dia gnosis approximated national rates of 50% and 34% at 3 and 6 months, respectively. However, even meeting these low national averages is insufficient, since this means that two thirds of patients with depression have not received sufficient antidepressant t reatment in the 6 months following their diagnosis. The implementation of the Mental Health Parity and Addiction Equity Act of 2008, for which final rules for Medicaid were published in March 2016, may signal a shift in focus towards behavioral health care although many remain skeptical that this will result in substantial change. 120 121 Our study offers several contributions to the literature. It focuses on non elderly adults in Medicaid qualifying due to disability -a historically understudied group. Second, most research on HCBS has examined health care expenditures or access to

PAGE 76

76 care rather than specific quality indicators, particularly for behavioral health. 5,17 Third, we study a program with mandatory enrollment, overcoming the limitation of potenti al selection bias in prior research on waiver programs with voluntary enrollment. Several considerations are necessary to contextualize these results. First, Medicaid managed care programs differ throughout the United States. For example, Duggan and Hayfo rd 122 discuss the importance of financing and organizational arrangements in moderating the effect of Medicaid managed care on expenditures. Therefore, it is possible that our findings may not be generalizable to other Medicaid programs. Even so, the infor mation gained will be directly relevant for more than 1 million adults in Texas Medicaid program. The population in STAR+ PLUS is also racially and ethnically diverse providing greater insight into the effects of a comprehensive HCBS program in a broad population. Further, information about the structure of the STAR+PLUS program is available, and policymakers and health care providers can examine the extent to which the program design characteristics are similar to existing or proposed programs. The possibility of residual confounding must also be considered, particularly as it relates to the appropriateness of the comparison group. The Texas legislature selected areas for initial implementation based on the presence of a strong health care infrastructure to increas e the likelihood of successful program implementation. As presented above, these were more likely to include larger metropolitan areas. However, few large metropolitan areas were present in the comparison sample. This was mainly due to the exclusion of the Dallas Fort Worth metropolitan area, as these counties were undergoing a separate health care delivery reform pilot from 2008 2010. If metropolitan

PAGE 77

77 size is related to differential change in measure performance over time, the limited overlap between the gr oups on this characteristic may predispose to bias. Finally, because we have focused on only one post transition year, it is possible that the results differ by measure timing, and may not reflect the average change due to the transition or shorter ter m differences (1 3 years post transition). However, review of weighted, unadjusted compliance suggests that large effects are unlikely. Furthermore, the absence of effect at the 4th year post transition is a meaningful resu lt and suggests that STAR+PLUS do es not have extended effects on quality of behavioral health care.

PAGE 78

78 CHAPTER 4 IMPACT OF STAR+PLUS ON RECEI PT OF PREVENTATIVE CARE Background and Significance Managed care has received continued attention from state policy makers as a model for delivering h ealth care services to individuals qualifying for Medicaid due to disability. 57,64 Increasingly these models have combined managed acute care with managed long term services and supports, including home and community based service alternatives, through dua l 1915(b)/(c) or 1115 waivers. 9 These trends continue despite limited evidence of the programs' impacts on access to care and preventive care provision. Managed care is premised on promoting more efficient utilization of health care services by encouraging uptake of high value preventive and ambulatory care to avert, attentuate or possibly substitute for costly emergency and inpatient based care. This health care delivery model holds particular promise for individuals with disabilities, who exper ience substantial gaps in receipt of guideline recommended screening and preventive care. 123 125 Those qualifying for Medicaid due to disability also experience well documented challenges in accessing ambulatory and special t y physican care. 59,66 The import ance of timely and routinue care is particulary pronounced given the high prevelence of multiple risk factors for cardiovascular and cancer morbidity and mortality in this population such as high rates of smoking, obesity, and limited physical activity. 12 6,127 Despite this promise, there is concern that commonly employed managed care tools (e.g. pre authorization including required referrals, provider networks, formularies) aimed to reduce ineffcient care will simulatanously limit access to and utilization of

PAGE 79

79 desired or high value services. Under fixed revenue streams (i.e. capitation), this organizational behavior might be particularly incentivized because already low provider rates restrict the ability of managed care organizations to negotiate down pric es. Furthermore, under a short term (i.e. monthly) capitation horizon, the increased costs associated with disease progression due to inadequate preventive care may not be borne by managed care organizations. Of the limited literature on managed care for a dults qualifying due to disability, the results on access and receipt of preventive care have been null or uncertain. Burns 59 found that beneficiaries in mandatory managed care were more likely to report waiting to see a provider, difficulty obtaining spec ialty care and not receiving a flu shot. Yet these same beneficiaries were also more likely to have a usual source of care compared to enrollees receiving care through Fee For Service (FFS), a traditional delivery model. Using the National Health Interview Survey, Coughlin, Long and Graves 62 explored access to care, flu shot receipt and use of primary care providers, specialty physicians and the emergency department for disabled Medicaid enrollees by ealed increases in the usual source of care reported by Medicaid adults with disabilities in managed care counties. A recent analysis by Caswell and Long 128 was unable to detect differences in access or expenditures for disabled enrollees, likely due to a relatively small overall sample (n = 1000). In a study by our group using enrollee level encounter data from Texas, we found mixed improvements in quality of chronic disease care resulting from managed care implementation, relative to FFS and Patient Cente red Case Management (PCCM).

PAGE 80

80 There were increases in quality of care following acute inpatient events, but no changes in ambulatory care for diabetes, asthma and cardiovascular conditions. 99 Responding to the call by Caswell and Long 128 and expanding on ou r previous work, this study seeks to provide information on the impacts of a compreshensive managed care program in Texas Medicaid the STAR+PLUS program on the receipt of screening for cancer and access to ambulatory care, for individuals qualifying fo r Medicaid due to disability. STAR+PLUS is more extensively described elsewhere. 98 In brief, the program delivers acute and long term services by contracting with managed care organizations (MCOs). These MCOs develop primary and specialty provider networks, employ service coordinators who develop individual care plans and assist enrollees in receiving needed medical and community services, implement disease management programs, and emphasize HCBS alternatives to institutional care. 98 Given the inte nded role of managed care on promoting higher value care, we hypothesized that receipt of recommended cancer screenings and access to care would increase after STAR+PLUS enrollment relative to a comparison group that remained enrolled in FFS or PCCM. Meth ods Overview In 1998, the state of Texas piloted a new Medicaid service delivery model in Harris county by contracting with managed care organizations to provide acute and long term supports to enrollees qualifying due to disability in lieu of the previous FFS and PCCM delivery models Following continued cost pressures, the Texas legislature mandated the expansion of t his program, STAR+PLUS to 28 additional counties in early 2007. 62 Since this time, STAR+PLUS has been expanded to all Texas counties

PAGE 81

81 We defined 6 STAR+PLUS implementation groups (A F) to aid in communication of these transitions and our ultimate analytic design. (Figure 1 3 provides more detail on implementation of the STAR+PLUS on the timing and counties affected by the transition ; App endix A provides specific detail on the counties included in each group). Note that t he phased manner of STAR+PLUS implementation serves as a unique natural experiment on which to base our analysis. In this paper, we focus on examining the impact of the first main expansion of STAR+PLUS occurring in January and February of 2007 We have chosen to focus on only this earlier transition to allow for construction of "treated" and comparison group s with sufficient post transition follow up since data availab ility lags by several years. Furthermore, this focus is motivated by gaining a more nuanced understanding of the findings presented previously in Wegman et al (2015) which were centered on the same 2007 transition. 99 Our design employs a generalization of a di fference in difference approach. 114 In this, we estimate the change in the outcome from the pre implementation period to the post implementation period among our target sample living in counties transitioned in 2007 adjusting for potential confounding variables. We then compare this change to that experienced by a similar sample during the same timeframe living in counties not transitioned to the STAR+PLUS delivery model during the study period. We sought to examine the intermediate effects of the 20 07 t ransition. Data from 2006 2010 was available from Texas HHSC for use in this analysis. Accordingly, data from 2006 provided information on pre implementation measure performance. A single post implementation year was selected to limit the computational complexity of an

PAGE 82

82 imbalanced longitudinal cluster design and based on previous research which demonstrated stability in measure performance durin g the post implementation years. 99 T he latest year for which data was available (2010) w as selected as the post implementation data period to ensure sufficient time for implementation of the program and maximal programmatic effect The transition group was comprised of selected enrollees living in the 28 counties in which STAR+PLUS was imple mented in early 2007 (i.e. group B) The comparison group was comprised of selected enrollees living in counties which were transition ed to STAR+PLUS after 2010 (i.e. groups D, E and F ) Enrollees in g roup C (comprised of Dallas and Tarrant service areas) were not included in the comparison group because Texas implemented an Integrated Care Management (ICM) program in these regions from February 2008 May 2009. The ICM program is similar in form to a non capitated managed care service delivery model 69 Selected e nrollees in a non transitioned county were receiving their health services from either FFS or PCCM healthcare delivery models. Our analyses do not distinguish between FFS and PCCM based on the similarity of these delivery models and previous research indicating similar quality of care among relatively healthier adult Medicaid enrollees receiving care from these models. 100 102 In addition, one previous study of Medicaid managed care for enrollees with disabilities found that PCCM does not appea r to differentially impact access to care 62, The University of Florida's Institutional Review Board approved this study. Sample and Data We began by selecting individuals enrolled in Medicaid during 2006 and/ or 2010, and living in the counties discussed above. We then restricted our sample to adults >21

PAGE 83

83 years old, as enrollees younger than 21 were not required to enroll in STAR+PLUS in the transitioned counties. We further fo cused our study el igibility on enro llees <65 and those who were not dual eligible as the purpose of this analysis was to understand the impact of managed care on non elderly adults who were not also receiving Medicare. We followed HEDIS definitions separately for each measure to arrive at m easure specific samples. We abstracted of enrollment and monthly enrollment information from person level enrollment and encounter data provided by the Texas Health and Human Services Commission. We also linked enrollment records to professional, inpatient and pharmacy encounter data that included International Classification of Diseases (ICD 9 CM) diagnosis codes, Current Procedural Terminology codes, and National Drug Codes for use with NCQA certified software to calculate enrollee level HEDIS measures for the calendar years 2006 and 2010 Using enrollee level address information when available, we incorporated county Zip Code Tabulation Area and census tract level data from the Area Health Resource File, U S Census Bureau, and US Department of Commerce. Outcome Measures We selected 4 outcome measures to examine the impact of STAR+PLUS on receipt of preventive health care: receipt of routine screening for colorectal cancer, for cervical cancer and for breast cancer (i.e. mammography) and access to ambulatory care. Centers for Medicare & Medicaid Services suggested these as initial core quality measures for adults enrolled in Medicaid. 104 Furthermore, s trong performance on these indicators is linked to improved health outcomes. 119

PAGE 84

84 Explanatory and Control Variables T he main explanatory variable was constructed as the amount of time a given individual was enrolled in STAR+PLUS during the evaluated post transition year (2010) Accordingly, this variable captured differential transition timing or partial Medicaid enrollment. W e controlled for enrollee and contextual characteristics which could confound the relationship between STAR+PLUS enrollment and the preventive care measures. Select ion of th ese variables was based on IOM's social risk factors framework and 106 107 Demographic control variables included a ge age squared, gender (male and female) and race/ethnicity (non Hispanic white, non Hispanic black, Hispanic and other). Health status was measured using the 3M Clinical Risk Groups (CRGs), which were assigned using health care encounter data 108 Individuals were assigned to one of the following categories: healthy, significant acute conditions (e.g., chest pain), minor chronic conditions (e.g., migraine), moderate chronic conditions (e.g., asthma and diabetes), or major chronic conditions (e.g., cystic fibrosis and cancer). These health status categories were then recoded as a polynomial functio n reflecting health status. Patients enrolled for less than the 6 months required for classification were assigned missing values. Many individuals qualifying for Medicaid due to disability require long term supports; some of these individuals live at assisted living facilit ies while others live in low occupancy homes in the community or group style independent living facilities. Access to care may be different in group living arrangements, particularly assisted living facilities where health care prov iders are often co located. However our data did not

PAGE 85

85 include a specific indicator d esignating whether an individual resided in such a facility. For this reason we constructed a proxy variable to indicate if 5 or more enrollees were residing at the same a ddress. Using data from the Area Health Resource File, U.S. Census Bureau and U.S. Department of Commerce, we controlled for several c ontextual variables in our analysis. These included the percentage of the population living in poverty, percentage of the population unemployed, and median household income, all assigned using the Zip Code Tabulation Area if valid census tract was unavailable). Median county income and metropolitan classification (recoded as a polynomial function) were also included. Additionally, w e included a variable measuring density of Medicaid health care utilization (Medicaid inpatient discharges per Medicaid enrollee), assigned by county in order to capture geographic differences in Medicaid health care ut ilization Finally, the county level average of the pre transition (2006) measure performance was computed and merged with the individual level data by county. The full models also included 2 and 3 way predictor interactions, between race/ethnicity, gen der and the main explanatory variable. Due to a high proportion of complete cases (>90%), list wise deletion was performed for enrollees with missing data. Empirical Model and Analyses For each measure separately, we employed a 2 stage, multilevel approach which used general linear models. 78 First, we computed person level models of the outcome during the post transition year, assuming an exchangeable correlation and taking county as the clustering unit We chose county as the clustering unit despite the fact

PAGE 86

86 that (supra county) service areas represent the natural administrative units in which STAR+PLUS was transitioned and administered. This was primarily motivated by the limited number of service areas, which would result in i nsufficient statistical power. Further, some service areas were transitioned in phases. Fortunately, the county level is likely representative of the context within which enrollees receive health care, including available health care resources that impact health service use. A second tier model was constructed using individually adjusted county specific averages and weights derived from computed cluster sizes and the intra class correlation coefficient We performed computation diagnostics and reduce d the predictor space when necessary. 109 Then we conducted f ixed order backwards selection of the predictor variables with = 0.05 used as the exclusion threshold, to arrive at the final, reduced models. Assumption diagnostics for the final models reveale d n o serious violations of linearity, homogeneity or Gaussian distribution of residuals Sensitivity analyses in which i nfluence points were removed were found not to substantially impact the results. Analyses were conducted using R (Version 3.3.0) and Rst udio (Version 0.99.902) using lme4 (Version 1.1.12) and lsmeans (Version 2.23) packages for the main analyses and model summarization. 110 11 3 After model fitting, we computed the predicted compliance rates for the STAR+PLUS and FFS/ PCCM counties using the overall covariate distributions and varying the enrollment in STAR+PLUS (i.e. predicted marginal means). The difference between these estimates for each group reflects the effect we attribute to the STAR+PLUS implementation.

PAGE 87

87 Results The overall study sample was comprised of unique enrollees meeting HEDIS eligibility criteria for any or multiple of the studied measures. This overall sample was comprised of 93,635 enrollees clustered across 237 Texas counties. Table G 1 demonstrates tha t ~32% of the study population met eligibility criteria for only the access to preventive/ambulatory health services measure, and 20% met eligibility criteria for all measures, with the large majority of the remaining sample qualifying for the access to pr eventive/ambulatory health services measure and either colorectal cancer screening or cervical cancer screening measures. Table 4 1 provides a summary of demographic, health and contextual characteristics for the study population. Table 4 1. Demographic h ealth and contextual characteristics for the study population Combined Access to Preventive/ Ambulatory Health Services Subgroup Breast Cancer Screening Subgroup Cervical Cancer Screening Subgroup Colon Cancer Screening Subgroup Unique enrollees 93635 93615 11505 51319 23048 Unique counties 237 237 218 237 233 Enrollees with complete data 89605 (95.7) 89585 (95.7) 11179 (97.2) 49336 (96.1) 22371 (97.1) Counties with complete data, by year 237 (100) 237 (100) 217 (99.5) 236 (99.6) 233 (100) Counties with complete data in post and pre years 237 (100) 217 (99.5) 236 (99.6) 233 (100) Age [mean (SD)] 45.5 (13) 45.5 (13) 58.3 (3.7) 48.2 (11.3) 57.6 (3.9) Age categories [n (%)]

PAGE 88

88 Table 4 1. Continued Combined Access to Preventive/ Ambulatory Health Services Subgroup Breast Cancer Screening Subgroup Cervical Cancer Screening Subgroup Colon Cancer Screening Subgroup 21 30 16271 (17.4) 16271 (17.4) 0 (0) 4654 (9.1) 0 (0) 30 39 13861 (14.8) 13859 (14.8) 0 (0) 7592 (14.8) 0 (0) 40 49 20290 (21.7) 20285 (21.7) 0 (0) 12244 (23.9) 0 (0) 50 59 28226 (30.1) 28217 (30.1) 6710 (58.3) 17402 (33.9) 14705 (63.8) 60 64 14987 (16) 14983 (16) 4795 (41.7) 9427 (18.4) 8343 (36.2) Female [n (%)] 54566 (58.3) 54553 (58.3) 11498 (99.9) 51298 (100) 14995 (65.1) Race/ethnicity [n (%)] Non Hispanic B lack 16530 (17.7) 16527 (17.7) 1394 (12.1) 8486 (16.5) 3283 (14.2) Hispanic 32059 (34.2) 32048 (34.2) 5123 (44.5) 18162 (35.4) 9255 (40.2) Other 6891 (7.4) 6890 (7.4) 1180 (10.3) 4326 (8.4) 2163 (9.4) Non Hispanic white 38155 (40.7) 38150 (40.8) 3808 (33.1) 20345 (39.6) 8347 (36.2) Health status [n (%)] Healthy 15840 (17.3) 15840 (17.3) 643 (5.7) 6130 (12.2) 1908 (8.4) Significant acute 2514 (2.8) 2514 (2.8) 127 (1.1) 1242 (2.5) 326 (1.4) Minor chronic 5331 (5.8) 5331 (5.8) 477 (4.2) 3092 (6.1) 1029 (4.5) Moderate chronic 22234 (24.3) 22233 (24.3) 1877 (16.6) 11536 (22.9) 4138 (18.2) Major chronic 45443 (49.7) 45424 (49.7) 8196 (72.4) 28290 (56.3) 15303 (67.4) Health status not available 2273 (2.4) 2273 (2.4) 185 (1.6) 1029 (2) 344 (1.5) County metro classification [n (%)]

PAGE 89

89 Table 4 1. Continued Combined Access to Preventive/ Ambulatory Health Services Subgroup Breast Cancer Screening Subgroup Cervical Cancer Screening Subgroup Colon Cancer Screening Subgroup Large metro (>1M) 24026 (25.7) 24018 (25.7) 3874 (33.7) 12427 (24.2) 7363 (31.9) Medium metro (250K 1M) 24641 (26.3) 24634 (26.3) 3677 (32) 13898 (27.1) 7025 (30.5) Small metro (<250k) 18743 (20) 18740 (20) 1583 (13.8) 10313 (20.1) 3450 (15) Adjacent to metro 18292 (19.5) 18291 (19.5) 1794 (15.6) 10306 (20.1) 3834 (16.6) Nonadjacent to metro (>20k) 3879 (4.1) 3878 (4.1) 279 (2.4) 2144 (4.2) 672 (2.9) Nonadjacent to metro (<20k) 4054 (4.3) 4054 (4.3) 298 (2.6) 2231 (4.3) 704 (3.1) Census tract poverty [mean (SD)] 23.9 (12.9) 23.9 (12.9) 26.2 (13.3) 24.2 (12.8) 25.7 (13.3) Census tract poverty categories [n (%)] 0.0% 4.9% 3032 (3.3) 3031 (3.3) 298 (2.6) 1430 (2.8) 581 (2.5) 5.0% 9.9% 8526 (9.2) 8525 (9.2) 779 (6.8) 4419 (8.7) 1648 (7.2) 10.0% 19.9% 29589 (32) 29584 (32) 3199 (28) 16287 (32.1) 6650 (29.1) Poverty area (20.0% 39.9%) 40100 (43.4) 40087 (43.4) 5367 (47) 22289 (44) 10542 (46.1) Extreme poverty area (> 40.0%) 11187 (12.1) 11187 (12.1) 1782 (15.6) 6245 (12.3) 3432 (15) Census tract not available 1201 (1.3) 1201 (1.3) 80 (0.7) 649 (1.3) 195 (0.8) Census tract unemployment level [n (%)] 0.0% 4.9% 18681 (20.2) 18680 (20.2) 1902 (16.7) 10078 (19.9) 3914 (17.1) 5.0% 9.9% 44287 (47.9) 44279 (47.9) 5492 (48.1) 24126 (47.6) 10915 (47.8) 10.0% 19.9% 27785 (30.1) 27775 (30.1) 3815 (33.4) 15536 (30.7) 7573 (33.1) > 20.0% 1663 (1.8) 1662 (1.8) 212 (1.9) 918 (1.8) 446 (2)

