Citation
An Evaluation of the Impact of Provider Service Networks in Florida Medicaid Managed Care on Healthcare Process and Outcomes

Material Information

Title:
An Evaluation of the Impact of Provider Service Networks in Florida Medicaid Managed Care on Healthcare Process and Outcomes
Creator:
Park, Sinyoung
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (134 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Health Services Research
Health Services Research, Management, and Policy
Committee Chair:
HARMAN,JEFFREY SCOTT
Committee Co-Chair:
HALL,ALLYSON G
Committee Members:
DUNCAN,R P
LEE,GWENDOLYN K
Graduation Date:
8/9/2014

Subjects

Subjects / Keywords:
Entitlement programs ( jstor )
Health care industry ( jstor )
Health care organizations ( jstor )
Health care services ( jstor )
Hispanics ( jstor )
Hospitals ( jstor )
Managed care ( jstor )
Medicaid ( jstor )
Physicians ( jstor )
Prescription drugs ( jstor )
Health Services Research, Management, and Policy -- Dissertations, Academic -- UF
expenditures -- healthcareoutcomes -- healthcareutilization -- integration -- providerservicenetworks
Duval County ( local )
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Health Services Research thesis, Ph.D.

Notes

Abstract:
In 2005, Florida was allowed to change its Medicaid program by the CMS. Broward and Duval counties were selected as pilot counties in July 2006 for the Medicaid Demonstration Program. In 2007, the demonstration added the rural counties of Baker, Clay, and Nassau counties. In reform counties, Medicaid enrollees were required to pick a managed care plan that was either a Health Maintenance Organization (HMO) or a Provider Service Network (PSN). PSNs are a form of managed care that provides health care services directly through a provider or network of organizations to a defined population without a middle man such as a third party insurance company and the health plan. There are two types of PSNs: Physician-based PSNs and Healthcare system-based PSNs. Physician-based PSNs are created and controlled mainly by physicians groups serving primarily the Medicaid population to provide health care services to their patients in the reform counties. Healthcare system-based PSNs are created by safety net hospitals to serve the Medicaid population. Health system-based PSNs may provide more efficient care, since they have access to more resources in terms of workforce, IT systems, and capital than physician groups. This study expected to find that utilization, expenditures on health services, and outcomes vary depending on the type of PSN in which Medicaid beneficiaries are enrolled. To assess the differences between physician-based PSNs and health system-based PSNs over time, this study used Structure-Process-Outcomes model and difference-in-difference study design with Medicaid claims, eligibility data (from FY0405 to FY0910), and CAHPS data (from 2006 to 2008). The study findings were that, compared to enrollees in physician-based PSNs, enrollees in health system-based PSNs were more likely to have higher utilization, expenditures, and satisfaction with health plans, overall healthcare, personal doctor, and specialist on average during the post period. However, the trends in utilization, expenditures, and outcomes for enrollees in health system-based PSNs over time were decreasing at a greater rate relative to the trends for enrollees in physician-based PSNs. While some hypotheses were not supported by findings, the impact of two different types of PSNs on utilization, expenditures, and outcomes can be demonstrated. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2014.
Local:
Adviser: HARMAN,JEFFREY SCOTT.
Local:
Co-adviser: HALL,ALLYSON G.
Statement of Responsibility:
by Sinyoung Park.

Record Information

Source Institution:
UFRGP
Rights Management:
Copyright Park, Sinyoung. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Classification:
LD1780 2014 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

AN EVALUATION OF THE IMPACT OF PROVIDER SERVICE NETWORKS IN FLORIDA MEDICAID MANAGED CARE ON HEALTHCARE PROCESS AND OUTCOMES By SINYOUNG PARK A DISSERTATION PRESENTED 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 2014

PAGE 2

© 2014 Sinyoung Park

PAGE 3

To my Dad and Mom, Chul Won Park and Kyung Hee Shin for your unconditional love and for unwavering faith, perseverance and spiritual support

PAGE 4

4 ACKNOWLEDGMENTS I would first like to thank my Lord, the guiding force in my life, who gives me strength and wisdom to accomplish this long journey. I do all this through Him, who gives me strength. I also want to thank my parents for their unconditional love, support, an d unwavering faith. They were always there for me, praying for me, and being my role models as parents. They taught me how to lean on the Lord, live for others, and share my love to others. My sincere thanks to my advisor and committee chair, Dr. Jeffrey S. Harman, who has been a source of encouragement, guidance, patience, and support throughout my time in the Health Services Research program. I had regular meetings with him every week for these several years. He is always patient and taught me how to con duct my research and share my knowledge with students. Dr. Harman is more than an academic advisor for me. He is my role model as a mentor and hopefully I can follow his way as a counselor, mentor, and trainer for students. I also wish to thank my other committee members. I thank Dr. R. Paul Duncan for his generosity and careful guidance regarding my doctoral program, dissertation, and future. I thank Dr. Allyson G. Hall for encouraging me through my entire time in the program, caring about me as a studen t and as a person, and suggesting the policy background and conceptual framework on the structure of my research. I also thank Dr. Gwendolyn K. Lee for her insightful feedback and suggestions on my dissertation, especially for the theoretical framework. Wh enever I have a meeting with her, I have learned how to be a successful woman as a researcher and a mom. I was so fortunate to have other faculty members and staff in our department and would like to thank them. Especially, I thank Dr. Christopher A. Harl e for sharing his

PAGE 5

5 experience on my career path and helping me to find the right place for my first academic career. I also thank Lilliana Bell , Donna Stilwell , and Patty Van Wert for their administrative help and support during my time at UF. Thank you guy s for everything you have done for me! I am also so blessed to have awesome friends and colleagues over the past thank you for your everlasting support and existence on my every step to grow up. Melody Schiaffino, Yeun Ji Jung, Jusil Lee, Heidi Kindsell, Cesar Escobar, Jon Mills, Damian Everhart, Shuo Yang, Wendy Zhong, Ara Jo, Sarah Bauer, Dr. Hal Lerch, Mrs. Judy Lerch, Pastor. Sungjung Kim, and other friends in this program and in my young our precious time sharing our life stories, and all the blessed times we had together. Thank you guys for always being there for me and prayi ng for me!

PAGE 6

6 TABLE OF CONTENTS P age ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 10 LIST OF ABBREVIATIONS ................................ ................................ ........................... 11 ABSTRACT ................................ ................................ ................................ ................... 12 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ ........ 14 Specif ic Aims ................................ ................................ ................................ .......... 15 Significance of the Study ................................ ................................ ........................ 15 2 BACKGROUND AND LITERATURE REVIEW ................................ ........................... 20 Medicaid: Issues and Challenges ................................ ................................ ........... 20 Medicaid Managed Care Organizations ................................ ................................ .. 22 Medicaid Reform Demonstration Program in Florida ................................ .............. 25 Provider Service Networks (PSNs) ................................ ................................ ......... 28 ACO like Organizations ................................ ................................ .................... 28 Physician led and Hospital led ACOs ................................ ............................... 29 Organizational Structures: Physician based and Health system based PSNs ................................ ................................ ................................ ............. 30 PSNs vs. Other Managed Care Plans ................................ .............................. 31 Integrated Delivery System ................................ ................................ ..................... 32 Vertical and Horizontal Integration ................................ ................................ ... 33 Impact of Integrated Delivery System ................................ ............................... 34 Conclusions ................................ ................................ ................................ ............ 35 3 CONCEPTUAL FRAMEWORK AND HYPOTHESIS ................................ .................. 39 Theoretical Implications ................................ ................................ .......................... 39 Agency Theory ................................ ................................ ................................ . 39 Transaction Cost Economics ................................ ................................ ............ 40 Resource Dependence Theory ................................ ................................ ......... 42 Co nceptual Framework for Provider Service Networks ................................ .......... 43 Hypotheses ................................ ................................ ................................ ............. 47 The Association of Organizational Structure and Utilization ............................. 47 The Association of Organizational Structure and Expenditures ........................ 48 The Association of Structure, Process, and Outcomes ................................ .... 49

PAGE 7

7 4 DATA AND METHODS ................................ ................................ .............................. 53 Study Design ................................ ................................ ................................ .......... 53 Data Sources and Empirical Models ................................ ................................ ....... 55 Florida Medicaid Claims and Eligibility Data ................................ ..................... 55 CAHPS Data ................................ ................................ ................................ .... 56 Empirical Models by Hypotheses ................................ ................................ ..... 57 Measures for Data ................................ ................................ ................................ .. 58 Outcome Va riables ................................ ................................ ........................... 58 Independent Variables ................................ ................................ ..................... 60 Control Variables ................................ ................................ .............................. 60 Analytical Data Set ................................ ................................ ................................ . 61 Specific Aim 1 and 2 ................................ ................................ ......................... 61 Specific Aim 3 ................................ ................................ ................................ ... 62 Statistical Analysis ................................ ................................ ................................ .. 63 Specific Aims 1 and 2 ................................ ................................ ....................... 63 Specif ic Aim 3 ................................ ................................ ................................ ... 67 5 RESULTS ................................ ................................ ................................ ................... 72 Specific Aim 1 and 2 ................................ ................................ ............................... 72 Descriptive Analysis ................................ ................................ ......................... 72 Univariate Analysis ................................ ................................ ........................... 73 Multivariate Analysis ................................ ................................ ......................... 74 Sensitivity Analysis ................................ ................................ ........................... 79 Specific Aim 3 ................................ ................................ ................................ ......... 80 Descriptive An alysis ................................ ................................ ......................... 80 Univariate Analysis ................................ ................................ ........................... 80 Multivariate Analysis ................................ ................................ ......................... 81 6 DISCUSSION AND CONCLUSIONS ................................ ................................ ....... 109 Specific Aims 1 and 2 Discussion ................................ ................................ ......... 109 Healthcare Utilization ................................ ................................ ..................... 110 PMPM Expenditures ................................ ................................ ....................... 111 Discussion ................................ ................................ ................................ ...... 111 Specific Aim 3 D iscussion ................................ ................................ ..................... 114 ................................ ................................ .................... 114 Discussion ................................ ................................ ................................ ...... 115 Policy Implicati ons ................................ ................................ ................................ 116 Limitations ................................ ................................ ................................ ............. 118 Future Research ................................ ................................ ................................ ... 120 Conclusions ................................ ................................ ................................ .......... 122 LIST OF REFERENCES ................................ ................................ ............................. 125 BIOGRAPH ICAL SKETCH ................................ ................................ .......................... 134

PAGE 8

8 LIST OF TABLES Table page 2 1 Characteristics of Physician based and Health System based PSNs ................ 38 4 1 List of variables ................................ ................................ ................................ ... 70 4 2 Study population by fiscal year (Aim 1 and 2) ................................ ..................... 71 4 3 Study population by year (Aim 3) ................................ ................................ ........ 71 4 4 Study population by plan (Aim 1 and 2) ................................ .............................. 71 4 5 Study population by plan (Aim 3) ................................ ................................ ........ 71 5 1 Sample characteristics (Aims 1 and 2)* ................................ .............................. 83 5 2 Univariate analysis (Aims 1 and 2) ................................ ................................ ..... 84 5 3 Multivariate model of ED visits (Aim 1) ................................ ............................... 85 5 4 Multivariate model of ED visits for SSI enrollees (Aim 1) ................................ .... 86 5 5 Multivariate model of ED visits for TANF enrollees (Aim 1) ................................ 86 5 6 GEE model of inpatient utilization (Aim 1) ................................ .......................... 87 5 7 Multivariate model of inpatient utilization for SSI enrollees (Aim 1) .................... 88 5 8 Multivariate model of inpatient utilization for TANF enrollees (Aim 1) ................. 88 5 9 GEE model for rate of prescription drug use (Aim 1) ................................ .......... 89 5 10 Multivariate model of prescription drug use for SSI enrollees (Aim 1) ................ 90 5 11 Multivariate model of prescription drug use for TANF enrollees (Aim 1) ............. 90 5 12 GEE model of PMPM expenditures (Aim 2) ................................ ....................... 91 5 13 GEE model of PMPM expenditures for SSI enrollees (Aim 2) ............................ 92 5 14 GEE model of PMPM expenditures for TANF enrollees (Aim 2) ......................... 92 5 15 Sensitivity analysis of ED visits with three months of observations (Aim 1) ........ 93 5 16 Sensitivity analysis of inpatient stays with three months of observations (Aim 1) ................................ ................................ ................................ ........................ 94

PAGE 9

9 5 17 Sensitivity analysis of prescription drugs with three months of observations (Aim 1) ................................ ................................ ................................ ................ 95 5 18 Sensitivity analysis of PMPM expenditures with three months of observations (Aim 2) ................................ ................................ ................................ ................ 96 5 19 Sensitivity analysis of ED visits with six months of observations (Aim 1) ........... 97 5 20 Sensitivity analysis of inpatient stays with six months of observations (Aim 1) .. 98 5 21 Sensitivity analysis of prescription drugs with six m onths of observations (Aim 1) ................................ ................................ ................................ ................ 99 5 22 Sensitivity analysis of PMPM expenditures with six months of observations (Aim 2) ................................ ................................ ................................ .............. 100 5 23 Sample characteristics (Aim 3) * ................................ ................................ ....... 101 5 24 Univariate analysis (Aim 3, unweighted) ................................ ........................... 102 5 25 Univariate analysis (Aim 3, weighted) ................................ ............................... 103 5 26 Multivariate analysis of satisfaction with health plans (Aim 3) .......................... 104 5 27 Multivariate analysis of satisfacti on with overall healthcare (Aim 3) ................. 105 5 28 Multivariate analysis of satisfaction with personal doctor (Aim 3) ..................... 106 5 29 Multivariate analysis of satisfaction with specialist (Aim 3) ............................... 107 5 30 Summary of Results (health system based over physician based PSNs) ........ 108 6 1 Discussion of Finding1 (health system based over physician based PSNs) .... 124 6 2 Discussion of Finding2 (health system based over physician based PSNs) .... 124

PAGE 10

10 LIST OF FIGURES Figure page Figure 2 1. Florida Medicaid Reform Areas ................................ ................................ ... 37 Figure 3 ................................ ................................ ........ 50 Figure 3 2. Florida PSNs Framework ................................ ................................ ............ 51 Figure 3 ................................ ................................ ....................... 52 Figure 4 1. Timeline of Florida Medicaid Reform Demonstration ................................ .. 69

PAGE 11

11 LIST OF ABBREVIATIONS ACO Accountable Care Organization AHCA Agency for Health Care Administration CAHPS Consumer Assessment of Healthcare Providers and Systems CMS Centers for Medicare & Medicaid Services DID Difference In Difference ED Emergency Department FFS Fee For Service GEE Generalized Estimating Equation HEDIS Healthcare Effectiveness Data and Information Set HMO Health Maintenance Organization IDS Integrated Delivery System IPA Independent Physician Association MPN Minority Physician Network PCCM Primary Care Case Management PMPM Per Member Per Month PPACA Patient Protection and Affordable Care Act PSN Provider Service Network PSO Provider Sponsored Organization RDT Resource Dependence Theory SFCCN South Florida Community Care Network SPO Structure Process Outcome SSI Supplemental Security Income TANF Temporary Assistance for Needy Families TCE Transaction Cost Economics

PAGE 12

12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Parti al Fulfillment of the Requirements for the Degree of Doctor of Philosophy AN EVALUATION OF THE IMPACT OF PROVIDER SERVICE NETWORKS IN FLORIDA MEDICAID MANAGED CARE ON HEALTHCARE PROCESS AND OUTCOMES By Sinyoung Park A ugust 2014 Chair: Jeffrey S. Harman Major: Health Services Research In 2005, Florida was allowed to change its Medicaid program by the CMS. Broward and Duval counties were selected as pilot counties in July 2006 for the Medicaid Demonstration Program. In 2007, the demonstration added the r ural counties of Baker, Clay, and Nassau counties. In reform counties, Medicaid enrollees were required to pick a managed care plan that was either a Health Maintenance Organization (HMO) or a Provider Service Network (PSN). PSNs are a form of managed care that provides health care services directly through a provider or network insurance company and the health plan. There are two types of PSNs: Physician based PSNs and H ealthcare system based PSNs. Physician based PSNs are created and controlled mainly by physicians groups serving primarily the Medicaid population to provide health care services to their patients in the reform counties. Healthcare system based PSNs are cr eated by safety net hospitals to serve the Medicai d population. Health system based PSNs may provide more efficient care, since they have access to more resources in terms of workforce, IT

PAGE 13

13 systems, and capital than physician groups. This study expected to find that utilization, expenditures on hea lth services, and outcomes vary depending on the type of PSN in which Medicai d beneficiaries are enrolled. To assess the differences between physician based PSNs and health system based PSNs over time, this study u se d Structure Process Outcomes model and difference in difference study design with Medicaid claims, eligibility data ( from FY0405 to FY0910) , and CAHPS data ( from 2006 to 2008) . The study findings were that, compared to enrollees in physician based PSNs, enrollees in health system based PSNs were more likely to have higher utilization, expenditures, and satisfaction with health plan s , overall healthcare, personal doctor, and specialist on average during the po st period . However, the trends in utilization, expenditures, and outcomes for enrollees in health system based PSNs over time were decreasing at a greater rate relative to the trends for enrollees in physician based PSNs. While some hypotheses were not sup ported by findings, the impact of two different types of PSNs on utilization, expenditures, an d outcomes can be demonstrated.

PAGE 14

14 CH APTER 1 INTRODUCTION Provider Service Networks (PSNs) is a type of Florida Medicaid managed care organization. The goals of a PSN are to provide coordinated care, enhance health outcomes in Medicaid enrollees, and improve quality of care (Duncan et al., 2007; Johnson et al., 2010). As Medicaid expands under federal law while individual state budg ets for Medicaid program shrink, there is an emerging recognition that redesigning a Medicaid managed care organization at the state level would be considered the best way to maximize efficiency of the Medicaid program. There are two types of PSNs evaluated in this study physician based and health system based PSNs. Physician based PSNs are created and controlled mainly by physicians groups serving primarily the Medicaid population to provide healthcare services to their patients . Health system ba sed PSNs are created within health systems that include safety net hospitals to serve the Medicaid population. However, little research has examined PSNs, as they are a relatively new model of managed care for Medicaid beneficiaries. In particular, there is limited research on classifying different types of PSNs physician based PSNs vs. health system based PSNs, studying their organizational characteristics, and evaluating the effect of different types of PSNs on healthcare outcomes. Identifying organiza tional characteristics and management structure of the PSNs may be helpful in explaining the differences in healthcare utilization, expenditures, and outcomes between these types of organizations. Moreover, a better understanding of these innovative health care delivery systems for Medicaid patients may help the states manage Medicaid programs efficiently. As such, this study evaluate d two different types of PSNs and compare their

PAGE 15

15 respective effects on utilization, expenditures of healthcare services, and ou tcomes for Medicaid enrollees. Specific Aims Based on this background, the study objectives are to : 1. Investigate the difference in emergency department (ED) utilization, inpatient hospital utilization, and prescription drug use by Medicaid enrollees between physician based PSNs and health system based PSNs in reform counties, 2. Assess the difference in per member per month (PMPM) expenditures of healthcare services between physician based PSNs and health system based PSNs in reform counties, and 3. Examine the difference in self reported satisfaction with health plans , overall healthcare, personal doctor, and specialist by Medicaid enrollees between physician based PSNs and health system based PSNs in reform counties. Significance of the Study This study addres ses a gap in the academic literature by examining the relationship between the types of PSNs and healthcare process and outcome for Medicaid enrollees. Despite the rapid pace of managed care expansion, little is known about the broad effects of new forms o f managed care organizations on healthcare utilization and expenditures at the individual and state level. A number of previous articles investigating the outcomes of Florida Medicaid reform were conducted with pre post study , which means most studies comp ared PSNs to traditional managed care including HMOs and PCCMs (Lemak et al., 2005; Harman et al., 2009 , 2011 ; Johnson et al., 2010; Schiller et al., 2010 ; Hall et al., 2013). However, at the same time, there is a need to evaluate how the two different ty pes of new managed care structures work for Medicaid enrollees by states over time. This study will be a significant contribution to the literature as prior studies have yielded inconsistent results about use and spending of healthcare services between dif ferent Medicaid managed care organization, especially

PAGE 16

16 HMOs and PSNs, and no information was available to determine the impact of the structural different types of PSNs in terms of structural integration on utilization, expenditure of health services, and o utcomes, including satisfaction with health plans , overall healthcare, personal doctor, and specialist . The significance of this study can be summarized in three areas. First of all, the study includes a longitudinal dataset that contains structure, proce ss, and outcome measures to evaluate the impact of the new form of Medicaid managed care on utilization, expenditures, and outcomes. Especially, this study focus es more on how structural difference between physician based PSNs and health system based PSNs, which are less and more integrated delivery system s , respectively influence on the healthcare process and outcomes. Most previous research on the impact of integrated delivery system has analyzed data from a single year (Wang et al., 2001; Wan et al., 200 1 , 2002; Newhouse et al., 2003; Kautz et al., 2007). Cross sectional design in the analysis limits causal inference. Longitudinal data allows for researchers to establish temporal precedence and make causal inferences about the relationship between new typ es of managed care and healthcare utilization and expenditures, as do statistical techniques that account for change in variation over time (Duncan & Duncan, 2004). Therefore, there is a need to examine the utilization, expenditures , and outcomes of differ ent organizational forms in PSNs with empirical evidence first , then decide whether or not the new organizational forms will be adopted for managing Medicaid utilization and expenditures at the state level. The longitudinal data allow s for an unprecedented opportunity to explore the association between

PAGE 17

17 longitudinally. The study proposes to analyze changes in utilization and expenditures by Medicaid beneficiaries between time periods and between two groups using longitudinal data and the difference in difference approach. This methodology isolates the relationship between treatment and effects from other confounding factors to provide an estimate of the effect of the establishment of new policy (Mortensen, 2010). Further, this study devel ops a conceptual framework by modifying an existing Process Outcomes (SPO) model. In components that relate to the organizational characterist ics of healthcare providers (structure), the clinical and administrative procedures involved in the delivery of healthcare to patients (process) , and the effect of care on the health status of patients (outcomes). Specifically, structure refers to organiza tional resources for delivery of healthcare services, such as several Medicaid managed care organizations, including physician based PSNs and health system based PSNs (Donabedian, 1966, 1988; Schiller et al., 2010). Therefore, the model suggests that organ izational structure influences the processes of patient care, which , in turn , affects patient outcomes. In addition, the theoretical framework developed in this study is rooted in the resource dependence theory (RDT), agency theory, and transaction cost ec onomics (TCE). The theories are used to characterize the organizational structures between the two types of PSNs. Specifically, RDT in the sociological literature is used as a lens for explaining why health system based PSNs have a more integrated organiza tional structure than physician based PSNs. According to RDT, organizations depend on other organizations to gain essential resources when they have limited resources and achieve

PAGE 18

18 their independenc e when they have sufficient resources in their own organizat ions (Pfeffer and Salancik, 1978). In addition, the agency theory (Jensen et al., 1976; Fama and Jensen, 1983) and TCE (Coase, 1937; Williamson, 1981) in economics literature are used as a lens for explaining how more integrated organizational structure ha s greater efficiency and greater quality of care than less integrated structure. Agency theory depicts that the principal delegates the agent who performs the specific work. But the problem arises when the two parties have different interests. The deviatio n from the delivery system, physicians and hospitals cooperate and have the same interests through tighter contracts in order to promote their performance and reduce ineff iciency in integrated delivery systems. According to TCE, vertically and horizontally integrated structure, which internalizes buyers and sellers within the same organizations, is a logical strategic response to reducing tra nsaction costs (Williamson, 1981 ). A majority of articles in the assessment of Medicaid managed care literature lack a conceptual or theoretical framework to explain the relationship among organizational structures of managed care and healthcare process and outcomes. This lack of concep tual and theoretical clarity limits the ability of researchers to determine how the different Medicaid managed care with different organizational structure s interact with Medicaid enrollees to impa ct their healthcare utilization, expenditures , and healthca re outcomes, such as enrollee satisfaction with health plans, overall healthcare, personal doctor, and specialist . In the previous literature, there have only been a few conceptual and theoretical models used that link managed care and healthcare outcomes for

PAGE 19

19 framework (1966). Therefore, a new model is constructed that addresses gaps in prior theoretical models and articulates a clear pathway between managed care structure and utiliza tion, expenditures , and healthcare outcomes for providing healthcare services to Medicaid beneficiaries.

