This item is only available as the following downloads:
1 EVALUATING THE EFFECT OF THE ENHANCED BENEFITS REWARD S PROGRAM PARTICIPATION ON FLORIDA MEDICAID UTILIZATIONS AND EXPENDITURES By SHUO YANG A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PA RTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013
2 20 13 Shuo Yang
3 To my parents for their unconditional love and support
4 ACKNOWLEDGMENTS It has been a long journey for me to get here I really do not know where to start because there are so many people I want to thank but I think I need to say thank you to Dandan first. I do not know if she would be able to read this but she was the most importan t person in that part of my life who kept me company on my way here. Without her support and faith in me I would not have had the courage to make this huge leap in my career and none of this would have happen ed I was so fortunate to have her in my life a nd to share some of my best memories. Even if we had to part our ways, I will always remember what she has done for me and I wish her all the best and success in her future. I also want to thank my parents for their unconditional love and support. They wer e always there for me, helping me, guiding me through some of my most difficult times. They taught me to believe in myself, respect others, and always stay humble. What they have given me is priceless and I will never be able to repay them. I still remembe r the day I met Dr. Jeffrey Harman in his office 5 years ago It was that conversation that helped me make up my mind to join the field of health services research and started this amazing journey. He is the mentor that I always wanted to have. Not only di d he help me so much in my stud ies and research, his encouragement and faith in me were something I will treasure for the rest of my life. Additionally, I thank my committee member s : Dr Paul Duncan, Dr. Allyson Hall, and Dr. Barbara Curbow. I remember the days I had just joined the program and was amazed by the fact that their doors were always be open to me. They care d about me as a student and as a person. During my dissertation work, they helped me tremendously with all those discussions and suggestio ns Without them, I cannot
5 imagine any of this would be possible. Dr. Curbow and I did not know each other for too long, but I enjoyed every conversation we had and her presence and advice for my dissertation were irreplaceable. Last but not least, I am so g rateful for the fun and supportive cohort I was fortunate to have over the past several years. Matt Kukla, Robert Flemming, Damian Everhart, Kimberly Elliott, Melody Schiaffino, Jon Ruwe, and other friends in this program, for all the classes we had togeth for the study sessions over coffee and beers, for sharing y our awesome life stories for always being there for me, and for all the good times we had together, thank you!
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 11 LIST OF ABBREVIATIONS ................................ ................................ ........................... 12 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 15 Overview of Medicaid ................................ ................................ .............................. 15 Florida Medicaid and Medicaid Reform Demonstration ................................ .......... 18 Enrollment/Eligibility ................................ ................................ ......................... 18 Financing ................................ ................................ ................................ .......... 19 Delivery ................................ ................................ ................................ ............ 1 9 Florida Medicaid Reform Introduction ................................ ............................... 20 EBR Program Introduction ................................ ................................ ................ 22 Significance of the Study ................................ ................................ .................. 24 2 LITERATURE REVIEW AND CONCEPTUAL MODEL ................................ ........... 27 Literature Review ................................ ................................ ................................ .... 27 Financial Incentives for Seeking Preventive Care ................................ ............ 30 Financial Incentives for More Complex Behavior Chan ging ............................. 33 Current Status of EBR Program Evaluation ................................ ...................... 37 Conceptual Model and Hypotheses ................................ ................................ ........ 38 3 METHODS ................................ ................................ ................................ .............. 44 Data Sources ................................ ................................ ................................ .......... 45 Description of Dependent and Independent Variables ................................ ............ 46 Dependent Variable for Research Question 1 ................................ .................. 46 Independent Variables for Research Question 1 ................................ .............. 46 Dependent Variable for Research Question 2a ................................ ................ 48 Independent Variables for Research Question 2a ................................ ............ 48 Dependent Variable for Research Question 2b ................................ ................ 48 Independent Variables for Research Question 2b ................................ ............ 49 Dependent Variable for Research Question 2c ................................ ................ 49 Independent Variables for Research Question 2c ................................ ............ 49 Statistical Analysis ................................ ................................ ................................ .. 49