PAGE 90

90 Table 4 1. Continued Combined Access to Preventive/ Ambulatory Health Services Subgroup Breast Cancer Screening Subgroup Cervical Cancer Screening Subgroup Colon Cancer Screening Subgroup Census tract not available 1219 (1.3) 1219 (1.3) 84 (0.7) 661 (1.3) 200 (0.9) Census tract household income [mean (SD)] 38676.3 (15739.5) 38676.8 (15739.7) 36629.5 (15075) 38120.2 (14909.1) 36895.8 (15109.5) County Medicaid discharge density [mean (SD)] 1 (0.6) 1 (0.6) 1.1 (0.6) 1 (0.6) 1.1 (0.6) Census tract determined from address [n (%)] 72466 (77.4) 72450 (77.4) 9118 (79.3) 39814 (77.6) 18136 (78.7) Facility residence [n (%)] 18847 (20.3) 18840 (20.3) 2601 (22.8) 10473 (20.6) 5263 (23.1) Months enrolled in Medicaid [n (%)] 12 91781 (98) 91761 (98) 11446 (99.5) 50303 (98) 22935 (99.5) 9 11 1824 (1.9) 1824 (1.9) 54 (0.5) 999 (1.9) 106 (0.5) 5 8 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 2 5 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 1 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) STAR+PLUS transition cohort [n (%)] A 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) B 29164 (31.1) 29152 (31.1) 5171 (44.9) 15293 (29.8) 9640 (41.8) C 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) D 8481 (9.1) 8481 (9.1) 466 (4.1) 4742 (9.2) 1273 (5.5) E 22115 (23.6) 22112 (23.6) 3518 (30.6) 12366 (24.1) 6605 (28.7) F 33875 (36.2) 33870 (36.2) 2350 (20.4) 18918 (36.9) 5530 (24)

PAGE 91

91 Notably the study population has poor health status with approximately three quarters belonging to the most severe health status category. Second, more than half of the sample lives in locations deemed impoverished or extremely impoverished A round 20% of the sample lives at addresses with 5 or more enrollees, suggesting a high degree of co location and perhaps utilization of assisted living services More than 90% of the sample lives in metro politan areas or areas adjacent to a recognized metropolis Close to 40% of the sample is white, non Hispanic and one third of the sample is Hispanic Several characteristics differ between the various subgroups qualifying for each measure Those qualifying for breast cancer screening and colorectal cancer screening measures te nd to be older, reflecting the later age at which these screenings are recommended. Additionally, the subpopulations eligible for breast cancer screening and colorectal cancer had more severe health status; these findings may be related to the increased ag e at which enrollees qualify for these measures As expected enrollees qualifying for breast cancer screening and cervical cancer screening were almost exclusively female ; male enrollees were not included in modelling these measures. Tables G2 5 present descriptive statistics for transition and comparison groups by measure. The large majority of characteristics were very similar between groups However, enrollees from transition counties were more likely to be Hispanic and less likely to be white, non His panic Enrollees in transition counties were more likely from large metro politan areas (80% vs 2% ) ; enrollees from comparison counties were typically living in medium, small and adjacent metro politan areas. Of note, characteristics largely demonstrated sta bility over time.

PAGE 92

92 The total sample size for each measure by group is presented in Figure F 1 and a summary of the number of observations per county is presented in Table G 6. Figure F 2 provides a spatial description of the sample sizes by group. Note th at the degree of clustering by county is similar between the groups. For ambulatory care access and cervical cancer screening measures, a decline in sample size was experienced temporarily in 2007 for the comparison group only, after which the sample size returned to baseline levels. Weighted, u nadjusted outcome measure performance is illustrated in Table 4 2 and Figure F 3. For cervical and colorectal cancer screening steady improvement was observed for both groups ; however actual measure adherence rema ined low at 40% and 30%, respectively. Breast cancer screening saw a more dramatic, nonlinear, increase in screening rates from < 2 % in 2006 to 50% by 2010 This trend was similar between groups. For ambulatory care access the transition group appeared to increase by 4 %, while the comparison group remained stable or slightly decreased. Table 4 2. Weighted, unadjusted measure performance Measure Group 2006 2007 2008 2009 2010 Access to Preventive/Ambulatory Health Services Comparison 75.9 74.6 73.3 73.4 74.0 Transition 80.2 78.2 80.5 82.6 84.5 Breast Cancer Screening Comparison 1.5 25.2 37.4 41.0 42.1 Transition 1.2 20.7 36.7 43.4 46.9 Cervical Cancer Screening Comparison 25.7 34.4 36.4 36.9 37.1 Transition 31.8 35.4 35.7 38.0 41.1 Colorectal Cancer Screening Comparison 17.5 20.8 24.3 27.0 29.2 Transition 18.9 20.1 24.5 28.2 31.2 Table 4 3 provides the final model predicted marginal means during 2010 for each measure. Of note, the STAR+PLUS transition was associated with a significant

PAGE 93

93 increase of 4% in rates of receipt of an annual preventive visit (.74 vs .78) For the remaining measures, no STAR+PLUS effect was found. Table 4 3. Predicted marginal means during 2010, by measure. Measure F FS/PCCM (Comparison) STAR+PLUS (Transition) Diff erence p value Access to Preventive/Ambulatory Health Services 0.74 (0.73 0.75) 0.78 (0.75 0.81) 0.04 0.011 Breast Cancer Screening 0.42 (0.40 0.43) 0.42 (0.39 0.46) 0.01 0.696 Cervical Cancer Screening 0.37 (0.36 0.38) 0.35 (0.33 0.37) 0.02 0.098 Colorectal Cancer Screening 0.29 (0.28 0.30) 0.29 (0.27 0.31) 0.00 0.881 Discussion Assessing the impact of Medicaid managed care and HCBS on receipt of preventive health care is critical, given the substantial gaps in preventive care provision and the cost effectiveness of such services 115 In this study, we found that the Texas STAR+PLUS program slightly increased the proportion of qualifying enrollees who received an annual ambulatory/preventive care visit. However, STAR+PLUS did not affect receipt of screening for breast, cervical or colo rectal cancer. Our finding of improved access to ambulatory care due to STAR+PLUS is consistent with results from Burns 59 and Coughlin, Long and Graves 62 which also suggested that Medicaid managed care enrollees were more likely to have a usual source of care. There are multiple explanations for this observed increase in rates of annual ambulatory care. First, in STAR+PLUS and many managed care programs, the enrollee's primary care provider serves as a gatekeeper for many medical and specialty

PAGE 94

94 services. Th us for patients with complex medical needs, who form the large majority of Medicaid patients with disabilities, a primary care provider visit is essential for accessing the added supports and services offered by the organization. Second, many managed care programs use annual visits as an opportunity to conduct health assessments to guide provision of services and added supports for the enrollee. To aid in this assessment goal, programs may provide reminders, convenient scheduling, convenient access points o r transportation supports -thereby increasing annual visit uptake. To our knowledge, no previous studies have examined the effects of managed care on cancer screening receipt. Given our finding of no effect, it is likely that these forms of screening were not targeted by the managed care organizations. Regardless of group (i.e. transition or comparison), we found unsatisfactory performance overall for rates of cancer screening, even after modest improvements over the study period. For example, less than one third of enrollees > 50 years of age were screened for colon cance r. These rates compare to national estimates of screening receipt for Medicaid adults 123 Such findings are significant, because colon cancer screening has been shown to decrease 11 to 12 year colorectal cancer mortality by 30% and with certain forms of scr eening, decrease the actual incidence of colorectal cancer by 20%. 124 125 Our study has several limitations. First, our findings may not be generalizable to other Medicaid managed care programs as these differ throughout the United States. Second, due to the non randomized nature of our design, our study is vulnerable to bias due to residual confounding, particularly as it relates to the appropriateness of the comparison group. Such a bias could result if an unmeasured external eff ect differentially impact ed our county groups during the studied time period Finally, we

PAGE 95

95 have focused on only one post transition year; our findings may not reflect the average change due to the transition over all post transition years. However, review of weighted, unadjusted me asure performance above suggests that large departures from our modeled results are unlikely. Furthermore, the absence of effects in the 4th year of program implementation are important findings and suggest the limited impact of STAR+PLUS on receipt of scr eening services. In summary, in this study of 1 large Medicaid program serving adults with disabilities, managed acute and long term services relative to FFS, does not appear to affect rates of cancer screening and only modestly impacts rates of annual pr imary care visits.

PAGE 96

96 CHAPTER 5 CONCLUSION Over the past several decades, states have continued to experiment with new models of health care delivery and financing in an attempt to reduce or contain the escalating costs of their Medicaid programs. Some of the most radical shifts have been from fee for service financing to global payment models, including managed care. Expansion of managed care delivery models to Medicaid enrollees with disabilities has occurred in more than half of all states covering the majority of this enrollee group. Yet, very few robust evaluations have been conducted to understand if these programs are reducing unnecessary utilization, while also not compromising access and quality of care. In this work, we have demonstrated that, ac ross most of our selected nationally validated quality of care measures, the effect of one of largest existing managed care programs is largely null. Several important exceptions exist, namely with large and sustained improvements in quality of care during discharge from the hospital after a heart attack and after COPD exacerbation, as well as to a lesser extent, receipt of an annual ambulatory/preventive health care visit and intermediate term engagement in substance abuse treatment. There are several po ssible explanations for managed care's limited impact on the explored set of measures. First, good performance on many of these measures is dependent on routinely accessing primary care. For example, receipt of asthma medications, antidepressant medication s, cholesterol screening, and diabetes sequelae screening are tasks which occur during or as a result of regular primary care visits. STAR+PLUS led to only a slight increase (3%) in the number of enrollees accessing a

PAGE 97

97 primary care provider in the previous year, with almost one quarter of patients who qualified for Medicaid due to disability having no evidence of a primary care visit during this time frame. An even larger proportion of patients is likely to have not seen their primary care provider at recomm end intervals (3 6 months), further explaining the lack of STAR+PLUS impact, and the poor performance overall, for measures which are clinically dependent on multiple visits yearly (e.g. longer term asthma medication adherence, screening for diabetes seque lae). For STAR+PLUS to impact most of the studied measures, it is likely that even greater rates of engagement in regular primary care visits will be necessary. Medicaid enrollees with disabilities face many barriers to accessing primary care. Among thes e are physical barriers, such as long distances between enrollees and providers, transportation difficulties, limited office hours, long provider wait times, and inaccessible facilities 126 128 Relatedly, Medicaid's history of repeated service payment decr eases has contributed to limited provider selection, insufficient networks to meet enrollee needs, and networks comprised of less preferred or under resourced organizations. 96 While STAR+PLUS made efforts to address some of these issues, including providi ng transportation supports and improving networks, the effects on access were likely minimal. For example, Medicaid enrollees have described the unreliability of medical transport and the long transportation times that such services require, discouraging r egular use. 129 Without substantially increasing payments, provider networks were unlikely to meaningfully change. Furthermore, STAR+PLUS did not address many more deeply rooted behavioral and social barriers to accessing care. For

PAGE 98

98 example, enrollees may b e reluctant or unable to leave their homes due to medical or behavioral conditions. Other enrollees may have mistrust of the health care system due to previous mistreatment. For patients living with repeated exposure to factors such as violence, housing in stability and food instability, routine or preventive health care is likely to lack salience. More than likely, most of these factors persisted with STAR+PLUS' implementation, limiting the program's impact. Second, beyond visiting a primary care provider, most of our selected measures are heavily dependent on the enrollee performing a prescribed task. Tasks span a spectrum from laboratory testing (e.g. cholesterol screening, A1c screening), to medication filling (e.g. asthma, depression), to returning for follow up visits (e.g. after psychiatric hospitalization, after substance abuse diagnosis), to more invasive screening testing (e.g. cervical cancer, colorectal cancer and breast cancer screenings). Multiple barriers inhibit enrollees from engaging in th ese tasks. To begin, many of the barriers to accessing primary care described above also apply to accessing screening services, follow up services, laboratory testing, and pharmacies. Furthermore, these activities often represent a component of a more com plex disease management plan which is, to a large extent, completed by the patient. Optimal engagement in this plan requires high levels of cognitive and social resources, both of which are, on average, substantially diminished in individuals qualifying fo r Medicaid due to disability, by virtue of their marginalized position in society and their serious medical conditions. 130 1 31 Even though many enrollees with disabilities have informal caregivers to assist with chronic disease management and health care e ngagement,

PAGE 99

99 research suggests that these caregivers have cognitive impairments and limited social capital themselves. 1 3 2 Third, several specific aspects of STAR+PLUS financing likely limited the model's impact. While Texas Medicaid shifted from paying serv ice based fees to providers to issuing capitation based payments to managed care organizations, providers were still largely paid in fee for service arrangements by the managed care organizations. Thus, despite small incentives and quality improvement proj ects aimed at improving the value of care, the resulting incentives were much more aligned with traditional volume based financing for the participating providers. Perhaps even more importantly, inpatient hospital services were carved out of the 2007 STAR +PLUS expansion, such that inpatient services were paid via fee for service financing. This structure removed one of the strongest incentives of the managed care model -that expected to drive more appropriate utilization and increased preventive services use in order to avoid expensive hospital care. Paradoxically, of the limited effects due to STAR+PLUS, the largest change we observed was improvement in quality of hospital based care following acute events. Thus, the hospital carve outs may have played a lesser role in the limited changes seen in the measures overall than we might have expected a priori. Providing high quality care for Medicaid patients with disabilities is challenging, given complex and interacting social and illness contexts. Thus the absence of change due to STAR+PLUS might reflect a local ceiling; further improvements may require even more substantial health care delivery redesign, or investments on par of those in the private market with a similar case mix. At the same time, recent literature suggests that

PAGE 100

100 even more pronounced risk bearing models in relatively more socially advantaged Medicare populations have failed to demonstrate a significant effect on quality of care. For example, in an analysis of early changes due to implementa tion of accountable care organizations in Medicare, only 1 of 11 quality measures demonstrated significance. 13 3 Towards this end, an expanding collection of research suggests that factors largely outside the traditional scope of health care play a substan tial, if not overwhelming, role in determining population health outcomes. 1 34 13 6 For example, enrollees' lifelong exposures to physically and psychologically unhealthy environments produced from chronic violence, housing instability, poor access to qualit y employment and educational opportunities, pollution, and barriers to healthy food consumption and exercise have direct negative impact on health status and optimal management of their clinical conditions. Accordingly, research suggests that services incl uding housing support, nutritional assistance, income support, and early childhood development support are linked to improvements in health and reduction in future health care costs. 13 7 Based on this work, the incremental benefit of investments in traditional health care services and supports is likely small, especially in relation to social service investments. The results of this work can also be viewed in alternative light. If STAR +PLUS is shown to have achieved cost savings in the context of unchanged or improved quality of care, then STAR+PLUS might be considered a success. However, we feel that the low levels of health care quality for an already vulnerable population must be gre atly improved before such a program would be deemed successful. To reach a definitive conclusion on the impact of STAR+PLUS, utilization, costs, satisfaction, mortality and

PAGE 101

101 other health outcomes should be studied. Future research is needed to explore the m echanism of the effe cts observed in these studies. In summary, managed care and HCBS, in its current formulation, does not appear to reliability impact the quality of care of Medicaid individuals with disabilities. Additional research is needed to furthe r evaluate and refine service delivery for this vulnerable population.

PAGE 102

102 APPEN DIX A TABLES FOR CHAPTER 1: INTRODUCTION Table A 1. STAR+PLUS transition history Bexar Service Area Bexar 1: January 2007: Atascosa, Bexar, Comal, Guadalupe, Kendall, Medina, Wilson Bexar 2: September 2011: Bandera Dallas Service Area February 2011: Collin, Dallas, Ellis, Hunt, Kaufman, Navarro, Rockwall El Paso Service Area March 2012: El Paso, Hudspeth Harris Service Area Harris 1: April 1998: Harris Harris 2: January 2007: Brazoria, Fort Bend, Galveston, Montgomery, Waller Harris 3: September 2011: Austin, Matagorda, Wharton Hidalgo Service Area March 2012: Cameron, Duval, Hidalgo, Jim Hogg, Maverick, McMullen, Starr, Webb, Willacy, Zapata Jefferson Service Area September 2011: Chambers, Hardin, Jasper, Jefferson, Liberty, Newton, Orange, Polk, San Jacinto, Tyler, Walker Lubbock Service Area March 2012: Carson, Crosby, Deaf Smith, Floyd, Garza, Hale, Hockley, Hutchinson, Lamb, Lubbock, Lynn, Potter, Randall, Swis her, Terry MRSA* Central September 2014: Bell, Blanco, Bosque, Brazos, Burleson, Colorado, Comanche, Coryell, DeWitt, Erath, Falls, Freestone, Gillespie, Gonzales, Grimes, Hamilton, Hill, Jackson, Lampasas, Lavaca, Leon, Limestone, Llano, Madison, McLennan, Milam, Mills, Robertson, San Saba, Somervell, Washington MRSA* Northeast September 2014: Anderson, Angelina, Bowie, Camp, Cass, Cherokee, Cooke, Delta, Fannin, Franklin, Grayson, Gregg, Harrison, Henderson, Hopkins, Houston, Lamar, Marion, Mon tague, Morris, Nacogdoches, Panola, Rains, Red River, Rusk, Sabine, San Augustine, Shelby, Smith, Titus, Trinity, Upshur, Van Zandt, Wood

PAGE 103

103 Table A 1. Continued MRSA Medicaid Rural Service Area. Table information obtained from: 1) Information Letter No. 12 26: Expansion of STAR in the Medicaid Rural Service Areas. Texas Health and Human Services Commission. Available at: http://www.dads.state.tx.us/providers/communications/2012/letters/IL2012 26.pdf Accessed March 16, 2015. 2) Managed Care Service Areas. Texas Health and Human Services Commission. Available at: https://www.hhsc.state.tx.us/medicaid/managed care/mmc/Managed Care Service Areas Map.pdf Accessed March 16, 2015. 3) Texas Medicaid and CHIP in Perspective: Ninth Edition. Texas Health and Human Services Commi ssion; 2013. Available at: http://www.hhsc.state.tx.us/medicaid/about/PB/PinkBook.pdf. Accessed June 16, 2014. MRSA* West September 2014: Andrews, Archer, Armstrong, Bailey, Baylor, Borden, Brewster, Br iscoe, Brown, Callahan, Castro, Childress, Clay, Cochran, Coke, Coleman, Collingsworth, Concho, Cottle, Crane, Crockett, Culberson, Dallam, Dawson, Dickens, Dimmit, Donley, Eastland, Ector, Edwards, Fisher, Foard, Frio, Gaines, Glasscock, Gray, Hall, Hansf ord, Hardeman, Hartley, Haskell, Hemphill, Howard, Irion, Jack, Jeff Davis, Jones, Kent, Kerr, Kimble, King, Kinney, Knox, La Salle, Lipscomb, Loving, Martin, Mason, McCulloch, Menard, Midland, Mitchell, Moore, Motley, Nolan, Ochiltree, Oldham, Palo Pinto, Parmer, Pecos, Presidio, Reagan, Real, Reeves, Roberts, Runnels, Schleicher, Scurry, Shackelford, Sherman, Stephens, Sterling, Stonewall, Sutton, Taylor, Terrell, Throckmorton, Tom Green, Upton, Uvalde, Val Verde, Ward, Wheeler, Wichita, Wilbarger, Winkle r, Yoakum, Young, Zavala Nueces Service Area Nueces 1: January 2007: Aransas, Bee, Calhoun, Jim Wells, Kleberg, Nueces, Refugio, San Patricio, Victoria Nueces 2: September 2011: Brooks, Goliad, Karnes, Kenedy, Live Oak Tarrant Service Area February 2011: Denton, Hood, Johnson, Parker, Tarrant, Wise Travis Service Area Travis 1: January 2007: Bastrop, Burnet, Caldwell, Hays, Lee, Travis, Williamson Travis 2: September 2011: Fayette