PAGE 20

20 CHAPTER 2 BACKGROUND AND LITERATURE REVIEW This chapter provide s a general overview of Medicaid, Medicaid managed care organizations, Florida Medicaid reform demonstration program , and PSNs. Especially, the chapter discusses the organizational structures of two different types of PSNs. T he struc tural differences between both are explained by the concept of integration . Finally, by reviewing previous research, the chapter summarize s the impact of the level of integration on healthcare utilization, expenditures, and outcomes. Medicaid: Issues and Challenges The Medicaid program was established by legislation signed into law on July 30, 1965, by President Lyndon B. Johnson under the Social Security Amendments of 1965 (Koch, 2002). Medicaid was enacted in 1965 as Title XVIII of the Social Security Act and is funded by both the federal and state government s (Landry et al., 2011). Its main focus i s on the welfare of the population, including: families with children receiving support under the Temporary Assistance for Needy Families (TANF) program, people receiving the Supplemental Security Income (SSI) benefits, which include many of the elderly, b lind, and disabled with low income s , and pregnant women and children with family income at or below 133 percent of the federal poverty level (Koch, 2002). It includes mandated services, such as hospital care, nursing home care, home health services, physic ian services, immunization and preventive medicine for children, and nurse practitioner services. Since 1965, Medicaid costs have increased annually. Federal and state expenditure is about $300 billion (Kaiser, 2008). According to Kenen (2012), the second most costly category of state spending is Medicaid related expenses. Drivers of Medicaid spending growth include overall healthcare costs,

PAGE 21

21 especially in acute care, prescription drugs, and enrollment growth over the years (Iglehart, 2012). There are sever al factors that increase the number of Medicaid enrollees and the amount of state budget needed for Medicaid programs. First, in recent years, as in previous recessions, states have experienced large increases in Medicaid spending as more people became une mployed and have enrolled in the program. About 6 million additional people enrolled in Medicaid during the previous recession (Iglehart, 2012). In addition, under the Patient Protection and Affordable Care Act (PPACA, 2010), Medicaid will be expanded to i nclude nearly all adults under age 65 with income below 133 percent of the federal poverty level until 2014 (Kenen, 2012; Rosenbaum et al., 2012; DeLeire et al., 2013), although states are not required to expand Medicaid coverage. The Patient Protection an d Affordable Care Act (PPACA) was signed into Federal law by President Obama and enacted on March 23, 2010 (Oberlander, 2010; Rosenbaum, 2011). It focuses on expanding access to insurance coverage, controlling healthcare costs, and improving quality of car e by healthcare system reform ( Keiser, 2013 ). It emphasizes the individual insurance market, small business insurance market, and the uninsured (Rosenbaum, 2011). To achieve these provisions, there are various approaches: the individual insurance mandate, increase in subsidies, the expansion of public programs, changes to private insurance, improving quality and health system performance, and developing a national strategy for prevention and wellness ( Keiser, 2013 ). The most controversial component of the A ct was the individual mandate 138 percent of the federal poverty level, they would actually qualify for Medicaid

PAGE 22

22 han 400 percent of the federal poverty level, they qualify for up to 100 percent subsidies to help them pay for mandated health insurance (Peterson et al., 2010). Under PPACA, Medicaid will cover an additional 16 million people who are currently uninsured and the total number of Medicaid beneficiaries will be around 50 million by 2019 (Sommers et al., 2010; Rieselbach et al., 2011). Third, the federal and state government funding on Medicaid will also increase (Rosenbaum, 2010). One consideration of this in crease is the concern that approximately 25 percent of existing uninsured persons are actually Medicaid eligible (Keiser, 2007). Medicaid Managed Care Organizations Medicaid managed care organizations are created to improve access to care for Medicaid beneficiaries and to slow the growth in Medicaid expenditures (Halstead et al., 1998; Holahan et al., 1998). There were two time periods for implementing Medicaid managed care. Before th e 1990s, managed care was implemented by several states primary care physician management (PCCM). During this period, states were focused more on using managed care as a means of improving access (Highsmith et al., 2000; by health maintenance organizations (HMOs) that were paid a capitated rate for providing a contractually specified set of s ervices (Highsmith et al., 2000; Zuckerman et al., 2002). The number of Medicaid beneficiaries enrolled in Medicaid managed care increased rapidly from 10 percent in 1991 to 56 percent (22.1million) in 2000 and currently, over 26 million Medicaid beneficia ries are enrolled in HMOs and another 8.8 million are enrolled in PCCM (Keiser, 2012).

PAGE 23

23 The general difference between Medicaid managed care and Medicaid Fee for S ervice (FFS) was that Medicaid enrollees in managed care are more likely to have a primary ca re provider and less likely to use hospital care, emergency room, have prescription drugs, and be refer red to specialists than Medicaid beneficiaries in Medicaid FFS (Hurley et al., 1993; Kirby et al., 2003). Also, managed care was able to generate savings by reducing use of hospital and other high cost care due to improved primary care access and care management, and lower unit prices relative to FFS payment rates (Keiser, 2012; Kirby et al., 2003). During the implementation of managed care, states have a dopted different types of managed care programs and some states vary the types of programs they implement across counties (Silberman et al., 2002). For this reason, they may have multiple categories of Medicaid managed care organizations , including mandatory PCCM only, mandatory HMOs only, and mandatory PCCM/HMO programs. PCCM programs include a primary care physician, groups of physicians, and primary care clinics as gatekeeper entities who are contracted by the state Medicaid agency and are paid on an FFS basis, plus a monthly c ase management fee per member (Hurley et al., 1993; Rawlings Sekunda et al. 2001; Garrett et al., 2003; Cook, 2006). HMO programs are comprised of existing HMOs, prepaid health plan, and other institutional healthc are providers who can provide PCCM services and are paid on a capitation rate (Hurley et al., 1993; Garrett et al., 2003). States Medicaid programs are available , but have managed care program variation s across and within states under a federal umbrella ( Garrett et al., 2003). These diverse Medicaid managed care organizations have different effects on utilization and

PAGE 24

24 program and the organizational characteristics of managed care health plans compared to Medicaid FFS. Some quantitative studies captured a major component of that variation by estimating separate effects for PCCM, HMOs, and PCCM/HMOs plans. According to Hurley et al. (1993), gatekeeper PCCM programs increased t he number of physician visits, and decreased inpatient stays, prescription drug use, and emergency room visits, while the HMO/PHP programs decreased physician and emergency room visits. One study focused on differential effects for woman and children at th e national level (Garrett et al, 2003). Findings were that mandatory PCCM programs increased the likelihood of having a usual source of care that were not emergency room visits; mandatory HMO programs decreased the number of physician and specialist visits and the length of inpatient stays; mandatory PCCM/HMO programs increased access and lead to more appropriate utilization, including preventive care as captured by immunization. The other study examined the difference in utilization between HMO, PCCM , and Medicaid FFS (Zuckerman et al., 2002). Medicaid beneficiaries in mandatory HMO programs were more likely to have access to a regular source of medical care and less likely to visit emergency rooms as a usual source of care than those in Medicaid FFS. In ma ndatory PCCM programs, children were less likely to depend on an emergency room and more compared to c hildren in FFS. Based on these results from several empirical studies, the effects of manage d Medicaid managed care programs.

PAGE 25

25 Medicaid Reform Demonstration Program in Florida In the S tate of Florida in 2012 , there were 3.4 million Medicaid enrollees (18% of Flor ida total population). Among those people, 43 percent were the disabled, 25 percent were the elderly, 20 percent were children, and 12 percent were healthy adults (Kaiser, 2012). Florida Medicaid expenditures are estimated at $4,168 per eligible enrollee f or a total of $17 billion in 2009 (Kaiser, 2013). Florida anticipated that Medicaid expenditures is not changed (Schiller et al., 2010). In 2005, under the Section 1 115 waiver program approved by CMS, Florida could change their Medicaid program to improve access to healt hcare services and anticipate Florida 2012). In 2006, only two countie s, Broward and Duval, were selected initially as reform pilot counties. In 2007, Baker, Clay, and Nassau counties, which are largely rural, were later chosen for the Medicaid Demonstration Program (Coughlin et al., 2008; Bragdon, 2011; Harman et al., 2011; Landry et al., 2011; Hall et al., 2012). The aims of implementing Medicaid Reform in F lorida responsibility and empowerment, make a market place competitive through choosing their health plans between HMOs and PSNs with different benefit packages, and improve the health status of Medicaid beneficiaries by providing financial incentives for their healthy behaviors (Coughlin et al., 2008; Harman et al., 2011; Landry et al., 2011; Hall et al., 2012). Medicaid enrollees are allowed to choose their health plan s among several managed care organizations, which included PCCM (also known as MediPass), PSNs, and HMOs in the market place under the F lorida Medicaid Reform Demonstration

PAGE 26

26 Program in 2006. Specifically , in reform counti es, Medicaid enrollees can choose between HMOs or PSNs. In non reform counties, Medicaid beneficiaries can choose between HMOs or PCCM, (Schiller et al., 2010; Bragdon, 2011; Harman et al., 2011). In 2006, 52,620 Medicaid enrollees (46.4%) choose PSNs and 60,701 Medicaid enrollees (53.6%) choose HMOs in Broward and Duval counties (Florida C ouncil for C ommunity M ental H ealth, 2007 , http://www.fccmh.org/resource s/docs/medicaid_reform_and_managed_care_ _4 16 07.pdf ). As mentioned above , regarding the PCCM program, the state Medicaid authority contracts with primary care physicians or groups of physicians who act like gatekeepers to slow the growth of expenditures of Medicaid, reduce unnecessary care , such as inappropriate emergency room visits, and ensure access to coordinated primary care as an alternative to Medicaid managed care through HMOs for Medicaid beneficiaries (Regestein et al., 1998; Hill et al., 1999; Johnson et al., 2010; Schiller et al., 2010). Florida established PCCM programs, called MediPass , in 1991 (Johnson et al., 2010; Schiller et al., 2010). Healthcare providers under MediPass are paid a per month per member (PMPM) case management fee and are reimbursed on a FFS basis (Garrett et al., 2003; Johnson et al., 2010; Schiller et al., 2010). In 1997, the F lorida Medicaid program introduced Provider Sponsored Organizations (PSOs) (Schiller et al., 2010) , which were originally established under the fe deral Balanced Budget Act of 1997 to authorize and formalize their role in contracting Medicare risk and organized by healthcare providers, including physicians, hospitals, and allied health professionals (Davis, 1997; Gleason et al., 1998). However, some

PAGE 27

27 states have passed their own statutes recognizing variations of PSOs for managing Medicaid risk (Davis, 1997). According to Davis (1997), PSOs are public or private provider networks organized by a variety of healthcare organizations , such as hospitals, ve rtically integrated delivery system (IDSs) with both physicians and hospitals, or physician only network such as independent practice association (IPAs). Providers could sponsor a managed care organization in several ways. Healthcare providers can create t heir own entity, have a joint venture with an existing managed care organization, or purchase an existing managed care entity (Clay, 1997). Also, to reduce transaction costs, providers can contract directly with employers and the Department of H ealth and H uman Services without third party insurance companies (Davis, 1997). These PSOs were the origin of Provider Service Networks (PSNs) for the F lorida Medicaid program in 2000 (Schiller et al., 2010). PSNs are a new type of managed care organizations, develo ped to be an efficient way to provide healthcare services to beneficiaries and manage Medicaid programs (Duncan et al., 2007; Johnson et al., 2010; Hall et al., 2013). At the first stage of implementation of PSNs, several safety net hospitals and the physi cians working with these hospitals elected to organize the delivery network in the south F lorida area compris ing Broward and Miami Dade counties. This was called the South Florida Community Care Network (SFCCN) (Davis, 1997; Duncan et al., 2007; Schiller e t al., 2010). This network was adopted as a health system based PSNs under the F lorida Medicaid demonstration program. The other physician based PSNs were derived from the new model of networks, known as the minority physician network (MPN) (Lemak, 2004; J ohnson et al., 2010). The aims of the MPN were to allow racial and ethnic minority physicians to participate in the Medicaid

PAGE 28

28 program, to provide local care management services to Medicaid enrollees, and to lower cost of healthcare services (Johnson et al., 2010). Provider Service Networks (PSNs) In 2006, Florida started the Medicaid Reform Demonstration program. The Florida Reform initiative included PSNs among the managed care health plans as a way to improve quality of care and patient outcomes at the low est possible cost (Duncan et al., 2007; Johnson et al., 2010). PSNs are provider led organizations, which are owned by a healthcare provider, group of affiliated providers, or a public agency, whose goals are to eliminate costs of a third party health plan , to manage utilization and expenditures of health services care without having to reduce needed care, and to improve quality of care (Duncan et al., 2007; Johnson et al., 2010; Schiller et al., 2010; Florida Statues, 2012). That is, healthcare providers w ill deliver care to beneficiaries efficiently through a coordinated continuum of care in PSNs (Davis, 1997; Schiller et al., 2010). As of 2010, 180,859 Medicaid beneficiaries were enrolled in PSNs and represented 6.5 percent of the total Medicaid program ( Florida S enate, 2010) . ACO like Organizations PSNs are all not for profit entities owned by physician only network s or health system s, including physicians, hospitals, and other healthcare providers and operated only in Florida. PSNs are ACO like organiza tions since both organizations have the same structure, ownership of organizations, and aims to create a healthcare delivery system. Like PSNs, ACOs , as outlined in the Patient Protection and Affordable Care Act of 2010 , are networks of providers and hospi tals, developed for controlling the costs of care, improving the coordinated continuum of care, and being accountable for the overall costs and quality of care ( Devers et al., 2009; Keiser, 2009; Kocher, 2010;

PAGE 29

29 Shortell et al., 2010; Salmon et al., 2012). A COs give healthcare providers financial incentives to cooperate and use a shared savings model in which providers receive regular fee for service payment, but qualify to share in any savings resulting from cost reduction and meeting predetermined performance and utilization targets to reduce healthcare expenditures by avoiding unnecessary healthcar e services (D evers et al., 2009; Keiser, 2013 ). Physician led and Hospital led ACOs Shortell et al. (2008) introduced five different types of existing organizations that could serve as an ACO. Those organizations are multispecialty group practice, hospit al medical staff organization, physician hospital organization, interdependent practice organization, and health plan provider network. This means ACOs will be mainly controlled by physicians and hospitals. The primary differences between the two ACOs are that in physician while in hospital led ACOs , hospitals employ physicians (Kocher et al., 2010). There may be different characteristics depending on who controls the ACOs. Physician led ACOs are based on an association of physicians in multiple practices with required clinical, administrative, and fiscal cooperation (Kocher et al., 2010). To organize this group efficiently, physician led ACOs need strong leadership, organizational culture of perfo rmance improvement, and enough patients aggregated across individual practices to support investments in information technology and care management systems (Shortell et al., 2008). Hospital led ACOs have more resources, including electronic medical records , healthcare providers, and quality improvement activities relative to physician led ACOs (Rittenhouse et al., 2004; Mehrotra et al., 2006). For this reason, they are more likely to redesign care processes, take

PAGE 30

30 advantages of economies of scale, and make t he changes needed to improve care (Crosson, 2005; Shortell et al., 2008). There is some evidence that the more integrated forms of ACOs provide higher quality of care on health promotion programs, selected preventive care , and evidence based care managemen t processes, including diabetes and asthma management , with lower healthcare expenditures when compared to the less integrated forms of ACOs (Fuchs, 2005; Gillies et al. 2006). Moreover, the more integrated ACOs have less cost of care and fewer number of h ospital days, number of intensive care days, hospital costs, and physician costs for Medicare patients than for patients in other settings (Shortell et al., 2010). Organizational Structures: Physician based and Health system based PSNs According to previ ous research on ACOs, the potential impact of PSNs can be anticipated. There are two different types of networks physician based and health system based PSNs included in Medicaid managed care organization under the Florida Medicaid Demonstration progra m. Safety net hospitals and large physician group s such MPNs that predominantly serve Medicaid recipients , were the origins of PSNs (Johnson et al., 2010). Both PSNs are reimbursed on a FFS basis and shared saving model (Hall et al., 2013) , which means bot h have similar characteristics in terms of financial and managerial mechanisms, including payment of providers and payment for quality improvement (Hall et al., 2013). However, organizational structures may be different depending on who will lead these org anizations. Physician based PSNs are physician only network s , which are similar to physician controlled ACOs and independent practice association s (IPA s ). IPAs are an association of independent physicians, or groups of practic ing physicians who contract w ith hospitals and other providers (Grumbach et al., 1998). Although IPAs hold

PAGE 31

31 traditional practice autonomy and manage their own offices , unlike integrated health plans, IPAs serve as a corporate structure for negotiating and administering HMO contracts for its physician members (Shenkin et al., 1995; Grumbach et al, 1998; Tollen, 2008). Health system based PSNs are network s of healthcare providers that arrange to provide a coordinated continuum of services (Shortell et al., 2000). Health system based PSNs are led by health care providers , including hospitals, physician groups, outpatient clinics, ambulatory care center s , and nursing home s ( Hall et al., 2013). system based PSNs strongly resemble the integrated delivery system (IDS) of the 1990s , which are the organizations that combine healthcare providers into a vertically and horizontally integrated organization (Robinson et al., 1996; Robinson, 1997). Integration is an organizational structure to achieve a continuum of the healthcare services and interrelationship of delivery system. PSNs and IDSs create a care continuum and involve horizontal consolidation of hospitals; both may also create vertical integration of hospitals, physicians, and providers of post acute care (Burns et al., 2012). In addition, health system based PSNs provide more efficient care, since they have access to more resources in terms of workforce, health information technology systems, and financial capital than physician based PSNs (Kocher et al., 2010). PSNs vs. Other Managed Care Plans There is existing literature that examin es the impact of PSNs on utilization, expenditures of health services, and sat isfaction with care relative to HMOs and MediPass. The results related to utilization depend on the types of health services. Enrollees in PSNs are less likely to use health services than enrollees in HMOs and

PAGE 32

32 MediPass (Vogel et al., 2004; Johnson et al., 2010). But PSNs have a higher level of inpatient use and ED visits compared to MediPass. For expenditures, the fiscal analysis in PSN plan s seem to result in reductions in expenditures of services, followed by reductions in office visits and prescription d rug uses, compared to MediPass and HMOs (Vogel e t al., 2004; Harman et al., 2013 ). According to Hall et al. (2013), parents of children who are in PSNs are more likely to have a better experience in their health plans and higher satisfaction with their car e when compared to parents of children in mixed as far as managing the uses, expenditures of services, and experience in health plans by Medicaid enrollees. In general, Medica id beneficiaries in PSNs have lower use and spending of health services and better experiences with their doctors, health plan, and specialty care relative to beneficiaries of HMOs and MediPass. Integrated Delivery System Integrated delivery systems (IDS) a network of healthcare providers and organizations that provides or arranges to provide a coordinated continuum of services to a defined population and is willing to be held clinically and fiscally accountable for the clinical outcomes and health sta tus of the population served According to Robinson (1996, 1997), IDS s that deal with physician hospital combine physicians and hospitals into a vertically or horizontally integrated organization with a single ownership structure, a single chain of authority, and a single bottom line. Physicians are employed through the IDS and receive patients through this single system. Patients can move from outpatient to inpatient to sub acute to home health services freely (Robinson, 1997). IDS s can also negotiate contracts with payers and act as a center for managed care delivery system s. IDS s are an open system that

PAGE 33

33 integrates administrative and governance structures, information systems, care coordination, and financing (Suter et al., 2009). Integration describes two strategies vertical and horizontal integration. Vertical and Horiz ontal Integration Vertical integration incorporates different stages of productions or services and involves the successive stages of service delivery by a single organizational entity. Vertically integrated organizations in a supply chain are united throu gh a common owner (Snail et al., 1998; Robinson, 2001; Wang et al., 2001; Singer et al., 2010). A provide different levels of care. Its goal is to increase efficienc y, enhance coordination of care, and provide the one stop services preferred by managed care payers and their enrollees (Conrad et al., 1996). Vertical integration decisions in healthcare are based on economies of scale, risk bearing ability, transaction c osts, and the capacity for innovation in a way to manage healthcare organizations (Robinson, 1996). Horizontal integration means to merge several locations of the same production stages and occurs when different medical specialists collaborate on a single patient population by mechanisms such as case managers, multi disciplinary guidelines , or one medical/nursing record (Snail et al., 1998; Robinson, 2001). Horizontal integration occurs when a firm is being taken over by, or merged with, another firm that i s in the same industry and in the same stage of production as the merged firm (Burns et al., 2002). Hospital mergers, chains, or alliances are prominent examples of horizontal integrations. System hospitals are more efficient than independent hospitals in terms of economics of scale and scope and they have greater benefits in reputation and cost reduction (Robinson, 1996; Snail et al., 1998).