7 Research Question 1 ................................ ................................ ........................ 50 Research Questions 2a, b, and c ................................ ................................ ..... 51 4 RESULTS ................................ ................................ ................................ ............... 54 Descriptive Statistics ................................ ................................ ............................... 54 Model Statistics ................................ ................................ ................................ ....... 55 Research Question 1 ................................ ................................ ........................ 55 Research Question 2 ................................ ................................ ........................ 58 Inpatient days ................................ ................................ ............................. 58 ED visits ................................ ................................ ................................ ..... 60 Avoidable hospitalizations ................................ ................................ .......... 61 5 DISCUSSION AND CONCLUSIONS ................................ ................................ ...... 97 Discussion ................................ ................................ ................................ .............. 97 Descriptive Results ................................ ................................ ........................... 97 Medicaid Expenditures ................................ ................................ ..................... 98 Acute Care Utilizations ................................ ................................ ................... 100 Inpatient days ................................ ................................ ........................... 100 ED visits ................................ ................................ ................................ ... 100 Avoidable hospitalizations ................................ ................................ ........ 102 Policy Implications ................................ ................................ ................................ 103 Limitations ................................ ................................ ................................ ............. 104 Future Study ................................ ................................ ................................ ......... 106 LIST OF REFERENCES ................................ ................................ ............................. 108 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 114
8 LIST OF TABLES Table page 2 1 Top 10 credit earning behaviors in the EBR program from 2006 to 2010 (Hall et al. 2013) ................................ ................................ ................................ .......... 41 3 1 PQI indicators for avoidable hospitalizations ................................ ...................... 53 4 1 Annual total EBR credits earning and counts of occurrence by behavior type ... 62 4 2 Composition of EBR credits earning by se rvice type and participation level for FY2006 07* ................................ ................................ ................................ ........ 63 4 3 Composition of EBR credits earning by service type and participation level for FY2007 08* ................................ ................................ ................................ ........ 63 4 4 Composition of EBR credits earning by service type and participation level for FY2008 09* ................................ ................................ ................................ ........ 63 4 5 Average personal annual total expenditures an d acute care utilizations by FY* ................................ ................................ ................................ ..................... 63 4 6 Sample characteristics by EBR program participation level for FY2007 08* ...... 64 4 7 Sample characteristics by EBR program participation level for FY2008 09* ...... 65 4 8 Sample characteristics by EBR program participation level for FY2009 10* ...... 66 4 9 GEE base model for Medicaid expenditures ................................ ...................... 67 4 10 GEE base model with two years of EBR data for Medicaid expenditures ........... 68 4 11 GEE model using EBR credits earning for Medicaid expenditures* .................... 69 4 12 GEE model with two years of EBR credits earn ing for Medicaid expenditures ... 70 4 13 GEE model for Medicaid expenditures (more than 6 months enrollment) .......... 71 4 14 GEE model with two years of EBR credits earning for Medicaid expenditures (more than 6 months enrollment) ................................ ................................ ........ 72 4 15 GEE model for the likelihood of having any inpatient services ........................... 73 4 16 GEE model for the likelihood of having any inpatient services (using EBR credits earned) ................................ ................................ ................................ .... 74 4 17 GEE model for the likelihood of having any inpatient services (more than 6 months of enrollment) ................................ ................................ ......................... 75
9 4 18 GEE model for the likelihood of having any inpatient services (using EBR credits earned and more than 6 months of enrollment) ................................ ...... 76 4 19 GEE model for annual total inpatient days ................................ ......................... 77 4 20 GEE model for annual total inpatient days (using annual EBR earnings) ........... 78 4 21 GEE model for annual total inpatient days (more than 6 months of enrollment) ................................ ................................ ................................ .......... 79 4 22 GEE model for annual total inpatient days (using annual EBR earnings and more than 6 months of enrollment) ................................ ................................ ..... 80 4 23 GEE model for the likelihood of having any ED visits ................................ ......... 81 4 24 GEE model for the likelihood of having any ED visits (using annual EBR earnings) ................................ ................................ ................................ ............ 82 4 25 GEE model for the likelihood of having any ED visits (more than 6 months of enrollment) ................................ ................................ ................................ .......... 83 4 26 GEE model for the likelihood of having any ED visits (using annual EBR earnings and more than 6 months of enrollment) ................................ ............... 84 4 27 GEE model for rates of ED visits ................................ ................................ ........ 85 4 28 GEE model for rates of ED visits (using annual EBR earnings) .......................... 86 4 29 GEE model for rates of ED visits (more than 6 months of enrollment) ............... 87 4 30 GEE model for rates of ED visits (using annual EBR earnings and more than 6 months of enrollment) ................................ ................................ ...................... 88 4 31 GEE model for the likelihood of having any avoidable hospitalizations .............. 89 4 32 GEE model for the likelihood of having any avoidable hospitalizations (using annual EBR earnings) ................................ ................................ ........................ 90 4 33 GEE mo del for the likelihood of having any avoidable hospitalizations (more than 6 months of enrollment) ................................ ................................ .............. 91 4 34 GEE model for the likelihood of having any avoidable hospitalizations (using annual EBR earnings and more than 6 months of enrollment) ........................... 92 4 35 GEE model for annual rates of avoidable hospitalizations ................................ .. 93 4 36 GEE model for annual rates of avoidable hospitalizations (using annual EBR earnings) ................................ ................................ ................................ ............ 94