PAGE 104

104 Table A 2 Comparing Fee for Service (FFS ) a nd Primary Care Case Management (PCCM) on measure c ompliance a t b aseline a nd control v ariables FFS PCCM Averaged measure compliance at baseline* (mean; std) 0.59 (0.37) 0.62 (0.37) Age (mean; std) 52.9 (10.1) 52.3 (9.9) 21 29 5.3% 4.8% 30 39 5.9% 7.2% 40 49 17.8% 20.0% 50 59 44.0% 44.5% 60 64 27.1% 23.5% Female (%) 62% 62% Race/Ethnicity (%) White, non Hispanic 38.9% 44.1% Black, non Hispanic 17.8% 17.6% Hispanic 32.6% 30.3% Other 10.6% 8.0% Health status # (mean; std) 4.81 (0.63) 4.72 (0.75) Healthy 1.7% 2.5% Significant Acute 0.4% 0.5% Minor Chronic 0.8% 1.2% Moderate Chronic 9.4% 14.1% Major Chronic 87.7% 81.6% Census tract poverty % (mean; std) 26.0 (13.9) 24.6 (12.7) 0.0% 4.9% 3.0% 1.8% 5.0% 9.9% 8.6% 8.1% 10.0% 19.9% 26.4% 33.2% Poverty Area (20.0% 39.9%) 45.0% 43.7% Extreme Poverty Area (> 40.0%) 17.0% 13.3% County level median income (mean; std) $37,100 ($15,000) $36,700 ($12,500) Facility residence (%) 26.0% 19.7% Years eligible for study, 2007 2010 (mean; std) 2.66 (1.21) 2.48 (1.11) Average compliance across the 6 outcome measures # 5 level Clinical risk group (3M)

PAGE 105

105 Table A 3 Abbreviated descriptions of outcome m easures Measure (Abbreviation) Numerator Denominator Use of Appropriate Medications for People with Asthma (ASM) Subset of denominator members who were dispensed at least one prescription for asthma controller medication during the measurement year Members 21 64 years old enrolled in the measurement year and the preceding year meeting one of the following criteria during the m easurement year and the preceding year: > 1 ED visit with asthma as principal diagnosis > 1 acute inpatient encounter with asthma as principal diagnosis > 4 outpatient visits with asthma as a diagnosis, including two asthma medication dispensing events > f our asthma medication dispensing events Exclude: members with a diagnosis of emphysema, COPD, chronic bronchitis, cystic fibrosis, or acute respiratory failure Medication Management for People With Asthma (MMA) Subset of denominator members who were d ispensed asthma controller medication covering at least 75% of the treatment period Members 21 64 years old enrolled in measurement year and preceding year with persistent asthma (as identified in ASM) Exclude: members with a diagnosis of emphysema, COP D, chronic bronchitis, cystic fibrosis, or acute respiratory failure Pharmacotherapy Management of COPD Exacerbation (PCE) Numerator 1 systemic corticosteroid: subset of denominator events for which members were dispensed a prescription for systemic corticosteroid < 14 days of the event date Numerator 2 bronchodilator: subset of denominator events for which members were dispensed a prescription for bronchodilator < 30 days of the event date Members 40 64 years old with an acute inpatient discharge or ED encounter with a principal diagnosis of COPD during the measurement year who were enrolled for > 30 days of the exacerbation event Exclude: ED visits that resulted in an inpatient admission event dates for which member was transferred directly to a n acute or non acute care facility for any diagnosis event dates for which the member was readmitted to an acute or non acute care facility, or had an ED visit for any diagnosis, within 14 days of the event date Cholesterol Management for Patients With C ardiovascular Conditions (CMC) Numerator 1 LDL C Screening: subset of denominator members who had an LDL C test in the measurement year Numerator 2 LDL C Control subset of denominator members whose most recent LDL C <100 mg/dL during the measurement year Members 21 64 years old enrolled in the measurement year and preceding year discharged for AMI, CABG or PCI in the year preceding the measurement year or with a diagnosis of IVD in the measurement year and the preceding year

PAGE 106

106 Table A 3. Continued Measure (Abbreviation) Numerator Denominator Persistence of Beta Blocker Treatment After a Heart Attack (PBH) Subset of denominator members who received treatment with beta supplied for the 180 days after discharge (> 75% of days) Members 21 64 years old discharged from an acute inpatient setting with an acute myocardial infarction (AMI) between July 1 of year preceding the measurement year through June 30 th of the measurement year and enrolled for > 6 months of discharge date Excl ude: hospitalizations in which member transferred directly to non acute care facility for any diagnosis Comprehensive Diabetes Care (CDC) HbA1c testing Subset of denominator members who had an HbA1c test during the measurement year Members 21 64 years old enrolled during the measurement year diagnosed with type 1 or type 2 diabetes (type 1 and type 2) in the measurement year or the preceding year through EITHER two face to face outpatient or non acute inpatient encounters with a d iagnosis of diabetes or one face to face ED or acute inpatient encounter with a diagnosis of diabetes OR dispensed insulin or hypoglycemics/antihyperglycemics on ambulatory basis Exclude: members discharged with CABG or PCI members with a diagnosis of IVD, CHF, MI, CRF/ESRD, dementia, blindness, or lower extremity amputation Eye exam Subset of denominator members who had an eye screening for diabetic retinal disease identified by a retinal or dilated eye exam by an eye care professional in the measurement year OR a negative retinal exam by an eye care professional in the year precedin g the measurement year Same as HbA1c testing LDL C screening Subset of denominator members who received an LDL C test during the measurement year Same as HbA1c testing

PAGE 107

107 Table A 3. Continued Measure (Abbreviation) Numerator Denominator Nephropathy screening Subset of denominator members who received a nephropathy screening test during the measurement year OR have evidence of nephropathy Same as HbA1c testing Source: HEDIS 2012 Technical Specifications for Health Plans, Volume 2. Washin gton, DC: National Committee on Quality Assurance; 2011. The following tables describe the patterns of measure eligibility for each study subject across the post baseline years (i.e. 2007 to 2010). indicates that the outcome was able to be calculated (i .e. the subject was eligible), while indicates the subject was not eligible for the measure for a given person year. For example, 53 subjects in the fee for service (FFS) or primary care case management (PCCM) program components were eligible for the ASM measure in the first two post baseline years, but were not eligible for the measure in 2009 or 2010. Note that both the counts of subjects with each measure eligibility pattern as well as the percentages of the sample attributable to a given pattern for e ach measure are provided.

PAGE 108

108 Table A 4 Member eligibility counts FFS/PCCM Pattern, 2007 2010 ASM CDC CMC MMA PBH PCE **** 88 5336 806 69 0 48 ***_ 48 1218 318 34 0 33 **_* 3 335 30 2 0 24 **__ 53 1685 620 44 4 87 *_** 0 206 19 2 0 32 *_*_ 2 67 5 2 1 45 *__* 9 215 132 7 0 33 *___ 202 3399 1355 163 161 682 _*** 34 2687 331 31 0 44 _**_ 33 1068 247 30 7 81 _*_* 1 153 26 1 1 63 _*__ 77 1872 519 52 127 500 __** 136 2730 889 94 5 153 __*_ 127 1132 708 110 105 635 ___* 216 4296 1355 160 122 866

PAGE 109

109 Table A 5 Member eligibility percentages FFS/PCCM Pattern, 2007 2010 ASM CDC CMC MMA PBH PCE **** 8.6 20.2 11.0 8.6 0 1.4 ***_ 4.7 4.6 4.3 4.2 0 1.0 **_* 0.3 1.3 0.4 0.2 0 0.7 **__ 5.2 6.4 8.4 5.5 0.8 2.6 *_** 0 0.8 0.3 0.2 0 1.0 *_*_ 0.2 0.3 0.1 0.2 0.2 1.4 *__* 0.9 0.8 1.8 0.9 0 1.0 *___ 19.6 12.9 18.4 20.3 30.2 20.5 _*** 3.3 10.2 4.5 3.9 0 1.3 _**_ 3.2 4.0 3.4 3.7 1.3 2.4 _*_* 0.1 0.6 0.4 0.1 0.2 1.9 _*__ 7.5 7.1 7.1 6.5 23.8 15.0 __** 13.2 10.3 12.1 11.7 0.9 4.6 __*_ 12.3 4.3 9.6 13.7 19.7 19.1 ___* 21.0 16.3 18.4 20.0 22.9 26.0

PAGE 110

110 Table A 6 Member eligibility counts STAR+PLUS Pattern, 2007 2010 ASM CDC CMC MMA PBH PCE **** 103 3999 354 86 0 15 ***_ 60 755 118 50 0 16 **_* 0 115 4 0 1 6 **__ 59 984 207 48 0 19 *_** 0 251 3 3 0 8 *_*_ 0 68 5 1 2 15 *__* 14 95 59 11 2 15 *___ 135 1359 400 112 46 159 _*** 146 1103 270 116 0 23 _**_ 74 418 117 67 5 35 _*_* 1 47 6 7 0 18 _*__ 128 417 252 114 76 189 __** 162 1676 380 131 3 56 __*_ 114 561 322 90 73 252 ___* 294 2275 668 231 69 343

PAGE 111

111 Table A 7 Member eligibility percentages STAR+PLUS Pattern, 2007 2010 ASM CDC CMC MMA PBH PCE **** 8.0 28.3 11.2 8.1 0 1.3 ***_ 4.7 5.3 3.7 4.7 0 1.4 **_* 0 0.8 0.1 0.0 0.4 0.5 **__ 4.6 7.0 6.5 4.5 0 1.6 *_** 0 1.8 0.1 0.3 0 0.7 *_*_ 0 0.5 0.2 0.1 0.7 1.3 *__* 1.1 0.7 1.9 1.0 0.7 1.3 *___ 10.5 9.6 12.6 10.5 16.6 13.6 _*** 11.3 7.8 8.5 10.9 0 2.0 _**_ 5.7 3.0 3.7 6.3 1.8 3.0 _*_* 0.1 0.3 0.2 0.7 0 1.5 _*__ 9.9 3.0 8.0 10.7 27.4 16.2 __** 12.6 11.9 12.0 12.3 1.1 4.8 __*_ 8.8 4.0 10.2 8.4 26.4 21.6 ___* 22.8 16.1 21.1 21.6 24.9 29.3 Reference groups are Female; Black non Hispanic ; and time=2010. Age is centered at 43.5 years. Coefficients for CDC are not reported because the model did not converge when using the county as the unit of clustering. The individual model coefficients are provided below for completeness. It is important to note that these do not represent the hyp othesis that is the primary focus of the study, and these coefficients cannot be meaningfully interpreted in isolation. Rather, it is the linear combinations of the coefficients that are of interest. Therefore, these coefficients have little meaning withou t concurrently applying the groups specific covariate distributions. The source of confusion lies in the dependence of the statistical results and interpretation on the choice of coding scheme (reference cell vs. cell mean), the choice of reference cells, and the fact that more than two categories are present for categorical predictors (i.e. health status). For these reasons, we focus on reporting the predicted means and associated confidence intervals in the main text, which directly address the hypothesis being tested.

PAGE 112

112 Table A 8 Coefficients for final reduced models ASM CMC MMA PBH PCE Intercept 0.3017 (0.1734) 0.1853 (0.2143) 0.8151 (0.2244) 0.1946 (0.3718) 0.0504 (0.1404) Male 0.0691 (0.1013) 0.2093 (0.1138) 0.0713 (0.1496) 0.3026 (0.1995) 0.1222 (0.0871) Hispanic 0.1316 (0.0747) 0.2871 (0.0838) 0.1076 (0.1007) 0.1024 (0.1711) 0.0258 (0.0701) Other 0.0346 (0.0986) 0.3505 (0.1095) 0.1682 (0.1303) 0.1769 (0.1902) 0.0294 (0.0813) White, non Hispanic 0.074 (0.0665) 0.0866 (0.08) 0.1148 (0.0839) 0.1893 (0.1667) 0.0709 (0.0611) Male* Hispanic 0.0886 (0.1602) 0.3499 (0.1446) 0.2794 (0.2254) 0.1633 (0.2559) 0.004 (0.1198) Male*Other 0.0583 (0.2069) 0.6357 (0.1804) 0.1507 (0.3231) 0.6265 (0.2901) 0.1418 (0.1499) Male*White, non Hispanic 0.0924 (0.1183) 0.1429 (0.1273) 0.0484 (0.1707) 0.4537 (0.2309) 0.1908 (0.0962) Age (Centered) 0.0013 (0.0017) 0.0125 (0.008) 0.005 (0.0023) 0.0104 (0.0077) 0.0038 (0.0096) Age (Centered, Quadratic) 0.0004 (0.0002) 0.0002 (0.0004) 0.0002 (0.0002) 0.0001 (0.0004) 0.0002 (0.0004) CRG 0.0331 (0.0383) 0.2002 (0.0649) 0.0088 (0.0475) 0.2237 (0.1412) 0.0224 (0.0683) CRG (Quadratic) 0.0162 (0.0129) 0.0024 (0.021) 0.0211 (0.019) 0.0133 (0.0502) 0.0039 (0.0236) Poverty % 0.0007 (0.0032) 0.0015 (0.0032) 0.0105 (0.004) 0.0028 (0.0059) 0.0039 (0.0021) Median HHI ($10,000) 0.016 (0.0261) 0.0251 (0.0256) 0.0677 (0.0295) 0.0645 (0.044) 0.0351 (0.016)

PAGE 113

113 Table A 8 Continued ASM CMC MMA PBH PCE Facility residence 0.0445 (0.0497) 0.0621 (0.0401) 0.0993 (0.0636) 0.0139 (0.0949) 0.038 (0.0317) Baseline Compliance Ratio (2006) 0.2657 (0.0607) 0.0474 (0.0478) 0.0776 (0.0619) 0.0776 (0.0892) 0.018 (0.0508) Months Enrolled in STAR+PLUS 0.0015 (0.0042) 0.0016 (0.0039) 0.0012 (0.0045) 0.0266 (0.0074) 0.0237 (0.0021) TIME: 2007 0.0437 (0.037) 0.0515 (0.0607) 0.0588 (0.0427) 0.1918 (0.1957) 0.0953 (0.0779) TIME: 2008 0.0305 (0.0392) 0.1769 (0.061) 0.052 (0.0461) 0.0258 (0.1928) 0.0373 (0.0751) TIME: 2009 0.065 (0.0304) 0.0877 (0.065) 0.0564 (0.0426) 0.0731 (0.1554) 0.0089 (0.0673) CRG*TIME: 2007 0.0429 (0.0402) 0.1228 (0.0755) 0.0912 (0.0529) 0.2437 (0.2218) 0.0155 (0.0889) CRG*TIME: 2008 0.0447 (0.0447) 0.2273 (0.0725) 0.0217 (0.0575) 0.0918 (0.2196) 0.0186 (0.085) CRG*TIME: 2009 0.0723 (0.0383) 0.1214 (0.0814) 0.1292 (0.0521) 0.1854 (0.1652) 0.0175 (0.0771)

PAGE 114

114 APPENDIX B TEXT DESCRIBING TREATMENT VARIABLE OPERATIONALIZATION AND BEHAVIOR The main predictor variable, time enrolled in STAR+PLUS in a given post baseline year, was measured in months. In theory, this variable could take values from 0 to 12, capturing any instances of differential transition timing and scale up or partial Medicaid enrollment. However, our data revealed that almost all transitio ns occurred quickly, that is during the first two months of 2007, and that HEDIS measure eligibility specifications precluded members enrolled for small portions of the year. As a result, this variable essentially behaved rather dichotomously. Even so, mor e accurate estimates of effect are obtained using the exact rather than the approximate enrollment information and thus we retained the original operationalization. As a result, the influence of measurements of individuals only enrolled for part of a y ear (typically 10 or 11 months of the year) are weighted appropriately accounting for their partial exposure to the intervention during the measured year.

PAGE 115

115 APPENDIX C TEXT DESCRIBING STATISTICAL MODEL VIS VIS STANDARD DIFFERENCE IN DIFFERENCE MODEL Frequently used in evaluations of policy implementation, a classic difference in difference (DD) model compares differences in outcomes between two groups at two time points to isolate and test the presence of an effect, assumed to be the deviation from the baseline difference. The unadjusted DD model is given as: where i is the observation unit for outcome Y during time j and for group k ; t is the dichotomous time indicator, before or after policy implementation; p gives the dichotomous group indicator ; and represents their interaction. Accordingly, the average baseline difference between the groups is given by represents the average change in the outcome experienced by the control group. The average deviation from the baseline difference between the groups is given by A mathematical generalization of the standard DD approach, our strategy allows th e simultaneous modeling of more than two time points, while allowing for different slopes between groups over time. This flexibility is accomplished by including the baseline response value as a model covariate and by assuming an unstructured covariance ma trix over time. In addition, we include time invariant and time varying control variables. Our model can be expressed as: where i is the observation unit for outcome Y during time j and for group k ; x ij represents the vector of person level and county level time invariant and time varying characteristics which we added as control variables; y i,j=0 mpliance in the baseline year ; p is the dichotomous group indicator; and t is a categorical variable representing time, which is allowed to have more than two levels The DD model is a specific case of this more generalized model. Our approach borrows str ength over time, thus increasing overall statistical power. Under the standard DD approach, this analysis would require four paired time comparisons; our analytic strategy allows us to compare all four post implementation time points simultaneously; subseq uently, we are not bound by the parallel trend assumption. The model that we use further generalizes the DD model by allowing the baseline difference to be any value, which is restricted to 1.0 in DD (also by the parallel trend assumption). To illustrate, consider a two time point study that has baseline compliance, y i,j=0 and a post baseline compliance, y i,j=1 for subject i in group k By way of the cell mean ANOVA model, the difference score y i,j=1 y i,j=0 can be expressed as: Rearranging terms: where the average change score for fixed group k, is given by equation (1) as:

PAGE 116

116 and substituting for gives the following result: This equation can be generalized to allow for varying group k : Allowing for more than one post baseline measure and n on zero intercept: This result reflects the special case in our model where = 1, and individual control variables, other than the baseline measure compliance, are removed. Model Fitting Details The predictor variables (as described in the article text) were placed into the full model for each outcome in a fixed order: main effects, time, interactions within main effects, and interactions between main effects and time. For each of these f ull models, fixed order backwards selection of predictor variables with alpha = 0.05 was used to arrive at the reduced models. Performing significance tests therefore required testing the higher order interaction variables first. Thus, in the backwards sel ection process, only those main effect variables that did not have a statistically significant interaction were tested; any other main effects were retained. Finding general agreement in terms retained between the models, we established one final reduced m odel form for consistency.