PAGE 34

34 Impact of Integrated Delivery System In PSNs, the providers (physicians, hospitals, and healthcare systems) coordin improvement activities (Hall et al., 2013). However, as mentioned above, it is expected that the two types of PSNs (physician based and health system based PSNs) have different organiza tional structures, so the different structures that are related to more or less integrated delivery systems contribute to the different impacts of both PSNs on utilization, expenditures , and outcomes by Medicaid beneficiaries. A number of studies have sho wn that the association between integrated delivery systems and utilization, expenditures, and patient outcomes is mixed (Lin et al., 1999; Wang et al., 2001; Wan et al., 2001; Kodner et al., 2002; Lee et al., 2002; Wan et al., 2002; Newhouse et al., 2003; Armitage et al., 2009; Crosson, 2009; Enthoven, 2009 ; Baker et al., 2014 ). Some studies found that integrated care is positively correlated with improved healthcare quality and efficiency, which is achieved through the coordination of care among specialis ts and the effective use of information technology (Kodner et al., 2002; Armitage et al., 2009; Crosson, 2009; Enthoven, 2009). On the other hand, previous study reported integrated delivery systems may be used in ways that do not benefit patients. The mor e integrated delivery systems have the potential to increase the market power of providers. Hospitals could employ or contract with physicians and encourage them to provide unnecessary admissions, diagnostic testing, and outpatient services (Baker et al., 2014) . Efficiency and Patient Outcomes. IDSs have more potential to provide access to a coordinated continuum of care and appear to be associated with higher efficiency due to economies of scale in purchasing, transacting, and managing activities than are

PAGE 35

35 found in less integrated systems (Lifton, 1996; Wang et al., 2002). Newhouse et al. (2003) researched the relationship between service integration and patient outcomes using the Health Service Cost Review Commission 1997 , plus the 1998 statewide inpatient data. The results were that service integration influenced lower readmission rates, length of stay, and cost of care, although only readmission rate wa s statistically significant. On the other hand, Lee et al. (2002) studied relationships between structur al integration and average total charge, and between average total charge and surgical structural integration had a higher average total charge and had a positive influence o n surgical outcomes, including the in hospital mortality ratio. Baker et al. (2014) investigated the impact of vertical integration on hospital prices, admissions, and spending for privately insured patients, using hospital claims from Truven Analytics Mar ketScan for the period of 2001 to 2007. The findings were that vertical integration was associated with increases in hospital prices and spending and related to a lower rate of hospital admissions. Therefore, the results regarding on the relationship betwe en integration and outcomes are inconsistent. Conclusions Florida implemented the Medicaid Reform Demonstration in 2006 to increase market competition, which would result in better access to care, higher quality of care , and efficiency. There are several components of the Florida Medicaid Reform plans, and an incentive program to encourage healthy b choice of health plans, beneficiaries in some reform counties, including Broward, Duval, Baker, Clay, and Nassau counties, can choose between PSNs and HMOs for their

PAGE 36

36 health plans, while enrollees in non reform counties can have their choice of health plans between MediPass and HMOs. Among several health plans, this study focuses on PSNs, especially the two different types of PSNs physician based PSNs and health system PSNs, and the impact of the structural differences on the healthcare process and outcomes. As men tioned above, the structural differences between both were explained by the concept of integration. Given the previous findings in the literature, there is reason to believe that healthcare service utilization, expenditure, and outcomes for Medicaid benef iciaries vary depending on the types of PSNs in which they were enrolled, but the mechanisms through which these PSNs play a role in the market place under the Florida Medicaid reform and how processes to serve healthcare services are influenced by structu ral changes are unclear. Previous insufficient and inconsistent findings related to the relationship between structural integration and outcome, including healthcare utilization, spending, and outcomes are subject to limitations due to (1) lack of research on distinguishing the types or level of integration and their outcomes, (2) variation of measurement and data , and (3) lack of consistent theoretical and conceptual frameworks to guide the research.

PAGE 37

37 Figure 2 1. Florida Medicaid Reform Areas ( http://www.fccmh.org/resources/docs/medicaid_reform_and_managed_care_ _4 16 07.pdf )

PAGE 38

38 Table 2 1. Characteristics of Physician based and Health S ystem based PSNs Physician based PSNs Health System based PSNs HMOs MediPass (PCCM) Financial Mechanism Payments Fee for Services and Shared Savings model Fee for Services and Shared Savings model Capitation Per member per month (PMPM) Managerial Structure Reform counties Yes Yes Yes No Area Served Florida only Florida only Multi state Multi state Geographic Orientation Local and Regional Local Local and Regional Local and Regional Ownership Not for profit Not for profit For profit/ Not for profit For profit/ Not for profit Mission Medicaid only Medicaid only Diversified, Medicaid, Govt payers Medicaid only Organizational Structure Provider led Physician network Health system Primary care physicians Primary care medial home base Yes No No Yes Plan Name in Broward and Duval Counties Access Health Solutions, Florida NetPass, Pediatrics Associates Better Health Medical Services, First Coast Advantages, South FL Community Care Network

PAGE 39

39 CHAPTER 3 CONCEPTUAL FRAMEWORK AND HYPOTHESIS This chapter discuss es the conceptual framework and hypotheses examined in this study. First, this chapter describe s theoretical implications of integrated delivery system, which is helpful to understand the structural differences between physician based PSNs and health system based PSNs. Organization theory was selected to describe the differences of the two types of PSNs in terms of organizational structure and characteristics. Next, the conc eptual framework for assessing the impact of PSNs on utilization, expenditure of healthcare services, and outcomes is introduced. Finally, this chapter discuss es hypotheses formulated from theoretical and conceptual frameworks. Theoretical Implications Agency Theory In economics, the principal agent problem treats the difficulties that arise under conditions of incomplete and asymmetric information when a principal hires an agent. Based on the seminal paper by Jensen (1976) and later Fama and Jensen (198 3), agency theory is directed at the principal agent relationship, in which one party (the principal) delegates work to another (the agent), who performs that work. Inherent in all contracts, the principal and agent often have conflicting interests in orde r to maximize their individual goals. Moreover, there often exists information asymmetries between the parties, where the principal has less information than the agent. Given that all individuals are bound in their rationality to make decisions, both parti es will undergo costs to align incentives. Agency costs are the sum of all costs to align these incentives.

PAGE 40

40 behavior), bonding costs (the costs an agent will incur to prove i ts honesty to the principal); and residual loss (the welfare loss to the principal by the misalignment of incentives). The traditional view of the physician hospital relationships (dual structure of authority) best explains the agency theory in that the h ospital (principal) contracts or hires the autonomous physician (agent) to recruit patients and perform high quality clinical services. To realign incentives and reduce inefficiencies, agency costs have been spent and various models have been developed. Th ese include Integrated Delivery Systems (IDS Vertical integration) and Contractual Networks (Virtual integration). According to Robinson (1997), IDS models take on a multi divisional structure, which includ e a corporate umbrella and a set of subsidy divi sions, including a physician division, one or more hospital divisions, and other divisions for services such as nursing homes or home healthcare. According to agency theory, as physicians become organizationally and financially dependent on the IDS and hos pital and are paid through tighter contracts, their incentives should better align and impel physicians to become more cost conscious. This would then lead to more efficient and profitable behavior within the hospital setting, ultimately leading to reduced patient costs and better quality. The logic behind PSNs is that physicians, hospitals , and other organizations are encouraged to coordinate care and grow by merging their operations with one another to align incentives and reduce inefficiency. Transactio n Cost Economics Coase (1937) and Williamson (1981) introduced TCE to describe why firms exist. The key decision relative to the cost of coordination, personnel and planning of any production, was to determine whether to make the transaction in house or bu y it in the

PAGE 41

41 market. These costs rise with transaction frequency, complexity, the degree of uncertainty in a market or environment, and asset specificity (site, physical and human asset s ). Firms will make a decision to buy a good or service on the market or produce it in house according to which provides the lowest transaction cost and greatest efficiency, thereby impacting the size, behavior, and governance structure of that firm. Robinson (1997) described the motives to vertically integrate through the TCE theory as a way to reduce transaction costs. A vertically integrated delivery system would have greater streamline efficiency , which would reduce costs and would result in better quality and performance. Vertical integration, which internalizes buyers and sellers within the same organizations, is a logical strategic response to rising tra nsaction costs (Williamson, 1981 ). In healthcare, given the increased intensity and complexity of exchange between doctors and hospitals and the increasing possibility of losing physicians to competing organizations, as often occurred in the 1990s, TCE could explain why hospitals seek to acquire admitting and referring physician practices. Under heightened uncertainty and the possibility of greater opportunistic behavior, T CE might also explain why providers and managed care companies would turn to vertical integration as a means by which to manage more efficient exchanges between them. The logic behind PSNs is that physicians, hospitals , and other organizations are encoura ged to coordinate care and grow by merging their operations with one another. TCE would explain such growth by suggesting that increased coordination and merging reduces the transaction costs of doing business. As providers become more financially accounta

PAGE 42

42 and better coordinating services. P promotes greater communication and ef ficiency among workers. However, by continuing to contract out processes and remain small in size, provider organizations would have both lower economies of scale and greater transaction costs. This essentially occurs due to reduced communication and coord ination, which are partially offset by transaction costs to reduce these inefficiencies. In short, the transaction costs for merging and producing services in house are far less than contracting out services to other providers. Resource Dependence Theory R ecourse Dependence Theory (RDT) (Pfeffer and Salancik, 1978) is concerned with how organizational survival is related to the extent that they effectively manage their environmental demands or acquire essential resources. RDT also proposes that organization s will depend on other organizations to gain essential resources when they have limited resources (Pfeffer and Salancik, 1978). RDT proposes that actors lacking in essential resources will seek to establish relationships with (i.e. , be dependent upon) othe rs in order to obtain needed resources. Also, organizations will attempt to alter their dependence relationships by minimizing their own dependence or by increasing the dependence of other organizations on them (Pfeffer and Salancik, 1978). RDT suggests th at managing the exchanges and relationships with interdependent organizations may be more important to survival than managing production efficiencies. One strategy to maximize the power of an organization is integration. Integration is defined by combining previously separate and independent functions, resources, and organizations into a new and united structure (Shortell et al., 1996 ; Lin et al., 2001). More specifically, vertical integration is viewed as a strategic mechanism organizations might need in o rder to cope with increasing interdependencies more than the better management of

PAGE 43

43 internal processes or the costs of exchange. Also, it allows for consolidation of resources, increased interdependency , and , ultimately , greater control over uncertainty beca use of cooptation and resource allocation control. In accordance with the RDT, organizations try to minimize their dependence and maximize their market share to be successful organizations (Pfeffer and Salancik, 1978). RDT would explain the benefits of PS Ns by suggesting that, given new incentives, cost and reimbursement structures, increased coordination , and merging will reduce organizational dependence and maximizes their independence. Compared to physician based PSNs, health system based PSNs are more integrated organizational structure s and tend to have more resources , such as a healthcare workforce, healthcare technology, and a larger financial budget. Therefore, health system based PSNs have more organizational independence than physician based PSNs, which means they are available to align healthcare facilities, programs, or services and offer a coordinated continuum of healthcare to Medicaid beneficiaries. Through these organization theories, this study makes the assumption that these two organizati ons have different organizational structures and characteristics, which impact the process and outcomes of delivering healthcare service to Medicaid beneficiaries. Conceptual Framework for Provider Service Networks Previous research investigating the distinguishing characteristics between the two types of PSNs was highly restricted in scope and depth. However, according to some recent literature on ACOs, PSNs resemble ACOs in terms of goals and focus, which was to reduce healthcare spending and maintai n quality of care at minimal cost, as well as in terms of the organizational structure and ownership (Shortell et al., 2008 ;

PAGE 44

44 Salmon et al., 2012 ). The outcomes of ACOs controlled by physicians or hospitals are different due to their internal resources and organizational structures (Rittenhouse et al., 2004; Crosson, 2005; Gillies et al., 2006; Mehrotra et al., 2006; Shortell et al., 2008; Kocher et al., 2010). For this reason, we expect that there will be significant differences in healthcare utilization, s pending, and health outcomes by Medicaid enrollees in the two different types of PSNs. As previously mentioned, however, there is limited research on the impact of PSNs on the process and outcomes of healthcare services delivery. To address the limitations in empirical studies, organization theoretical perspectives utilizing the resource dependence theory, agency theory, and transaction cost economics are used to identify the organizational structures between the two models of PSNs as structur al indicators. Process Outcomes (SPO) model to investigate the impact of specific Medicaid health plans on utilization and expenditure of healthcare services in Florida reform counties and the association between the proces s and outcomes of healthcare services within the two different types of PSNs. Therefore, this study use s theoretical and conceptual explanations for how these two models of PSNs are different and how they ultimately affect the healthcare utilization, expenditures, and health outcomes of Medicaid beneficiaries enrolled in physician based PSNs and health system based PSNs . This study use s managed care plan is related to utilization, expenditures of healthcare services, and , overall healthcare, personal

PAGE 45

45 doctor, and specialist . Donabedi an (1966) posited three categories of organizational assessment: structure, process, and outcomes. The SPO model is a function of the components that relate to the organizational characteristics of healthcare providers (structure), the clinical and adminis trative procedures involved in the delivery of healthcare to patients (process) and the effect of care on the health status of hospital patients (outcomes) (Donabedian, 1966, 1988). These three categories of measures are interdependent and are linked in an underlying framework. The relationship between the new health plan structure and the process of healthcare services is important because processes can affect patient outcomes, such as health status and consumer satisfaction. Good processes should be drive n by good structure and good outcome should be derived from good processes (Donabedian, 1998; Zinn et al., 1998). For Medicaid managed care organization, Schiller et al. (2010) adopted the SPO model to examine the effect of Provider Sponsored Organizations (PSOs) on patient assessment of care and self reported utilization relative to HMOs and PCCMs in Florida. This model illustrates the derived conceptual framework and describes the mechanisms through which the structural characteristics of the organization impact s unit level utilization and Structure f actors. According to the SPO model from previous research (Donabedian, 1966, 1988; Schiller et al., 2010), structure refers to organizational resources for delivery of healthcare services, such as several Medicaid managed care (Figure 3 3) , the organizational structures examined are physician based and health system based PSNs. Medic may

PAGE 46

46 expenditures in reform counties. Process f actors. The processes refer to the actual process of delivering ca re, 1966, 1988, 1992). Indicators that measure the actual process of delivering healthcare services are inpatient days/utilization, outpatient utilization, prescription dr ugs use, and PMPM expenditures (Donabedian, 1966, 1988). Process measures need to be adjusted by case mix since the process in different types of organizational structures will depend race/ethnicity, residence in rural or urban areas, and health status (Zuckerman et a l., 2002; Schiller et al., 2010 ). These variables influence experiences of healthcare utilization and expenditures in this study. In the d contro l driving factors that affect their use of healthcare services and their health outcomes, factors are environment al , predisposing, and enabling characteristics (Philip e t al., 1998, Figure 3 1). The Andersen (1995) behavioral model for this study was used to adjust the association between the new Medicaid managed care plans and Medicaid process and outcomes, so the traditional population characteristics, includ ing predisposing and enabling factors, will influence the outcomes indicators. For this study, the predisposing and enabling factors are age, gender, race/ethnicity, geographic residence, type of eligibility, and health status (risk score) . Outcomes f actors. Finally, outcomes consist of measures of patient health, including patient outcome measures (Donabedian, 1966, 1988, 1992). Outcome

PAGE 47

47 quality of life (Donabedian, 1 988; Zinn et al., 1998). For this study, outcome measures are with their health plans , overall healthcare, personal doctor, and specialist , which is consi stent wi th prior empirical studies that examined the influence of Medicaid managed care organization s outcomes and investigated the differences in rating and report of care between enrollees in HMOs and those in PSNs (Schiller et al., 2010; Hall et al., 2013). Hypotheses Hypotheses were formulated to evaluate the impact of two different models of utilization, spending of healthcare and satisfaction with health plans , overall healthcare, personal doctor, and specialist in reform counties using the SPO conceptual model to describe the relationships between these concepts. Using the Florida PSNs framework , s SPO model, hypotheses were developed as follows. The Association of Organizational Structure and Utilization Like IDS, health system based PSNs are networks established and operated by groups of affiliated healthcare providers . Based on arguments by Kautz et al. (2007), Robinson et al. (1996), and Shortell et al. (1993), we expect ed that there would be variations in Medicaid enro based and health system based PSNs. Particularly, networks that are more integrated will be more likely to improve quality of care and reduce unnecessary healthcare utilization and spending. Therefore, S pecific Aim 1 assesses the differences in use of healthcare services between enrollees in physician based PSNs and those in health system based PSNs over time. Specifically, this study will measure emergency department (ED)

PAGE 48

48 utilization, inpatient hospital utilization, and prescription drug use separately as healthcare utilization indicators. According to previous systematic review (Armitage et al., 2009), several studies reported a reduction i n emergency department visits and inpatient length of stay after integrating delivery systems. Integrated medical groups in California reported a higher level of clinical quality and more health promotion programs when compared with independent physician associations (IPAs) (Robinson et al., 2010). Therefore, the follow ing hypotheses are postulated. Hypothesis 1 . ED utilization by Medicaid enrollees in health system based PSNs will be less than ED utilization by Medicaid en rollees in physician based PSNs . Hypothesis 2 . Inpatient utilization by Medicaid enrollees in health system based PSNs will be less than inpatient utilization by Medicaid enrollees in physician based PSNs. Hypothesis 3 . Prescription drug utilization for Medicaid enrollees in health system based PSNs will be more than prescription drug utilization for Medicaid enrollees in physician based PSNs. The Association of Organizational Structure and Expenditures Given prior findings in the literature review that more integrated organizations are more likely to reduce healthcare spending compared to less in tegrated organizations (Lifton, 1996; Backer et al., 2014 ), this study expects to find variations in expenditures between the two models of PSNs. We also anticipate that PMPM Medicaid expenditures in physician based PSNs will be more than expenditures in h ealth system based PSNs , since less integrated organizations have higher expenditures and costs based on agency theory and transaction cost economics . Therefore, Specific Aim 2 investigate s the differences in PMPM Medicaid expenditure between physician bas ed and health system based PSNs. The hypothesis is postulated as follows.

PAGE 49

49 Hypothesis 4: PMPM expenditures for Medicaid enrollees in health system based PSNs will be less than PMPM expenditures for Medicaid enrollees in physician based PSNs. The Associati on of Structure, Process, and Outcomes processes of patient care, which in turn affects patient outcomes. We anticipate d that the different organizational structure s would affect patient outcomes, such as specialist. In addition, given the different types of PSNs in Medicaid managed care, depending on use and spending of healthcare services between physician based and health system based PSNs. Therefore, Specific Aim 3 investigate s how process of care affects outcomes in the health system based PSNs compared to physician based PSNs over ti me. Therefore, it is postulated from this aim that: Hypothesis 5 : Medicaid enrollees in health system based PSNs will have a higher satisfaction with health plans , overall healthcare, personal doctor, and specialist when compared to those in physician bas ed PSNs.

PAGE 50

50 Figure 3

PAGE 51

51 Figure 3 2. Florida PSNs Framework

PAGE 52

52 Figure 3 Model

PAGE 53

53 CHAPTER 4 DATA AND METHODS T his chapter first describe s the study design and the data source. The next step is to define and measure the variables. Finally, this chapter include s a discussion on econometric methods for statistical analysis. This proposal is aimed at defining the effect of the new Medicaid delivery system on healthcare utilization and expenditures amon g Medicaid enrollees. To calculate access to care and expenditures, this study was conducted using a retrospective, longitudinal study design over a 72 month period of time from July 2004 through June 2010 . Also, this study analyze d the association between healthcare process and outcome indicators using a survey data set from 2006 to 2008 . Study Design For Specific Aims 1 and 2, this analysis use s a quasi experimental and pre post study design, which in economic terms is a difference in difference approach. This difference in difference (DID) methodology, which was similarly employed by Tai Se ale et al. (2001), measures the comparison of changes in utilization and expenditures by Medicaid enrollees before and after redesign ed Florida Medicaid managed care , which utiliz ed the two types of PSNs applied in Broward and Duval counties. Under the assumption that there are no factors other than the Demonstration of both types of PSNs or that the effect of other factors do not vary between the two different types of PSNs, the result is a DID estimate that can be identified as the effect of the Demonstration on use and expenditures of services between Medicaid enrollees in health system based PSNs and those in physician based PSNs. In the comparisons between pre and post reform , MediPass (PCCM in Florida) enrollment is a reasonable

PAGE 54

54 proxy of PSN enrollment for the pre reform periods, since the two reform counties had minimal PSN market penetration before Medicaid reform and both are paid on a FFS basis . For this study , the changes in utilization and expenditures between the baseline periods, which is defined as the two fiscal years before policy implementation, and the six years of the demonstration in the health system based PSNs are measured and compared with changes in the physician based PSNs in Broward and Duval counties over the same time period. By measuring the DID, we can begin to examine whether the Demonstration program had a meaningful impact on the difference in health service use and healthcare expenditure s between health system based PSNs and physician based PSNs. In addition, for Specific Aim 3, this study again u ses a DID approach to assess with their health plans, healthcare, personal doctor, and specialist between enrollees in physician based PSNs and health system based PSNs in Broward and Duval County from 2006 to 2008 . The changes in was in 2006 , before implementation of reform, and two years of the demonstration in the health system based PSNs were calculated and compared with changes in enrollees satisfaction in the physician based PSNs over the same time period.