10 4 37 GEE model for annual rates of avoidable hospitalizations (more than 6 mo nths of enrollment) ................................ ................................ ......................... 95 4 38 GEE model for annual rates of avoidable hospitalizations (using annual EBR earnings and more than 6 months of enrollment) ................................ ............... 96
11 LIST OF FIGURES Figure page 1 1 A list of approved EBR credits earning activities (AHCA 20 06) ......................... 25 1 2 A list of selected redeemable products and supplies with EBR credits (AHCA 20 06) ................................ ................................ ................................ .................. 26 2 1 utilizations and outcomes (Bragdon 20 1 1) ................................ ........................ 42 2 2 Stock of health, health depreciation, and health investment ............................... 43 2 3 EBR program, health behaviors changing, and deman d of medical care. .......... 43
12 LIST OF ABBREVIATIONS AHCA Agency for Health Care Administration AHRQ Agency for Healthcare Research and Quality CDC Centers for Disease Control and Prevention CHIP m CMS Centers for Medicare and Medicaid Service s EBA Enhanced Benefits Account EBR Enhanced Benefits Rewards FFS Fee for service FGA Foundation for Government Accountability FMAP Federal Medical Assistance Percentage GEE Generalized Estimating Equations HH S Department of Health and Human Services ICU Intensive Care Unit LIP Low income Pool MCO Managed Care Organizations OTC Over the counter PCCM Primary Care Case Management PCP Primary Care Provider PMPM Per member per month PQI Preventive Quality Indicator s PSN Provider Service Network RCT Randomized Clinical Trials SSI Supplemental Security Income TANF Temporary Assistance for Needy Families
13 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial F ulfillment of the Requirements for the Degree of Doctor of Philosophy EVALUATING THE EFFECT OF THE ENHANCED BENEFITS REWARD S PROGRAM PARTICIPATION ON FLORIDA MEDICAID UTILIZATIONS AND EXPENDITURES By Shuo Yang December 20 13 Chair: Jeffrey S. Harman Maj or: Health Services Research Medicaid provides coverage for more than 62 million of the poorest Americans who cannot get health insurance elsewhere. Sharing the same dilemma as the federal health care system, Medicaid program s have experienc ed unsustainab le rates of spending growth over the past decade. As a state with one of the largest Medicaid programs, the Medicaid Reform One major element is the Enhanced Benefits Reward (EBR) program that uses financial incentives to promote healthy behaviors among Medicaid beneficiaries However, to date few studies have been conducted to systematically examine the effect of financial incentive programs in the public sector. This study examines the di fference in health care utilization and expenditures associated with different levels of EBR participation The study peri od is from fiscal year 2007 08 to 2009 10 The study population included Medicaid Reform participants in urban reform counties (Browar d and Duval) who enrolled in a Provider Services Network plan. Medicaid claims data and hospital discharge data were used to assess would be affected by the EBR prog ram. The amount of reward credits earned by
14 Medicaid beneficiaries participating in the program was extracted from Enhanced Benefits Information System data and used to categorize different levels of EBR program participation. For each research question, a time series model using a person year approach was used to ex amine the association between EBR program participation level and Med icaid utilizations/expenditures. Although only three years of data were available for this study, the results were able to sh ow that compared to non participants in the reform counties, EBR program participants were more likely to have lower health expenditures and lower odds of requiring acute care utilization The findings from s ensitivity analyses based on different enrollme nt length s of time and approaches to categorize the EBR program participation level s generally agree d with the main study. Despite certain issues that may negatively affect the the findings from this study suggest that the effectiveness of the EBR program can be
15 CHAPTER 1 INTRODUCTION Overview of Medicaid Medicaid and Medicare are governmental programs that provide health coverage to specific gr oups of people in the United States. While these two programs are very different, both are managed by the Centers for Medicare and Medicaid Services (CMS), a division of the Department of Health and Human Services (HHS). Medicare provides coverage to senio rs age 65 and older and younger beneficiaries with disabilities. As of June 20 10, the program covers more than 50 million people across the nation a nd in fiscal year ( FY ) 20 10 it cost about $389 billion. For Medicaid, since it was enacted in 1965 under T itle XIX of the Social Security Act, it has become an integral part of the United States health care system, covering medical services and long term care for millions of the sickest and poorest individuals of the population (Morone et al. 20 08). Among the individuals who rely on this program are children and their parents, people with disabilities and the elderly. Medicaid is the dominant payer for long term care and the single largest funding source for safety net care. Although Medicaid was initially es pressures and the rising uninsured rate have caused Congress to incrementally expand Medicaid eligibility to reach more Americans (Kaiser Family Foundation 20 08). W hen considering the history of Medicaid, the major events that led to its passing include the New Deal in 1935, State Vendor Payments in the 1950s and the Kerr Mills Act of 1960 (Morone et al. 20 08). As a result of these historical events, as well as the passed in 1965 as part of the Social Security Act. At the time of enactment, Medicaid