PAGE 117

117 APPENDIX D FIGURES FOR CHAPTER 3: BEHAVIORAL HEALTH Figure D 1. Number of enrollees that qualify for a given measure, by year and by group Figure D 2. Sample spatial distribution

PAGE 118

118 Figure D 3. Weighted, unadjusted measure performance over time, by group

PAGE 119

119 Figure D 4. Regress ion assumption diagnostics for AMM A cute (final and full models)

PAGE 120

120 Figure D 5. Regression assumption diagnostics for AMM CONT (final and full models)

PAGE 121

121 Figure D 6. Regression assumption diagnostics for FUH 7 (final and full models)

PAGE 122

122 Figure D 7. Regression assumption diagnostics for FUH 30 (final and full models)

PAGE 123

123 Figure D 8. Regression assumption diagnostics for IET EGMT (final and full models)

PAGE 124

124 Figure D 9. Regression assumption diagnostics for IET INIT (final and full models)

PAGE 125

125 APPENDIX E TABLES FOR CHAPTER 3: BEHAVIORAL HEALTH CARE Table E 1. Percentage of the sample qualifying for a given measure n [%] AMM only 4827 (14.3) FUH only 5863 (17.3) IET only 16251 (48) Only AMM and FUH 506 (1.5) Only AMM and IET 932 (2.8) Only FUH and IET 4753 (14.1) All measures 691 (2) Table E 2. AAM eligible enrollees: Characteristics of enrollees in transition and comparison counties during the baseline and post baseline periods STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Unique enrollees 554 741 737 1058 Unique counties 26 27 120 126 Enrollees with complete data 510 (92.1) 713 (96.2) 686 (93.1) 935 (88.4) Counties with complete data, by year 26 (100) 26 (96.3) 114 (95) 90 (71.4) Counties with complete data in post and pre years 26 (96.3) 90 (71.4) Age [mean (SD)] 45.2 (11.5) 46.3 (11.6) 46.7 (11.5) 46.1 (11.9) Age categories [n (%)] 21 30 74 (13.4) 91 (12.3) 82 (11.1) 139 (13.1) 30 39 98 (17.7) 111 (15) 101 (13.7) 164 (15.5) 40 49 153 (27.6) 214 (28.9) 215 (29.2) 260 (24.6) 50 59 169 (30.5) 226 (30.5) 226 (30.7) 347 (32.8) 60 64 60 (10.8) 99 (13.4) 113 (15.3) 148 (14) Age not available 0 (0) 0 (0) 0 (0) 0 (0) Female [n (%)] 368 (66.4) 529 (71.4) 536 (72.7) 748 (70.7) Sex not available 0 (0) 0 (0) 0 (0) 0 (0) Race/ethnicity [n (%)] Non H ispanic black 67 (12.1) 94 (12.7) 102 (13.8) 208 (19.7) Hispanic 291 (52.5) 401 (54.1) 253 (34.3) 298 (28.2) Other 22 (4) 39 (5.3) 65 (8.8) 92 (8.7) Non H ispanic white 174 (31.4) 207 (27.9) 317 (43) 460 (43.5) Race/ethnicity not available 0 (0) 0 (0) 0 (0) 0 (0) Health status [n (%)] Healthy 41 (7.8) 30 (4.1) 28 (4) 65 (6.5) Significant acute 10 (1.9) 7 (1) 4 (0.6) 7 (0.7) Minor chronic 39 (7.4) 41 (5.7) 44 (6.3) 66 (6.6)

PAGE 126

126 Table E 2. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Moderate chronic 112 (21.4) 162 (22.4) 162 (23.3) 229 (23) Major chronic 322 (61.5) 484 (66.9) 457 (65.8) 630 (63.2) Health status not available 30 (5.4) 17 (2.3) 42 (5.7) 61 (5.8) County metro classification [n (%)] Large metro (>1M) 441 (79.6) 606 (81.8) 20 (2.7) 23 (2.2) Medium metro (250K 1M) 78 (14.1) 89 (12) 308 (41.8) 394 (37.2) Small metro (<250k) 16 (2.9) 17 (2.3) 199 (27) 328 (31) Adjacent to metro 19 (3.4) 29 (3.9) 135 (18.3) 221 (20.9) Nonadjacent to metro (>20k) 0 (0) 0 (0) 36 (4.9) 45 (4.3) Nonadjacent to metro (<20k) 0 (0) 0 (0) 39 (5.3) 47 (4.4) Census tract poverty [mean (SD)] 24.4 (12.5) 24 (12.7) 26.7 (14.1) 25.1 (13.4) Census tract poverty categories [n (%)] 0.0% 4.9% 22 (4.1) 30 (4.1) 11 (1.5) 22 (2.1) 5.0% 9.9% 41 (7.6) 65 (8.9) 56 (7.7) 97 (9.3) 10.0% 19.9% 159 (29.3) 215 (29.3) 208 (28.5) 326 (31.1) Poverty area (20.0% 39.9%) 262 (48.3) 344 (46.9) 317 (43.4) 431 (41.1) Extreme poverty area (> 40.0%) 59 (10.9) 79 (10.8) 138 (18.9) 172 (16.4) Census tract not available 11 (2) 8 (1.1) 7 (0.9) 10 (0.9) Census tract unemployment level [n (%)] 0 (NA) 0 (NA) 0 (NA) 0 (NA) 0.0% 4.9% 76 (14) 100 (13.6) 146 (20) 225 (21.5) 5.0% 9.9% 259 (47.7) 369 (50.3) 358 (49) 472 (45.1) 10.0% 19.9% 203 (37.4) 253 (34.5) 218 (29.9) 326 (31.1) > 20.0% 5 (0.9) 11 (1.5) 8 (1.1) 24 (2.3) Census tract not available 11 (2) 8 (1.1) 7 (0.9) 11 (1) Census tract household income [mean (SD)] 38148.8 (16672.5) 39465.3 (17563) 35429.8 (13311.6) 37165.2 (13508.2) County Medicaid discharge density [mean (SD)] 1.1 (0.5) 1.1 (0.5) 1.1 (0.6) 1.1 (0.6) Census tract determined from address [n (%)] 466 (84.1) 632 (85.3) 579 (78.6) 833 (78.7) Facility residence [n (%)] 106 (19.4) 126 (17.1) 167 (22.8) 203 (19.3)

PAGE 127

127 Table E 2. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison ) Pre Post Pre Post Months enrolled in Medicaid [n (%)] 12 469 (84.7) 664 (89.6) 640 (86.8) 952 (90) 9 11 43 (7.8) 44 (5.9) 60 (8.1) 54 (5.1) 5 8 19 (3.4) 18 (2.4) 17 (2.3) 31 (2.9) 2 5 21 (3.8) 11 (1.5) 20 (2.7) 18 (1.7) 1 2 (0.4) 2 (0.3) 0 (0) 2 (0.2) STAR+PLUS transition cohort [n (%)] A 0 (0) 0 (0) 0 (0) 0 (0) B 554 (100) 741 (100) 0 (0) 0 (0) C 0 (0) 0 (0) 0 (0) 0 (0) D 0 (0) 0 (0) 87 (11.8) 188 (17.8) E 0 (0) 0 (0) 318 (43.1) 340 (32.1) F 0 (0) 0 (0) 332 (45) 530 (50.1) Table E 3. FUH eligible enrollees: Characteristics of enrollees in transition and comparison counties during the baseline and post baseline periods STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Unique enrollees 1420 1875 1730 2223 Unique counties 27 28 146 150 Enrollees with complete data 1282 (90.3) 1718 (91.6) 1493 (86.3) 1851 (83.3) Counties with complete data, by year 26 (96.3) 27 (96.4) 139 (95.2) 128 (85.3) Counties with complete data in post and pre years 27 (96.4) 128 (85.3) Age [mean (SD)] 41.2 (11.3) 41.1 (11.9) 40.9 (11.5) 40 (11.9) Age categories [n (%)] 21 30 287 (20.2) 445 (23.7) 377 (21.8) 558 (25.1) 30 39 307 (21.6) 376 (20.1) 376 (21.7) 538 (24.2) 40 49 459 (32.3) 520 (27.7) 522 (30.2) 572 (25.7) 50 59 305 (21.5) 425 (22.7) 363 (21) 437 (19.7) 60 64 62 (4.4) 109 (5.8) 92 (5.3) 118 (5.3) Age not available 0 (0) 0 (0) 0 (0) 0 (0) Female [n (%)] 780 (54.9) 1027 (54.8) 1074 (62.1) 1242 (55.9) Sex not available 0 (0) 0 (0) 0 (0) 0 (0) Race/ethnicity [n (%)] Non H ispanic black 233 (16.4) 286 (15.3) 328 (19) 358 (16.1) Hispanic 600 (42.3) 719 (38.3) 359 (20.8) 545 (24.5)

PAGE 128

128 Table E 3. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Other 68 (4.8) 153 (8.2) 114 (6.6) 230 (10.3) Non H ispanic white 519 (36.5) 717 (38.2) 929 (53.7) 1090 (49) Race/ethnicity not available 0 (0) 0 (0) 0 (0) 0 (0) Health status [n (%)] Healthy 0 (0) 1 (0.1) 3 (0.2) 2 (0.1) Significant acute 0 (0) 1 (0.1) 0 (0) 0 (0) Minor chronic 0 (0) 0 (0) 2 (0.1) 4 (0.2) Moderate chronic 335 (25.3) 450 (25.5) 424 (27.6) 466 (24.1) Major chronic 988 (74.7) 1310 (74.3) 1106 (72.1) 1464 (75.6) Health status not available 97 (6.8) 113 (6) 195 (11.3) 287 (12.9) County metro classification [n (%)] Large metro (>1M) 1113 (78.4) 1576 (84.1) 44 (2.5) 48 (2.2) Medium metro (250K 1M) 249 (17.5) 224 (11.9) 470 (27.2) 750 (33.7) Small metro (<250k) 17 (1.2) 21 (1.1) 605 (35) 717 (32.3) Adjacent to metro 41 (2.9) 54 (2.9) 424 (24.5) 510 (22.9) Nonadjacent to metro (>20k) 0 (0) 0 (0) 119 (6.9) 96 (4.3) Nonadjacent to metro (<20k) 0 (0) 0 (0) 68 (3.9) 102 (4.6) Census tract poverty [mean (SD)] 23.5 (13.3) 22.1 (12.5) 23 (12.3) 23.6 (12.8) Census tract poverty categories [n (%)] 0.0% 4.9% 63 (4.5) 122 (6.6) 34 (2) 61 (2.8) 5.0% 9.9% 141 (10.2) 184 (10) 183 (10.8) 217 (10) 10.0% 19.9% 419 (30.2) 564 (30.5) 605 (35.7) 742 (34.1) Poverty area (20.0% 39.9%) 601 (43.3) 815 (44.1) 691 (40.8) 902 (41.5) Extreme poverty area (> 40.0%) 164 (11.8) 162 (8.8) 181 (10.7) 254 (11.7) Census tract not available 32 (2.3) 28 (1.5) 36 (2.1) 47 (2.1) Census tract unemployment level [n (%)] 0 (NA) 0 (NA) 0 (NA) 0 (NA) 0.0% 4.9% 202 (14.6) 292 (15.8) 401 (23.7) 512 (23.6) 5.0% 9.9% 651 (46.9) 941 (50.9) 809 (47.8) 967 (44.5) 10.0% 19.9% 508 (36.6) 586 (31.7) 461 (27.2) 669 (30.8)

PAGE 129

129 Table E 3. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post > 20.0% 27 (1.9) 28 (1.5) 22 (1.3) 26 (1.2) Census tract not available 32 (2.3) 28 (1.5) 37 (2.1) 49 (2.2) Census tract household income [mean (SD)] 39813 (17933.7) 42586.4 (19399.4) 38559.1 (13764.6) 38331.8 (13790) County Medicaid discharge density [mean (SD)] 1.1 (0.5) 1.1 (0.5) 1 (0.6) 1 (0.6) Census tract determined from address [n (%)] 1194 ( 84.1) 1605 (85.6) 1279 (73.9) 1695 (76.2) Facility residence [n (%)] 292 (20.9) 364 (19.7) 424 (24.8) 474 (21.7) Months enrolled in Medicaid [n (%)] 12 1051 (74) 1380 (73.6) 1357 (78.4) 1592 (71.6) 9 11 164 (11.5) 260 (13.9) 186 (10.8) 253 (11.4) 5 8 117 (8.2) 135 (7.2) 110 (6.4) 279 (12.6) 2 5 88 (6.2) 100 (5.3) 77 (4.5) 99 (4.5) 1 0 (0) 0 (0) 0 (0) 0 (0) STAR+PLUS transition cohort [n (%)] A 0 (0) 0 (0) 0 (0) 0 (0) B 1420 (100) 1875 (100) 0 (0) 0 (0) C 0 (0) 0 (0) 0 (0) 0 (0) D 0 (0) 0 (0) 239 (13.8) 370 (16.6) E 0 (0) 0 (0) 464 (26.8) 671 (30.2) F 0 (0) 0 (0) 1027 (59.4) 1182 (53.2) Frequency of times qualifying for measure 1 1404 (74.9) 1771 (79.7) 2 335 (17.9) 347 (15.6) 3 99 (5.3) 81 (3.6) > 3 37 (2) 24 (1.1)

PAGE 130

130 Table E 4. IET eligible enrollees: Characteristics of enrollees in transition and comparison counties during the baseline and post baseline periods STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Unique enrollees 2553 3422 3976 4994 Unique counties 28 28 177 177 Enrollees with complete data 2354 (92.2) 3268 (95.5) 3539 (89) 4411 (88.3) Counties with complete data, by year 28 (100) 28 (100) 173 (97.7) 160 (90.4) Counties with complete data in post and pre years 28 (100) 160 (90.4) Age [mean (SD)] 44.6 (10.8) 45.4 (11.2) 45.1 (11) 44.9 (11.4) Age categories [n (%)] 21 30 313 (12.3) 428 (12.5) 465 (11.7) 692 (13.9) 30 39 428 (16.8) 579 (16.9) 678 (17.1) 827 (16.6) 40 49 875 (34.3) 973 (28.4) 1293 (32.5) 1442 (28.9) 50 59 755 (29.6) 1140 (33.3) 1176 (29.6) 1630 (32.6) 60 64 182 (7.1) 302 (8.8) 364 (9.2) 403 (8.1) Age not available 0 (0) 0 (0) 0 (0) 0 (0) Female [n (%)] 1113 (43.6) 1615 (47.2) 1799 (45.2) 2331 (46.7) Sex not available 0 (0) 0 (0) 0 (0) 0 (0) Race/ethnicity [n (%)] Non H ispanic black 476 (18.6) 616 (18) 863 (21.7) 1115 (22.3) Hispanic 1029 (40.3) 1320 (38.6) 944 (23.7) 1018 (20.4) Other 115 (4.5) 242 (7.1) 222 (5.6) 375 (7.5) Non H ispanic white 933 (36.5) 1244 (36.4) 1947 (49) 2486 (49.8) Race/ethnicity not available 0 (0) 0 (0) 0 (0) 0 (0) Health status [n (%)] Healthy 67 (2.8) 110 (3.3) 144 (4) 189 (4.2) Significant acute 20 (0.8) 34 (1) 57 (1.6) 69 (1.5) Minor chronic 30 (1.2) 48 (1.4) 70 (1.9) 106 (2.3)

PAGE 131

131 Table E 4. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Moderate chronic 642 (26.5) 703 (21) 882 (24.3) 1045 (23) Major chronic 1665 (68.7) 2454 (73.3) 2479 (68.3) 3126 (68.9) Health status not available 129 (5.1) 73 (2.1) 344 (8.7) 459 (9.2) County metro classification [n (%)] Large metro (>1M) 1979 (77.5) 2869 (83.8) 118 (3) 140 (2.8) Medium metro (250K 1M) 408 (16) 399 (11.7) 1023 (25.7) 1142 (22.9) Small metro (<250k) 41 (1.6) 46 (1.3) 1324 (33.3) 1730 (34.6) Adjacent to metro 125 (4.9) 108 (3.2) 1065 (26.8) 1404 (28.1) Nonadjacent to metro (>20k) 0 (0) 0 (0) 233 (5.9) 305 (6.1) Nonadjacent to metro (<20k) 0 (0) 0 (0) 213 (5.4) 273 (5.5) Census tract poverty [mean (SD)] 23.3 (12.3) 22.5 (12.3) 24.4 (12.8) 23.1 (12.3) Census tract poverty categories [n (%)] 0.0% 4.9% 99 (4) 170 (5) 71 (1.8) 101 (2.1) 5.0% 9.9% 216 (8.6) 329 (9.7) 334 (8.5) 473 (9.6) 10.0% 19.9% 801 (32) 1113 (32.9) 1314 (33.6) 1737 (35.4) Poverty area (20.0% 39.9%) 1149 (45.9) 1458 (43.1) 1685 (43.1) 2087 (42.5) Extreme poverty area (> 40.0%) 240 (9.6) 315 (9.3) 503 (12.9) 508 (10.4) Census tract not available 48 (1.9) 37 (1.1) 69 (1.7) 88 (1.8) Census tract unemployment level [n (%)] 0 (NA) 0 (NA) 0 (NA) 0 (NA) 0.0% 4.9% 366 (14.6) 497 (14.7) 847 (21.7) 1180 (24.1) 5.0% 9.9% 1248 (49.8) 1787 (52.8) 1878 (48.1) 2234 (45.6) 10.0% 19.9% 842 (33.6) 1062 (31.4) 1115 (28.5) 1399 (28.5) > 20.0% 49 (2) 39 (1.2) 66 (1.7) 91 (1.9)

PAGE 132

132 Table E 4. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Census tract household income [mean (SD)] 40176.6 (17580.8) 41989.7 (18154) 36534.8 (13296.1) 37679.2 (13012.7) County Medicaid discharge density [mean (SD)] 1.1 (0.6) 1.1 (0.5) 0.9 (0.6) 0.9 (0.6) Census tract determined from address [n (%)] 2122 (83.1) 2977 (87) 3068 (77.2) 3850 (77.1) Facility residence [n (%)] 510 (20.3) 606 (18) 813 (20.7) 913 (18.5) Months enrolled in Medicaid [n (%)] 12 1985 (77.8) 2787 (81.4) 3174 (79.8) 3889 (77.9) 9 11 271 (10.6) 387 (11.3) 443 (11.1) 434 (8.7) 5 8 180 (7.1) 188 (5.5) 238 (6) 576 (11.5) 2 5 117 (4.6) 60 (1.8) 121 (3) 95 (1.9) 1 0 (0) 0 (0) 0 (0) 0 (0) STAR+PLUS transition cohort [n (%)] A 0 (0) 0 (0) 0 (0) 0 (0) B 2553 (100) 3422 (100) 0 (0) 0 (0) C 0 (0) 0 (0) 0 (0) 0 (0) D 0 (0) 0 (0) 547 (13.8) 803 (16.1) E 0 (0) 0 (0) 1114 (28) 1053 (21.1) F 0 (0) 0 (0) 2315 (58.2) 3138 (62.8)

PAGE 133

133 Table E 5 Summary of sample size, by group and by measure STAR+PLUS (Transition) FFS+PCCM (Comparison) Measure Type Min Q1 Media n Q3 Max Min Q1 Median Q3 Max AMM 1 4.00 7.00 14.00 387 1 1.00 3.00 8.00 196 FUH 2 6.00 16.00 30.00 872 1 2.00 4.00 13.00 284 IET 4 19.0 0 37.00 76.00 1246 1 4.00 11.00 29.00 325 Table E 6 Backward step wise regression results for AMM Acute L abels R 2 F p A ction Full Model NA NA NA Gender X Enrolled In STAR + PLUS 0.0017 1.6 0.2143 Removed Medicaid Discharge Density 0.1390 10.7 0.0015 Kept Metro Classification 2 0.0007 1.0 0.3098 Removed Metro Classification 0.0048 2.5 0.1150 Removed Unemployment 0.0859 8.8 0.0039 Kept Poverty 0.0009 1.1 0.2924 Removed County Income 0.0020 1.6 0.2036 Removed Median Household Income 0.0431 6.6 0.0118 Kept Group Facility Proxy 0.0001 0.4 0.5360 Removed Health Status 2 0.0695 8.0 0.0055 Kept Health Status NA NA NA Kept Age 2 0.5421 16.4 0.0001 Kept Age NA NA NA Kept Gender X Race/Ethnicity 0.4495 5.2 0.0024 Kept Race/Ethnicity NA NA NA Kept Gender NA NA NA Kept Baseline 0.2529 12.9 0.0005 Kept Table E 7 Backward step wise regression results for AM M Cont L abels R 2 F p A ction Full Model NA NA NA Gender X Enrolled In STAR + PLUS 0.0157 4.4 0.0378 Kept Medicaid Discharge Density 0.0323 6.1 0.0156 Kept