PAGE 55

55 Data Sources and Em pirical Models The purpose of this study was to determine whether unnecessary healthcare utilization and expenditures for Medicaid enrollees decrease and whether Medicaid , overall healthcare, personal doctor, and specialist improve after the implementation of PSNs in Florida Medicaid managed care. The study use s secondary data Florida Medicaid medical, facility, and pharmaceutical cl aims data from 2004 through 2010 (FY0405 FY0910 ) and Consumer Assessment of Healt hcare Providers and Systems (CAHP S) survey data from 2006 to 2008 , with Medicaid beneficiaries from Broward and Duval counties who were enrolled in at least one of the two types of PSNs after the policy change and MediPass before the policy implementation. These data were then manipulated to produce the analytic datasets required for the statistical analyses designed to measure the effect between the two types of PSNs on utilization and expenditures, and to investigate the impact on ion. were taken from the Medicaid Enrollment database. For Specific Aim 1 and 2, the unit of analysis is a member month, which is equal to one member enrolled in a Medicaid managed care for o ne month, whether or not the member actually receives any services during the period. For Specific Aim 3, the unit of analysis is a member year. Florida Medicaid Claims and Eligibility Data Medicaid administrative data were obtained from the Florida Agenc y for Healthcare Administration (AHCA). The claims data result from administering healthcare delivery, enrollment into health insurance plans, and reimbursement for services (Iezzoni, 1997). The study data set covers a 72 month p eriod from July 2004 June 2010 to examine the influence of the policy. The study population includes all Medicaid

PAGE 56

56 beneficiaries who lived at least 1 month in Broward and Duval counties during the study period. The study samples are enrolled in MediPass (PCCM) between July 2004 and June 2006 (FY0405 FY0506) and were enrolled in physician based PSNs and health system based PSNs betw een July 2006 and June 2010 (FY0607 FY0910 ). Inclusion criteria were families with children receiving support under the Temp orary Assistance for Needy Families (TANF) program and individuals receiving the Supplemental Security Income (SSI) benefits , which include many of the elderly, blind, and disabled with low income, and pregnant women. Individuals with eligibility based on TANF and SSI are required to take part in the demonstration , while individuals, including dually eligible and pregnant women , were not considered mandatory participants and could choose between HMOs and PSNs (Harman et al., 2011 ). Exclusion criteria were v oluntary participants and children who received services through a special program for children with special healthcare needs. Voluntary participants were pregnant women and all persons 65 years and older due to dual enrollment with Medicare . CAHPS Data T he CAHPS survey measures health plan performance in several dimensions , including ratings of overall healthcare, health plan, primary doctor or nurse, specialty care , and reports of experiences with using a health plan and healthcare services (Farley et al ., 2002; Quilgley et al., 2003; Keller et al., 2005). The aims of CAHPS were viewpoint (Fox et al., 2001; Farley et al., 2002). The response scales of each rating range d from 0 to 10. For this study, we measured outcome indicators using

PAGE 57

57 overall healthcare, personal doctor, and specialist. Health plan rating: Using any number from 0 to 10, where 0 is the worst and 10 is the best health plan possible, what number would you use to rate your health plan? Health care rating: Using any number from 0 to 10, where 0 is the worst and 10 is the best health care possible, what number would you use to rate your health care? Personal doctor rating: Using any number from 0 to 10, where 0 is the worst and 10 is the best personal doctor possible, what number would you use to rate your personal doctor? Specialist rating: Using any number from 0 to 10, where 0 is the wors t and 10 is the best specialist possible, what number would you use to rate your specialist? The enrollee list generated from AHCA member month and recipient eligibility files were randomized and used to create the survey sample. The dataset from AHCA included Medicaid ID, enrollee name, demographics, county, eligibility status, and health plan na me (Duncan et al., 2007). Plan members were randomly sampled using probability sampling to confirm that members were a representative sample from each Medicaid managed care plan (MediPass, HMOs, and PSNs) (Hall et al., 2013). Medicaid enrollees in the orig inal Demonstration counties, Broward and Duval, for at least six consecutive months were randomly selected to participate in a 20 minute telephone based CAHPS survey. The survey was completed by the University of eau of Economic and Business Research. Empirical Models by Hypotheses Based on hypotheses 1, 2, 3, 4, and 5 in Chapter 3, there are five empirical models to examine the impact of the new healthcare plan on utilization, spending, and n .

PAGE 58

58 Model 1 . ED visits = f [ time, pre /post reform, physician based/health system based PSNs, demographic characteristics, geographic location, and risk scores] Model 2 . Inpatient hospital utilization = f [ time, pre /post reform, physician based/health sys tem based PSNs, demographic characteristics, geographic location, and risk scores] Model 3 . Prescription drugs = f [ time, pre /post reform, physician based/health system based PSNs, demographic characteristics, geographic location, and risk scores] Model 4 . Per Member p er Month (PMPM) expenditures = f [ time, pre /post reform, physician based/health system based PSNs, demographic characteristics, geographic location, and risk scores] Model 5 . Satisfaction with health plan, healthcare, personal doctor, and specialist = f [ time, pre /post reform, physician based/health system based PSNs, demographic characteristics, geographic location, and risk scores] Measures for Data This section outline s the operationalization of variables incorporated in the study model, including the outcome measures, PSN measures, and control variables that includ e population characteristics and health status of Medicaid enrollees. The dependent and independent variabl es and covariates are derived from the Florida Medicaid facility, medical, and pharmacy claims data and CAHPS survey data . Table 4 1 summarizes the measurement and operationalization of these variables in our study. Outcome Variables Ultimate outcome varia bles . To assess ultimate outcomes indicators, two questions in the CAHPS were their health plans and healthcare services (healthcare, personal doctor, and specialist) . Ratings of satisfaction with health plans and healthcare services are 0 to 10, which represents 0 if beneficiaries are not satisfied with their health plans and 10 are most satisfied. Because of the skewed distribution of CAHPS scores, the responses were

PAGE 59

59 collapsed into three categorie s using the cut points identified by Hall et al. (2013). R atings were categorized into the following groups: high score (9 10), medium (7 8), and low score (6 and below). Inter mediator outcome variables . Utilization included per member per month ED , phar macy utilization, and inpatient hospital utilization. This study measure d those variables, following previous literature (Lemak et al., 2005). ED visits were analyzed by the total number of emergency room visits, while inpatient utilization was analyzed by the total number of inpatient admissions and the total number of inpatient days. Prescription drug uses were measured in two ways: total number of pharmacy claims and number of generic pharmacy claims. To calculate pre reform baseline expenditures, we collect ed all inpatient, outpatient, medical, and pharmacy claims paid amounts in MediPass per member per month (PMPM) payment amounts for all Medicaid enrollees who lived at least one month in Broward and Duval County and were in an eligibility category that would have made them participate mandatorily in the demonstration from July 2004 to Ju ne 2006 (Harman et al., 2009 , 2011). To calculate refo amounts in physician based PSNs and health system based PSNs for individuals who were enrolled for at least one month from September 2006 to June 2010 were included. Fo r PSNs, PMPM expenditures were the sum of all paid amounts for claims in a given month, including a monthly patient case management fee paid to PS N providers (Harman et al., 2011, 2013 ). For this study, the expenditures do not include admi nistrative costs incurred by the Florida Agency for Healthcare Administration.

PAGE 60

60 Independent Variables The primary explanatory variables were the two different types of PSNs in Florida Medicaid managed care under the Florida Medicaid Reform Demonstration program. These varia bles were coded as physician based PSNs and health system based PSNs based on their plan name. In physician based PSNs, there were several health plans, including Pediatric Associations, Access Health Solution, and Florida NetPass. Health system based PSNs were Advantages, and South F lorida Community Care Network. This variable is a binary indicator; with Medicaid beneficiaries enrolled in physician based PSNs indicat ing a value of 0 and Medicaid beneficiaries enrolled in health system based PSNs indicat ing a value of 1. Co ntrol Variables Control variables include d age, gender, race/ethnicity, residence (Broward or Duval), eligibility category, and health status. Demographic c haracteristics. Demo graphic characteristics included age, gender, race/ethnicity, and eligibility type. Age was a continuous variable, whereas m ost characteristics were categorical variables. For gender, male was the reference group in the multivariate models. The following c ategories were used to describe an race/ethnicity: Latino, Black, White , or other. White was the reference group. We included two eligibility types: Temporary Assistance for Needy Families (TANF) and Supplementary Social Security Income (SSI). T his variable was a binary variable, with TANF indicated a value of 0 and SSI indicated a value of 1. Geographic l ocation. Medicaid enrollees, who were enrolled in PSNs in Broward or Duval counties, were selected as study samples. However, there were

PAGE 61

61 potent ial culture, population characteristics, and socioeconomic status differences between Broward and Duval counties due to geographic location. Broward County is located in the southeast part of Florida , which includes Fort Lauderdale , while Duval County is l ocated in the northeast part of Florida and includes Jacksonville. Broward County has more persons who are Hispanic or Latino origin (25.8 percent vs. 7.9 percent), are foreign born (31.2 percent vs. 9.1 percent), and speak other languages than English at home (37.2 percent vs. 12.8 percent), compared to Duval County (US Census Bureau, 2011). Risk s cores. The Florida Agency for Healthcare Administration (AHCA) used the Medicaid Prescription Drug risk adjustment model to calculate risk scores to risk adjust premiums for each HMO and PSN recipient. Medicaid Prescription Drug risk adjustment However, PSNs were paid on a FFS basis. AHCA co mpared the amount the plan using the FFS payment to the amount being paid by a risk adjustment rate, since payment of PSNs change from FFS to risk adjusted capitation (Harman et al., 2009). In this analysis, risk scores are used to account for differences encounters (Iezzoni, 2003). We used data from FY0405 and 0506 as baseline and data fro m FY0607 to 0910 as follow up. Analytical Data Set Specific Aim 1 and 2 These are a total of 12,554,322 (FY0405: 2,561,004, FY0506: 2,591,865, FY0607: 1,054,817, FY0708: 1,923,457, FY0809: 2,016,745, FY0910: 2,389,434) member months included in the analytic dataset (Table 4 2) . The population from

PAGE 62

62 Broward and Duval counties for pre refor m period and for post reform period included 5,017,028 member months and 7,393,453 member months. For this study, the study population were enrollees in MediPass, physician based PSNs, and health system based PS Ns . The final cohort in MediPas s for the pre reform period included 2,544,281 member months with 15.48% eligible through SSI and 84.52% eligible through TANF (Table 4 3) . T he final cohort in physician based PSNs and health system based PS Ns for post reform period included 838,254 member months and 1, 073,434 member months , with 14.76% and 18.27% eligible through SSI and 85.24% and 81.73% eligible through TANF. Specific Aim 3 CAHPS surveys from 2006 to 2008 were used to analyze the differences in healthcare outcomes between physician based PSNs and heal th system based PSNs. The analytic sample is limited to enrollees in Broward and Duval counties. For the S pecific A im 3, the study population were enrollees in physician based PSNs and health system based PSNs, totaling 6,483. For the period prior to the r eform demonstration, 1,288 enrolled in P PSNs and 428 enrolled in health system based PSNs. For the demonstration period, 2,414 enrolled in physician based PSNs and 2,353 enrolled in health system based PSNs. Therefore, among the study population (6,483 pe rson years), 57.1% (3,702 person years) of beneficiaries enrolled in physician based PSNs and 42.9% (2,781 person years) of beneficiaries enrolled in health system based PSNs from 2006 to 2008. Survey data were weighted to enable appropriate calculation of population estimates based on plan size and non response in each county (Hall et al., 2013).

PAGE 63

63 Statistical Analysis Statistical analyses were conducted using Statistical Analysis Software (SAS). Statistical significance for these analyses was set at p < 0.05. First, descriptive statistics were conducted to determine the means, standard deviations , and ranges of all measures in our model. Next, for the difference in difference approach, differences in utilization, expenditures, and outcomes by Medicaid enr ollees before and after policy implementation were calculated for each type of PSN without considering confounding variables. We measure d the variation between differences in per member per month ED visits, inpatient hospital utilization, and prescription drug use, PMPM expenditures, and outcomes between the two types of PSNs. Based on the findings from these statistics, we conduct ed multivariate analysis to test the specific aims. Specific Aims 1 and 2 To analyze the relationship between the structural cha nge in Medicaid managed care and healthc are efficiency, this study use d Florida Medicaid claims and eligibility data from July 2004 to June 2010 , which means the autocorrelation will be problematic. Autocorrelation refers to errors in the next time period and errors in the current time period were correlated and could be tested using the Durbin Watson test statistics (Woodbridge, 2002). With longitudinal data, it is possible to have error terms that are partially due to cross sectional and time effects. Poi sson Random effect with Maximum Likelihood Estimators or negative binominal regression models can correct the error terms from both cross sectional and time effects. First, univariate analysis was conducted to find differences in utilization and expenditu res between health system based PSNs and physician based PSNs with the full sample population. These differences were calculated separately for SSI and TANF

PAGE 64

64 enrollees. These calculations for expenditures were done on the full sample and with the top 5 perc ent of observations removed to delete the influence of outlier s . However, this process was not used percent of observations removed, since the unit of analysis is monthly based, suggesting there were few ED visits, inpatient hospital utilization, and prescription drug use. After conducting descriptive analysis and univariate analysis, we use d multivariate analyses to account for changes in utilization of ED, inpat ient, and prescription drug use, and claims expenditures before and after the Demonstration while controlling for demographic characteristics (age, gender, race/ethnicity, and type of eligibility), geographic location (residence area), and risk scores between health system based and phys ician based PSNs. Thus, the overall differences in the effect of Medicaid Reform on healthcare utilization and PMPM expenditures for enrollees in physician based PSNs and health system based PSNs could be more precisely assessed. The nature of the dependent variable should be considered. Specific Aim 1 . The first research objective was to assess the differences in utilization of healthcare services ED visits, inpatient utilization, and prescription drug use between physician based PSNs and health syst em based PSNs. Hypotheses 1, 2 are ED utilization and inpatient hospital utilizatio n by Medicaid enrollees in health system based PSNs will be less than ED utilization by Medicaid enrollees in physician based PSNs, ceteris paribus. Hypothesis 3 is prescription drug use by Medicaid enrollees in health system based PSNs will be more than p rescription drug use by enrollees in physician based PSNs. Dependent varia bles of hypotheses 1 3 were the

PAGE 65

65 number of ED visits, inpatient hospital utilization, and prescription drug use s . These dependent variables are nonnegative and count dependent variabl es. There are some problems, including heteroskedasticity, if we try to estimate the model using OLS regressions, because the distribution of the count data was severely skewed with the long tail on the right side. A solution to this problem was to use a d istribution that fits the data and conduct nonlinear regression analysis. According to Woodbridge (2002), we analyze d changes in utilization between time periods and between the two groups using the negative binominal regression model. This model appropria tely reflects the distribution of utilization. This model include d , s months 1 through 72 , a dummy variable for whether the observation was from the reform period from September 2006 to June 2010 , variable for whether the observation was from health system based PSNs , which is one of options for health plan s , interaction of time, post period, and enrollment in a health system based PSN (time*post*HPSN). The estimated equation is: log(# of ED visits/inpatient utilization/ Rx drugs uses) = 0 + 1 *Time + 2 *Post 3 *HPSN + 4 *(Time*Post) + 5 *(Time*Post*HPSN) + 6 Covariates + , (4 1) 5 indicates the variation in the slope pre and post Medicaid reform was significantly different for beneficiaries enrolled in health system based PSNs from those enrolled in physician based PSNs. An estimate 5 that is less than 0 mean s that ut ilization will be declining at a greater rate among Medicaid enrollees in health system -

PAGE 66

66 based PSNs relative to the changes experienced by beneficiaries in physician based PSNs. Specific Aim 2 . The second research objective was to examine the difference in PMPM expenditures between physician based PSNs and health system based PSNs. Hypothesis 4 was that PMPM expenditures for Medicaid enrollees in health system based PSNs will be less than PMPM expenditures for Medicaid enrollees in physician based PSNs, cet eris paribus. The dependent variable of H ypotheses 4 was PMPM expenditure. Expenditure was a common example of a censored variable because it was impossible to observe expenditures that were less than zero , but , in theory , someone could have negative spending. The two part model developed for the Rand Health Insurance Experiment was accurate in accommodating expenditure data where there was a large proportion of the population that ha d no expenditures and when the distribut ion of the non zero expenditures was skewed to the right. The first equation in the two part model was estimated with a logit model to estimate the probability of the expenditure value being zero versus not being zero. However, previous literature (Harman et al., 2011 ) using the same dataset confirmed that the one part Generalized Estimating E quations (GEE) model using a gamma family displayed adequate model fit. GEE s using a gamma family with a log link were used for the second part of the model. This mode l include d 72 , a dummy variable for whether the observation was from the reform period from September 2006 to June 2010 , from health syst em based PSNs , which is one of options for health plan s choices

PAGE 67

67 , time, post period, and enrollment in a health system based PSN (time*post*HPSN). The estimated equati on is: PMPM E 0 1 2 3 *HPSN 4 * (Time *Post) + 5 *(Time*Post*HPSN) + 6 *Covariates , (4 2) 3 indicates the difference in the intercept for the period after Medicaid reform for observations from health system based PSNs compared to observations from physician based PSNs. 5 reports whether the change in the slope pre and post policy implementation was significantly different for beneficiaries enrolled in health system based PSNs than those enrolled in physician based PSNs. Sensitivity a nalysis . Two sensitivity analyses are presented in this study. These analyses are based on previous research, studying the variation in expenditures of Florida Medicaid managed care between pre reform and post reform (Ha rman, et al., 2011 ). The first sensitivity analysis was conducted on individuals who enrolled in MediPass, physician based PSNs, and health system based PSNs and contributed at least three person months over the study period . T he second analysis include d the sample who contributed at least six person months in MediPass, physician based PSNs, and health system based PSNs. T hrough these two steps of analyses, we expect ed to examine the differences in healthcare utilization and PMPM expenditures between enro llees in physician based PSNs and enrollees in health system based PSNs with more stable Medicaid enrollment. Specific Aim 3 Last, we analyze d the association between healthcare process es and outcomes indicators. Hypothesis 5 states that Medicaid enrollee s in health system based PSNs

PAGE 68

68 will have higher satisfaction with health plans and healthcare services when compared t o those in physician based PSNs. Dependent variables of H ypotheses 5 were satisfaction with health plans, healthcare, personal doctor, and specialist. Those variables were considered ordered variables, so that ordered logistic regression would be used to estimate the cumulative probability of being in one category versus all lower or higher categories. Regression models control for age, gende r, race/ethnicity, eligibility, geographic location, and risk score. This model include d a dummy variable for whether the observation was from the reform period from 2007 to 2008, r eferred to as , health system based PSNs , one of options for health plan s , post and enrollment in a health system based PSN ( post* HPSN). The functional forms are as follows: Satisfaction with health plans, healthcare, per sonal doctor, and specialist 0 1 2 3 * (Post*HPSN) + 4 (4 3) 3 show ed that the and post reform demonstration over time were significantly different for the enrollees in health system based PSNs than those in physician based PSNs .

PAGE 69

69 Figure 4 1. Timeline of Flori da Medicaid Reform Demonstration

PAGE 70

70 Table 4 1 . List of v ariables Variable Definition Type Category Data Independent v ariable PSNs Type of PSNs Categorical Physician based PSNs=0, Health s ystem based PSNs=1 Eligibility Dependent v ariable Ultimate o utcomes Satisfaction Satisfaction with health plans, healthcare, personal doctor, and specialist Categorical High=2 Medium=1 Low=0 CAHPS Inter m ediator o utcomes E D visits Number of emergency department visits Continuous Claims Inpatient Number of inpatient admissions and inpatient days Continuous Claims Prescription d rugs Number of pharmacy claims and generic pharmacy claims Continuous Claims PMPM e xpenditures Total healthcare expenditures / Number of member months Continuous Claims Covariates Age Age Continuous Eligibility Gender Male, Female Categorical Female=0, Male=1 Eligibility Race/Ethnicity Categorical White=0, Black=1, Latino=2, Other Race=3 Eligibility Geographic a rea Two reform counties Categorical Duval=0, Broward=1 Eligibility Eligibility Eligibility category Categorical TANF=0, SSI=1 Eligibility Health s tatus Risk Score Continuous Health s tatus Self reported health status Categorical Excellent = 4, Very good = 3, Good = 2, Fair = 1, Poor = 0 CAHPS

PAGE 71

71 Table 4 2. Study population by fiscal year (Aim 1 and 2) Aim 1 and 2 ( p erson m onths) FY0405 2,561,004 FY0506 FY0607 FY0708 FY0809 FY0910 2,591,865 1,054,817 1,923,457 2,016,745 2,398,434 Table 4 3. Study population by year (Aim 3) Aim 3 (Person Year) 2006 2007 2008 Unweighted N 5,767 6,209 6,152 Weighted N 139,801 131,809 116,182 Table 4 4 . Study population by plan (Aim 1 and 2) Pre reform Post reform MediPass (N=2,544,281) HMO (N=2,472,747) P PSNs (N= 838,254) H PSNs (N=1,073,434) HMO (N=5,481,765) Eligibility status SSI 15.48% 9.65% 14.76% 18.27% 11.27% TANF 84.52% 90.35% 85.24% 81.73% 88.73% Table 4 5. Study population by plan (Aim 3) Pre Reform Post Reform P PSNs H PSNs MediPass HMO P PSNs H PSNs HMO Unweighted N 1,288 428 907 2,281 2,414 2,353 6,641 Weighted N 22,655 2,247 34,045 73,031 30,482 36,728 164,007

PAGE 72

72 CHAPTER 5 RESULTS The results of this study are presented in the following sections, including descriptive analysis, univariate analysis, multivariate analysis, and results for each specific aim. The aims of this study were to investigate the difference s in emergency department (ED) utilization, inpatient hospital utilization, prescription drug use, and PMPM expenditures by Medicaid enrollees between physician based PSNs and health system based PSNs in reform counties, and t o assess the difference in self reported satisfaction with health plans, overall healthcare, personal doctor, and specialist by Medicaid enrollees between physician based PSNs and health system based PSNs in Broward and Duval counties. Specific Aim 1 and 2 Descriptive Analysis The characteristics of the study sample across the three plan types are summarized in Table 5 1. The study population was compared based on their plan types. Characteristics of the samples were significantly different across plan type s ( p <0.001). For enrollees before the reform period, among the 2,544,281 member months in MediPass, over half of enrollees indicated they were male (54.05%) and lived in Broward County (57.60%). Almost half of MediPass enrollees were black (47.48%) and mo st of MediPass beneficiaries were enrolled through TANF (84.52%). Risk scores were calculated for enrollees during the Demonstration period, which means a risk before refo rm implementation. Therefore, scores for enrollees before the reform period was assigned 0 as baseline. Among 838,254 and 1,073,434 member months in