16 was thought to be a very minimal part of the 1965 Social Security Legislation and exp enditure estimates for the future were grossly underestimated (Morone et al. 20 08). Today Medicaid has grown to cover a large percent of the population and expenditures continue to rise. At present Medicaid is the third largest domestic program in the fe deral budget, behind only Social Security and Medicare (needs reference) The federal government and the states jointly finance Medicaid, and the states administer the program within broad federal guidelines (Morone et al., 20 08). The federal share of Med icaid spending is at least 50% in every state I t varies based on state per capita income relative to the national average and can be as high as 76% in the poorest state (Kaiser Family Foundation 20 09). Medicaid accounts for roughly one sixth of the natio term care (Kaiser Family Foundation 20 09). Medicaid spending is concentrated among the group of enrollees who demand more health care than others. Although children and their parents account for about 75% of Medicaid enrollees, they account for only about one third of the total spending each year. The other two thirds of spending goes to the elderly and people with disabilities who mak e up roughly 25% of enrollees. Regarding the coverage le vel, t he federal government sets minimum requirements for coverage, and the states can elect to apply for additional benefits through the 1115 waiver ( Kaiser Family Foundation 20 09) To qualify for Medicaid, a person must meet financial criteria and also belong to one of the groups that are eligible for the program. order to receive any federal matching funds, and these mandatory groups are pregnant women and children under age 6 with famil y income below 133% FPL, children age 6 to
17 below 50% FPL), and most elderly and persons with disabilities receiving Supplemental Security Income (SSI), for which income el igibility equates to 74% FPL for an individual (Kaiser Family Foundation 20 09). Medicaid normally covers the health services that are covered by private health insurance, but may also includes many additional services, such as dental and vision care and t ransportation, as well as long term care services, to better meet the needs of the enrollees that can be easily distinguished from individuals covered by general private insurance (Morone et al. 20 08). Since its enactment, Medicaid has undergone many cha nges, including the revisions during the years under the Reagan a dministration and the use of managed care in attempt to address uncontrolled costs that for many states threaten the viability of the program. Take the care delivery system of Medicaid as an example. Initially, most Medicaid enrollees received care on a fee for service (FFS) basis which means states reimburse hospitals, physicians, and other providers for each service provided. In the 1980s, many state Medicaid agencies started to steer away from FFS and adopted managed care models for health care delivery and financing. It was believed that the change would lead to better access to care, improved quality, and lower costs. There are generally two approaches to Medicaid managed care (Kaiser Fam ily Foundation 20 12). One is primary care case management (PCCM). Under PCCM, the enrollee is required to choose a primary care provider (PCP) who is responsible for his/her care coordination. Medicaid will pay an additional monthly fee on top of the FFS payment for medical services the PCP provided to reimburse them for the time and effort associated with care coordination and management. Many would argue that this model is
18 essentially FFS with little modification, but federal Medicaid law acknowledges PC CM as a form of managed care nonetheless. Recently, more and more state Medicaid agencies adopted risk based managed care as a delivery and financing model. Under this model, a fixed per member per month (PMPM) premium is paid to managed care organizations (MCOs) to cover a pre determined set of medical services to Medicaid enrollees. As of calendar year (CY) 20 11, more than 30 states in the United States use this model, at least in part, for Medicaid care delivery and fin ancing (Palmer and Pettit 20 12) Florida Medicaid and Medicaid Reform Demonstration The role of Medicaid in the health care system in Florida is critical. As of 20 10, thirds of nursing home days, and pays for about 51% of a ll health care (Roberta K. Bradford, Florida Medicaid: Program Overview 2010 ). Florida Medicaid and its companion program, Florida KidCare, cover a projected 3.19 million Florida residents for state fiscal year (SFY) 20 12 2013. It is the fourth largest M edicaid population in the nation, only behind California, New York, and Texas. The total spending of the program is estimated to be over $21 billion in FY 20 11 12 Enrollment/Eligibility In the state of Florida, Medicaid services are administered by the A gency for Health Care Administration (AHCA). Eligibility to receiv e Medicaid services is determined by the Department of Children and Families (DCF) and the Social Security Administration. In more detail, DCF determines Medicaid eligibility for a) low inco me families with children; b) children only; c) pregnant women; d) non citizens with medical emergencies; and e) aged and/or disabled individuals not currently receiving SSI
19 (Florida Department of Children and Families, http://www.myflfamilies.com/service programs/access florida food medical assistance cash/medicaid ). The Social Security Administration determine s the eligibility of SSI recipients. Financing The Medicaid program is jointly funded by federal and state governments. The federal share of the Florida Medicaid spending varies from year to year based on state per capita income relative to the national average and the economy in general. Th is matching funds system means that every dollar the state government spends on Medicaid provides a much greater benefit to Medicaid enrollees. Federal Medical Assistance Percentage (FMAP) for Florida Medicaid remained relatively stable at about 59% from F Y2005 07 and decreased to 56.8% in FY 20 08. To help states cope with the ir budget ary pressure s during the recession, the federal government increased its share of the costs. FMAP was increased to 67.6% in FY2009 and FY 20 10. In FY2011, the federal match drop ped back to 55.5% and in FY2013 the federal government will cover 58.1% of the Florida Medicaid spending which means $1 in Florida Medicaid funds yields $2.38 in benefits for Medicaid enrollees ( Kaiser Family Foundation 20 12) Delivery Excluding the coun ties participating in the demonstration, also as known as Florida Medicaid Reform, Florida currently covers 47 services either on a FFS basis or through managed care. Enrollees can receive care through FFS Provider Service Network (PSN), the statewide PCCM system (MediPass), or MCOs contracted with AHCA. Approximately 43% of all Florida Medicaid enrollees are covered under managed care with the remaining 57% enrolled in FFS programs (Appendix B of the