PAGE 134

134 Table E 7 Continued Labels R 2 F p Action Metro Classification 2 0.0039 2.3 0.1303 Removed Metro Classification 0.0008 1.1 0.3038 Removed Unemployment 0.0000 0.1 0.7963 Removed Poverty 0.0000 0.0 0.8833 Removed County Income 0.0002 0.6 0.4466 Removed Median Household Income 0.0057 2.8 0.0956 Removed Group Facility Proxy 0.0000 0.1 0.7253 Removed Health Status 2 0.0000 0.0 0.8777 Removed Health Status 0.0122 3.0 0.0849 Removed Age 2 0.2060 8.6 0.0042 Kept Age NA NA NA Kept Gender X Race/Ethnicity 1.4760 5.0 0.0028 Kept Race/Ethnicity NA NA NA Kept Gender NA NA NA Kept Baseline 0.0026 1.3 0.2522 Removed Table E 8 Backward step wise regression results for FUH 7 L abels R 2 F p A ction Full Model NA NA NA Gender Enrolled In STAR + PLUS 0.0000 0.1 0.8101 Removed Medicaid Discharge Density 0.0344 11.5 0.0009 Kept Metro Classification 2 0.0357 11.7 0.0008 Kept Metro Classification NA NA NA Kept Unemployment 0.0009 2.1 0.1480 Removed Poverty 0.0029 3.7 0.0562 Removed County Income 0.0100 6.5 0.0118 Kept Median Household Income 0.0253 9.8 0.0021 Kept Group Facility Proxy 0.0001 0.8 0.3820 Removed Health Status 2 0.0244 9.7 0.0023 Kept Health Status NA NA NA Kept Age 3 0.0012 2.4 0.1242 Removed Age 0.0000 0.3 0.5702 Removed Gender X Race/Ethnicity 0.1132 6.0 0.0007 Kept Race/Ethnicity NA NA NA Kept Gender NA NA NA Kept Baseline 0.1094 17.7 0.0000 Kept Table E 9 Backward step wise regression results for FUH 30 L abels R 2 F p A ction Full Model NA NA NA Gender X Enrolled In STAR + PLUS 0.0000 0.1 0.7989 Removed Medicaid Discharge Density 0.0094 3.8 0.0536 Removed Metro Classification 2 0.0000 0.0 0.9238 Removed

PAGE 135

135 Table E 9 Continued Labels R 2 F p Action Metro Classification 0.0345 6.6 0.0111 Kept Unemployment 0.0046 2.7 0.1026 Removed Poverty 0.0000 0.2 0.6774 Removed County Income 0.1388 11.4 0.0009 Kept Median Household Income 0.2665 14.3 0.0002 Kept Group Facility Proxy 0.0002 0.5 0.4746 Removed Health Status 0.1565 12.0 0.0007 Kept Age 2 0.0022 1.9 0.1711 Removed Age 0.0000 0.0 0.8358 Removed Gender X Race/Ethnicity 0.0293 2.1 0.1084 Removed Race/Ethnicity 11.6010 10.6 0.0000 Kept Gender 0.0178 4.9 0.0290 Kept Baseline 0.0137 4.3 0.0391 Kept Table E 10 Backward step wise regression results for IET EGMT V ariable R 2 T otal R 2 F value p value A ction Full Model 0.3193 NA NA NA Medicaid Discharge Density 0.3178 0.0015 0.4 0.5443 Removed Metro Classification 2 0.3144 0.0034 0.8 0.3636 Removed Metro Classification 0.3134 0.0010 0.2 0.6280 Removed Unemployment 0.3134 0.0000 0.0 0.9968 Removed Poverty 0.3103 0.0031 0.8 0.3799 Removed County Income 0.3102 0.0001 0.0 0.9107 Removed Median Household Income 0.3082 0.0021 0.5 0.4760 Removed Group Facility Proxy 0.3037 0.0045 1.1 0.2930 Removed Health Status 2 0.2923 0.0115 2.9 0.0918 Removed Health Status 0.2781 0.0142 3.5 0.0626 Removed Age 3 0.2751 0.0027 0.7 0.4163 Removed Age 2 0.2747 0.0004 0.1 0.7560 Removed Age 0.2722 0.0025 0.6 0.4355 Removed Gender X Race/Ethnicity 0.2721 0.0002 0.0 0.9977 Removed Race /Ethnicity 0.2431 0.0290 2.4 0.0667 Removed Gender 0.2431 0.0687 16.9 0.0001 Kept Baseline 0.2431 0.1105 27.2 0.0000 Kept Table E 11 Backward step wise regression results for IET INIT V ariable R 2 T otal R 2 F value p value A ction Full Model 0.2436 NA NA NA Medicaid Discharge Density 0.2436 0.0197 4.3 0.0393 Kept Metro Classification 2 0.2431 0.0005 0.1 0.7502 Removed Metro Classification 0.2359 0.0072 1.6 0.2086 Removed Unemployment 0.2342 0.0016 0.4 0.5506 Remov ed Poverty 0.2241 0.0101 2.2 0.1372 Removed

PAGE 136

136 Table E 11 Continued Variable R 2 Total R 2 F value p value Action County Income 0.2181 0.0060 1.3 0.2526 Removed Median Household Income 0.2164 0.0017 0.4 0.5400 Removed Group Facility Proxy 0.2147 0.0016 0.4 0.5481 Removed Health Status 0.2151 0.0569 12.6 0.0005 Kept Age 3 0.2151 0.0000 0.0 0.9440 Removed Age 2 0.2150 0.0001 0.0 0.8925 Removed Age 0.2147 0.0003 0.1 0.8046 Removed Gender X Race/Ethnicity 0.2098 0.0050 0.4 0.7728 Removed Race /Ethnicity 0.1874 0.0224 1.7 0.1690 Removed Gender 0.1874 0.0483 10.9 0.0012 Kept Baseline 0.1874 0.0496 11.2 0.0010 Kept Table E 12 Full model coefficients for AMM Acute Variable Value (std error) 95% C.I. (Intercept) 0.7173 (0.1128) (0.4933 0.9412) Male 0.0695 (0.2375) ( 0.4021 0.5412) Hispanic 0.0595 (0.1070) ( 0.2719 0.1528) Other 0.0492 (0.2019) ( 0.4501 0.3517) Whi te 0.1001 (0.1096) ( 0.1176 0.3177) Group Facility Proxy 0.1658 (0.1160) ( 0.0646 0.3962) Male Hispanic 0.0069 (0.0325) ( 0.0576 0.0715) Male Other 0.0468 (0.0509) ( 0.1478 0.0542) Male White 0.0070 (0.0269) ( 0.0604 0.0464) Age 0.0051 (0.0044) ( 0.0036 0.0139) Age 2 0.0005 (0.0005) ( 0.0014 0.0004) Health Status 0.0747 (0.0639) ( 0.2016 0.0523) Health Status 2 0.0602 (0.0344) ( 0.1285 0.0082) County Income 0.0221 (0.0362) ( 0.0939 0.0498) Median Household Income 0.0508 (0.0364) ( 0.0215 0.1231) Poverty 0.0034 (0.0050) ( 0.0066 0.0133) Unemployment 0.0025 (0.0101) ( 0.0175 0.0225) Metro Classification 0.0155 (0.0168) ( 0.0489 0.0180) Metro Classification 2 0.0026 (0.0089) ( 0.0151 0.0203) Medicaid Discharge Density 0.0377 (0.0283) ( 0.0186 0.0940) Baseline 0.0407 (0.0804) ( 0.2002 0.1189) Months Enrolled in STAR+PLUS 0.0016 (0.1069) ( 0.2139 0.2107) Male Months Enrolled in STAR+PLUS 0.1719 (0.2722) ( 0.7124 0.3687) Table E 13 Full model coefficients for AMM Cont Variable Value (std error) 95% C.I. (Intercept) 0.4070 (0.1168) (0.1751 0.6390) Male 0.0959 (0.2369) ( 0.3746 0.5664) Hispanic 0.0301 (0.1113) ( 0.2511 0.1910)

PAGE 137

137 Table E 13 Continued Variable Value (std error) 95% C.I. Other 0.1498 (0.2085) ( 0.2643 0.5639) White 0.2089 (0.1131) ( 0.0157 0.4335) Group Facility Proxy 0.1872 (0.1152) ( 0.0415 0.4158) Male Hispanic 0.0176 (0.0333) ( 0.0838 0.0485) Male Other 0.0336 (0.0508) ( 0.1344 0.0672) Male White 0.0050 (0.0266) ( 0.0579 0.0479) Age 0.0038 (0.0044) ( 0.0050 0.0125) Age 2 0.0001 (0.0004) ( 0.0010 0.0008) Health Status 0.0465 (0.0645) ( 0.1745 0.0815) Health Status 2 0.0256 (0.0344) ( 0.0938 0.0426) County Income 0.0110 (0.0366) ( 0.0836 0.0617) Median Household Income 0.0247 (0.0395) ( 0.1032 0.0538) Poverty 0.0014 (0.0051) ( 0.0116 0.0088) Unemployment 0.0020 (0.0105) ( 0.0189 0.0228) Metro Classification 0.0223 (0.0177) ( 0.0575 0.0130) Metro Classification 2 0.0097 (0.0091) ( 0.0085 0.0278) Medicaid Discharge Density 0.0153 (0.0301) ( 0.0446 0.0751) Baseline 0.0124 (0.0734) ( 0.1581 0.1334) Months Enrolled in STAR+PLUS 0.0685 (0.1080) ( 0.1459 0.2829) Male Months Enrolled in STAR+PLUS 0.2133 (0.2726) ( 0.7548 0.3281) Table E 14 Full model coefficients for FUH 7 Variable Value (std error) 95% C.I. (Intercept) 0.1818 (0.0575) (0.0680 0.2956) Male 0.2172 (0.2252) ( 0.6626 0.2283) Hispanic 0.0429 (0.0839) ( 0.2088 0.1230) Other 0.0720 (0.1488) ( 0.3663 0.2223) White 0.1230 (0.0834) ( 0.2880 0.0420) Group Facility Proxy 0.2003 (0.0808) ( 0.3602 0.0405) Male Hispanic 0.0267 (0.0277) ( 0.0281 0.0815) Male Other 0.0484 (0.0357) ( 0.0223 0.1190) Male White 0.0347 (0.0246) ( 0.0138 0.0833) Age 0.0034 (0.0040) ( 0.0113 0.0044) Age 2 0.0001 (0.0003) ( 0.0004 0.0007) Health Status 0.0525 (0.1139) ( 0.2779 0.1728) Health Status 2 0.0253 (0.1163) ( 0.2554 0.2048) County Income 0.0424 (0.0223) ( 0.0017 0.0864) Median Household Income 0.0061 (0.0307) ( 0.0668 0.0547) Poverty 0.0017 (0.0035) ( 0.0087 0.0053) Unemployment 0.0051 (0.0072) ( 0.0193 0.0091) Metro Classification 0.0029 (0.0118) ( 0.0262 0.0204) Metro Classification 2 0.0126 (0.0054) (0.0019 0.0233) Medicaid Discharge Density 0.0008 (0.0183) ( 0.0369 0.0354)

PAGE 138

138 Table E 14 Continued Variable Value (std error) 95% C.I. Baseline 0.2425 (0.0743) (0.0955 0.3895) Months Enrolled in STAR+PLUS 0.0714 (0.1000) ( 0.2692 0.1264) Male Months Enrolled in STAR+PLUS 0.2072 (0.2107) ( 0.2096 0.6239) Table E 15 Final model coefficients for 30 Variable Value (std error) 95% C.I. (Intercept) 0.3574 (0.0741) (0.2108 0.5040) Male 0.0141 (0.2898) ( 0.5591 0.5873) Hispanic 0.0237 (0.1074) ( 0.2361 0.1887) Other 0.0867 (0.1891) ( 0.4607 0.2873) White 0.1839 (0.1067) ( 0.3949 0.0271) Group Facility Proxy 0.1236 (0.1051) ( 0.3314 0.0843) Male Hispanic 0.0020 (0.0358) ( 0.0688 0.0729) Male Other 0.0321 (0.0463) ( 0.0594 0.1237) Male White 0.0007 (0.0318) ( 0.0621 0.0636) Age 0.0058 (0.0051) ( 0.0159 0.0043) Age 2 0.0001 (0.0004) ( 0.0006 0.0009) Health Status 0.0834 (0.1456) ( 0.2046 0.3714) Health Status 2 0.0842 (0.1439) ( 0.2005 0.3689) County Income 0.0129 (0.0286) ( 0.0693 0.0436) Median Household Income 0.0345 (0.0387) ( 0.0420 0.1110) Poverty 0.0022 (0.0045) ( 0.0111 0.0067) Unemployment 0.0126 (0.0092) ( 0.0308 0.0056) Metro Classification 0.0051 (0.0150) ( 0.0246 0.0349) Metro Classification 2 0.0088 (0.0069) ( 0.0049 0.0225) Medicaid Discharge Density 0.0120 (0.0232) ( 0.0338 0.0579) Baseline 0.0014 (0.0734) ( 0.1438 0.1467) Months Enrolled in STAR+PLUS 0.0088 (0.1280) ( 0.2445 0.2621) Male* Months Enrolled in STAR+PLUS 0.0459 (0.2698) ( 0.4878 0.5795) Table E 16 Full model coefficients for IET EGMT Variable Value (std error) 95% C.I. (Intercept) 0.0328 (0.0187) ( 0.0042 0.0698) Male 0.0907 (0.0775) ( 0.0623 0.2436) Hispanic 0.0120 (0.0244) ( 0.0360 0.0601) Other 0.0052 (0.0596) ( 0.1228 0.1124) White 0.0380 (0.0256) ( 0.0885 0.0125) Group Facility Proxy 0.0278 (0.0284) ( 0.0838 0.0282) Male Hispanic 0.0048 (0.0200) ( 0.0443 0.0347) Male Other 0.0152 (0.0265) ( 0.0676 0.0372) Male White 0.0034 (0.0171) ( 0.0371 0.0303) Age 0.0019 (0.0032) ( 0.0083 0.0045) Age 2 0.0001 (0.0001) ( 0.0002 0.0004)

PAGE 139

139 Table E 16 Continued Variable Value (std error) 95% C.I. Age Cu 0.0000 (0.0000) ( 0.0000 0.0000) Health Status 0.0262 (0.0347) ( 0.0948 0.0424) Health Status 2 0.0192 (0.0110) ( 0.0410 0.0026) County Income 0.0005 (0.0064) ( 0.0132 0.0122) Median Household Income 0.0036 (0.0095) ( 0.0223 0.0151) Poverty 0.0012 (0.0011) ( 0.0033 0.0009) Unemployment 0.0002 (0.0021) ( 0.0044 0.0039) Metro Classification 0.0006 (0.0034) ( 0.0073 0.0061) Metro Classification 2 0.0015 (0.0016) ( 0.0017 0.0046) Medicaid Discharge Density 0.0035 (0.0058) ( 0.0079 0.0150) Baseline 0.2540 (0.0558) (0.1439 0.3642) Months Enrolled in STAR+PLUS 0.0131 (0.0126) ( 0.0117 0.0380) Table E 17 Full model coefficients for I ET INIT Variable Value (std error) 95% C.I. (Intercept) 0.3646 (0.0565) (0.2531 0.4760) Male 0.4672 (0.2028) (0.0667 0.8677) Hispanic 0.0201 (0.0674) ( 0.1532 0.1130) Other 0.2390 (0.1561) ( 0.0691 0.5471) White 0.0812 (0.0701) ( 0.0573 0.2196) Group Facility Proxy 0.0252 (0.0743) ( 0.1718 0.1214) Male Hispanic 0.0506 (0.0519) ( 0.1531 0.0518) Male Other 0.0662 (0.0695) ( 0.2034 0.0709) Male White 0.0454 (0.0448) ( 0.1338 0.0430) Age 0.0017 (0.0083) ( 0.0180 0.0146) Age 2 0.0001 (0.0004) ( 0.0009 0.0006) Age 3 0.0000 (0.0000) ( 0.0001 0.0001) Health Status 0.1390 (0.0413) (0.0574 0.2206) County Income 0.0200 (0.0181) ( 0.0557 0.0158) Median Household Income 0.0243 (0.0275) ( 0.0300 0.0786) Poverty 0.0038 (0.0030) ( 0.0020 0.0097) Unemployment 0.0052 (0.0058) ( 0.0166 0.0061) Metro Classification 0.0120 (0.0093) ( 0.0303 0.0063) Metro Classification 2 0.0014 (0.0043) ( 0.0099 0.0072) Medicaid Discharge Density 0.0341 (0.0164) (0.0017 0.0666) Baseline 0.1922 (0.0594) (0.0750 0.3094) Months Enrolled in STAR+PLUS 0.0204 (0.0354) ( 0.0494 0.0903) Table E 18 Final model coefficients for AMM Acute Variable Value (std error) 95% C.I. (Intercept) 0.5411 (0.0136) (0.5142 0.5680) Months Enrolled in STAR+PLUS 0.0433 (0.0276) ( 0.0977 0.0112)

PAGE 140

140 Table E 19 Final model coefficients for AMM Cont Variable Value (std error) 95% C.I. (Intercept) 0.3864 (0.0148) (0.3571 0.4157) Months Enrolled in STAR+PLUS 0.0037 (0.0326) ( 0.0606 0.0681) Table E 20 Final model coefficients for FUH 7 Variable Value (std error) 95% C.I. (Intercept) 0.2111 (0.0135) (0.1845 0.2377) Group Facility Proxy 0.1799 (0.0763) ( 0.3306 0.0292) County Income 0.0429 (0.0164) (0.0105 0.0754) Metro Classification 0.0021 (0.0096) ( 0.0211 0.0170) Metro Classification 2 0.0125 (0.0050) (0.0026 0.0224) Baseline 0.2320 (0.0653) (0.1030 0.3611) Months Enrolled in STAR+PLUS 0.0243 (0.0352) ( 0.0452 0.0938) Table E 21 Final model coefficients for FUH 30 Variable Value (std error) 95% C.I. (Intercept) 0.4736 (0.0124) (0.4491 0.4980) Median Household Income 0.0321 (0.0170) ( 0.0015 0.0657) Months Enrolled in STAR+PLUS 0.0543 (0.0325) ( 0.0099 0.1185) Table E 22 Final model coefficients for IET EGMT Variable Value (std error) 95% C.I. (Intercept) 0.0377 (0.0030) (0.0318 0.0435) Male 0.0992 (0.0242) (0.0516 0.1469) Baseline 0.2758 (0.0529) (0.1714 0.3802) Months Enrolled in STAR+PLUS 0.0290 (0.0070) (0.0152 0.0428) Table E 23 Final model coefficients for IET INIT Variable Value (std error) 95% C.I. (Intercept) 0.3609 (0.0122) (0.3368 0.3849) Male 0.2183 (0.0662) (0.0877 0.3488) Health Status 0.1242 (0.0380) (0.0492 0.1992) Medicaid Discharge Density 0.0280 (0.0141) (0.0003 0.0558) Baseline 0.1896 (0.0567) (0.0777 0.3015) Months Enrolled in STAR+PLUS 0.0251 (0.0213) ( 0.0170 0.0671)

PAGE 141

141 APPENDIX F FIGURES FOR CHAPTER 4: PREVENTATIVE CARE Figure F 1. Number of enrollees that qualify for a given measure, by year and by group Figure F 2. Sample spatial distribution

PAGE 142

142 Figure F 3. Weighted, unadjusted measure performance over time, by group

PAGE 143

143 Figure F 4. Regression assumption diagnostics for AAP (full and final models)

PAGE 144

144 Figure F 5. Regression assumption diagnostics for BCS (full and final models)

PAGE 145

145 Figure F 6. Regression assumption diagnostics for CCS (full and final models)

PAGE 146

146 Figure F 7. Regression assumption diagnostics for COL (full and final models)

PAGE 147

147 APPENDIX G TABLES FOR CHAPTER 4: PREVENTATIVE CARE Table G 1. Percentage of the sample q ualifying for a given measure n [%] AAP only 49566 (31.6) BCS only 0 (0) CCS only 4 (0) COL only 9 (0) Only AAP and BCS 0 (0) Only AAP and CCS 49966 (31.9) Only AAP and COL 21952 (14) Only BCS and CCS 0 (0) Only BCS and COL 1 (0) Only CCS and COL 3 (0) Only AAP, BCS and CCS 78 (0) Only AAP, BCS and COL 103 (0.1) Only AAP, CCS and COL 4319 (2.8) Only BCS, CCS and COL 6 (0) All measures 30627 (19.6) Table G 2. AAP eligible enrollees: Characteristics of enrollees in transition and comparison counties during the baseline and post baseline periods STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Unique enrollees 29152 34701 64463 62888 Unique counties 28 28 209 209 Enrollees with complete data 28288 (97) 33992 (98) 61297 (95.1) 58975 (93.8) Counties with complete data, by year 28 (100) 28 (100) 209 (100) 208 (99.5) Counties with complete data in post and pre years 28 (100) 208 (99.5) Age [mean (SD)] 44.5 (13.1) 45.6 (12.7) 46 (12.9) 45.7 (13) Age categories [n (%)] 21 30 5701 (19.6) 5809 (16.7) 10570 (16.4) 10842 (17.2) 30 39 4464 (15.3) 5257 (15.1) 9395 (14.6) 9411 (15) 40 49 6491 (22.3) 7464 (21.5) 13794 (21.4) 12705 (20.2)