PAGE 73

73 physician based PSNs and health system based PSNs , respectively, over half of enrollees were female (53.94% and 56.13%), black (51.94% and 59.52%), and were enrolled in TANF (85.24% and 81.73%). Enrollees in health system based PSNs were older compared to physician based PSNs (17.68 vs. 14.95) and had lower risk scores (0.10 vs. 0.12). Also, more enrollees in h ealth system based PSNs lived in Duval County than in Broward County, while more enrollees in physician based PSNs lived in Broward County than in Duval County. Univariate Analysis Table 5 2 shows the unadjusted differences in healthcare utilization and PMPM expenditures by Medicaid beneficiaries enrolled in physician based PSNs and health system based PSNs. Enrollees for the period prior to the Demonstration had higher healthcare utilization, except for prescription drugs, and had higher PMPM expenditure s relative to those in the Demonstration period. F or the period immediately after implementation of reform, average inpatient hospital utilization and ED visits for Medicaid enrollees in health system based PSNs were 0.01 higher, compared to those in physi cian based PSNs. Average prescription drug use for beneficiaries enrolled in health system based PSNs were 0.09 higher, relative to those enrolled in physician based PSNs. Enrollees in health system based PSNs were more likely to have higher average PMPM e xpenditures compared to enrollees in physician based PSNs ($155.67 vs. $133.27). Thus, there were variations in healthcare utilization and expenditures by enrollees between physician based PSNs and health system based PSNs in reform counties. Relative to e nrollees in physician based PSNs, enrollees in health system based PSN s had higher healthcare service utilization and Medicaid PMPM expenditures. When unadjusted results were estimated separately for enrollees in

PAGE 74

74 health system based and physician based PSN s in SSI and TANF, the results did n o t change. Average inpatient hospital utilization and prescription drug use for SSI enrollees in health system based PSNs were 0.01 and 0.27 higher than those for SSI enrollees in physician based PSNs. Average PMPM expen ditures for SSI enrollees also were $46 higher in health system based PSNs. For TANF enrollees, average inpatient hospital utilization and ED visits were equal to enrollees between the two different PSNs, while average prescription drug use and PMPM expend itures were 0.01 and $0.01 higher for beneficiaries enrolled in health system based PSNs, relative to those enrolled in physician based PSNs. Therefore, healthcare utilization and Medicaid expenditures for SSI and TANF by Medicaid enrollees in between heal th system based PSNs and physician based PSNs were about the same. In addition, we calculated the differences in expenditures across three different plans without the top 5 percent of expenditures to reduce the effect of outliers. The univariate results a fter removing the top 5 percent of observations were similar to the previous results using the full sample (Table 5 2). Enrollees in the pre reform period were more likely to have higher expenditures compared to beneficiaries in the post reform period. For the Demonstration period, enrollees in health system based PSNs were more likely to have higher PMPM expenditure compared to those in physician based PSNs ($35.39 vs $30.38). Multivariate Analysis This section describes the results of the four models. For the S pecific A im s 1 and 2, SAS® PROC GENMOD was used to analyze the Generalized Estimating Equations (GEE) using a gamma family with a log link and negative binomial model for the four models. The multivariate analysis used a difference in difference appr oach , as indicated

PAGE 75

75 in Chapter 4. Multivariate analyses controlling for the impact of socio demographic factors such as age, gender, race/ethnicity, geographic location, eligibility status, and risk score on healthcare utilization and PMPM expenditures were conducted. The parameter estimates of the model were estimated by GEE model clustered on the person month level. The coefficient for health system based PSNs (HPSN) shows the shift in the intercept that occurred for enrollees in health system based PSNs a fter implementation of reform. The coefficient for Time × Post × HPSN indicates whether the change in the trends in utilization and PMPM expenditures for enrollees in health system based PSNs after Medicaid reform was significantly different from the chang e in utilization and PMPM expenditures for enrollees in physician based PSNs over the same time period. ED visits. The first GEE model estimated the differences in ED visits by enrollees between physician based PSNs and health system based PSNs in reform c ounties. The results of GEE model of ED visits are shown in Table 5 3. Similar to the univariate analysis, beneficiaries in health system based PSNs had significantly higher ED visits than those in physician based PSNs during the Medicaid reform ( p < 0.001). However, the coefficient for Time × Post × HPSN was 0.0047 and was statistically significant ( p <0.001), which means ED visits for beneficiaries in health system based PSNs was reduced by 0.47% each month more than ED visits for beneficiaries in p hysician based PSNs. Additional ly, two more GEE models were estimated for enrollees in SSI and TANF separately (Table 5 4 and Table 5 5). For SSI enrollees, the coefficient for health system based PSNs (HPSN) was 0.2066 ( p = 0.1844) and the coefficient for Time × Post × HPSN was 0.0031 and was also not statistically significant

PAGE 76

76 ( p = 0.3023). On the other hand, for TANF enrollees, the coefficient for HPSN was 0.3942 and was statistically significant ( p <0.001), while the coefficient for Time × Post × HPSNs was 0.0057 and was also statistically significant ( p <0.001). Thus, there were no variations in ED visits by SSI enrollees between in physician based PSNs and in health system based PSNs, while there were differences in ED visits between TANF enrollees in physician based PSNs and health system based PSNs. ED utilization for TANF enrollees in health system based PSNs declined more over time than ED utilization for those in physician based PSNs. Inpatient u tilization. The second GEE model estimated the diffe rence in inpatient hospital utilization by Medicaid enrollees between physician based PSNs and health system based PSNs in Broward and Duval Counties (Table 5 6). The multivariate analysis indicates that enrollees in health system based PSNs had 102% highe r inpatient utilization compared to enrollees in physician based PSNs ( p <0.001). However, the trend of inpatient hospital utilization over time for enrollees in health system based PSNs decreased 1.9% each month more than enrollees in physician based PSNs , as shown by the negative coefficient for Time × Post × HPSN ( p <0.001). Table 5 7 and Table 5 8 summarize the GEE models of inpatient utilization for SSI and TANF enrollees. For SSI enrollees, the coefficients for health system based PSNs (HPSN) was 0.22 97, but SSI enrollees in health system based PSNs did not differ in number of inpatient admissions and inpatient days compared to those in physician based PSNs ( p = 0.3852). The coefficient for Time × Post × HPSN was 0.004, but was not statistically signi ficant ( p = 0.9377). For TANF enrollees, the coefficient for health system based PSNs (HPSN) was 1.1151 and was statistically significant ( p <0.001),

PAGE 77

77 while the coefficient for Time × Post × HPSN was 0.0222 and was also statistically significant ( p <0.001) , suggesting TANF enrollees in healthsystem based PSNs had higher use on average during the post period than those in physician based PSNs, but the trend of inpatient hospital utilization over time for TANF enrollees in healthsystem based PSNs decreas ed mo re than the trend for TANF enrollees in physician based PSNs. Prescription d rug u se. The results of GEE model of prescription drug utilization are shown in Table 5 9. The multivariate analysis showed that beneficiaries in health system based PSNs had 10. 8% higher prescription drug use , but the trend in pr escription drug use was decreasing at a greater rate for beneficiaries enrolled in health system based PSNs compared to those enrolled in physician based PSNs. Beneficiaries enrolled in health system based PSNs appeared to be reducing prescription drug us e by 0.38% per month compared to those enrolled in physician based PSNs. When examining the impact of the different organizational structures separately for beneficiaries who enrolled in SSI and TANF, it app eared that the coefficient for HPSN was 0.1070 ( p =0.0926) and the coefficient for Time × Post × HPSN was 0.0016 and was also not statistically significant ( p =0.2112) for SSI enrollees (Table 5 10). For TANF enrollees, the coefficient for HPSN was 0.1634 and was statistically significant ( p <0.001), while the coefficient for Time × Post × HPSN was 0.0049 and was also statistically significant ( p <0.001) (Table 5 11), suggesting there was a similar result with the GEE model of prescription drug use with f ull sample. PMPM e xpenditures. The second aim of this study was to examine the effect of the two different types of PSNs on PMPM expenditures. As reported in Table 5 12, the

PAGE 78

78 coefficient for health system based PSNs (HPSN) was 0.9726 and was statistically significant ( p <0.001), while the coefficient for Time × Post × HPSN was 0.0180 and was also statistically significant ( p <0.001). This result indicated there were 97.26% higher PMPM expenditures for enrollees in health system based PSNs on average during the Medicaid Demonstration, but the trend in expenditures over time decreased more for enrollees in health system based PSNs compared to enrollees in physician based PSNs. PMPM expenditures for enrollees in health system based PSNs reduced by 1.80% every month relative to expenditures for those in physician based PSNs. The two other GEE models estimated the differences in PMPM expenditures by SSI and TANF enrollees between physician based PSNs and health system based PSNs in reform counties separately (Tab le 5 13 and Table 5 14). SSI enrollees in health system based PSNs had higher PMPM expenditures ( p <0.001), but had a downward trend in expenditures over time compared to SSI enrollees in physician based PSNs ( p <0.001). Similar to SSI enrollees, TANF enro llees in health system based PSNs had higher expenditures relative to those in physician based PSNs ( p <0.001). However, the trend in PMPM expenditures over time for TANF enrollees in health system based PSNs decreased more than those in physician based PS Ns ( p <0.001). Therefore, PMPM expenditures for enrollees in health system based PSNs were higher on average during the reform period , however, expenditures for them were getting lower over time relative to expenditures for enro llees in physician based PSN s. And th e s e findings were consistent with SSI and TANF enrollees in physician based PSNs and health system based PSNs in reform counties.

PAGE 79

79 Sensitivity Analysis The purpose of the sensitivity analysis was to assess the effect of the two different types of PSNs on utilization and expenditures with continuous Medicaid enrollments. The sensitivity analyses were limited to only enrollees in health system based PSNs and physician based PSNs with at least 3 or 6 months of observation in both the pre reform and po st reform periods (Table 5 16). With the analysis limited to enrollees with at least 3 months of observations in the pre and post Demonstration period, the differences in healthcare utilization were similar with the base model, while the differences in PMPM expenditures between physician based PSNs and health system based PSNs were less than the results of the full sample ($4.83 vs. $59.96). Analysis of enrollees with at least 6 months of observations showed that the differences in ED visits, inpat ient, and prescription drug use did n o t change from the base model, whereas the differences in PMPM expenditures were still lower than the differences in expenditures with the full sample ($25.70 vs. $59.96). In the multivariate analysis limited to enrollees wi th at least 3 months or 6 months of observations, enrollees in health system based PSNs had higher healthcare utilization and expenditures on average during the post period when compared to enrollees in physician based PSNs. However, healthcare utilization and expenditures decreased more for beneficiaries enrolled in health system based PSNs than those enrolled in physician based PSNs over time (Table s 5 17 5 24). Both sensitivity analyses showed similar results with the main findings.

PAGE 80

80 Specific Aim 3 Descr iptive Analysis The characteristics of the study sample between the two types of PSNs through CAHPS survey data set are summarized in Table 5 25. Characteristics of the sample were significantly different across plan types ( p <0.001). For enrollees in the reform counties from 2006 to 2008, 51.35%, 21.39%, and 50.39% were black, Hispanic, and Male , respectively, in health system based PSNs compared to 45.07%, 26.41%, and 47.90% , respectively, in physician based PSNs. Enrollees in health system based PSNs wer e younger than those in physician based PSNs (15.47 vs 18.73) and had higher risk scores (1.12 vs 0.44). Over half of enrollees physician based PSNs and health system based PSNs lived in Broward County and were enrolled through TANF. Univariate Analysis Th their health plans, overall healthcare, personal doctor, and specialist are presented in Table 5 26 and Table 5 27. According to results of the unweighted analysis for the non D emonstration period, over 50% of the population provided either a 9 or 10 for their ratings of health plans and aspects of their healthcare services. Whe n compar isons were made between based PSNs and health system based PSNs , enrollees in health system based PSNs were more likely to have high er ratings of satisfaction with their health plans than enrollees in physician based PSNs (65.40% vs. 56.43%). With regard to satisfaction with personal doctor and specialist, enroll ees in health system based PSNs had 9.28% and 9.57% higher ratings, compared to those in physician based PSNs. For the Demonstration period from 2007 to 2008, the results were consistent with the period prior to the Demonstration.

PAGE 81

81 Therefore, relative to en rollees in physician based PSNs, enrollees in health system based PSNs rated their health plans and aspects of healthcare services higher. However, beneficiaries in health system based PSNs rated their overall healthcare lower than physician based PSNs. Ac cording to results of the weighted analysis, over 50% of the population provided either a 9 or 10 for their satisfaction with health plans and healthcare services. Also, enrollees in health system based PSNs were more likely to have higher satisfaction with their plans and specialist relative to enrollees in physician based PSNs for the pre reform and post reform periods. However, when compared to satisfaction with overall healthcare and personal doctor for enrollees in physician based PSNs, enrollees in health system based PSNs had a lower satisfaction with their overall healthcare and personal doctor during the Medicaid reform period. Weighted results about all means and percent distribution for ratings were significantly different ( p <0.0001). Multiva riate Analysis For Specific Aim 3, SAS® PROC LOGISTIC was used to analyze the ordered logit model. The multivariate analysis for Specific Aim 3 used a difference in difference approach , as indicated in Chapter 4. Multivariate analyses controlling for the i mpact of such socio demographic factors as age, gender, race, ethnicity, geographic location, eligibility status, health status, and risk score on satisfaction with health plans, overall healthcare, personal doctor, and specialist were conducted. The coeff icient for Post × HPSN indicates whether the change in the trends in satisfaction with health plans, overall healthcare, personal doctor, and specialist for enrollees in health system based

PAGE 82

82 PSNs after Medicaid reform was significantly different from the ch ange in trends in satisfaction for enrollees in physician based PSNs over the same time period. The difference in difference estimates satisfaction with health plans, overall healthcare, personal doctor, and specialist between beneficiaries enrolled in hea lth system based PSNs and physician based PSNs over time are presented in Table s 5 28 5 31. Beneficiaries enrolled in health system based PSNs were more likely to have a positive rating for their health plans, overall healthcare, personal doctor, and spec ialist relative to beneficiaries enrolled in physician based PSNs. Particularly, the odds of H PSNs enrollees rati ng their health plans highly were 72%, 53%, 87%, and 58% higher than P PSNs during the post period (AOR= 1.72, 1.53, 1.87, and 1.58; p <0.0001 ) . However, these odds were reduced by 15%, 37%, 34%, and 24% per year, meaning H PSNs enrollee satisfaction with their health plan decreased at a greater rate over time (AOR= 0.846, 0.632, 0.663, and 0.761; p <0.05).

PAGE 83

83 Table 5 1. Sample c haracteristics ( Aims 1 and 2)* Pre reform Post reform MediPass (N= 2 , 544 , 281 ) P PSNs (N= 838,254 ) H PSNs (N= 1,073,434 ) Age 13.34 14.95 17.68 <1 10.91% 3.76% 3.85% 1 5 26.95% 22.31% 16.50% 6 13 28.25% 32.62% 31.71% 14 20 14.58% 20.93% 19.66% 21 54 16.19% 17.29% 23.91% 55 64 2.71% 2.67% 3.91% >65 0.36% 0.33% 0.30% Gender Female 45.95 % 53.94 % 56.13 % Male 54.05 % 46.06 % 43.87 % Race/Ethnicity White 23.84 % 20.61% 19.55% Black 47.48 % 51.94% 59.52% Hispanic 15.84 % 16.87% 11.87% Other 12.84 % 10.47% 8.95% County Duval 42.39 % 35.42 % 57.59 % Broward 57.60 % 64.58 % 42.41 % Eligibility Status S SI 15.48 % 14.76 % 18.27 % TANF 84.52 % 85.24 % 81.73 % Risk s core 0 0.12 0.10 * All characteristics percentage or means are significantly different across groups .

PAGE 84

84 Table 5 2. Univariate analysis (Aims 1 and 2) Pre reform Post reform Full Sample MediPass (N= 2 , 544 , 281 ) P PSNs (N= 838,254 ) H PSNs (N= 1,073,434 ) ED visits 0.04 0.03 0.04 Inpatient 0.09 0.03 0.04 Prescription d rugs 0.42 0.59 0.68 PMPM e xpenditures $247.78 $133.27 $155.67 SSI TANF SSI TANF SSI TANF ED visits 0.06 0.04 0.05 0.03 0.05 0.03 Inpatient 0.35 0.05 0.13 0.01 0.14 0.01 Prescription d rugs 1.58 0.21 1.55 0.42 1.82 0.43 PMPM e xpenditures $898.30 $128.60 $439.29 $80.29 $486.32 $80.30 Top 5 % deleted MediPass P PSNs H PSNs PMPM Expenditures $63.96 $30.38 $35.39

PAGE 85

85 Table 5 3. Multivariate model of ED visits (Aim 1) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.0414 0.0007 0.0427 0.0400 <.0001 Post reform 0.4521 0.0377 0.3782 0.5259 <.0001 HPSN 0.3272 0.0515 0.2262 0.4281 <.0001 Time × Post 0.0168 0.0010 0.0148 0.0188 <.0001 Time × Post × HPSN 0.0047 0.0010 0.0067 0.0028 <.0001 Age 0.0111 0.0004 0.0104 0.0119 <.0001 Gender (Female) Male 0.1911 0.0098 0.2104 0.1718 <.0001 Race/ e thnicity (White) Black 0.2978 0.0545 0.1910 0.4046 <.0001 Hispanic 0.2459 0.0555 0.1372 0.3547 <.0001 Other 0.0726 0.0570 0.1843 0.0392 0.2033 County (Broward) Duval 0.3364 0.0100 0.3167 0.3560 <.0001 Eligibility (SSI) TANF 0.1753 0.0182 0.2110 0.1397 <.0001 Risk s core 0.1227 0.0046 0.1138 0.1317 <.0001

PAGE 86

86 Table 5 4. Multivariate model of ED visits for SSI enrollees (Aim 1) Estimate S.E. 95% Confidence Limits Pr > |Z| Time 0.0357 0.0020 0.0396 0.0319 <.0001 Post reform 0.4451 0.1174 0.2149 0.6752 0.0002 H PSNs 0.2066 0.1556 0.0985 0.5116 0.1844 Time × Post 0.0126 0.0031 0.0065 0.0186 <.0001 Time × Post × HPSN 0.0031 0.0030 0.0090 0.0028 0.3023 Age 0.0002 0.0008 0.0017 0.0013 0.7869 Gender (Female) Male 0.3127 0.0343 0.3799 0.2455 <.0001 Race/ e thnicity (White) Black 0.7497 0.1954 0.3666 1.1327 0.0001 Hispanic 0.5192 0.2035 0.1204 0.9180 0.0107 Other 0.4512 0.1957 0.0676 0.8348 0.0211 County (Broward) Duval 0.0463 0.0296 0.0116 0.1043 0.1169 Risk s core 0.1934 0.0093 0.1752 0.2117 <.0001 Table 5 5. Multivariate model of ED visits for TANF enrollees (Aim 1) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.0435 0.0007 0.0448 0.0421 <.0001 Post reform 0.4873 0.0370 0.4148 0.5598 <.0001 H PSNs 0.3942 0.0502 0.2957 0.4926 <.0001 Time × Post 0.0173 0.0010 0.0152 0.0193 <.0001 Time × Post × HPSN 0.0057 0.0010 0.0076 0.0038 <.0001 Age 0.0189 0.0004 0.0181 0.0197 <.0001 Gender (Female) Male 0.1446 0.0088 0.1617 0.1274 <.0001 Race/ e thnicity (White) Black 0.2256 0.0566 0.1147 0.3366 <.0001 Hispanic 0.2098 0.0576 0.0968 0.3227 0.0003 Other 0.3200 0.0608 0.4391 0.2009 <.0001 County (Broward) Duval 0.3999 0.0094 0.3816 0.4183 <.0001 Risk s core 0.0855 0.0050 0.0758 0.0952 <.0001

PAGE 87

87 Table 5 6. GEE model of inpatient utilization (Aim 1) Estimate S.E. 95% Confidence Limits Pr > |Z| Time 0.0111 0.0018 0.0076 0.0145 <.0001 Post reform 0.6941 0.1187 0.9268 0.4614 <.0001 HPSN 1.0234 0.1589 0.7120 1.3347 <.0001 Time × Post 0.0193 0.0029 0.0249 0.0137 <.0001 Time × Post × HPSN 0.0193 0.0030 0.0252 0.0134 <.0001 Age 0.0100 0.0010 0.0080 0.0120 <.0001 Gender (Female) <.0001 Male 0.1720 0.0261 0.2231 0.1209 <.0001 Race/ e thnicity (White) Black 0.0708 0.1881 0.2978 0.4395 0.7065 Hispanic 0.0207 0.1922 0.3974 0.3559 0.9141 Other 0.0884 0.1943 0.2923 0.4692 0.6490 County (Broward) Duval 0.1178 0.0269 0.1706 0.0650 <.0001 Eligibility (SSI) TANF 1.7727 0.0392 1.8496 1.6958 <.0001 Risk s core 0.1881 0.0146 0.1594 0.2167 <.0001

PAGE 88

88 Table 5 7. Multivariate model of inpatient utilization for SSI enrollees (Aim 1) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.0131 0.0028 0.0076 0.0186 <.0001 Post reform 0.3527 0.2033 0.0458 0.7511 0.0828 H PSNs 0.2297 0.2645 0.2887 0.7480 0.3852 Time × Post 0.0432 0.0050 0.0530 0.0334 <.0001 Time × Post × HPSN 0.0004 0.0052 0.0107 0.0098 0.9377 Age 0.0109 0.0011 0.0087 0.0131 <.0001 Gender (Female) Male 0.0682 0.0419 0.1504 0.0141 0.1042 Race/ e thnicity (White) Black 1.2988 0.3009 0.7091 1.8884 <.0001 Hispanic 0.9995 0.3098 0.3923 1.6068 0.0013 Other 1.0264 0.3013 0.4359 1.6169 0.0007 County (Broward) Duval 0.3333 0.0404 0.4124 0.2542 <.0001 Risk s core 0.2552 0.0171 0.2216 0.2887 <.0001 Table 5 8. Multivariate model of inpatient utilization for TANF enrollees (Aim 1) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.0105 0.0020 0.0066 0.0145 <.0001 Post reform 0.9912 0.1418 1.2691 0.7133 <.0001 H PSNs 1.1151 0.1951 0.7327 1.4974 <.0001 Time × Post 0.0130 0.0033 0.0194 0.0065 <.0001 Time × Post × HPSN 0.0222 0.0036 0.0292 0.0152 <.0001 Age 0.0095 0.0016 0.0064 0.0127 <.0001 Gender (Female) Male 0.2067 0.0308 0.2670 0.1464 <.0001 Race/ e thnicity (White) Black 0.0026 0.1954 0.3857 0.3804 0.9892 Hispanic 0.0640 0.2001 0.4562 0.3282 0.7492 Other 0.1746 0.2084 0.2340 0.5831 0.4023 County (Broward) Duval 0.0463 0.0323 0.1095 0.0170 0.1521 Risk s core 0.1350 0.0254 0.0852 0.1849 <.0001