20 Florida Medicaid Program YEAR ). The FFS care delivery s ystem in Florida Medicaid is fragmented and does not have strong incentives for providers to provide coordinated and cost efficient care. In recent years, Medicaid spending has consistently outgrown the state revenue s that supports the program. From state fiscal year (SFY) 20 06 07 to SFY 20 11 12, Medicaid expenditure growth for Florida has averaged 6.5% per year. And total Medicaid expenditures will exceed $21 billion dollars in SFY 20 11 12 which represent about 33% of the state budget (Florida Social Serv ice Estimating Conference Report 20 11). To address these issues, on May 6 20 05, the Florida legislature Florida Medicaid Reform Introduction On October 19 20 05, the Florida Med icaid Reform Demonstration was approved by CMS and implemented in Broward and Duval counties starting July 1 20 06. One year later, the demonstration was expanded to Baker, Clay, and Nassau counties. The s application for a 1115 Research and Demonstration Waiver (Florida Medicaid Reform Application for 1115 Research and Demonstration Waiver 20 05). The state intends to change its role in chaser that ensures delivery of high quality and cost efficient care to its beneficiaries. Under the reform, managed care plans are paid risk adjusted premiums and have the flexibility to design customized benefit packages for different beneficiary groups. The state expects this change to encourage more market competition that will result in increased access to care and better efficiency of the Medicaid program. More importantly, patient responsibility and empowerment are emphasized in the reform. Measures are taken in
21 the reform to facilitate and encourage people to take a more active role in their own health care. Public reporting on performance measurement and consumer satisfaction on health plans and providers, along with other assistance from the system is expected to allow individuals to more actively take control of their health care management. Also, they will be rewarded if they assume more responsibility and exercise health ier behaviors. Overall, the goal of the reform is to increase access to care improve quality, enhance efficiency of care delivery, introduce more individual choice, and contain cost s to ensure the pro term sustainability There are four key elements under Florida Medicaid Reform: Risk adjusted premiums that are actuari ally equivalent to all services covered managed care plans on a monthly premium basis. It is expected that risk d that high medical spending due to untreated or poor managed chronic conditions can be reduced or avoided. Low income pool (LIP) is established and managed by the state to p ay safety net providers for their coverage of Medicaid, uninsured, and underinsured populations. eligible individuals who are willing to use their premiums to purchase insurance through their employers or pay for a private plan if self employed. behaviors by providing them financial incentives. By participating in the program, such as disease management, weight contr ol, or smoking cessation program participation, preventive care utilizations, and adherence to prescribed medications are rewarded. Earned credits are deposited into their Enhanced Benefits Accounts (EBAs) and can be used to purchase over the counter (OTC) health related goods at any Medicaid participating pharmac y Details of the program will be elaborated in the following section.
22 EBR Program Introduction healthy behaviors and perso nal responsibility, by rewarding good choices with financial a 1115 Research an active role in mana Reform Application for 1115 Research and Demonstration Waiver 20 05). This is what is anticipated to happen following the implementation of the EBR program. By rewarding enrollees to actively eng age in healthy behaviors, we will see improved health among participants and eventually lower Medicaid costs. Florida is one of the first states to attempt to incorporate financial incentives for enrollees in its state Medicaid program. Idaho, Michigan, We st Virginia, and Wisconsin are all implementing or contemplating a similar approach for their own Medicaid programs ( Commonwealth Fund 20 07). All Florida Medicaid Reform enrollees who enroll in a managed care plan are eligible participants of the EBR prog ram. For each eligible enrollee, an EBA is assigned. A seven member Enhanced Benefits Panel established by AHCA designates activities that generate rewards credits toward an n confirmed participation. Figure 1 1 show s the approved behavio rs that will earn credits for participants ( http://www.fdhc .state.fl.us/medicaid/Enhanced_Benefits/EB_Welcome_Letter_WEB_E NG_SPAN.pdf ). AHCA contracted with Affiliated Computer Services and Imager Software to manage and maintain the information system earning activities and processes credit deposit s
23 eligible to receive up to $125 rewards credits annually in his/her EBA. For easy to identify activities such as a wellness visi t, credits can be automatically deposited into that a provider with a matching diagnostic code. For more complicated behaviors such as participating in a disease management program, weight loss pr ogram, or smoking cessation, an individual need to submit a form to the health plan signed by both the provider and enrollee to qualify for EBR credits. The credits in an redeemed for cash under any circumstance but the beneficiar y can spend earned credits in Medicaid participating pharmacies for specific OTC health related produc ts. Figure 1 2 is a list of products beneficiary can purchase using their EBR credits (Georgetown University Healthy Policy Institute 20 08). Once credits are deposited into ended. The credits can still be used to purchase insurance, such as employer sponsored insurance, private insurance, or COBRA. However, if the r emaining credits 20 0% of FPL, the credits become unavailable to the individual and are returned to the state (Florida Medicaid Reform Application for 1115 Research and Demonstration Waiver 20 05). Through May 20 10, over 170,000 Florida Medicaid enrollees received EBR credits through participating in approved activities. Total credits distributed to the enrollees have exceeded $28.3 million. Among that total approxi mately $14.2 million have been redeemed for approved products (AHCA 20 10).