PAGE 148

148 Table G 2. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post 50 59 8413 (28.9) 11163 (32.2) 19804 (30.7) 20064 (31.9) 60 64 4083 (14) 5008 (14.4) 10900 (16.9) 9866 (15.7) Age not available 0 (0) 0 (0) 0 (0) 0 (0) Female [n (%)] 16569 (56.8) 20126 (58) 37984 (58.9) 35920 (57.1) Sex not available 0 (0) 0 (0) 0 (0) 0 (0) Race/ethnicity [n (%)] Non H ispanic black 4357 (14.9) 5462 (15.7) 12170 (18.9) 12578 (20) Hispanic 12327 (42.3) 14657 (42.2) 19721 (30.6) 17640 (28) Other 1772 (6.1) 2719 (7.8) 5118 (7.9) 5101 (8.1) Non H ispanic white 10696 (36.7) 11863 (34.2) 27454 (42.6) 27569 (43.8) Race/ethnicity not available 0 (0) 0 (0) 0 (0) 0 (0) Health status [n (%)] Healthy 5356 (18.6) 5740 (16.6) 10484 (16.8) 10031 (16.7) Significant acute 783 (2.7) 821 (2.4) 1731 (2.8) 1625 (2.7) Minor chronic 1678 (5.8) 1593 (4.6) 3653 (5.8) 3552 (5.9) Moderate chronic 7207 (25) 7702 (22.3) 15026 (24.1) 13982 (23.3) Major chronic 13846 (48) 18694 (54.1) 31578 (50.5) 30708 (51.3) Health status not available 282 (1) 151 (0.4) 1991 (3.1) 2990 (4.8) County metro classification [n (%)] Large metro (>1M) 22544 (77.3) 28199 (81.3) 1474 (2.3) 1623 (2.6) Medium metro (250K 1M) 4243 (14.6) 4209 (12.1) 20391 (31.6) 16473 (26.2) Small metro (<250k) 837 (2.9) 823 (2.4) 17903 (27.8) 19580 (31.1) Adjacent to metro 1528 (5.2) 1470 (4.2) 16763 (26) 17365 (27.6) Nonadjacent to metro (>20k) 0 (0) 0 (0) 3878 (6) 3935 (6.3) Nonadjacent to metro (<20k) 0 (0) 0 (0) 4054 (6.3) 3912 (6.2) Census tract poverty [mean (SD)] 21.7 (12.2) 21.8 (12.2) 25 (13.1) 24.3 (12.8) Census tract poverty categories [n (%)]

PAGE 149

149 Table G 2. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post 0.0% 4.9% 1670 (5.8) 1901 (5.5) 1361 (2.1) 1385 (2.2) 5.0% 9.9% 3206 (11.1) 3789 (11) 5319 (8.4) 5603 (9) 10.0% 19.9% 9573 (33.3) 11387 (33.1) 20011 (31.4) 20459 (32.8) Poverty area (20.0% 39.9%) 12017 (41.8) 14561 (42.3) 28070 (44.1) 26774 (43) Extreme poverty area (> 40.0%) 2300 (8) 2774 (8.1) 8887 (14) 8064 (12.9) Census tract not available 386 (1.3) 289 (0.8) 815 (1.3) 603 (1) Census tract unemployment level [n (%)] 0.0% 4.9% 4716 (16.4) 5526 (16.1) 13964 (21.9) 14196 (22.8) 5.0% 9.9% 14735 (51.2) 17654 (51.3) 29544 (46.4) 27917 (44.8) 10.0% 19.9% 8848 (30.8) 10710 (31.1) 18927 (29.7) 18861 (30.3) > 20.0% 467 (1.6) 522 (1.5) 1195 (1.9) 1294 (2.1) Census tract not available 386 (1.3) 289 (0.8) 833 (1.3) 620 (1) Census tract household income [mean (SD)] 42708 (19126.8) 42673.1 (18808.6) 36854.7 (13553.6) 37198.3 (13185.4) County Medicaid discharge density [mean (SD)] 1 (0.5) 1 (0.5) 1 (0.6) 1 (0.6) Census tract determined from address [n (%)] 24054 (82.5) 29496 (85) 48396 (75.1) 47846 (76.1) Facility residence [n (%)] 5722 (19.8) 5985 (17.4) 13118 (20.5) 12083 (19.4) Months enrolled in Medicaid [n (%)] 12 28538 (97.9) 32886 (94.8) 63223 (98.1) 61083 (97.1) 9 11 585 (2) 1815 (5.2) 1239 (1.9) 1799 (2.9) 5 8 0 (0) 0 (0) 0 (0) 0 (0) 2 5 0 (0) 0 (0) 0 (0) 0 (0) 1 0 (0) 0 (0) 0 (0) 0 (0)

PAGE 150

150 Table G 2. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post STAR+PLUS transition cohort [n (%)] A 0 (0) 0 (0) 0 (0) 0 (0) B 29152 (100) 34701 (100) 0 (0) 0 (0) C 0 (0) 0 (0) 0 (0) 0 (0) D 0 (0) 0 (0) 8481 (13.2) 9563 (15.2) E 0 (0) 0 (0) 22112 (34.3) 17920 (28.5) F 0 (0) 0 (0) 33870 (52.5) 35405 (56.3) Table G 3. BCS eligible enrollees: Characteristics of enrollees in transition and comparison counties during the baseline and post baseline periods STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Unique enrollees 5171 6819 6334 12423 Unique counties 28 28 190 199 Enrollees with complete data 5067 (98) 6708 (98.4) 6112 (96.5) 11793 (94.9) Counties with complete data, by year 28 (100) 28 (100) 189 (99.5) 185 (93) Counties with complete data in post and pre years 28 (100) 185 (93) Age [mean (SD)] 58 (3.7) 58 (3.7) 58.6 (3.6) 58.1 (3.7) Age categories [n (%)] 21 30 0 (0) 0 (0) 0 (0) 0 (0) 30 39 0 (0) 0 (0) 0 (0) 0 (0) 40 49 0 (0) 0 (0) 0 (0) 0 (0) 50 59 3174 (61.4) 4252 (62.4) 3536 (55.8) 7476 (60.2) 60 64 1997 (38.6) 2567 (37.6) 2798 (44.2) 4947 (39.8) Age not available 0 (0) 0 (0) 0 (0) 0 (0) Female [n (%)] 5165 (99.9) 6813 (99.9) 6333 (100) 12423 (100) Sex not available 0 (0) 0 (0) 0 (0) 0 (0) Race/ethnicity [n (%)] Non H ispanic black 594 (11.5) 902 (13.2) 800 (12.6) 2132 (17.2) Hispanic 2411 (46.6) 3047 (44.7) 2712 (42.8) 3786 (30.5) Other 418 (8.1) 646 (9.5) 762 (12) 1485 (12) Non H ispanic white 1748 (33.8) 2224 (32.6) 2060 (32.5) 5020 (40.4) Race/ethnicity not available 0 (0) 0 (0) 0 (0) 0 (0) Health status [n (%)] Healthy 381 (7.4) 403 (5.9) 262 (4.2) 840 (7)

PAGE 151

151 Table G 3. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Significant acute 81 (1.6) 65 (1) 46 (0.7) 160 (1.3) Minor chronic 279 (5.4) 257 (3.8) 198 (3.2) 554 (4.6) Moderate chronic 1002 (19.5) 1041 (15.3) 875 (14.1) 2112 (17.6) Major chronic 3386 (66) 5033 (74) 4810 (77.7) 8340 (69.5) Health status not available 42 (0.8) 20 (0.3) 143 (2.3) 417 (3.4) County metro classification [n (%)] Large metro (>1M) 3799 (73.5) 5372 (78.8) 75 (1.2) 332 (2.7) Medium metro (250K 1M) 828 (16) 917 (13.4) 2849 (45) 3518 (28.3) Small metro (<250k) 201 (3.9) 177 (2.6) 1382 (21.8) 3532 (28.4) Adjacent to metro 343 (6.6) 353 (5.2) 1451 (22.9) 3443 (27.7) Nonadjacent to metro (>20k) 0 (0) 0 (0) 279 (4.4) 757 (6.1) Nonadjacent to metro (<20k) 0 (0) 0 (0) 298 (4.7) 841 (6.8) Census tract poverty [mean (SD)] 22.8 (12) 22.4 (12) 28.9 (13.8) 25.4 (12.9) Census tract poverty categories [n (%)] 0.0% 4.9% 213 (4.1) 297 (4.4) 85 (1.4) 210 (1.7) 5.0% 9.9% 459 (8.9) 704 (10.4) 320 (5.1) 952 (7.7) 10.0% 19.9% 1701 (33.1) 2227 (32.9) 1498 (23.8) 3862 (31.3) Poverty area (20.0% 39.9%) 2312 (45) 2996 (44.2) 3055 (48.6) 5523 (44.8) Extreme poverty area (> 40.0%) 448 (8.7) 553 (8.2) 1334 (21.2) 1794 (14.5) Census tract not available 38 (0.7) 42 (0.6) 42 (0.7) 82 (0.7)

PAGE 152

152 Table G 3. Continued STAR+PLUS (Transition ) FFS+PCCM (Comparison) Pre Post Pre Post Census tract unemployment level [n (%)] 0.0% 4.9% 755 (14.7) 1007 (14.9) 1147 (18.2) 2722 (22.1) 5.0% 9.9% 2610 (50.8) 3458 (51) 2882 (45.8) 5345 (43.3) 10.0% 19.9% 1678 (32.7) 2194 (32.4) 2137 (34) 3997 (32.4) > 20.0% 90 (1.8) 118 (1.7) 122 (1.9) 273 (2.2) Census tract not available 38 (0.7) 42 (0.6) 46 (0.7) 86 (0.7) Census tract household income [mean (SD)] 40383.1 (16879.2) 41239.9 (17451.8) 33567.4 (12624.7) 36097.7 (12321.4) County Medicaid discharge density [mean (SD)] 1 (0.5) 1 (0.5) 1.1 (0.6) 1 (0.6) Census tract determined from address [n (%)] 4267 (82.5) 5704 (83.6) 4851 (76.6) 9328 (75.1) Facility residence [n (%)] 1136 (22.1) 1388 (20.5) 1465 (23.3) 2569 (20.8) Months enrolled in Medicaid [n (%)] 12 5143 (99.5) 6657 (97.6) 6303 (99.5) 12320 (99.2) 9 11 23 (0.4) 162 (2.4) 31 (0.5) 102 (0.8) 5 8 0 (0) 0 (0) 0 (0) 0 (0) 2 5 0 (0) 0 (0) 0 (0) 0 (0) 1 0 (0) 0 (0) 0 (0) 0 (0) STAR+PLUS transition cohort [n (%)] A 0 (0) 0 (0) 0 (0) 0 (0) B 5171 (100) 6819 (100) 0 (0) 0 (0) C 0 (0) 0 (0) 0 (0) 0 (0) D 0 (0) 0 (0) 466 (7.4) 1880 (15.1) E 0 (0) 0 (0) 3518 (55.5) 4060 (32.7) F 0 (0) 0 (0) 2350 (37.1) 6483 (52.2)

PAGE 153

153 Table G 4. CCS eligible enrollees: Characteristics of enrollees in transition and comparison counties during the baseline and post baseline periods STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Unique enrollees 15293 18850 36026 33279 Unique counties 28 28 209 207 Enrollees with complete data 14850 (97.1) 18457 (97.9) 34486 (95.7) 31301 (94.1) Counties with complete data, by year 28 (100) 28 (100) 208 (99.5) 207 (100) Counties with complete data in post and pre years 28 (100) 207 (100) Age [mean (SD)] 47.3 (11.4) 47.5 (11.4) 48.5 (11.3) 47.9 (11.5) Age categories [n (%)] 21 30 1552 (10.1) 1875 (9.9) 3102 (8.6) 3305 (9.9) 30 39 2430 (15.9) 2995 (15.9) 5162 (14.3) 5071 (15.2) 40 49 3829 (25) 4498 (23.9) 8415 (23.4) 7520 (22.6) 50 59 5030 (32.9) 6408 (34) 12372 (34.3) 11451 (34.4) 60 64 2452 (16) 3074 (16.3) 6975 (19.4) 5932 (17.8) Age not available 0 (0) 0 (0) 0 (0) 0 (0) Female [n (%)] 15273 (99.9) 18832 (99.9) 36025 (100) 33279 (100) Sex not available 0 (0) 0 (0) 0 (0) 0 (0) Race/ethnicity [n (%)] Non H ispanic black 2124 (13.9) 2797 (14.8) 6362 (17.7) 6144 (18.5) Hispanic 6885 (45) 8349 (44.3) 11277 (31.3) 9543 (28.7) Other 985 (6.4) 1556 (8.3) 3341 (9.3) 2950 (8.9) Non H ispanic white 5299 (34.6) 6148 (32.6) 15046 (41.8) 14642 (44) Race/ethnicity not available 0 (0) 0 (0) 0 (0) 0 (0) Health status [n (%)] Healthy 1975 (13.1) 2215 (11.8) 4155 (11.8) 3827 (12) Significant acute 371 (2.5) 402 (2.1) 871 (2.5) 759 (2.4) Minor chronic 954 (6.3) 938 (5) 2138 (6.1) 1929 (6.1) Moderate chronic 3568 (23.6) 3837 (20.4) 7968 (22.7) 7018 (22.1) Major chronic 8265 (54.6) 11373 (60.6) 20025 (57) 18268 (57.4) Health status not available 160 (1) 85 (0.5) 869 (2.4) 1478 (4.4)

PAGE 154

154 Table G 4. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post County metro classification [n (%)] Large metro (>1M) 11609 (75.9) 15324 (81.3) 818 (2.3) 889 (2.7) Medium metro (250K 1M) 2308 (15.1) 2241 (11.9) 11590 (32.2) 8929 (26.8) Small metro (<250k) 486 (3.2) 459 (2.4) 9827 (27.3) 10125 (30.4) Adjacent to metro 890 (5.8) 826 (4.4) 9416 (26.1) 9146 (27.5) Nonadjacent to metro (>20k) 0 (0) 0 (0) 2144 (6) 2113 (6.3) Nonadjacent to metro (<20k) 0 (0) 0 (0) 2231 (6.2) 2077 (6.2) Census tract poverty [mean (SD)] 22.2 (12) 22 (12.1) 25.1 (13.1) 24.4 (12.8) Census tract poverty categories [n (%)] 0.0% 4.9% 720 (4.8) 931 (5) 710 (2) 684 (2.1) 5.0% 9.9% 1555 (10.3) 2006 (10.7) 2864 (8.1) 2862 (8.7) 10.0% 19.9% 5063 (33.5) 6207 (33.2) 11224 (31.6) 10936 (33.2) Poverty area (20.0% 39.9%) 6539 (43.3) 8056 (43.1) 15750 (44.3) 14172 (43) Extreme poverty area (> 40.0%) 1237 (8.2) 1491 (8) 5008 (14.1) 4293 (13) Census tract not available 179 (1.2) 159 (0.8) 470 (1.3) 332 (1) Census tract unemployment level [n (%)] 0.0% 4.9% 2375 (15.7) 2942 (15.7) 7703 (21.7) 7392 (22.4) 5.0% 9.9% 7649 (50.6) 9551 (51.1) 16477 (46.4) 14706 (44.7) 10.0% 19.9% 4836 (32) 5903 (31.6) 10700 (30.1) 10151 (30.8) > 20.0% 254 (1.7) 295 (1.6) 664 (1.9) 687 (2.1)

PAGE 155

155 Table G 4. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Census tract not available 179 (1.2) 159 (0.8) 482 (1.3) 343 (1) Census tract household income [mean (SD)] 41557.2 (17790.6) 42128.8 (18117.1) 36659.1 (13231.8) 37064.1 (12846.7) County Medicaid discharge density [mean (SD)] 1 (0.5) 1 (0.5) 1 (0.6) 1 (0.6) Census tract determined from address [n (%)] 12659 (82.8) 16005 (84.9) 27155 (75.4) 25322 (76.1) Facility residence [n (%)] 3072 (20.3) 3496 (18.7) 7401 (20.7) 6515 (19.7) Months enrolled in Medicaid [n (%)] 12 14986 (98) 17916 (95) 35317 (98) 32391 (97.3) 9 11 290 (1.9) 934 (5) 709 (2) 884 (2.7) 5 8 0 (0) 0 (0) 0 (0) 0 (0) 2 5 0 (0) 0 (0) 0 (0) 0 (0) 1 0 (0) 0 (0) 0 (0) 0 (0) STAR+PLUS transition cohort [n (%)] A 0 (0) 0 (0) 0 (0) 0 (0) B 15293 (100) 18850 (100) 0 (0) 0 (0) C 0 (0) 0 (0) 0 (0) 0 (0) D 0 (0) 0 (0) 4742 (13.2) 5139 (15.4) E 0 (0) 0 (0) 12366 (34.3) 9477 (28.5) F 0 (0) 0 (0) 18918 (52.5) 18663 (56.1) Table G 5. COL eligible enrollees: Characteristics of enrollees in transition and comparison counties during the baseline and post baseline periods STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Unique enrollees 9640 12548 13408 23054 Unique counties 28 28 205 204 Enrollees with complete data 9437 (97.9) 12329 (98.3) 12934 (96.5) 21995 (95.4) Counties with complete data, by year 28 (100) 28 (100) 205 (100) 202 (99) Counties with complete data in post and pre years 28 (100) 202 (99)

PAGE 156

156 Table G 5. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Age [mean (SD)] 57.5 (4) 57.4 (3.9) 57.7 (3.9) 57.6 (3.9) Age categories [n (%)] 21 30 0 (0) 0 (0) 0 (0) 0 (0) 30 39 0 (0) 0 (0) 0 (0) 0 (0) 40 49 0 (0) 0 (0) 0 (0) 0 (0) 50 59 6200 (64.3) 8309 (66.2) 8505 (63.4) 14763 (64) 60 64 3440 (35.7) 4239 (33.8) 4903 (36.6) 8291 (36) Age not available 0 (0) 0 (0) 0 (0) 0 (0) Female [n (%)] 6012 (62.4) 7747 (61.7) 8983 (67) 14041 (60.9) Sex not available 0 (0) 0 (0) 0 (0) 0 (0) Race/ethnicity [n (%)] Non H ispanic black 1228 (12.7) 1818 (14.5) 2055 (15.3) 4505 (19.5) Hispanic 4163 (43.2) 5231 (41.7) 5092 (38) 6356 (27.6) Other 713 (7.4) 1036 (8.3) 1450 (10.8) 2224 (9.6) Non H ispanic white 3536 (36.7) 4463 (35.6) 4811 (35.9) 9969 (43.2) Race/ethnicity not available 0 (0) 0 (0) 0 (0) 0 (0) Health status [n (%)] Healthy 1001 (10.5) 1198 (9.6) 907 (6.9) 2307 (10.3) Significant acute 168 (1.8) 162 (1.3) 158 (1.2) 381 (1.7) Minor chronic 513 (5.4) 481 (3.8) 516 (3.9) 1092 (4.9) Moderate chronic 1979 (20.7) 2135 (17.1) 2159 (16.4) 4204 (18.9) Major chronic 5911 (61.8) 8533 (68.2) 9392 (71.5) 14308 (64.2) Health status not available 68 (0.7) 39 (0.3) 276 (2.1) 762 (3.3) County metro classification [n (%)] Large metro (>1M) 7151 (74.2) 9923 (79.1) 212 (1.6) 644 (2.8) Medium metro (250K 1M) 1577 (16.4) 1721 (13.7) 5448 (40.6) 6047 (26.2) Small metro (<250k) 315 (3.3) 287 (2.3) 3135 (23.4) 6704 (29.1)