PAGE 89

89 Table 5 9. GEE model for rate of prescription drug use (Aim 1) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.1389 0.0008 0.1406 0.1372 <.0001 Post reform 0.6351 0.0202 0.6747 0.5955 <.0001 HPSN 0.1082 0.0273 0.0546 0.1617 <.0001 Time × Post 0.1356 0.0010 0.1337 0.1375 <.0001 Time × Post × HPSN 0.0038 0.0005 0.0048 0.0029 <.0001 Age 0.0302 0.0003 0.0296 0.0308 <.0001 Gender (Female) Male 0.1658 0.0071 0.1797 0.1519 <.0001 Race/ e thnicity (White) Black 0.2983 0.0290 0.3553 0.2414 <.0001 Hispanic 0.1504 0.0301 0.2095 0.0913 <.0001 Other 0.2878 0.0318 0.3501 0.2255 <.0001 County (Broward) Duval 0.1855 0.0077 0.1705 0.2005 <.0001 Eligibility (SSI) TANF 0.9198 0.0137 0.9466 0.8930 <.0001 Risk s core 0.1843 0.0028 0.1789 0.1898 <.0001

PAGE 90

90 Table 5 10. Multivariate model of prescription drug use for SSI enrollees (Aim 1) Estimate S.E. 95% Confidence l imits Pr > |Z| Time0 0.1171 0.0016 0.1203 0.1140 <.0001 Post reform 0.2917 0.0504 0.3904 0.1929 <.0001 H PSNs 0.1070 0.0636 0.2316 0.0177 0.0926 Time × Post 0.0995 0.0019 0.0957 0.1033 <.0001 Time × Post × HPSN 0.0016 0.0013 0.0009 0.0041 0.2112 Age 0.0317 0.0004 0.0309 0.0324 <.0001 Gender (Female) Male 0.3566 0.0160 0.3879 0.3253 <.0001 Race/ e thnicity (White) 0.6025 0.1101 0.3867 0.8183 Black 0.4027 0.1101 0.1869 0.6185 0.0003 Hispanic 0.1655 0.1153 0.0606 0.3915 0.1514 Other 0.4141 0.1112 0.1962 0.6321 0.0002 County (Broward) Duval 0.1015 0.0165 0.0693 0.1337 <.0001 Risk s core 0.2884 0.0044 0.2798 0.2970 <.0001 Table 5 11. Multivariate model of prescription drug use for TANF enrollees (Aim 1) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.1555 0.0009 0.1574 0.1537 <.0001 Post reform 0.7622 0.0228 0.8069 0.7176 <.0001 H PSNs 0.1634 0.0308 0.1031 0.2238 < .0001 Time × Post 0.1546 0.0011 0.1524 0.1567 <.0001 Time × Post × HPSN 0.0049 0.0006 0.0060 0.0039 <.0001 Age 0.0289 0.0005 0.0279 0.0299 <.0001 Gender (Female) Male 0.1150 0.0074 0.1294 0.1005 <.0001 Race/ e thnicity (White) 0.0316 0.0303 0.0910 0.0279 Black 0.2792 0.0296 0.3373 0.2212 <.0001 Hispanic 0.0950 0.0307 0.1553 0.0348 0.0020 Other 0.2403 0.0341 0.3071 0.1734 <.0001 County (Broward) Duval 0.2151 0.0083 0.1988 0.2313 <.0001 Risk s core 0.1646 0.0036 0.1575 0.1716 <.0001

PAGE 91

91 Table 5 12. GEE model of PMPM expenditures (Aim 2) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.0061 0.0009 0.0079 0.0043 <.0001 Post reform 2.0935 0.0438 2.1793 2.0077 <.0001 HPSN 0.9726 0.0572 0.8604 1.0848 < .0001 Time × Post 0.0492 0.0013 0.0466 0.0517 <.0001 Time × Post × HPSN 0.0180 0.0012 0.0203 0.0157 <.0001 Age 0.0068 0.0005 0.0058 0.0078 <.0001 Gender (Female) Male 0.0539 0.0118 0.0770 0.0307 <.0001 Race/ e thnicity (White) Black 0.2715 0.0782 0.4248 0.1182 0.0005 Hispanic 0.2318 0.0793 0.3873 0.0763 00035 Other 0.2116 0.0814 0.3711 0.520 0.0093 County (Broward) Duval 0.1388 0.0124 0.1632 0.1145 <.0001 Eligibility (SSI) TANF 1.5720 0.0208 1.6128 1.5312 <.0001 Risk s core 0.118 0.0053 0.1015 0.1221 <.0001

PAGE 92

92 Table 5 13. GEE model of PMPM expenditures for SSI enrollees (Aim 2) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.0015 0.0014 0.0043 0.0013 0.2968 Post reform 1.0861 0.1043 1.2905 0.8817 <.0001 H PSNs 0.5471 0.1244 0.3033 0.7909 <.0001 Time × Post 0.0169 0.0025 0.0120 0.0219 <.0001 Time × Post × HPSN 0.0082 0.0024 0.0130 0.0034 0.0007 Age 0.0064 0.0006 0.0051 0.0076 <.0001 Gender (Female) Male 0.0178 0.0244 0.0657 0.0301 0.4667 Race/ e thnicity (White) Black 0.5715 0.1254 0.3257 0.8173 <.0001 Hispanic 0.4503 0.1327 0.1901 0.7105 0.0007 Other 0.3782 0.1263 0.1307 0.6258 0.0027 County (Broward) Duval 0.4365 0.0238 0.4832 0.3898 <.0001 Risk s core 0.1701 0.0082 0.1539 0.1862 <.0001 Table 5 14. GEE model of PMPM expenditures for TANF enrollees (Aim 2) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.0074 0.0011 0.0096 0.0052 <.0001 Post reform 2.2975 0.0489 2.3933 2.2016 < .0001 H PSNs 0.9779 0.0648 0.8509 1.1048 <.0001 Time × Post 0.0558 0.0015 0.0529 0.0588 <.0001 Time × Post × HPSN 0.0185 0.0013 0.0211 0.0159 <.0001 Age 0.0070 0.0008 0.0055 0.0085 <.0001 Gender (Female) Male 0.0637 0.0132 0.0896 0.0378 <.0001 Race/ e thnicity (White) Black 0.2595 0.0817 0.4195 0.0994 0.0015 Hispanic 0.1794 0.0828 0.3417 0.0172 0.0302 Other 0.0653 0.0874 0.2366 0.1060 0.4548 County (Broward) Duval 0.0568 0.0143 0.0848 0.0289 <.0001 Risk s core 0.0973 0.0072 0.0831 0.1115 <.0001

PAGE 93

93 Table 5 15. S ensitivity analysis of ED visits with three months of observations (Aim 1) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.0448 0.0018 0.0482 0.0413 <.0001 Post reform 0.1885 0.1215 0.0496 0.4266 0.1207 HPSN 0.2591 0.1622 0.0589 0.5771 0.1103 Time × Post 0.0274 0.0028 0.0219 0.0328 <.0001 Time × Post × HPSN 0.0037 0.0028 0.0092 0.0018 0.1917 Age 0.0162 0.0007 0.0148 0.0175 <.0001 Gender (Female) Male 0.2805 0.0217 0.3230 0.2380 <.0001 Race/ e thnicity (White) Black 0.2594 0.1052 0.0533 0.4655 0.0136 Hispanic 0.1435 0.1083 0.0687 0.3557 0.1851 Other 0.2019 0.1141 0.4255 0.0218 0.0769 County (Broward) Duval 0.2518 0.0215 0.2096 0.2940 <.0001 Eligibility (SSI) TANF 0.1197 0.0406 0.1993 0.0401 0.0032 Risk s core 0.1336 0.0179 0.0985 0.1686 <.0001

PAGE 94

94 Table 5 16. S ensitivity analysis of inpatient stays with three months of observations (Aim 1) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.0237 0.0049 0.0140 0.0333 <.0001 Post reform 0.1538 0.4538 1.0433 0.7356 0.7346 HPSN 1.1714 0.6161 0.0361 2.3789 0.0572 Time × Post 0.0334 0.0086 0.0503 0.0164 0.0001 Time × Post × HPSN 0.0238 0.0100 0.0434 0.0042 0.0171 Age 0.0092 0.0026 0.0042 0.0143 0.0004 Gender (Female) Male 0.2365 0.0700 0.3738 0.0992 0.0007 Race/ e thnicity (White) Black 0.7915 0.3419 0.1214 1.4616 0.0206 Hispanic 0.6476 0.3485 0.0354 1.3305 0.0631 Other 0.9162 0.3658 0.1992 1.6332 0.0123 County (Broward) Duval 0.0342 0.0744 0.1801 0.1117 0.6456 Eligibility (SSI) TANF 2.1265 0.0962 2.3151 1.9380 <.0001 Risk s core 0.1343 0.0576 0.0214 0.2472 0.0197

PAGE 95

95 Table 5 17. S ensitivity analysis of prescription drugs with three months of observations (Aim 1) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.1292 0.0016 0.1324 0.1260 <.0001 Post reform 0.1600 0.0500 0.2580 0.0621 0.0014 HPSN 0.4975 0.0671 0.3659 0.6291 <.0001 Time × Post 0.1246 0.0019 0.1210 0.1283 <.0001 Time × Post × HPSN 0.0093 0.0011 0.0114 0.0071 <.0001 Age 0.0257 0.0005 0.0247 0.0267 <.0001 Gender (Female) Male 0.2398 0.0144 0.2680 0.2116 <.0001 Race/ e thnicity (White) Black 0.1628 0.0487 0.2583 0.0673 0.0008 Hispanic 0.1421 0.0516 0.2433 0.0409 0.0059 Other 0.2197 0.0555 0.3285 0.1110 <.0001 County (Broward) Duval 0.1670 0.0147 0.1381 0.1959 <.0001 Eligibility (SSI) TANF 1.0535 0.0254 1.1033 1.0037 <.0001 Risk s core 0.1025 0.0138 0.0755 0.1296 < .0001

PAGE 96

96 Table 5 18. S ensitivity analysis of PMPM expenditures with three months of observations (Aim 2) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.0074 0.0027 0.0128 0.0021 0.0060 Post reform 2.1387 0.1359 2.4051 1.8722 <.0001 HPSN 1.1447 0.1990 0.7548 1.5347 <.0001 Time × Post 0.0472 0.0034 0.0405 0.0539 <.0001 Time × Post × HPSN 0.0217 0.0033 0.0281 0.0153 <.0001 Age 0.0044 0.0014 0.0018 0.0071 0.0012 Gender (Female) Male 0.1191 0.0362 0.1901 0.0482 0.0010 Race/ e thnicity (White) Black 0.0248 0.0994 0.1701 0.2196 0.8034 Hispanic 0.0950 0.1082 0.3072 0.1171 0.3799 Other 0.0043 0.1225 0.2357 0.2444 0.9718 County (Broward) Duval 0.0366 0.0380 0.1111 0.0378 0.3347 Eligibility (SSI) TANF 1.8611 0.0561 1.9710 1.7513 <.0001 Risk s core 0.0908 0.0187 0.0542 0.1275 <.0001

PAGE 97

97 Table 5 19. S ensitivity analysis of ED visits with six months of observations (Aim 1) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.0447 0.0018 0.0482 0.0413 <.0001 Post reform 0.1467 0.1092 0.0674 0.3607 0.1792 HPSN 0.2917 0.1469 0.0037 0.5797 0.0471 Time × Post 0.0279 0.0026 0.0228 0.0331 <.0001 Time × Post × HPSN 0.0041 0.0026 0.0092 0.0010 0.1126 Age 0.0159 0.0007 0.0145 0.0173 <.0001 Gender (Female) Male 0.2750 0.0213 0.3167 0.2333 <.0001 Race/ e thnicity (White) 0.3314 0.1046 0.1264 0.5364 0.0015 Black 0.2768 0.1036 0.0737 0.4798 0.0076 Hispanic 0.1582 0.1066 0.0508 0.3672 0.1380 Other 0.1619 0.1123 0.3821 0.0583 0.1496 County (Broward) Duval 0.2473 0.0212 0.2058 0.2888 <.0001 Eligibility (SSI) TANF 0.1294 0.0396 0.2071 0.0517 0.0011 Risk s core 0.1407 0.0150 0.1113 0.1702 <.0001

PAGE 98

98 Table 5 20. S ensitivity analysis of inpatient stays with six months of observations (Aim 1) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.0239 0.0050 0.0142 0.0336 <.0001 Post reform 0.1664 0.4670 1.0818 0.7489 0.7216 HPSN 1.1068 0.6003 0.0696 2.2833 0.0652 Time × Post 0.0337 0.0089 0.0512 0.0163 0.0001 Time × Post × HPSN 0.0225 0.0098 0.0417 0.0033 0.0218 Age 0.0097 0.0026 0.0046 0.0147 0.0002 Gender (Female) Male 0.2338 0.0701 0.3712 0.0963 0.0009 Race/ e thnicity (White) 0.8149 0.3392 0.1502 1.4797 0.0163 Black 0.8385 0.3380 0.1760 1.5010 0.0131 Hispanic 0.6549 0.3444 0.0201 1.3300 0.0572 Other 0.9466 0.3612 0.2387 1.6546 0.0088 County (Broward) Duval 0.0405 0.0743 0.1861 0.1051 0.5859 Eligibility (SSI) TANF 2.1142 0.0947 2.2998 1.9286 <.0001 Risk s core 0.0979 0.0545 0.0088 0.2046 0.0722

PAGE 99

99 Table 5 21. S ensitivity analysis of prescription drugs with six months of observations (Aim 1) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.1287 0.0016 0.1319 0.1255 <.0001 Post reform 0.3738 0.0464 0.4647 0.2829 <.0001 HPSN 0.6163 0.0616 0.4956 0.7371 <.0001 Time × Post 0.1278 0.0019 0.1241 0.1314 <.0001 Time × Post × HPSN 0.0113 0.0010 0.0133 0.0093 <.0001 Age 0.0254 0.0005 0.0244 0.0264 <.0001 Gender (Female) Male 0.2335 0.0140 0.2609 0.2060 <.0001 Race/ e thnicity (White) 0.0614 0.0487 0.1568 0.0340 0.2073 Black 0.1843 0.0475 0.2775 0.0911 0.0001 Hispanic 0.1603 0.0504 0.2590 0.0616 0.0015 Other 0.2300 0.0540 0.3359 0.1241 <.0001 County (Broward) Duval 0.1633 0.0144 0.1351 0.1916 <.0001 Eligibility (SSI) TANF 1.0435 0.0247 1.0920 0.9951 <.0001 Risk s core 0.1134 0.0122 0.0895 0.1374 <.0001

PAGE 100

100 Table 5 22. S ensitivity analysis of PMPM expenditures with six months of observations (Aim 2) Estimate S.E. 95% Confidence l imits Pr > |Z| Time 0.0074 0.0027 0.0127 0.0020 0.0067 Post reform 2.2643 0.1523 2.5628 1.9657 <.0001 HPSN 1.1282 0.2016 0.7330 1.5233 <.0001 Time × Post 0.0493 0.0036 0.0422 0.0563 <.0001 Time × Post × HPSN 0.0215 0.0033 0.0280 0.0150 <.0001 Age 0.0048 0.0014 0.0021 0.0074 0.0004 Gender (Female) Male 0.1092 0.0365 0.1808 0.0375 0.0028 Race/ e thnicity (White) 0.0005 0.0984 0.1933 0.1924 0.9962 Black 0.0369 0.0970 0.1531 0.2270 0.7032 Hispanic 0.0992 0.1048 0.3046 0.1063 0.3442 Other 0.0102 0.1191 0.2232 0.2437 0.9315 County (Broward) Duval 0.0410 0.0378 0.1151 0.0330 0.2774 Eligibility (SSI) TANF 1.8539 0.0552 1.9621 1.7458 <.0001 Risk s core 0.0823 0.0181 0.0467 0.1178 <.0001

PAGE 101

101 Table 5 23. Sample c haracteristics (Aim 3) * P PSNs (N= 4,040 ) H PSNs (N= 4,185 ) Age 18.73 15.47 <1 0.22 % 0.24 % 1 5 25.10 % 30.99 % 6 13 30.50 % 32.71 % 14 20 18.29 % 18.18 % 21 54 16.21 % 11.35 % 55 64 7.33 % 5.52 % >65 2.25 % 0.86 % Gender Female 52.10% 49.61 % Male 47.90 % 50.39 % Race White 34.93 % 31.28 % Black 45.07 % 51.35 % Other 18.84 % 15.79 % Ethnicity Hispanic 26.41% 21.39% Non Hispanic 73.59% 78.61% County Duval 23.74 % 48.00 % Broward 76.26 % 52.00 % Eligibility Status S SI 28.59 % 46.19 % TANF 71.41 % 53.81 % Health Status Excellent 32.62% 24.73% Very good 23.29% 23.27% Good 23.74% 27.60% Fair 13.59% 17.59% Poor 6.16% 6.21% 0.40% 0.48% Refused 0.20% 0.12% Risk s core 0.44 1.12 * All characteristics percentage or means are significantly different across groups .

PAGE 102

102 Table 5 24. Univariate analysis (Aim 3, unweighted) 2006 2007 2008 Weighted P PSNs H PSNs P value P PSNs H PSNs P value Satisfaction with h ealth plans (rating) 1.40 1.55 <0.0001 1.35 1.47 <0.0001 Low 16.24% 10.43 % 18.87 % 14.32% Medium 27.32% 24.17 % 27.27 % 24.45% High 56.43% 65.40 % 53.86 % 61.23% Healthcare rating 1.52 1.60 0.2204 1.50 1.49 0.0004 Low Rating 11.61 % 8.09 % 11.66 % 12.44% Medium 25.14 % 24.02 % 26.91 % 26.60% High 63.25 % 67.89 % 61.43 % 60.96% Personal d octor (rating) 1.61 1.73 0.0075 1.67 1.70 <0.0001 Low 8.50 % 5.18 % 7.73 % 7.03% Medium 22.28 % 16.32 % 17.27 % 16.13% High 69.22 % 78.50 % 75.00 % 76.83% Specialist (rating) 1.48 1.64 0.0022 1.51 1.61 <0.0001 Low 14.57 % 7.59 % 12.84 % 9.08% Medium 23.28 % 20.69 % 23.59 % 20.66% High 62.15 % 71.72 % 63.56 % 70.27%

PAGE 103

103 Table 5 25. Univariate analysis (Aim 3, weighted) 2006 2007 2008 Weighted P PSNs H PSNs P value P PSNs H PSNs P value Satisfaction with h ealth plans (rating) 1.42 1.54 <0.0001 1.37 1.49 <0.0001 Low 15.01% 10.31% 17.97% 13.54% Medium 27.53% 25.08% 26.57% 24.07% High 57.47% 64.61% 55.46% 62.40% Healthcare rating 1.54 1.58 <0.0001 1.53 1.47 <0.0001 Low Rating 10.26% 8.83% 10.58% 14.12% Medium 25.02% 24.50% 26.23% 25.09% High 64.73% 66.67% 63.19% 60.79% Personal d octor (rating) 1.62 1.72 <0.0001 1.69 1.68 <0.0001 Low 7.74% 5.45% 7.07% 7.63% Medium 22.48% 16.74% 17.12% 16.43% High 69.78% 77.81% 75.81% 75.93% Specialist (rating) 1.45 1.64 <0.0001 1.49 1.57 <0.0001 Low 15.42% 7.64% 13.33% 10.75% Medium 24.21% 21.00% 23.91% 21.35% High 60.37% 71.35% 62.76% 67.90%

PAGE 104

104 Table 5 26. Multivariate analysis of satisfaction with health plans (Aim 3) Point Estimate 95% Confidence l imits Pr > ChiSq Post reform 0.836 0.801 0.872 <.0001 PSN (PPSN) HPSN 1.722 1.565 1.895 <.0001 Post reform*HPSN 0.846 0.762 0.939 0.0017 Age 1.000 0.999 1.001 0.6514 Gender (Female) Male 0.932 0.902 0.963 <.0001 Race (White) Black 1.148 1.104 1.194 <.0001 Other 0.825 0.785 0.867 <.0001 Ethnicity (Non Hispanic) Hispanic 1.687 1.609 1.768 <.0001 County (Broward) Duval 1.046 1.007 1.086 0.0193 Eligibility (SSI) TANF 1.046 0.998 1.096 0.0581 Health d tatus (Excellent) Very g ood 0.685 0.656 0.716 <.0001 Good 0.542 0.519 0.566 <.0001 Fair 0.459 0.434 0.485 <.0001 Poor 0.250 0.231 0.270 <.0001 Risk s core 1.033 1.018 1.049 <.0001

PAGE 105

105 Table 5 27. Multivariate analysis of satisfaction with overall healthcare (Aim 3) Point e stimate 95% Confidence l imits Pr > ChiSq Post reform 0.897 0.855 0.940 <.0001 PSN (PPSN) HPSN 1.533 1.382 1.700 <.0001 Post reform*HPSN 0.632 0.565 0.708 <.0001 Age 1.001 1.000 1.003 0.0299 Gender (Female) Male 0.951 0.917 0.986 0.0065 Race (White) Black 1.030 0.986 1.076 0.1834 Other 0.733 0.694 0.774 <.0001 Ethnicity (Non Hispanic) Hispanic 1.456 1.383 1.534 <.0001 County (Broward) Duval 1.098 1.054 1.144 <.0001 Eligibility (SSI) TANF 1.191 1.132 1.252 <.0001 Health s tatus (Excellent) Very Good 0.661 0.628 0.696 <.0001 Good 0.429 0.408 0.451 <.0001 Fair 0.366 0.345 0.389 <.0001 Poor 0.182 0.167 0.198 <.0001 Risk s core 1.028 1.014 1.043 0.0001