24 Significance of the Study Using financial incentives to promote an drawing growing interest, especially for the Medicaid population which is generally considered to have higher prevalence rates of unhealthy behaviors and more chronic illness than the population as a whole (Healthy People 20 10 20 00). Because the concept behind the EBR program is somewhat intuitive its effect on the public sector, including Medicaid, is difficult to predict. Prior to Florida Medicaid Reform, most of the programs using financial incentives to promote personal health behaviors occurred in the private sector as insurers or employers looked for approaches to co ntrol health care costs ( Kane et al. 20 04 ; Sutherland, Christianson, and Leatherman 20 08 ) While detailed di scussion will be provided in the next chapter, past studies only offer mixed behavior, lifestyle, and health care utilizations and expenditures. The EBR program in Florida Me dicaid Reform has generated great interest nationwide. A well structured study of the s expenditures will be one of the first to examine the value of this concept in the public sector a nd provide evidence and guidance for policymakers on both the federal and state level s
25 Figure 1 1. A list of approved EBR credits earning activities (AHCA 20 06)
26 Figure 1 2. A list of selected redeemable products and supplies with EBR credits (AHCA 20 06)
27 CHAPTER 2 LITERATURE REVIEW AND CONCEPTUAL MODEL Literature Review The concept of u sing financial incentives to promote healthy individual behaviors and preventive care utilization is not new. Although fewer attempts have been carried out in the p ublic sector, it has been adopted in the private sector for quite some time to encourage health y behavior chang es ( Higgins et al. 20 12 ) It is expected that with people taking more responsibility for their health, maki ng better choices regarding their health care, and using more preventive services, they will become healthier and lower the costs of medical care in the long run. Today, in the US and most other industrialized countries, chronic conditions have become the major threat to population health. And many of these conditions, including type 2 diabetes, coronary heart disease, and certain types of cancer, are all related to such unhealthy lifestyle and behaviors as poor diet, physical inactivity, substance abuse, u nder use of preventive services, and non adherence to disease treatment and medications. It is estimated that about 40% of annual total premature death in the US can be attributed to unhealthy behaviors ( Schroeder 20 07 ) More importantly, the prevalence of the unhealthy behaviors are higher among low income population s ( Cutler and Lleras Muney 20 10 ; Higgins and Chilcoat 20 09 ; Isaacs and Schroeder 20 04 ) For example, Medicaid benefi ciaries have high prevalence rates of unhealthy behaviors that lead to higher rates of morbidity and mortality than the general population (Healthy People 20 10 20 00). T he consequences of t hese unhealthy behaviors are a significant economic burden on the health care system. The Centers for Disease Control and Prevention (CDC) has estimated that being overweight or obes e can cost
28 about $150 billion of direct medical costs annually (Center s for Disease Control and Prevention 20 11). Promoting preventive ser vices is considered a viable solution for the continuously rising health care spending associated with chronic conditions and health behavior related risks despite some researcher s argu ing that offering more preventive care may add to health care costs rat her than saving s (Russell 20 07). It is true that, for example, cancer screening costs could exceed the savings when screenings are given to a general population and only a very small proportion of the people would have actually developed cancer without th e early screenings. However, when preventive services are provided to high risk population, the consensus is that they will be able to reduce acute care utilization including ED visits and preventable hospitalizations, and lower total health care expendit ures. A report on potential Medicare savings through preventive services and reducing health risks estimated that, with 10% of upward risk transitions prevented, FFS Medicare could save more than $65 billion every year ( Rula, Pope, and Hoffman 20 11 ) As for the Medicaid population, spending on acute care amounted to 64% of total Medicaid spending in FY2010 in the US and 70% in the state of Florida (Kaiser Family Foun dation 20 11). By ensuring access to such preventive services as vaccinations, screening tests, and other preventi on programs, evidence suggests that the rates of avoidable acute care utilizations and expenditures are expected to be reduced (Perlino 20 07) With the growing concern about unhealthy behaviors and the ir impact on population health, professionals have been trying to find a way to encourage behavior chang e and active participation in preventive services for their potential effect s on
29 improving h ealth and controlling health care costs. Among other measures, financial incentives targeting individuals have generated broad interest in the recent years and showed some promising results in encouraging people to engage in more healthy behaviors and prev entive activities. Theoretically, why do we use financial incentives to psychological perspectives. Economic incentives provided by these programs can be deemed an approach to sti mulate positive externality from demanding more preventive care or adopting more healthy behaviors. Past studies demonstrated that people are generally price sensitive toward preventive cares. For example, co payments are associated with less utilization o f preventive services and physical checkups ( Cherkin, Grothaus, and Wagner 1990 ; Lillard et al. 1986). And having health insurance has been linked to more mammogram and cervical cancer screenings (Kenkel, 1994). Therefore, financial incentives should be able to shift consumer demand curve for preventive cares and behavior changing programs like weight control or smoking cessation. From a making process es can be skewed by certain biases that would contribute to an individual s ( Berns, Laibson, and Loewenstein 20 07 ; Loewenstein, Brennan, and Volpp 20 07 ) rewards over the one s with greater long term gains, especially when higher initial costs are involved. Offering financial incentives should help people weigh the options and adjust the balance between their preferences and actions. This paragraph only serves the purpose of gi ving a very brief introduction to the theoretical background of using