PAGE 157

157 Table G 5. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Adjacent to metro 597 (6.2) 617 (4.9) 3237 (24.1) 6670 (28.9) Nonadjacent to metro (>20k) 0 (0) 0 (0) 672 (5) 1440 (6.2) Nonadjacent to metro (<20k) 0 (0) 0 (0) 704 (5.3) 1549 (6.7) Census tract poverty [mean (SD)] 22.8 (12.1) 22.5 (12) 27.9 (13.7) 24.9 (12.7) Census tract poverty categories [n (%)] 0.0% 4.9% 392 (4.1) 535 (4.3) 189 (1.4) 388 (1.7) 5.0% 9.9% 874 (9.1) 1247 (10) 774 (5.8) 1825 (8) 10.0% 19.9% 3209 (33.6) 4122 (33.1) 3441 (25.9) 7316 (31.9) Poverty area (20.0% 39.9%) 4210 (44.1) 5490 (44.1) 6332 (47.6) 10241 (44.7) Extreme poverty area (> 40.0%) 872 (9.1) 1069 (8.6) 2560 (19.3) 3130 (13.7) Census tract not available 83 (0.9) 85 (0.7) 112 (0.8) 154 (0.7) Census tract unemployment level [n (%)] 0.0% 4.9% 1370 (14.3) 1828 (14.7) 2544 (19.1) 5046 (22) 5.0% 9.9% 4860 (50.9) 6351 (51) 6055 (45.6) 10049 (43.9) 10.0% 19.9% 3145 (32.9) 4078 (32.7) 4428 (33.3) 7295 (31.9) > 20.0% 182 (1.9) 206 (1.7) 264 (2) 503 (2.2) Census tract not available 83 (0.9) 85 (0.7) 117 (0.9) 161 (0.7) Census tract household income [mean (SD)] 40511.7 (17140.3) 41202.1 (17600.9) 34296.6 (12848.3) 36290 (12390.9) County Medicaid discharge density [mean (SD)] 1 (0.5) 1 (0.5) 1.1 (0.6) 1 (0.6)

PAGE 158

158 Table G 5. Continued STAR+PLUS (Transition) FFS+PCCM (Comparison) Pre Post Pre Post Census tract determined from address [n (%)] 7890 (81.8) 10540 (84) 10246 (76.4) 17317 (75.1) Facility residence [n (%)] 2147 (22.5) 2419 (19.5) 3116 (23.5) 4764 (20.8) Months enrolled in Medicaid [n (%)] 12 9591 (99.5) 12211 (97.3) 13344 (99.5) 22861 (99.2) 9 11 42 (0.4) 337 (2.7) 64 (0.5) 191 (0.8) 5 8 0 (0) 0 (0) 0 (0) 0 (0) 2 5 0 (0) 0 (0) 0 (0) 0 (0) 1 0 (0) 0 (0) 0 (0) 0 (0) STAR+PLUS transition cohort [n (%)] A 0 (0) 0 (0) 0 (0) 0 (0) B 9640 (100) 12548 (100) 0 (0) 0 (0) C 0 (0) 0 (0) 0 (0) 0 (0) D 0 (0) 0 (0) 1273 (9.5) 3559 (15.4) E 0 (0) 0 (0) 6605 (49.3) 6907 (30) F 0 (0) 0 (0) 5530 (41.2) 12588 (54.6) Table G 6 Summary of sample size, by group and by measure STAR+PLUS (Transition) FFS+PCCM (Comparison) Measure Type Min Q1 Median Q3 Max Min Q1 Median Q3 Max AAP 55 221.00 406.00 830.00 14260 1 32.00 106.00 292.00 7553 BCS 6 46.00 88.00 186.00 2907 1 9.00 21.00 56.00 1993 CCS 26 120.00 216.00 433.00 7884 1 18.00 57.00 152.00 4179 COL 12 81.00 162.00 333.00 5171 1 14.00 43.00 111.00 3180

PAGE 159

159 Table G 7 Backward step wise regression results for AAP V ariable R 2 total R 2 F value p value A ction Full Model 0.5741 NA NA NA Medicaid Discharge Density 0.5725 0.0016 0.8 0.3737 Removed Metro Classification 2 0.5698 0.0027 1.4 0.2411 Removed Metro Classification 0.5698 0.0254 12.9 0.0004 Kept Unemployment 0.5696 0.0002 0.1 0.7320 Removed Poverty 0.5644 0.0052 2.6 0.1054 Removed County Income 0.5635 0.0009 0.5 0.4935 Removed Median Household Income 0.5621 0.0014 0.7 0.3993 Removed Group Facility Proxy 0.5620 0.0001 0.1 0.8066 Removed Health Status 0.5620 0.0499 25.5 0.0000 Kept Age 3 0.5591 0.0028 1.4 0.2316 Removed Age 2 0.5589 0.0002 0.1 0.7502 Removed Age 0.5569 0.0020 1.0 0.3079 Removed Race /Ethnicity 0.5527 0.0042 0.7 0.5435 Removed Gender 0.5525 0.0002 0.1 0.7408 Removed Baseline 0.5525 0.2420 124.9 0.0000 Kept Table G 8 Backward step wise regression results for BCS V ariable R 2 total R 2 F value p value A ction Full Model 0.2766 NA NA NA Medicaid Discharge Density 0.2766 0.0255 6.9 0.0093 Kept Metro Classification 2 0.2765 0.0001 0.0 0.8989 Removed Metro Classification 0.2758 0.0008 0.2 0.6467 Removed Unemployment 0.2619 0.0139 3.8 0.0526 Removed Poverty 0.2488 0.0131 3.5 0.0618 Removed County Income 0.2456 0.0032 0.9 0.3566 Removed Median Household Income 0.2438 0.0018 0.5 0.4941 Removed Group Facility Proxy 0.2412 0.0026 0.7 0.4058 Removed Health Status 2 0.2305 0.0107 2.9 0.0923 Removed Health Status 0.2305 0.0390 10.3 0.0015 Kept Age 0.2282 0.0024 0.6 0.4308 Removed Race /Ethnicity 0.2282 0.0538 4.8 0.0031 Kept Baseline 0.2171 0.0111 2.9 0.0880 Removed

PAGE 160

160 Table G 9 Backward step wise regression results for CCS V ariable R 2 total R 2 F value p value A ction Full Model 0.3874 NA NA NA Medicaid Discharge Density 0.3870 0.0004 0.1 0.7130 Removed Metro Classification 2 0.3869 0.0000 0.0 0.9104 Removed Metro Classification 0.3869 0.0152 5.4 0.0209 Kept Unemployment 0.3861 0.0008 0.3 0.5960 Removed Poverty 0.3856 0.0006 0.2 0.6532 Removed County Income 0.3844 0.0012 0.4 0.5167 Removed Median Household Income 0.3788 0.0056 2.0 0.1590 Removed Group Facility Proxy 0.3788 0.0136 4.9 0.0284 Kept Health Status 2 0.3787 0.0002 0.1 0.8081 Removed Health Status 0.3717 0.0070 2.5 0.1139 Removed Age 3 0.3665 0.0051 1.8 0.1782 Removed Age 2 0.3665 0.0000 0.0 0.9612 Removed Age 0.3665 0.0404 14.4 0.0002 Kept Race /Ethnicity 0.3665 0.0500 5.9 0.0006 Kept Baseline 0.3665 0.0994 35.4 0.0000 Kept Table G 10 Backward step wise regression results for COL V ariable R 2 total R 2 F value p value A ction Full Model 0.2171 NA NA NA Medicaid Discharge Density 0.2072 0.0099 2.7 0.1034 Removed Metro Classification 2 0.2059 0.0013 0.4 0.5500 Rem oved Metro Classification 0.2027 0.0031 0.8 0.3596 Removed Unemployment 0.1997 0.0030 0.8 0.3666 Removed Poverty 0.1922 0.0075 2.0 0.1568 Removed County Income 0.1898 0.0025 0.7 0.4175 Removed Median Household Income 0.1893 0.0005 0.1 0.7228 Removed Group Facility Proxy 0.1811 0.0082 2.2 0.1372 Removed Health Status 2 0.1753 0.0058 1.6 0.2144 Removed Health Status 0.1753 0.0260 7.0 0.0089 Kept Race /Ethnicity 0.1719 0.0033 0.9 0.3452 Removed Age 0.1719 0.0341 3.0 0.0297 Kept Gender 0.1719 0.0000 0.0 0.9848 Removed Baseline 0.1719 0.0910 24.5 0.0000 Kept

PAGE 161

1 61 Table G 11 Full model coefficients for AAP Variable Value (std error) 95% C.I. (Intercept) 0.7142 (0.0488) (0.6181 0.8103) Male 0.0039 (0.0390) ( 0.0808 0.0729) Hispanic 0.0229 (0.0170) ( 0.0106 0.0563) Other 0.0467 (0.0685) ( 0.0883 0.1818) White 0.0134 (0.0183) ( 0.0226 0.0494) Group Facility Proxy 0.0244 (0.0378) ( 0.0501 0.0990) Age 0.0105 (0.0076) ( 0.0044 0.0255) Age 2 0.0001 (0.0003) ( 0.0006 0.0007) Age 3 0.0000 (0.0000) ( 0.0001 0.0000) Health Status 0.9681 (0.2028) (0.5684 1.3678) County Income 0.0030 (0.0096) ( 0.0159 0.0220) Median Household Income 0.0183 (0.0140) ( 0.0458 0.0092) Poverty 0.0021 (0.0015) ( 0.0050 0.0007) Unemployment 0.0012 (0.0024) ( 0.0035 0.0058) Metro Classification 0.0145 (0.0041) ( 0.0226 0.0064) Metro Classification 2 0.0023 (0.0019) ( 0.0060 0.0015) Medicaid Discharge Density 0.0077 (0.0087) ( 0.0094 0.0249) Baseline 0.5643 (0.0527) (0.4605 0.6682) Months Enrolled in STAR+PLUS 0.0454 (0.0177) (0.0104 0.0803) Table G 12 Full model coefficients for BCS Variable Value (std error) 95% C.I. (Intercept) 0.3120 (0.0703) (0.1733 0.4507) Hispanic 0.0023 (0.0247) ( 0.0510 0.0465) Other 0.0883 (0.0607) ( 0.2079 0.0314) White 0.0518 (0.0281) ( 0.1071 0.0035) Group Facility Proxy 0.0389 (0.0597) ( 0.1567 0.0789) Age 0.0038 (0.0117) ( 0.0192 0.0269) Health Status 0.9725 (0.4081) (0.1676 1.7773) Health Status 2 0.3590 (0.2760) ( 0.1854 0.9033) County Income 0.0105 (0.0143) ( 0.0388 0.0177) Median Household Income 0.0422 (0.0222) ( 0.0016 0.0860) Poverty 0.0035 (0.0022) ( 0.0009 0.0079) Unemployment 0.0067 (0.0038) ( 0.0008 0.0141) Metro Classification 0.0029 (0.0065) ( 0.0157 0.00 98) Metro Classification 2 0.0004 (0.0031) ( 0.0057 0.0065) Medicaid Discharge Density 0.0343 (0.0130) (0.0085 0.0600) Baseline 0.1523 (0.1327) ( 0.1095 0.4141) Months Enrolled in STAR+PLUS 0.0217 (0.0259) ( 0.0728 0.0295)

PAGE 162

162 Table G 13 Full model coefficients for CCS Variable Value (std error) 95% C.I. (Intercept) 0.3918 (0.0693) (0.2551 0.5284) Hispanic 0.0013 (0.0149) ( 0.0281 0.0306) Other 0.0374 (0.0558) ( 0.0726 0.1474) White 0.0257 (0.0167) ( 0.0586 0.0072) Group Facility Proxy 0.0727 (0.0394) ( 0.1505 0.0050) Age 0.0196 (0.0080) ( 0.0355 0.0038) Age 2 0.0000 (0.0003) ( 0.0005 0.0006) Age 3 0.0000 (0.0000) ( 0.0000 0.0001) Health Status 0.2814 (0.3214) ( 0.3522 0.9149) Health Status 2 0.0293 (0.2311) ( 0.4262 0.4848) County Income 0.0059 (0.0081) ( 0.0219 0.0101) Median Household Income 0.0053 (0.0134) ( 0.0317 0.0212) Poverty 0.0006 (0.0014) ( 0.0021 0.0033) Unemployment 0.0012 (0.0023) ( 0.0057 0.0034) Metro Classification 0.0087 (0.0038) ( 0.0161 0.0013) Metro Classification 2 0.0002 (0.0018) ( 0.0038 0.0033) Medicaid Discharge Density 0.0028 (0.0075) ( 0.0121 0.0176) Baseline 0.3700 (0.0684) (0.2351 0.5048) Months Enrolled in STAR+PLUS 0.0108 (0.0145) ( 0.0394 0.0178) Table G 14 Full model coefficients for COL Variable Value (std error) 95% C.I. (Intercept) 0.2099 (0.0466) (0.1181 0.3017) Male 0.0110 (0.0340) ( 0.0560 0.0779) Hispanic 0.0027 (0.0178) ( 0.0377 0.0323) Other 0.0669 (0.0559) ( 0.0434 0.1771) White 0.0464 (0.0183) ( 0.0825 0.0104) Group Facility Proxy 0.0421 (0.0447) ( 0.1303 0.0461) Age 0.0056 (0.0096) ( 0.0246 0.0134) Health Status 1.0484 (0.3945) (0.2708 1.8260) Health Status 2 0.3513 (0.2201) ( 0.0825 0.7851) County Income 0.0124 (0.0096) ( 0.0314 0.0065) Median Household Income 0.0181 (0.0154) ( 0.0486 0.0123) Poverty 0.0029 (0.0016) ( 0.0060 0.0002) Unemployment 0.0030 (0.0026) ( 0.0081 0.0020) Metro Classification 0.0033 (0.0045) ( 0.0122 0.0055) Metro Classificatio n 2 0.0012 (0.0021) ( 0.0030 0.0054) Medicaid Discharge Density 0.0147 (0.0090) ( 0.0030 0.0324) Baseline 0.2899 (0.0580) (0.1756 0.4042) Months Enrolled in STAR+PLUS 0.0117 (0.0178) ( 0.0467 0.0234)

PAGE 163

163 Table G 15 Final model coefficients for AAP Variable Value (std error) 95% C.I. (Intercept) 0.7457 (0.0046) (0.7366 0.7547) Health Status 1.0864 (0.1806) (0.7306 1.4421) Metro Classification 0.0110 (0.0035) ( 0.0179 0.0040) Baseline 0.5489 (0.0491) (0.4522 0.6457) Months Enrolled in STAR+PLUS 0.0375 (0.0146) (0.0087 0.0663) Table G 16 Final model coefficients for BCS Variable Value (std error) 95% C.I. (Intercept) 0.3788 (0.0163) (0.3468 0.4109) Hispanic 0.0048 (0.0210) ( 0.0461 0.0365) Other 0.0613 (0.0565) ( 0.1726 0.0500) White 0.0620 (0.0256) ( 0.1125 0.0115) Health Status 0.8809 (0.2482) (0.3918 1.3700) Medicaid Discharge Density 0.0388 (0.0115) (0.0161 0.0616) Months Enrolled in STAR+PLUS 0.0067 (0.0172) ( 0.0272 0.0407) Table G 17 Final model coefficients for CCS Variable Value (std error) 95% C.I. (Intercept) 0.3957 (0.0091) (0.3777 0.4137) Hispanic 0.0115 (0.0131) ( 0.0142 0.0372) Other 0.0141 (0.0520) ( 0.0883 0.1165) White 0.0302 (0.0149) ( 0.0595 0.0009) Group Facility Proxy 0.0875 (0.0354) ( 0.1573 0.0177) Age 0.0094 (0.0025) ( 0.0143 0.0045) Metro Classification 0.0083 (0.0033) ( 0.0148 0.0018) Baseline 0.3841 (0.0645) (0.2570 0.5113) Months Enrolled in STAR+PLUS 0.0195 (0.0117) ( 0.0426 0.0036) Table G 18 Final model coefficients for COL Variable Value (std error) 95% C.I. (Intercept) 0.2925 (0.0060) (0.2808 0.3042) Hispanic 0.0143 (0.0139) ( 0.0416 0.0130) Other 0.0550 (0.0514) ( 0.0462 0.1563) White 0.0416 (0.0164) ( 0.0738 0.0093) Health Status 0.5242 (0.1942) (0.1415 0.9069) Baseline 0.2761 (0.0558) (0.1662 0.3860) Months Enrolled in STAR+PLUS 0.0018 (0.0117) ( 0.0248 0.0213)

PAGE 164

164 LIST OF REFERENCES 1. Squires D, Anderson C. U.S. Health Care from a Global Perspective: Spending, Use of Services, Prices, and Health in 13 Countries. The Commonwealth Fund; 2015 2. Bradley EH, Books24x I, Fineberg H, et al. The American health care paradox why spending more is getting us less. New York: PublicAffairs; 2013 3. National Research Council, Institute of Medicine. U.S. Health in International Perspective: Shorter Lives, Poorer Health. Washington, DC: The National Academies Press; 2013 4. Berwi ck DM, Hackbarth AD. Eliminating waste in US health care. JAMA. 2012;307:1513 1516 5. Institute of Medicine. Crossing the Quality Chasm: A New Health System for the 21st Century. Washington, DC: The National Academies Press; 2001 6. Berwick DM, Nolan TW, Whitt ington J. The Triple Aim: Care, Health, And Cost. Health Affairs. 2008;27:759 769 7. Shortell SM, Gillies R, Wu F. United States innovations in healthcare delivery. Public Health Reviews. 2010;32:190 212 8. Texas Medicaid and CHIP in Perspective: Ninth Edition Texas Health and Human Service Commission; 2013 9. Report to Congress on Medicaid and CHIP. Washington, DC: Medicaid and CHIP Payment and Access Commission; 2012 10. Medicaid Managed Care for People with Disabilities: Policy and Implementation Consideration s for State and Federal Policymakers. National Council on Disability; 2013 11. Kronick RG, Bella M, Gilmer TP. The Faces of Medicaid III: Refining the Portrait of People with Multiple Chronic Conditions. Center for Health Care Strategies, Inc; 2009 12. Wallace E, Salisbury C, Guthrie B, et al. Managing patients with multimorbidity in primary care. BMJ. 2015;350 13. Boyd C, Leff B, Weiss C, et al. Clarifying Multimorbidity Patterns to Improve Targeting and Delivery of Clinical Services for Medicaid Populations. C enter for Health Care Strategies, Inc; 2010

PAGE 165

165 14. Melek S, Norris D. Chronic conditions and comorbid psychological disorders. 2008:S21 S22 15. Mitchell AJ, Malone D, Doebbeling CC. Quality of medical care for people with and without comorbid mental illness and substance misuse: systematic review of comparative studies. Br J Psychiatry. 2009;194:491 499 16. Druss BG, Rosenheck RA, Desai MM, et al. Quality of preventive medical care for patients with mental disorders. Med Care. 2002;40:129 136 17. Wells KB, Stewart A, H ays RD, et al. The functioning and well being of depressed patients. Results from the Medical Outcomes Study. JAMA. 1989;262:914 919 18. Druss BG, Zhao L, Cummings JR, et al. Mental comorbidity and quality of diabetes care under Medicaid: a 50 state analysis. Med Care. 2012;50:428 433 19. Dickerson F, Brown CH, Fang L, et al. Quality of life in individuals with serious mental illness and type 2 diabetes. Psychosomatics. 2008;49:109 114 20. Cook EL, Harman JS. A comparison of health related quality of life for indivi duals with mental health disorders and common chronic medical conditions. Public Health Rep. 2008;123:45 51 21. Sareen J, Jacobi F, Cox BJ, et al. Disability and poor quality of life associated with comorbid anxiety disorders and physical conditions. Arch In tern Med. 2006;166:2109 2116 22. Druss BG, Schlesinger M, Allen HM. Depressive symptoms, satisfaction with health care, and 2 year work outcomes in an employed population. Am J Psychiatry. 2001;158:731 734 23. Eaton WW, Martins SS, Nestadt G, et al. The burden of mental disorders. Epidemiol Rev. 2008;30:1 14 24. Colton CW, Manderscheid RW. Congruencies in increased mortality rates, years of potential life lost, and causes of death among public mental health client s in eight states. Prev Chronic Dis. 2006;3:A42 25. Shenkman E, Muller K, Vogel B, et al. The wellness incentives and navigation project: design and methods. BMC Health Services Research. 2015;15:1 13 26. Yost KJ, DeWalt DA, Lindquist LA, et al. The association between health literacy and indicators of cognitive impairment in a diverse sample of primary care patients. Patient Education and Counseling. 2013;93:319 326