PAGE 106

106 Table 5 28. Multivariate analysis of satisfaction with personal doctor (Aim 3) Point e stimate 95% Confidence l imits Pr > ChiSq Post reform 1.192 1.132 1.256 <.0001 PSN (PPSN) HPSN 1.871 1.666 2.101 <.0001 Post reform*HPSN 0.663 0.585 0.753 <.0001 Age 0.997 0.996 0.999 0.0004 Gender (Female) Male 0.922 0.886 0.959 <.0001 Race (White) Black 1.067 1.018 1.118 0.0073 Other 0.852 0.803 0.904 <.0001 Ethnicity (Non Hispanic) Hispanic 1.448 1.368 1.533 <.0001 County (Broward) Duval 0.907 0.868 0.949 <.0001 Eligibility (SSI) TANF 1.119 1.058 1.183 <.0001 Health s tatus (Excellent) Very Good 0.757 0.717 0.799 <.0001 Good 0.539 0.511 0.568 <.0001 Fair 0.602 0.563 0.644 <.0001 Poor 0.409 0.372 0.449 <.0001 Risk s core 1.081 1.058 1.104 <.0001

PAGE 107

107 Table 5 29. Multivariate analysis of satisfaction with specialist (Aim 3) Point e stimate 95% Confidence l imits Pr > ChiSq Post reform 0.927 0.856 1.004 0.0640 PSN (PPSN) HPSN 1.581 1.389 1.800 < .0001 Post reform*HPSN 0.761 0.656 0.883 0.0003 Age 0.997 0.995 0.999 0.0009 Gender (Female) Male 0.909 0.858 0.963 0.0012 Race (White) Black 1.217 1.139 1.301 <.0001 Other 0.939 0.865 1.019 0.1333 Ethnicity (Non Hispanic) Hispanic 1.524 1.413 1.645 <.0001 County (Broward) Duval 1.130 1.056 1.209 0.0004 Eligibility (SSI) TANF 0.734 0.683 0.787 <.0001 Health s tatus (Excellent) Very Good 0.647 0.590 0.710 <.0001 Good 0.580 0.532 0.633 <.0001 Fair 0.442 0.402 0.485 <.0001 Poor 0.411 0.366 0.463 <.0001 Risk s core 1.069 1.047 1.092 <.0001

PAGE 108

10 8 Table 5 30. Summary of Results (health system based over physician based PSNs) Average for p ost r eform Trends over time Hypothesis ED visits Higher Decreasing Supported Inpatient h ospital u tilization Higher Decreasing Supported Prescription d rug u ses Higher Decreasing Not Supported PMPM e xpenditures Higher Decreasing Supported Satisfaction Higher Decreasing Not Supported

PAGE 109

109 CHAPTER 6 DISCUSSION AND CONCLUSION S This chapter discuss es each specific aim based on the study results. The second part of this chapter discusses policy implications, study limitations, and possible future studies. Last, this chapter present s conclusions regarding the impact of the two different structures of PSNs on Florida Medicaid managed care on healthcare process and outcomes. Specific Aim s 1 and 2 Discussion The Florida Medicaid Reform Demonstration program introduced PSNs as an option HMOs, physician based PSNs and health system based PSNs. This study focused on the two different types of PSNs, since they have different organizational structures and characteris tics. Compared to physician based PSNs, health system based PSNs are more integrated healthcare delivery systems. The first and second purpose of this study was to examine the differences in utilization (ED visits, inpatient hospital utilization, and presc ription drug use) and PMPM expenditures by Medicaid enrollees between physician based PSNs and health system based PSNs. Before adjusting for socio demographic factors, utilization and expenditures among enrollees in health system based PSNs was higher. After controlling for socio demographic factors, for enrollees in health system based PSNs there were decreasing trends in ED visits, inpatient utilization, prescription drug use, and PMPM expenditures at a greater rate, while having higher ED visits, inpa tient hospital utilization, prescription drug use, and PMPM expenditures for enrollees in health system based PSNs compared to those in physician based PSNs on average during the post reform.

PAGE 110

110 Healthcare Utilization Hypothesis 1. ED utilization by Medicaid enrollees in health system based PSNs will be less than ED utilization by Medicaid enrollees in physician based PSNs. This hypothesis was supported by the study result s . Overall, controlling for socio demographic factors, enrollees in health system based PSNs had higher ED visits on average, however, the number of ED visits will be lower for enrollees in health system based PSNs compared to enrollees in physician based PSNs 69.6 months after the post reform period. This number was derived by the coefficien t of HPSN (0.3272) divided by the rate of decline per month ( 0.0047). Hypothesis 2. Inpatient utilization by Medicaid enrollees in health system based PSNs will be less than inpatient utilization by Medicaid enrollees in physician based PSNs. This hypothe sis was supported by the study result s . Beneficiaries enrolled in health system based PSNs had higher inpatient hospital utilization in the beginning during the post reform period. However, they will have lower inpatient days and inpatient claims from 54 m onths after implementation of Medicaid reform. This number was derived by the coefficient of HPSN (1.0234) divided by the coefficient of Time*Post*HPSN ( 0.0193). Hypothesis 3. Prescription drug utilization for Medicaid enrollees in health system based PSNs will be more than prescription drug utilization for Medicaid enrollees in physician based PSNs. This hypothesis was not supported by the study finding s . We found that enrollees in health system based PSNs had higher p rescription drug use on average d uring the implementation of Medicaid reform, but enrollees in health system based PSNs will have lower p rescription drug use than enrollees in physician based PSNs 29 months after initiation of the Medicaid reform demonstration.

PAGE 111

111 This number was derived by the coefficient of HPSN (0.1082) divided by the rate of decline per month ( 0.0038). PMPM Expenditures Hypothesis 4 . PMPM expenditures for Medicaid enrollees in health system based PSNs will be less than PMPM expenditures for Medicaid enrollees in physician based PSNs. This hypothesis was supported by the study finding s . Controlling for demographic factors, the trend in PMPM expenditures over time were declining for enrollees in health system based PSNs more than those in physician based PSNs, even though enrollees in health system based PSNs were more likely to have higher PMPM expenditures compared to those in physician based PSNs on average. The result indicated that PMPM expenditures will be lower for beneficiaries in health syst em based PSNs than those in physician based PSNs 55 months into demonstration. This number was derived by the coefficient of HPSN (0.9726) divided by the rate of s hrinkage per month ( 0.180). Discussion The study found there were statistically significant differences in healthcare utilization and PMPM expenditures by Medicaid enrollees between health system based PSNs and physician based PSNs in reform counties over time. Initial findings showed higher levels of utilization and expenditures over physician b ased PSNs, but utilization and expenditures decreased at a greater rate over time. It was interesting to find that average healthcare utilization and expenditures were higher for enrollees in health system based PSNs during the post reform period. There a re some potential reasons for this result. First of all, two of the primary safety net hospitals, one in Broward County and one in Duval County, elected to create health

PAGE 112

112 system based PSNs at the moment of Medicaid reform (Landry et al., 2011). Safety net h ospitals, which are the origin of health system based PSNs, are more likely to have Medicaid enrollees with worse health status, more healthcare utilization, and higher expenditures. Therefore, in theory, the initial results from the model are fully consis tent with this supposition. Physician groups serving primarily the Medicaid population, including a large pediatric practice and minority physician network, also chose to create physician based PSNs in the reform counties (Landry et al., 2011). Previous re search (Lemak et al., 2004) that management, sophisticated information technology and medical management expertise, and reduce d healthcare expenditures. In addition, as hospitals position themselves to become integrated systems, many are joining forces and purchasing physician practices, leaving fewer independent hospitals and physicians. Greater market share gives these health systems more control, which can drive up health costs with increasing hea lthcare utilizations. These reasons can explain the variations in utilization and expenditures between two PSNs after the post reform period. The other finding was that enrollees in health system based PSNs had more decreasing trends in utilization and ex penditures over time. There are several possible explanations. First, the effect of the level of integration will not be immediate and its influence on healthcare utilization and expenditures could take time to have an effect. The finding s related to ED vi sits and inpatient utilization are also consistent with the result of the study by Gabow et al. (2003), which tested hospital utilization using comparative analysis among Denver Health and Hospital Authority, urban public hospitals, and urban community hea lth centers. They found that average length of stay

PAGE 113

113 and emergency room visits in Denver Health were reduced. The authors defined Denver Health as a fully integrated delivery system for vulnerable populations. In addition, Baker et al. (2014) found vertical integration was associated with a lower rate of hospital admissions. The authors confirmed that the integrated delivery system benefits the patient and the system due to the improvement of the coordination of care and serves as a model for the US safety n et. This argument can describe more integrated delivery systems available to align healthcare facilities, programs, or services and offer a coordinated continuum of healthcare to Medicaid beneficiaries over time , as explained in the theoretical framework. Second, this result may be associated with the type s of healthcare services between two PSNs. It is possible that health system based PSNs provide more preventive care to their beneficiaries than physician based PSNs, resulting in higher costs in the short run , but reduction in the need for ED visits and inpatient stays in the long run. Third, reduction in ED visits and inpatient hospital utilization is driving the observed changes in PMPM expenditures in the long run for more integrated healthcare delivery systems. Previous studies resulted in reduction of expenditures of healthcare services, followed by decreases in utilization of ED visits and inpatient services (Vogel et al., 2004). The result s on prescription drug use were opposite from the study hypothesis. While initially prescription drug utilization was higher, this trend reversed over time and frame. One possible explanation is that having health information systems, such as electronic health records (EHRs), in a more integrated delivery system may be a significant tool to improve communications between healthcare providers and health organizations across care settings , thus

PAGE 114

114 reduce wasteful duplication of healthcare services. Baicker et al. (2 013) argued that efforts to promote integrated and coordinated care surrounding the implementation of the ACO initiative can generate incentives for potentially anticompetitive horizontal and vertical integration. However, the meaningful use of health IT m ay be an example of such an opportunity to improve the allocation of health care resources while keeping costs down. According to their study, coordinated care using health IT may help shrink the number of unnecessary prescription drugs in more integrated delivery systems. Another possible explanation is that hospital based PSNs may utilize a restricted formulary of prescription drugs that prescribers within that PSN can prescribe under general use. They are selected by drug class, efficacy, safety profile, and price and can act as a cost containment mechanism and reduction in duplication of therapy within drug classes. Therefore, these reasons can explain the declining trend in prescription drug use for health system based PSNs over time. Specific Aim 3 Di scussion satisfaction with their health plans, overall healthcare, personal doctor, and specialists between physician based PSNs and health system based PSNs. atisfaction Hypothesis 5. Medicaid enrollees in health system based PSNs will have higher satisfaction with health plans, healthcare, personal doctor, and specialist when compared to those in physician based PSNs. This hypothesis was not supported by the study findings, since the trends in satisfaction with health plans, overall healthcare, personal doctor, and specialist for beneficiaries enrolled in health system based PSNs

PAGE 115

115 were declining at a greater rate over time compared to beneficiaries enrolled in physician based PSNs. Discussion This study compared the experiences Medicaid enrollees had regarding their health plan and healthcare services between health system based PSNs and physician based PSNs with the underlying hypothesis that differences in the degree of integration The first finding showed that enrollees in health system based PSNs had higher satisfaction w ith their health plans and healthcare. It may confirm that the more integrated organization s have higher coordination of care during the post reform period. Relatively, beneficiaries can have more positive experience in this kind of organization, since the y have higher availability and accommodation in terms of access to care. The other finding, which showed that the reduction in satisfaction trend for enrollees in health system based PSNs, is consistent with the results of some previous studies. The stud y by Schiller et al. (2010) investigated the variations in ratings and reports of care between MediPass and provider sponsored organizations (PSOs), using CAHPS data for Medicaid beneficiaries in Florida. MediPass is the least integrated healthcare deliver y system in terms of organizational structure among the Florida Medicaid managed care plans, since this plan is organized by primary care physicians or groups of physicians who acted as gatekeepers. PSOs are the origins of Provider Service Networks (PSNs), which are the most integrated delivery system. They found that enrollees in PSO s had lower specialists rating than those in Medipass, while enrollees in PSOs did n o t rate their health plans, personal doctor, and overall healthcare differently from enrolle es in MediPass. The study by Roland (2013) also

PAGE 116

116 found that patients in the more integrated organizations were less positive about their experience than those in less integrated ones. Therefore, the previous research showed that Medicaid enrollees and patie nts in the more integrated delivery system provided lower ratings of satisfaction than those in the less integrated delivery system. The authors suggested that the findings could be attributed to the fact that patients in the more integrated systems felt t hat they were less likely to have the same primary care physician than they expected. Along with the results of the previous research, enrollees in more integrated systems fel t they are being managed, therefore, they were less satisfied with their overall care. Also, reduction in utilization would be directly linked to decrease in satisfaction, since patient may not feel they are getting enough healthcare services and hav e barriers to access care. Last, it may not be easier to find the differences in enroll ees satisfaction than expected, since it may also have been too soon for beneficiaries enrolled in PSNs to assess their experience. A longer time frame can be added into the analysis to monitor more precisely any long term relationship satisfaction and the two types of PSNs. Policy Implications Florida is one of several states that remodeled their managed care programs to reduce unnecessary healthcare services and expenditures by Medicaid enrollees. PSNs were introduced as a new type of Florida Medicaid managed care and classified into physician based PSNs and health system based PSNs , depending on the organizational structure and characteristics. According to the result of this study, physician based PSNs showed promise in overall cost savings in the short to mid term, whereas health system based PSNs will show their benefits and strengths later. The integration allows health system based

PAGE 117

117 PSNs to reduce healthcare utilization and expenditures. Findings presented here should prove helpfu l to Florida Medicaid program directors and managed care organization administrators who have an intimate knowledge of their own program, but who lack an understanding of how healthcare utilization, spending, and outcomes are different between the two type s of PSNs. This study may help the state of Florida to assess the impact of these organizations on utilization and expenditure at the state level and drive future policy decisions to deploy this form of managed care statewide. In addition, it gives other states, facing similar decisions to reform their Medicaid managed care system, information to decide whether to adopt a similar plan or to consider other interventions to reduce unnecessary utilization and expenditures of health services by Medicaid benefi ciaries, help manage their public program s more efficiently within their individual budgetary constraints, and to improve Medicaid However, this study was conducted with enrollees in only two urban counties. Urban counties already have existing models that are ready to serve Medicaid recipients in a more integrated delivery system compared to rural counties. This will be a particular challenge for PSNs in rural areas, since the ability of PSNs in rural areas will be limited in terms of more integrated delivery system s and PSNs in rural area s will require considerable technical assistance, pooled resources, and local leadership. So, we encourage policy makers to strive to better understand why health process es and outcomes are different for the two PSNs and how to establish a more integrated delivery system of Medicaid managed care in rural area s . It may also be necessary for policy

PAGE 118

118 makers to consider carefully the differences in healthcare process and outcomes by enrollees bet ween a more or less integrated system in urban and rural counties. Last, we expect to see the role of a ccountable c are like organizations in a community, since PSNs resemble ACOs , which emerged from the Patient Protection and Affordable Care Act in 2010 in terms of organizational characteristics and objectives of organizations. In the first national survey of ACOs (Colla et al., 2014), it was found that 75% of ACOs were either physician led or jointly led by physicians and hospitals. The other recent study (Epstein et al., 2014) found only modest baseline differences in the use of inpatient services between ACOs with a hospital and those without one, so they suggested it would be necessary to track performance over time to see how ACOs with hospitals compare d to ACOs without hospitals. Therefore, findings from this study may help inform whether there are advantages or opportunities to ACOs jointly led by physicians and hospitals compared to physician led and hospital led ACOs. Limitations There are some pote ntial limitations in this study. First, the study design is a nonequivalent comparison group design, which is also a quasi experimental design. This study design consist of a treatment group (health system based PSNs) and one comparison group (physician based PSNs). Unlike the groups in the true experiment, however, observations in nonequivalent comparison groups are not randomly assigned, so causal statements about treatment effects may be substantia lly weakened. The study also used separate samples measured at two time periods. Using the same individuals in the pre and post reform periods would allow us to better control for th status.

PAGE 119

119 We were unable to control for possible threats to internal validity of selection, known as selection bias , since we used Medicaid claims and eligibility data set to identify Medicaid recipients and measure their utilization, expenditures, and ou tcomes. However, for this study, we used the same exclusion criteria for the study population and the difference in difference approach to measure the outcome differences between two groups , not because of other possible risk factors , and this design still can offer valid comparison of population based treatment effects. In addition, this study compared the population mean of utilization and expenditures by Medicaid enrollees before and after Medicaid reform utilizing the two types of PSNs in Broward and Duval counties over time. This study did n o t measure changes in use and expenditures of health services for individual enrollees. However, if individual levels were a unit of analysis for this study, this would threaten the external validity because Medica id beneficiaries can drop out of the Medicaid program and may not be enroll ed continuously. Continuously enrolled would lead to too much of the sample being dropped. For this reason, this study measured uses and expenditures of health services with a membe r month approach. Also, we conducted the sensitivity analysis with three and six month observations to solve this limitation and found the same trends with the base models. Third, we assumed that changes in utilization, expenditures, and outcomes between enrollees in physician based and health system based PSNs in reform counties were not due to unobserved factors that also affected the healthcare process and outcomes, such as different behaviors of physicians and patients between physician based and heal th system based PSNs. To make this assumption as plausible

PAGE 120

120 location, eligibility status, and health status. However, the risk scores are not a perfect reflection of enrolle Medicaid enrollees that impact healthcare utilization, expenditures, and outcomes. Finally , this study limited the study population to Medicaid Reform participants from two urban counties , Broward and Duval , who enrolled in MediPass and PSN plans. MediPass and PSN enrollees in rural counties, Baker, Clay, and Nassau, were not included in the study because of their incomplete data. Therefore, it is not advised to generalize the findings fro m this study to the general Medicaid Reform population, including beneficiaries enrolled in rural counties without further validation. Future Research Although there are limitations to this study, this study was able to present significant findings reg arding variations in utilization, expenditures, and outcomes between health system based and physician based PSNs . It also points to some future research opportunities that would investigate the program in more depth or help address some limitations mentio ned in the previous section. First, a study over a longer period of time can be carried out when more recent Medicaid data become available. This study using the data from FY2006 07 to FY2009 10 reported over time the trends in utilization and expenditures for enrollees in health system based PSNs decreased at a greater rate compared to those in physician based PSNs. A future study with a longer data set can confirm whether or not this finding is consistent, which may show that a more integrated delivery sy stem will have better performance in terms of Medicaid healthcare utilization and expenditures.

PAGE 121

121 Second, we found that the trends in utilization and expenditures declined more for enrollees in health system based PSNs compared to those in physician based PSNs. However, this study could n o t identify the factors that are associated with the variations in the healthcare process and outcomes between enrollees in physician based PSNs and in health system based PSNs. There are several potential reasons, includi ng organizational factors, market factors, and geographic factors. We can conduct the future study to find driving factors leading to the observed variations in utilization, expenditures, and outcomes between the less and more integrated healthcare deliver y system s organizational culture, meaningful use of HIT, health system based PSNs and physician based PSNs penetration rates, number of health system based PSNs and physician based PSNs, and market co mpetition of Medicaid managed care markets. Also, it will be important to investigate the attitude or culture of healthcare providers in health system based PSNs and physician based PSNs using in depth interview s or survey instrument s . Third, in the near future, it will be important to see the impact of process of care on outcomes in terms of patient perspectives by health plans, since their experience in the process of care would have direct links to their satisfaction. We can measure t he process of care as performance on important dimensions of care and service using HEDIS (Healthcare Effectiveness Data and Information Set) indicators, which include asthma medication use, comprehensive diabetes care, and breast cancer screening , and com as satisfaction by health plans.

PAGE 122

122 In addition, Medicaid enrollees can switch or switch back their health plans between Medicaid managed care types, especially between less and more integrated systems. However, this study could n o t capture the change between Medicaid managed care types, such as how many picked physician based PSNs or health system based PSNs and stayed, or switch ed to health system based PSNs or physician based PSNs or switch back to their original plans, and the effect of switching is another avenue for future study. Finally , outcome indicators for this study were more interpers onal. These measurements were report. Self reported data is good for measuring health status, functional status determination, and personal health practices since it has a solid underlying logic in that it is transparent and simple to understand. However, there is some concern about accuracy, including recall bias, overly positive or negative bias, and interpretation problems. Outcome indicators that can be measured by technical outcomes, such as the absence of postsurgical complicat ions or the successful management of chronic disease based on current standards of practice will be valuable for future research . Conclusion s To our knowledge, this study represents one of the first attempts to provide empirical evidence on the differenc es in healthcare utilization, expenditures, and outcomes by Medicaid enrollees between physician based and health system based PSNs using a difference in difference approach. The differences in types of PSNs were explained by the concept of integration and several organizational theories to identify the structural difference between both. Based on those approaches, health system -

PAGE 123

123 based PSNs are a more integrated healthcare delivery system, while physician based PSNs are a less integrated one. Although there were several limitations to this study, such as selection bias, external validity, and omitted variable bias, this study was able to conclude that health system based PSNs had higher levels of utilization, expenditures, and satisfaction compared to physici an based PSNs, but decreased at a greater rate over time. This result was interpret ed to mean that there were variations in healthcare utilization, expenditures, and outcomes between the more and less integrated healthcare delivery system s . However, why th e more and less integrated delivery systems perform differently is not clear in the data and statistical analysis . It will be important to monitor the different performance between two PSNs using more data and a longer study period and to determine how the se variations are being achieved. T his evaluation regarding the variations in utilization, expenditures, and outcomes between physician based PSNs and health system based PSNs in urban reform counties would enable policymakers and researchers to make appro priate changes to the healthcare delivery system and accept a specific Medicaid managed care organization as the best way to effectively administer their Medicaid program.