30 financial incentives to encourage health y behavior chang es Discussion on why we choose financial incentives for the task can go further beyond. However, the proposed study aims to analy z e the effect of the EBR program rather than why the EBR program was chosen. Thus, for the rest of the chapter, we will be focusing on reviewing the evidence of the effectiveness of incentive programs implemented in the past. Financial incentive programs t argeting individuals can be categorized in different ways based on the design of the incentives (cash, credits, coupons, gifts, discounted goods and services, among others ), target population, or target behaviors. The EBR program is a credit earning progra m targeting Florida Medicaid beneficiaries with credits earned by participating in certain AHCA approved activities ranging from a simple child wellness check or having a flu shot to more complex behaviors like participation in a s top smoking or weight lo ss program. For more complex activities, individuals will be rewarded for both participat ion and final results (see Figure 1 1). As mentioned earlier, using financial incentives to encourage health behavior change is not anything new in the US, especially in the private sector. Based on different target behaviors, we will review evidence from past studies on the effectiveness of financial incentive programs. Financial Incentives for Seeking Preventive Care Vaccination is critical to population health and of ten a focal point of public health policies. Despite the success of using such vaccines against preventable diseases like influenza, measles, hepatitis A and B, and others in the U S ensuring adherence to recommended vaccination s especially for childre n and high risk population, is still an important component in public health strategy (Atkinson et al., 20 12). Previous studies
31 show that most incentive programs for vaccination were used primarily on children, low income families, and high risk adults ( Kane et al. 20 04 ) The structure of the incentives mostly includes voucher, lottery for small cash pri z es ( Moran et al. 1996 ) or free vaccin ation ( Nexoe, Kragstrup, and Ronne 1997 ; Satterthwaite 1997 ) Some of these programs were administered by the Women Infants, and Children (WIC) Supplemental Nutrition Program or Aid to Families with Dependent Children (AFDC) ( Birkhead et al. 1995 ; Hutchins et al. 1999 ; Kerpelman, Connell, and Gunn 20 00 ) Most of these studies found evidence that support the effectiveness of using financial incentives to increase receipt of pocket (OOP) costs. In the U S it was estimated that over 1.6 million new cancer cases would be diagnosed in 20 12 and about 577,190 Americans would die because of cancer in the ye ar. Regular cancer screenings by a health care professional can result in the detection and diagnosis of cancers at an early stage. Breast, colon, cervi cal and prostate cancers can be diagnosed early through screenings. Actually, cancers that can be detec ted or prevented by screening exams account for at least half of all new cancer cases (American Cancer Society 20 12 Cancer Facts and Figures). Both mortality and morbidity can be reduced by early detection and treatment through screening examinations. For mammography, two randomized clinical trials (RCTs) that reduced OOP costs of screening for low income and rural farm women showed effectiveness in achieving more completion of exams while two other studies using small gifts as incentives presented mixed results with o ne study suggest ing the incentive program was effective but the other not ( Jepson et al. 20 00 ) Another study tested the effect of
32 mailing incentiv es to a population of low income, underinsured women in the U S for mammography. One group was given free access to the test and the second group was offered an additional $10 with the free exam. Both groups had significantly higher rates of mammography w ithin one year than the control group ( Slater et al. 20 05 ) For cervical cancer screening, three studies examined the effect of using financial incentives to impr ove follow up visits after Pap smear test s among women from low income families. One provided free bus transportation to the patients and a second offered a $20 to $ 25 voucher off the clinic visit fee, and both studies found incentives effective ( Marcus et al. 1992 1998 ) A third study provided free bus passes to one group of women and a $15 voucher to the other g roup, but no significant differences were found between these two groups regarding their follow up visits ( Kaplan et al. 20 00 ) Stone and his colleagues conducted a meta analysis on the interventions that increase d adult vaccination s and cancer screenings ( Stone et al. 20 02 ) It included 33 mammography studies (2 using fina ncial incentives), 27 cervical cancer screening studies (3 with financial incentives), and 19 colorectal screening studies (5 with financial incentives). Based on the adjusted odds ratio from meta regression, patient financial incentives ranked as the firs t, second, and fourth most effective measures for improving the use of mammography, Pap test, and colorectal screening respectively Prenatal and postnatal care is essential to reduc e the risks associated with pregnancy and childbirth. Economically disadv antaged women usually have lower rates of regular prenatal and postnatal care. Paired with other conditions such as smoking, alcohol addiction, or young age of the mother makes them a particularly vulnerable population. Measures have been taken to improv e adherence to prenatal and postnatal
33 visits. In one study 20 5 Medicaid eligible women from Michigan were randomly assigned to three groups. Women in two test groups received either a $5 gift certificate or a $5 gift certificate plus a $100 raffle ticket for each prenatal appointment they kept. However, the researchers found no significant differences among the test groups and the control group in the number of prenatal and postnatal visits ( Laken and Age r 1995 ) Another study of prenatal care with 104 low income women used either a taxicab voucher or a gift of a baby blanket to promote prenatal care utilization. The result s showed a significant effect of the taxi voucher on attending the first prenatal visit when compared with the control group, but no effect was found for the group who received baby blankets ( Me lnikow, Paliescheskey, and Stewart 1997 ) Other studies on financial incentives and prenatal/postnatal care can also be found in the field ( Burr et al. 20 07 ; Kane et al. 20 04 ) incentive programs show a significant effect on prenatal and postnatal visits more often than not, but many findings are based on relatively small sample size s and this limitation has to be co nsidered when interpreting the results. Nonetheless, programs and trials using financial incentives to encourage using more preventive services showed promising results and evidence suggest s that on many occasions, a small incentive could make an impact o n the level of compliance Financial Incentives for More Complex Behavior Changing In most cases, p reventive care like vaccination or cancer screening require s only a one time action. Conversely behavior chang es, like smoking cessation or weight loss can be much more complex. They require long term efforts and significantly greater investments, both financially and psychologically. Adopting and sustaining these