PAGE 166

166 27. Berkman ND, Sheridan SL, Donahue KE, et al. Low Health Literacy and Health Outcomes: An Updated Systematic Review. Annals of Internal Medicine. 2011;155:97 107 28. America's Health Literacy: Why We Need Accessible Health Information. An Issue Brief From the U.S. Department of Health and Human Services. 2008 29. Social Security Administration. SSI Annual Statistical Report, 2014. 2015 30. Social Security Administration. SSI Annual Statistical Report, 2002. 2003 31. Commission on Social Determinants of Health. Achieving Health Equity: from root causes to fair outcomes: Interim Statement. Geneva, Switzerland2007 32. Phelan JC, Link BG, Tehranifar P. Social Conditions as Fundamental Causes of Health Inequalities: Theory, Evidence, and Policy Implications. Journal of Health and Social Behavior. 2010;51:S28 S40 33. Farmer P. An Anthropology of Structural Violence. Current Anthropology. 2004;45:305 325 34. Institute of Medicine, National Academies of Sciences E, and Medicine,. Acco unting for Social Risk Factors in Medicare Payment: Identifying Social Risk Factors. Washington, DC: The National Academies Press; 2016 35. Andersen R, Newman JF. Societal and Individual Determinants of Medical Care Utilization in the United States. Milbank Q uarterly. 2005;83 36. Gettings RM, American Association on I, Developmental D, et al. Forging a Federal state Partnership: A History of Federal Developmental Disabilities Policy. American Association on Intellectual and Developmental Disabilities; 2011 37. Schne ider A, Elias R, Garfield R, et al. The Medicaid Resource Book. 2002 38. Klees BS, Wolfe CJ, Curtis CA. Brief Summaries of Medicare & Medicaid. Centers for Medicare and Medicaid Services; 2015 39. Kongstvedt PR. Essentials of managed health care, sixth edition. Burlington, Mass: Jones & Bartlett Learning; 2013 40. Paradise J. Medicaid Moving Forward. Kaiser Family Foundation; 2015 41. Cunningham P, May J. Medicaid Patients Increasingly Concentrated Among Physicians. Washington, DC: Center for Health Systems Change; 2006

PAGE 167

167 42. Coughlin T, Long S, Clemans Cope L, et al. What Difference Does Medicaid Make?: Kaiser Family Foundation; 2013 43. Vestal C. How Severe is the Shortage of Substance Abuse Specialists? : The Pew Charitable Trusts; 2015 44. Fingar KR, Smith MW, Davies S, et al. Medicaid Dental Coverage Alone May Not Lower Rates Of Dental Emergency Department Visits. Health Affairs. 2015;34:1349 1357 45. Rhodes KV, Kenney GM, Friedman AB, et al. Primary care access for new patients on the eve of health car e reform. JAMA Internal Medicine. 2014;174:861 869 46. Zuckerman S, Goin D. How Much Will Medicaid Physician Fees for Primary Care Rise in 2013? Evidence from a 2012 Survey of Medicaid Physician Fees. Urban Institute and Kaiser Commission on Medicaid and the Uninsured; 2012 47. Polsky D, Richards M, Basseyn S, et al. Appointment Availability after Increases in Medicaid Payments for Primary Care. New England Journal of Medicine. 2015;372:537 545 48. Chandra A, Holmes J, Skinner J. Is This Time Different? The Slowdown in Healthcare Spending. National Bureau of Economic Research Working Paper Series. 2013;No. 19700 49. Hasan O, Orav EJ, Hicks LS. Insurance status and hospital care for myocardial infarction, stroke, and pneumonia. Journal of Hospital Medicine. 2010;5:452 45 9 50. McDavid K, Tucker TC, Sloggett A, et al. Cancer survival in kentucky and health insurance coverage. Archives of Internal Medicine. 2003;163:2135 2144 51. Reaves EL, Musumeci M. Medicaid and Long Term Services and Supports: A Primer. The Kaiser Commission o n Medicaid and the Uninsured; 2015 52. National Council on Disability. Medicaid Managed Care for People with Disabilitiess: Policy and Implementation Considerations for State and Federal Policymakers. 2013 53. Gifford K, Smith VK, Snipes D, et al. A Profile of Medicaid Managed Care Programs in 2010: Findings from a 50 State Survey. Kaiser Commission on Medicaid and the Uninsured; 2011

PAGE 168

168 54. Rom MC. State Health and Welfare Programs. In: Gray V, Hanson RL, Kousser T, eds. Politics in the American States: A Comparative Analysis. Thousand Oaks, California: CQ Press; 2013 55. Skinner J. Chapter Two Causes and Consequences of Regional Variations in Health Care1. In: Mark V. Pauly TGM, Pedro PB, eds. Handbook of Health Economics: Elsevier; 2011:45 93 56. Goodell S, D russ BG, Walker ER. Mental disorders and medical comorbidity. Robert Wood Johnson Foundation. 2011 57. Sparer M. Medicaid Managed Care: Costs, Access and Quality of Care. Robert Wood Johnson Foundation; 2012 58. Ng T, Harrington C, Musumeci M, et al. Medicaid Ho me and Community Based Services Programs: 2010 Data Update. The Kaiser Commission on Medicaid and the Uninsured; 2014 59. Burns ME. Medicaid managed care and health care access for adult beneficiaries with disabilities. Health Serv Res. 2009;44:1521 1541 60. Rep ort to Congress on Medicaid and CHIP. Washington, DC: Medicaid and CHIP Payment and Access Commission; 2011 61. Duckett MJ, Guy MR. Home and Community Based Services Waivers. Health Care Financing Review. 2000;22:123 125 62. Coughlin TA, Long SK, Graves JA. Does managed care improve access to care for Medicaid beneficiaries with disabilities? A national study. Inquiry. 2008;45:395 407 63. Saucier P, Kasten J, Burwell B, et al. The Growth of Managed Long Term Services and Supports (MLTSS) Programs: A 2012 Update. Cen ters for Medicare and Medicaid Services and Truven Health Analytics; 2012 64. Musumeci M. Key Themes in Capitated Medicaid Managed Long Term Services and Supports Waivers. The Kaiser Commission on Medicaid and the Uninsured; 2014 65. Davies S, Schmidt E, Shultz E, et al. Home and Community Based Services Quality Indicators: A Review of Literature Related to HCBS Populations. Rockville, MD: Agency for Healthcare Research and Quality (US); 2010 66. Burns ME. Medicaid managed care and cost containment in the adult disabled population. Med Care. 2009;47:1069 1076

PAGE 169

169 67. Harman JS, Scholle SH, Ng JH, et al. Association of Health Plans' Healthcare Effectiveness Data and Information Set (HEDIS) performance with outcomes of enrollees with diabetes. Med Care. 2010;48:217 223 68. Oversight of Quality of Care in Medicaid Home and Community Based Services Waiver Programs: OEI 02 08 00170. Rockville, MD: Department of Health and Human Services, Office of Inspector General; 2012 69. Institute of Medicine (US) Roundtable on Value & Science Driven Health Care. Redesigning the Clinical Effectiveness Research Paradigm: Innovation and Practice Based Approaches: Workshop Summary. Washington (DC): National Academies Press (US); 2010 70. Kronick RG, Bella M, Gilmer TP, et al. The Faces of Medicaid II : Recognizing the Care Needs of People with Multiple Chronic Conditions. Center for Health Care Strategies, Inc; 2007 71. Total Medicaid Enrollment in Managed Long Term Services and Supports (MLTSS): Kaiser Family Foundation; 72. Quast T, Sappington DEM, Shenkma n E. Does the quality of care in Medicaid MCOs vary with the form of physician compensation? Health Econ. 2008;17:545 550 73. Wysocki A, Butler M, Kane RL, et al. Long term care for older adults: a review of home and community based services versus institutio nal care. Rockville, MD: Agency for Healthcare Research and Quality (US); 2012. 74. Bubolz T, Emerson C, Skinner J. State spending on dual eligibles under age 65 shows variations, evidence of cost shifting from Medicaid to Medicare. Health Aff (Millwood). 2012;31:939 947. 75. Walsh EG, Wiener JM, Haber S, et al. Potentially avoidable hospitalizations of dually eligible Medicare and Medicaid beneficiaries from nursing facility and home and community based services waiver programs. Journal of the Am erican Geriatrics Society 2012;60:821 829. 76. Sands LP, Xu H, Weiner M, et al. Comparison of resource utilization for Medicaid dementia patients using nursing homes versus home and community based waivers for long term care. Med Care 2008;46:449 453. 77. Ande rsen RM. Revisiting the behavioral model and access to medical care: does it matter? J Health Soc Behav. 1995;36:1 10.

PAGE 170

170 78. Hughes JS, Averill RF, Eisenhandler J, et al. Clinical Risk Groups (CRGs): a classification system for risk adjusted capitation based payment and health care management. Med Care 2004;42:81 90. 79. Edwards LJ, Muller KE, Wolfinger RD, et al. An R2 statistic for fixed effects in the linear mixed model. Stat Med 2008;27:6137 6157. 80. Shadish WR, Cook TD, Campbell DT. Experimental and Quasi Experimental Designs for Generalized Causal Inference. New York, NY: Houghton Mifflin Company; 2002 81. Quality of care externa l quality review (EQR). Baltimore, MD: Centers for Medicare & Medicaid Services Available at: http://www.medicaid.gov/Medicaid CHIP Program Information/By Topics/Quality of Care/Quality of Care External Quality Review.html. Accessed December 1, 2014. 82. Wagn er EH, Austin BT, Davis C, et al. Improving chronic illness care: translating evidence into action. Health Aff (Millwood). 2001;20:64 78. 83. Medicaid Managed Care Market Tracker. Kaiser Family Foundation Available at: http://kff.org/data collection/medicaid managed care market tracker/. Accessed March 16, 2015. 84. Managed Care Profiles. Centers for Medicare & Medicaid Services Available at: http://medicaid.gov/medicaid chip program information/by topics/delivery systems/managed care/managed care profiles.html. Accessed March 16, 2015 85. Whiteford HA, Degenhardt L, Rehm J, et al. Global burden of disease attributable to mental and substance use disorders: findings from the Global Burden of Disease Study 2010. Lancet 2013;382(9904):1575 86. 86. Bloom B, Cohen RA, and Freeman G. Summary health statistics for U.S. children: National Health Interview Survey, 2010 Vital Health Stat. 10. 2011 ; (250):1 80. 87. Mechanic D. Mental health and social policy: beyond managed care. 5th ed. Boston (MA): Pearson Education; 2008 88. National Association of State Mental Health Program Directors (NASMHPD) Medical Directors Council. Morbidity and mortality in people with serious mental illness http://www.nasmhpd.org/sites/default/files/Mortality%20and%20Morbid ity %20Final%20Report%208.18.08.pdf Accessed June 1, 2015. 89. Wang PS, Aguilar Gaxiola S, Alonso J, et al. Use of mental health services for anxiety, mood, and substance disorders in 17 countries in the WHO world mental health surveys. Lancet 2007 ;370(9590):841 50.

PAGE 171

171 90. Institute of Medicine. Improving the quality of health care for mental and substance use conditions. Washington (DC): National Academie s Press;2006. (Quality Chasm Series). 91. Wang PS, Lane M, Olfson M, Pincus HA, Wells KB, Kessler RC. Twelve month use of mental health services in the United States: results from the National Comorbidity Survey Replication. Arch Gen Psychiatry 2005;62(6):629 640. 92. McGlynn EA, Asch SM, Adams J, et al. The quality of health care delivered to adults in the United States. NEJM 2003;348:2635 2645. 93. SAMSHA. Results from the 2010 national survey on drug use and health: summary of national findings. https://www.samhsa.gov/data/sites/default/files/NSDUHNationalFindingsResults2 010 web/2k10ResultsRev/NSDUHresultsRev20 10.pdf Accessed June 1, 2017. 94. Busch SH, Meara E, Huskamp HA, et al. Characteristics of adults with substance use disorders expected to be eligible for Medicaid under the ACA. Psychiatr Serv 2013;64(6):520 526. 95. Druss BG & Walker ER. Mental disorders and medical comorbidity. The Synthesis Project 2011;21:1 32. 96. Bindman AB. Managing the future of Medicaid. JAMA 2015;314(4):345 6. 97. Bachman SS, Drainoni ML, Tobias C. Medicaid managed care, substance abuse treatment, and people with disabilities: review of the literature. Health Soc Work 2004 Aug;29(3):189 96. 98. Texas Medicaid and CHIP In Perspective: 10th Edit ion February 2015 (The http://www.hhsc.state.tx.us/medicaid/about/PB/PinkBook.pdf. In. 99. Wegman MP, Herndon JB, Muller KE, et al. Quality of care for chronic conditions among disabled Medicaid enrollees: an evaluation of a 1915 (b) and (c) Waiver Program. Med Care. 2015;53(7):599 606. 100. Zuckerman S, Brennan N, Yemane A. Has Medicaid managed care affected beneficiary access and use? Inquiry. 2002;39(3):221 242. 101. G arrett B, Davidoff AJ, Yemane A. Effects of Medicaid managed care programs on health services access and use. Health Serv Res. 2003;38(2):575 594.

PAGE 172

172 102. Garret B, Zuckerman S. National estimates of the effects of mandatory Medicaid managed care programs on health care access and use, 1997 1999. Med Care. 2005;43(7):649 657. 103. Medical Expenditure Panel Survey Topics. In. Priority Conditions -General : Agency for Healthcare Research and Quality (US). 104. Department of Health and Human Services (US). Medicaid Program: Initial Set of Health Care Quality Measures for Medicaid Eligible Adults. Final Notice. Fed Regist. 2012;77(2):286 291. 105. National Committee for Quality Assurance. 2000. HEDIS 2001 Vol. 1. Washington, D.C.: NCQA 106. Institute of Medicine and E. National Academies of Sciences, and Medicine,, Accounting for Social Risk Factors in Medicare Payment: Identifying Social Risk Factors 2016, Washington, DC: The National Academies Press. 107. Andersen, R. and J.F. Newman, Societal and Individual Determinants of Medical Care Utilization in the United States. Milbank Quarterly, 2005;83(4). 108. Lo Sasso AT, Freund DA. A longitudinal evaluation of the effect of Medi Cal managed care on supplemental security income and aid to families with dependent children enrollees in two California counties. Med Care. 2000;38(9):937 947. 109. Muller KE, Fetterma BA. Regression and ANOVA: An integrated approach using SAS software New York: SAS Institute; 2002. 110. R [Computer software] 111. Rstudio [Computer software] 112. Ime4 [Computer software] 113. Ismeans [Computer software] 114. Dimick JB, Ryan AM. Methods for evaluating changes in health care policy: The difference in differences approach. JAMA. 2014;312(22):2401 2402. 115. de Oliveira C, Cheng J, Vigod S, et al. Patients with high mental health costs incur over 30 percent more costs than other high cost patients. Health Aff Jan 2016;35:136 43; doi:10.1377/hlthaff.2015.0278.

PAGE 173

173 116. Busch AB, Huskamp HA, McWilliams JM. Early efforts by Medicare accountable care organizations have limited effect on mental illness care and management. Health Aff 2016;35(7):1247 1256. 117. Barry CL, Stuart EA, Donohue JM, et al. The early impact of the Quality on mental health service use and spending in Massachusetts. Health Aff Dec 2015; 34 : 12 2077 12 2085 118. Cummings JR. Declining psychiatrist participation in health insurance networks: Where do we go from here? JAMA. Jan 2015:13(2):190 191. 119. National Committee for Quality Assurance. 2016. HEDIS 2016 Vol. 1. Washington, D.C.: NCQA 120. Center for Medicaid and Medicare Services (CMS), HHS. Medicaid and children's insurance programs; mental health parity and addiction equity act of 2008; The application of mental health parity requirements to coverage offered by Medicaid managed care organ izations, the Children's Health Insurance Program ( CHIP), and Alternative Benefit Plans. Final rule. Fed Regist 2016;81(61):18389 18445. 121. Huskamp HA, Iglehart JK. Mental health and substance use reforms milestones reached and challenges ahead. NEJM 2016;375:688 695. 122. Duggan M & Hayford T. Has the shift to managed care reduced Medicaid expenditures? Evidence from state and local level mandates. J of Pol Anal & Manag 2013;32(3):505 5 35. 123. Medicaid and CHIP Payment and Acces s Commission. Access in brief: Using cervical, breast, and colon cancer tests among adult Medicaid enrollees. Nov 2016 124. Lin JS, Piper MA, Perdue LA, et al. Screening for colorectal cancer: Updated evidence report and systematic review for the US Preventive Services Task Force. JAMA 2016;315(23):2576 2594. doi:10.1001/jama.2016.3332 125. Pinsky PF, Doroudi M. Colorectal cancer screening. JAMA 2016;316(16):1715. doi:10.1001/jama.2016.13849 126. Cheung PT, Wiler JL, Lowe RA. National study of barriers to timely primary care and emergency department utilization among Medicaid beneficiaries. Ann Emerg Med 2012;60(1):4 10. 127. Silver D, Bluste in J, Weitzman BC. Transportation to clinic: findings from a pilot clinic based survey of low income suburbanites. J Immigr Minor Health 2012;14(2):350 355.

PAGE 174

174 128. Syed ST, Gerber BS, Sharp LK. Traveling towards disease: transportation barriers to health care access. J Community Health. 2013;38(5):976 993. 129. Rees SA, Williams A. Promoting and supporting self management for adults living in the community with physical chronic illness: A systematic review of the effectiveness and meaningfulness of the patient practitioner encounter. 2009;7(13):492 582. 130 Shipee ND, Shah ND, May CR. Cumulative complexity: a functional, patient centered model of patient complexity can improve research and practice. J of Clin Epidem 2012;65(10):1041 1051. 131 Wegman M, Patel AGM, Sacks PK, et al. Measurement of cognitive functioning in clinical trials of the chronically ill: a systematic review. 2017. Unpublished manuscript. 132. Wegman M. WIN Care: characteristics of adults with chronic illness and their care givers. 2017. Unpublished manuscript. 13 3 McWilliams JM, Hatfield LA, Chernew ME, et al. Early performance of accountable care organizations in Medicare. N Engl J Med 2016;374:2357 2366. 134 WHO Commission on Social Determinants of Health, World Health Organization. Closing the gap in a generation: health equity through action on the social determinants of health: Commission on social determinants of health final report. 2008 135 Bradley EH, Taylor LA. The American Health Care Paradox 2013. Print. 136 McGinnis JM, Williams Russo P, Knickman JR. The case for more active policy attention to health promotion. Health Aff 2002 ;21(2):78 93. 137 Taylor LA, Coyle CE, Ndumele C, et al. Leveraging the social determinants of health: What works? PLoS ONE;(8): e0160217. https://doi.org/10.1371/journal.pone.016021

PAGE 175

175 BIOGRAPHICAL SKETCH Martin Wegman received his PhD degree from the University of Florida while enrolled in the MD PhD program His PhD was concentrated in e pidemiology, under the mentorship of Dr. Elizabeth Shenkman professor and chair of the Department of Health Outcomes a nd Policy He was formerly an NIH Ruth L. Kirschstein National Research Service Awardee and served as a Doris Duke International Clinical Research Fellow at Yale School of Medicine from 2014 to 2015 He also holds a Bachelor of Science degree with Highest Distinction in B iomedical E ngineering from the University of Rochester. work focuses on analyzing and developing policies and systems to improve population health and social outcomes, with at tention to individuals living at the margins of society, including those affected by drug use and mental illness. His research has taken him to Malaysia and South Africa, and has spanned topics ranging from harm reduction, managed care and patient engageme nt. His achievements have include serving as principal investigator or lead author on over $450,000 in grant funding and publishing more than 20 articles, including work featured in the Lancet Global Health, JAMA and Health Affairs. He has received nationa l recognition as a National Quality Scholar from the American College of Medical Quality and received the Excellence in Medicine Award from the American Medical Association F oundation. While at the University of Florida, he was inducted in the Gold Humanis m Society. After his training, Martin plans to continue to influence patient and population health as a clinician scientis t.