PAGE 124

124 Table 6 1. Discussion of Finding1 ( h ealth system based over p hysician based PSNs ) Average for Post Reform Possible e xplanations Utilization and PMPM e xpenditures Higher The origin of health system based PSNs: Safety net hospitals having more patients with worse health status The origin of physician based PSNs: Minority physician networks having utilization management, sophisticated health information system, and medical management expertise More integrated system having more resources and assets Provision of more preventive cares in health system based PSNs Satisfaction Higher Higher coordination of care Higher availability and accommodation in terms of access to care Table 6 2. Discussion of Finding2 ( h ealth system based over p hysician based PSNs) Trends over time Possible e xplanations Utilization and PMPM e xpenditures Decreasing More integrated delivery system: More complicated organizational structure taking time to have an effect Provision of more preventive care in health system based PSNs Available to align healthcare facilities, programs or services and offer a coordinated continuum of care Prescription d rug u se Decreasing Having health information system More restricted formulary of prescription drugs Satisfaction Decreasing Reduction in utilization patient can feel they are getting enough services and having access barriers to care. Greater management of enrollees Too soon for enrollees to assess their experience and perception

PAGE 125

125 LIST OF REFERENCES Andersen, R.M. 1995 . Revisiting the Behavioral Model and Access to Medical Care: Does it matter?. Journal of Health and Social Behavior 36: 1 10. Armitage, G.D., Suter , E., Oelke, N.D., Adair, C.E. 2009 . Health System Integration: State of the Evidence. International Journal of Integrated Care 9(17) . Baker, L.C., Bun dorf, M.K., Kessler, D.P. 2014 . Vertical Integration: Hospital Ownership of Physician Practices Is Associated with Higher Prices and Spending. Health Affairs 33(5): 756 763. Burns, L.R., Pauly, M.V. 2002. Integrated Delivery Networks: a Detour on the Road to Integrated Health Care?. Health Affairs 21(4): 128 43. Coase, R. H. 1937 . The Nature of the Firm. Economica 4(16): 386 405. Colla, C.H., Lewis, V.A. , Shortell, S.M., Fisher, E.S. 2014 . First National Survey of ACOs Finds that Physicians are Playing Strong Leadership and Ownership Roles. Health Affairs 33(6): 964 971. Conrad, D.A., Shortell, S.M. 1996 . Integrated Health Systems: Promise and Performance. Frontiers of Health Services Managemen t 13(1): 3 40. Cook, B.L. 2006 . Effect of Medicaid Managed Care on Racial Disparities in Health Care Access. Health Services Research 42(2): 124 145. Coughlin, T.A., Long, S.K., Triplett, T., Artiga, S., Lyons, B., Duncan, R.P., Hall, A.G. 2008 . Health Affairs 27(6): w523 532 . Clay, S.B. 1997 . Provider sponsored HMOs: make, buy, or joint venture?. Health Care Financial Management 51(3): 46 48. Crosson, F.J. 2005 . The Delivery System matters . Health Affairs 24(6): 1543 1548. Crosson, F.J. 2009 . 21 st centrury health care the case for integrated delivery system. New England Journal of Medicine 361(14): 1324 1325. Davis, G.S. 1997 . Provider sponsored Organizations: The Next Generation of Managed Care?. Managed Care . Available at: http://www.managedcaremag.com/archives/9709/9709.pso.html DeLeire, T., Dague, L., Leininger, L., Voskuil, K., Friedsam, D. 2013 . Wisconsin Experience Indicates that Expanding Public Insurance to Low Income Chil dless Adults has Health Care Impacts. Health Affairs 32(6): 1037 1045.

PAGE 126

126 Devers, K.J., Shortell, S.M., Gillies, R.R., Anderson, D.A., Mitchell, J.B., Erickson, K.M. 2009 . Implementing organized delivery systems. International Journal of Integrated Care 9(17). Donabedian, A. 1966 . Evaluating the quality of medical care. Milbank Mem orial Fund Q. 44(1): 166 203. Donabedian, A. 1988 . The quality of care: How can it be assessed? JAMA 260: 1743 1748. Donabedian, A. 1992 . The role of outcomes in quality assessment and assurance. Quality Review Bulletin 11: 356 360. Duncan, R.P., Hall, A.G., Brumbac k, B., Zhang, J., Chorba, L.P. 2007 . Medicaid Reform Enrollee Satisfaction: Baseline CAHPS Survey in Broward and Duval Counties. Available at: http://mre.ph hp.ufl.edu/publications/Medicaid%20Reform%20Enrollee%20Satisfa ction%20Baseline%20CAHPS%20Survey%20in%20Broward%20and%20Duval %20Counties%20.pdf Duncan, R.P., Lemak, C.H., Vogel, W.B., Johnson, C.E., Hall, A.G., Porter, C.K. 2007 . caid Provider Service Network Demonstration. Health Services Research 43(1): 384 400. Duncan, T ., & Duncan, S. 2004 . An introduction to latent growth curve modeling. Behavioral Therapy 35 : 333 363. Enthoven, A.C. 2009 . Integrated Delivery Systems: The Cur The American Journal of Managed Care 15(10): 284 290. Epstein, A.M., Jha, A.K., Orav, E.J., Liebman, D.L., Audet, A .J., Zezza, M.A., Guterman, S. 2014 . Analysis of Early Accountable Care Organizations Defines Patient, Structural , Cost, and Quality of Care Characteristics. Health Affairs 33(1): 95 102. Fama, E.F., Jensen, M. C. 1983 . Separation of Ownership and Control. Journal of Law and Economics 26(2): 301 325. Farley, D.O., Elliott, M.N., Short, P.F., Damiago, P., Kanouse, D .E., Hays, R.D. 2002 . Effects of CAHPS Performance Information on Health Plan Choices by Iowan Medicaid Beneficiaries. Medical Care Research and Review 59(3): 319 336. Florida S enate. 2010 . Overview of Medicaid Managed Care Programs in Florida. Available at: http://www.flsenate.gov/Committees/InterimReports/2011/2011 221hr.pdf Florida Statutes. 2012 . Available at: http://www.flsenate.gov/Laws/Statutes/2012/409.962

PAGE 127

127 Fox, M.H., Moore, J., Zimmerm an, M., Hill, S., Foster, C.H. 2001. The Effectivenes s of CAHPS among Women Enrolling in Medicaid Managed Care. Journal of Ambulatory Care Management 24(4): 76 91. Fuchs, V.R. 2005 . Health Care Expenditures Reexamined. Annuls of Internal Medicine 143(1): 76 78. Garrett, B., Davidoff, A.J., Yemane, A. 2003 . Effects of Medicaid Managed Care Programs on Health Services Access and Use. Health Services Research 38(2): 575 594. Gillies, R.R., Chenok, K.E., Shortell, S. M., Pawlson, G., Wimbush, J.J. 2006 . The Impact of Health Delivery System Organization on Clinical Quality and Patient Satisfaction. Health Services Research 41(1): 1181 1199. Gleason, S. C., Sokolov, J.J., Henshaw, C. 19 98 . Provider Sponsored Organizations: A Golden Opportunity in Medicare Managed Care. Family Practice Management 5(3): 34 4 5. Grumbach, K., Coffman, J., Vraniza . Independent Practice Association Physician Groups in California. Health Affairs 17(3): 227 237. Hall, A.G., Lemak, C. H., Landry, A.Y., Duncan, R.P. 2012 . Incentives for Healthy Be Program. Journal of Primary care and Community Health 4(2): 112 118. Hall, A.G., Landry, A.Y., Lemak, C .H., Boyle, E.L., Duncan, R.P. 2013 . Reported Experiences with Medicaid Managed Ca re Models Among Parents of Children. Maternal Child Health J ournal 18(3): 544 553 . Halstead, D., Becherer, R.C. 1998 . Surveying Patient Satisfaction among Disadvantaged Managed Care Customers. Health Care Strategic Management 16 (11): 15 9. Harman, J.S. , Duncan, R.P. 2009 . An Analysis of Medicaid Expenditures before and Center for Medicaid and the Insured. counties.com/docs/pdfs/an analysis of med icaid expenditures before and after implementation of florida%27s medicaid reform pilot demonstration june 2009.pdf?sfvrsn=0 Harman, J.S., Lemak, C.H., Al Amin , M., Hall, A.G., Duncan, R.P. 2011 . Changes in Per Member Per Month Expenditures after Imple Reform Demonstration. Health Services Research 46(3): 787 804. Harman, J.S., Hall, A.G., Lemak, C.H., Duncan, R.P. 2013 . Do Provider Service Networks Result in Lower Expenditures Compared to HMOs and Primary Care

PAGE 128

128 Case Florida Center for Medicaid and the Insured. Health Services Research 49(3): 858 877. Highsmith, N., Somers, S.A. 2000 . Medicaid Managed Care: From Cost Savings to Accountability and Quality Improvement. Evalua tion and the Health Professions 23(4): 385 396. Hill, I., Schwalberg, R., Zimmerman, B., T ilson, W. 1999 . Achieving Service Integration for Children with Special Health Care Needs. The Division of Children with Special Health Care Needs, Maternal and Child Health Bureau, Health Resources and Services Administration, Rockville, MD. Available at: http://www.jhsph.edu/research/centers and institutes/womens and childrens health policy center/publications/cshcn vI 1.pdf Holahan, J., Zuckerman, S., Evans, A., Rangarajan, S. 1998 . Medicaid Managed Care in Thirteen States. Health Affairs 17(1): 43 63. Holahan, J., Headen, I. 2010. Medicaid Coverage and Spending in Health Reform. Keiser Commission on Medicaid and the Uninsured. http://kaiserfam ilyfoundation.files.wordpress.com/2013/01/medicaid coverage and spending in health reform national and state by state results for adults at or below 133 fpl.pdf Hurley, R.D., Paul, J. 1993 . Managed Care in Medicaid: Lessons for Policy and Program Design . Ann Arbor, Michigan: Health Administration Press. Iezzoni, L.I. 1997 . Assessing Quality using Administrative Data. Annuals of Internal Medicine 127(8): 666 674. Iezzoni, L.I. 2003 . Risk Adjustment for Measuring Healthcare Outcomes . 3 RD edition. Ann Arbor, Michigan: Health Administration Press. Iglehart, J.K. 2012 . Ex p anding Eligibility, Cutting Costs A Medicaid Update. New England Journal of Medicine 366: 105 107. Jensen, M.C., Meckling, W.H. 1976. Theory of the Firm: Managerial Be havior, Agency Costs and Ownership Structure. Journal of Financial Economics 3(4): 305 360. Johnson, C.E., Lemak, C.H., Hall, A.G., Harman, J.S., Zhang, J., Duncan, R.P. 2010 . Outsourcing Administrative Functions: Service Organization Demonstrations an d Florida Medicaid PCCM Program Costs. Journal of Health Care Finance 37(1): 1 12. Kautz, C.M., Gittell, J.H., Weinberg, D. B., Lusenhop, R.W., Wright, J. 2007 . Patient Benefits from Participating in an Integrated Delivery System: Impact on Coordination of Care. Health Care Management Review 32(3): 284 294 .

PAGE 129

129 Keiser Commission on Medicaid and the Uninsured. 2007 . Medicaid Enrollment and Spending Trends. h ttp://kaiserfamilyfoundation.files.wordpress.com/2013/01/7523_02.pdf Keiser Commissio n on Medicaid and the Uninsured. 2008 . Florida Medicaid Reform Waiver: Early Findings and Current Status. http://www.dhcs.ca.gov/provgovpart/Documents/Waiver%20Renewal/Florida%20 Medicaid%20Reform%20Waiver%20Oct2008%20(2).pdf Keis er Commission on M edicaid and the Uninsured. 2012 . Medicaid Managed Care: Key Data, Trends, and Issues. http://kaiserfamilyfoundation.files.wordpress.com/ 2013/01/8046 02.pdf Keiser Commission on Medicaid and the Uninsured. 2013 . Medicaid and Its Role in State/Federal Budgets and Health Reform. http://kaiser familyfoundation.files.wordpress.com/2012/05/8139 03.pdf Keller, S., Epstein J., Hays, R.D., Matthew, R.A., Zaslavsky, A. M., Hepner, K.A., Cleary, P.D. 2005 . Methods Used to Streamline the CAHPS Hospital Survey. Health Services Research 40(6): 2057 2077. Kenen, J. 2012 . Health Policy Briefs: Medicaid Reform. Health Affairs . Kirby, J.B., Machlin, S.R., Co hen, J.W. 2003 . Has the Increase in HMO Enrollment within the Medicaid Population Changed the Pattern of Health Service Use and Expenditures?. Medica l Care 41(7): III24 34. Koch, A.L. 2002 . Financing Health Services. Introduction to Health Services . 6 th edition. Albany, New York: Delmar. Kocher, R., Sahni, N.R. 2010 . Physicians versus Hospitals as Leaders of Accountable Care Organizations. New England Journal of Medicine 363(27): 2579 2582. K odner, D.L., Spreeuwenberg, C. 2002 . Integrated Care: Meaning, Logic, Applications, and Implications. International Journal of Integrated Care 2(14). Landr y, A.Y., Lemak, C.H., Hall, A. 2011 Successful Implementation in the Public Journal of Public Health Management Practice 72(2): 154 163. Lee, K., Wan, T,T. 2002 . and pat ient outcome. Health Services Management Research 15(4):234 44. Lemak, C.H., Hall, A.G., Johnson, C.E., Saxen a, P., Aftuck, C., Johnson, C. 2004 . Medicaid and the Uninsured. Un iversity of Florida. Available at: http://fcmu.phhp.ufl.edu/research/focus/pdf/Report05 04 2004 EvalOfFlMinorityPhysicianNetworkProg.pdf

PAGE 130

130 Lemak, C.H., Hall, A.G., Saxeda, P., Zhang, J., A lbertie, M., Chorba, L.P. 2005 . Utilization Analysis of the Minority Physician Network and Provider Service Network Programs. Florida Center for Medicaid and the Uninsured. University of Florida. Available at: http://fcmu.phhp.ufl.edu/research/focus/pdf/Report15 12 2005 NetworkProgramsUtilization.pdf Lifton, J.G. 1996 . Assessing Options for Developing the Continuum of Care. Healthcare Financial Management 50(10). Lin, B.Y.J, Wan, T.T.H. 1999 . Performance: A Contingency Strategic Management Perspective. Journal of Medical Systems 23(6): 467 484. Mehrotra, A., Epstein, A.M., Rosenthal, M.B. 2006 . Do Integrated Medical Groups Provide Higher Quality Medical Care than Individual Practice Associations? Annals of Internal Medicine 145: 826 833. Mortensen, K. 2010 . Copayments did not Reduce Medicaid Enrollees' Nonemergency Use of Emergency Departments. Health Affairs 29(9): 1643 1650. Newhouse, R.P., Mills, M.E., Johantgen, M., Pronovost, P.J. 2003 . Is there a Relationship between Service Integration and Differentiation and P atient Outcomes?. International Journal of Integrated Care 3(10). Oberlander, J. 2010 . Long time coming: why health reform finally passed. Health Affairs 29(6): 1112 1116. Pfeffer, J., Salancik, G.R. 1978 . The External Control of Organizations: A Resource Dependence Perspective . New York: Harper and Row. Peterson, C.L., Gabe, T. 2010 . Health Insurance Premium Credits in the Patient Protection and Affordable Care Act. Congressional Research Service. at: http://www.ncsl.org/documents/heal th/hlthinspremcredits.pdf Philip, K.A., Morrison, K.R., Andersen, R.M., Aday, L. 1998 . Understanding the Context of Healthcare Utilization: Assessing Environmental and Provider related Variables in the Behavioral Model of Utilization. Health Services Res earch 33(3): 517 596. Quilgley, D.D., Scanlon, D .P., Farley, D.O., Vries, H.D. 2003 . The Utility of CAHPS for Health Plans. RAND. http://www.rand.org/content/dam/rand/pubs/working_papers/2004/RAND_WR11 5.pdf Roland, M. et al. 2013 . Does Integrated Care Deliver the Benefits Expected?. RAND. Available at: http://www.rand.org/content/dam/rand/pubs/research_briefs/RB9700/RB9703/RA ND_RB9703 .pdf

PAGE 131

131 Rawlings Sek unda, J., Curtis, D., Kaye, N. 2001 . Emerging Practices in Medicaid Primary Care Case Management Programs. National Academy for State Health Policy. Available at: http://aspe.hhs.gov/health/reports/pccm/report.pdf Regestein, M., Anthon y, S .E. 1998 . Medicaid Managed Care for Persons with Disabilities. Economi c and Social Research Institute, Urban Institute. Available at: http://www.urban.org/UploadedPDF/occa11.pdf Ries elbach, R.E., Kellermann, A.L. 2011 . A Model Health Care Delivery Syst em for Medicaid. New England Journal of Medicine 364(26): 2476 2478. 2004 . Physician Organization and Care Management in California: From Cottage to Keiser. Health Affairs 23(6): 51 6 2 . Robins on, J.C., Casalino, L.P. 1996 . Vertical Integration and Organizational Networks in Health Care. Health Affairs 15(1): 7 22. Robinson, J.C. 1997 . Physician Hospital Integration and the Economic Theory of the Firm. Medical Care Research and Re view 54(1): 3 24. Robinson, J.C. 2001 . Organizational Economics and Health Care Markets. Health Services Research 36(1): 177 189. Robinson, J.C., Dolan, E.M. 2010 . Accountable Organizations in California. Integrated Health Care Association. Available at: http://www.iha.org/pdfs_documents/home/ACO_whitepaper_final.pdf Rosenbaum, S. 2010 . A Customary and Necessary Program Medicaid and Health Care Reform. New England Journal of Medicine 362(21): 1952 1953. Ros enbaum, S., Westmoreland, T.M. 2012 . T on the Medicaid Expansion: How will the Federal Government and States Proceed?. Health Affairs 31(8): 1663 1672. Salmon, R.B., Sanderson, M.I., Walters, B.A., Kennedy, K. , Flores, R., Muney, A.M. 2012 . A Collaborati ve Accountable Care Model in Three Practices Showed Promising Early Results on Costs and Quality of Care. Health Affairs 31(11): 2379 2387. Schiller, K.C., W eech Maldonado, R., Hall, A.G. 2010 . Patient Assessments of Care and Utilization in Medicaid Mana ged Care: PCCMs vs. PSOs. Journal of Health Care Finance 36(3): 13 23. Shenkin, B.N. 1995. The Independent Practice Association in Theory and Practice: Lessons from Experience. Journal of the American Medical Association 273:1937 42 .

PAGE 132

132 Shortell , S.M., Anderson, D.A., Gillies, R.R., Mitchell, J.B., Morgan, K.L. 1993 . Building integrated systems: The holographic organization. The Healthcare Forum Journal 36: 20 6. Shortell, S.M., Gillies, R.R., Anderson, D.A. 1996 . Remaking Health Care in Americ a , San Francisco:Josey Bass. Shortell, S.M., Casalino, L.P. 2008 . Health Care Reform Requires Accountable Care Systems. Journal of American Medical Association 300(1): 95 97. Shortell, S.M., Casalino, L.P., Fisher, E.S. 2010 . How the Center for Medicare and Medicaid Innovation should test Accountable Care Organizations. Health Affairs 29(7): 1293 1298. Shortel l, S.M., Gillies, R.R., Wu, F. 20 10 . United States Innovations in Healthcare Delivery. Public Health Review 32(1): 190 212. Silberman, P., Pol ey , S., James, K., Slifkin, R. 2002 . Tracking Medicaid Managed Care in Rural Communities: A Fifty State Follow up. Health Affairs 21(4): 255 263. Singer, S.J., Burgers, J., Friedberg, M., Rosenthal, M.B., Leape, L., Schneider, E. 2010 . Defining and Measur ing Integrated Patient Care: Promoting the Next Frontier in Health Care Delivery. Medical Care Research and Review 68(1): 112 127. Snail, T., Robinson, J.C. 1998 . Organizational Diversification in the American Hospital. Annual Review of Public Health 1 9:41 453. Sommmers, B.D., Baicker, K., Epstein, A.M. 2010 . Mortality and Access to Care among Adults after State Medicaid Expansions. New England Journal of Medicine 367: 1025 1034. Suter, E., Oelke, N.D ., Adair, C.E., Armitage, G.D. 2009 . Principles for S ucces sful Health Systems I ntegration. Healthcare Quarterly 13: 16 23. Tai Se ale, M. , Freund, D., LoSasso, A. 2001 . Racial Disparities in Service Use among Medicaid Beneficiaries after Mandatory Enrollment in Managed Care: A Difference in Difference Approach. Inquiry 38: 49 59. Tollen, L. 2008 . Physician Organization in Relation to Quality and Efficiency of Care. Available at: http://www.commonwealthfund.org/usr_doc/Tollen_physician_org_quality_efficie ncy_ 1121.pdf Vogel, B., Johnson, C., Radcli ff, T., Veazie, P., Dewald, L. 2004 Medicaid Provider Service Networks Demonstration Project. Florida Center for Med icaid and the Uninsured. University of Florida.

PAGE 133

133 Wan, T. T.H., Ma, A., Lin, B.Y.J. 2001 . Integration and the Performance of Healthcare Networks: do Integration Strategies Enhance Efficiency, Profitability, and Image?. International Journal of Integrated Care 1(1). Wa n, T.T.H., Lin, B.Y.J., Ma, A. 2002 . Integration Mechanisms and Hospital Efficiency in Integrated Health Care Delivery Systems. Journal of Medical Systems 26(2): 127 143. Wang, B.B., Wan , T.T., Clement, J., Begun, J. 2001 . Vertical Integration Strategies and Hospital P erformance. Health Care Management Science 4(3): 181 91. Williamson, O.E. 1981 . The Economics of Organization: The Transaction Cost Approach. American Journal of Sociology 87(3): 548 577. Wooldridge, J. (20 06). Introductory Econometrics: A Modern Approach . 3 rd Ed. Thomson South Western. Zinn, J.S., Mor, V. 1998 . Organizational Structure and the Delivery of Primary Care to Older Americans. Health Services Research 33(2): 354 380. Zuckerma n, S., Brennan, N., Yemane, A. 2002 . Has Medicaid Managed Care affected Beneficiary Access and Use?. Inquiry 39: 221 242.

PAGE 134

134 BIOGRAPHICAL SKETCH Sinyoung Park received her M aster of P ublic H ealth in health policy and m anagement from Seoul National U niversity, Korea in 2007 and her Master of Science in health administration from the University of Colorado in 2009. she was actively involved in research projects, specifically in cross border utilization of healthcare services in dev eloping countries. These research experiences led her to pursue an internship at the WHO headquarters in Genève, Switzerland to study foreign policy and global health, which led her to pursue a PhD in Health Services Research. She joined the Department of Health Services Research, Management, and Policy at the University of Florida in August, 2010. She received rigorous training in the area of health policy, healthcare organizations as well as healthcare administration. Beyond coursework, she served as a gr aduate research and teaching assistant for four years. Her research field is focused on healthcare organization, healthcare delivery systems, Evaluation of Impact of Provider Service Networks in Florida Medicaid Managed Care She begins a tenure track appointment at Indiana University Purdue University in Ft. Wayne immediately following graduation.