34 lifestyle changes can be difficult for many people and financial incentives are often used to e ncourage consumers to pursue these changes. Financial incentives for smoking cessation have been widely used in workplace settings and sometimes in community programs. Over the years, there have been a substantial number of trials and programs designed to facilitate smoking cessation and evaluate the effectiveness of financial incentives for this particular purpose. A Cochrane review published in 20 08 examined 17 financial incentive programs for smoking cessation targeting adults in workplace settings from 1980 to 20 07 ( Cahill and Perera 20 08a ) The authors selected only RCTs using financial incentive scheme s, including cash rewards free nicotine patches, lotteries, and contingent payments to promote smoking cessation. In most of these studies, participants were required to reach lab verified abstinence over a certain period of time to be rewarded. The results suggest that most of these incentive based programs did increase participation rates and cessation rates while the incentives were in place, but the normal pattern of relapse was observed when the incentives were withdrawn. The effectiveness of financial incentives on long ter m cessation rates in workplaces is still in doubt. Quit and Win Contests is a population based smoking cessation program initiated by the Minnesota Heart Health Program in three Minnesota communities in the 1980s. The winning prize was a holiday to Disneyw orld in a raffle for people who could reach one month post program lab verified abstinence. The program was successful in increasing the quit rate at one month but the relapse rate at one year was still high ( Lando, Pechacek, and Fruetel 1994 ) The relative success of the program made it a model that was extended to national and international applications in the following years. A nother 20 08 Cochrane
35 review examined some RCTs based on Quit and Win model in community settings and failed to find conclusive proof on its effectiveness on smoking cessation ( Cahill and Perera 20 08b ) While there are many other programs and trials using certain types of financial incentives to help stop smoking in both workplaces and community settings, all suffered from trial de sign, small sample size, or other limitations, so current literature cannot provide sufficient evidence on how effective financial incentives are on helping smoking cessation, especially on the long term quit rate. Obesity and overweight are associated wit h higher level of cardiovascular diseases, diabetes, and some cancers. With the current trend, it has been projected that there will be 65 million more obese adults in the US by 20 30. The medical costs of obesity and treatment of associated diseases are es timated to increase by between $48 billion and $ 66 billion per year in the US by 20 30 ( Wang et al. 20 11 ) As it is one of the biggest concerns in population health, health professionals have been seeking an effective tool to help control weight and using financial incentives has been a popular choice. It is noteworthy that, although many programs used such usual form of incentives as lotteries for cash or gift s some programs chose a group or individual monetary contract that required participants to deposit a certain amount of money and the return of the money was contingent on weight loss ( Jeffery 20 12 ) It can be seen as a typ e of negative incentive so is less commonly used ( Kane et al. 20 04 ) Also, for weight control, financial incentives were often combined with other intervention approaches like health education classes, newsletter, or behavioral counseling (G oodman and Anise 20 06). Results from these studies somewhat agree on the short term benefit from financial incentives, especially at the beginning of the weight loss
36 process, but few show promising results on long term efficacy ( Sutherland et al. 20 08 ) Most of these studies were limited by small sample size and they varied widely in incentive size and program desi gn. Combined with the fact that many financial incentives were bound to other intervention methods, the effectiveness, especially long term effectiveness, of financial incentives and their value in addressing the overweight and obesity problem in public he alth is still unclear. There were also programs designed to use financial incentives to improve exercise program participation. These programs more often take place in schools and workplace settings ( Blue and Conrad 1995 ; Matson Koffman et al. 20 05 ) Studies show mixed results on whether providing financial incentives could increase exercise program participat ion or physical activity level ( Bloch et al. 20 06 ; Harland et al. 1999 ; Herman et al. 20 06 ; Sutherland et al. 20 08 ; Wing et al. 1996 ) To summarize, for complex behavior changes such as weight loss, smoking cess ation, or increasing physical activity level, we cannot find clear cut evidence to support the long term efficacy of using financial incentives although in most cases short term effectiveness can be expected. The reason s of the lack of conclusive evidence can be multi faceted. First, financial incentives are sometimes combined with other interventions in the program ; this makes it difficult to single out the effect of financial incentives. Second, many trials were limited by their target population and sam ple size which makes the results more vulnerable to bias. Third, considering the natural differences between these complex behaviors and simple one time action of preventive services, long term commitment and significantly more financial and psychological investments are needed from participants. Although financial incentives can increase individual adherence while they are in place,
37 it is uncertain if a regular pattern of relapse would take over once the incentives are withdrawn. In addition to the diffic ulty of assessing the effectiveness of financial incentive programs on helping health behavior change, very few studies were able to directly link incentive programs to health care expenditures and acute care utilization, probably due to the short time spa n and fragmented nature of the programs studied in the past. One study that examined the impact of financial incentives for prenatal care on childbirth outcomes and spending found that incentive program participation was significantly associated with lower rates of neonatal intensive care unit ( N ICU) admission s and health care expenditures in the first year of a ( Rosenthal et al. 20 09 ) As the resul ts seem encouraging, we should note that, in this program, financial incentives were provided to both patients and their providers. It remains unknown how much of the positive outcomes can be attributed to financial incentives to individuals. Simply put, c urrent evidence base d on individual financial incentives is, by no means, sufficient or conclusive. Despite all the above questions, using financial incentives to encourage health y behavior chang e and utilizing preventive care remains a common choice in pu blic health strategy. Current Status of EBR Program Evaluation As a relatively new incentive program, study results on the EBR program are scarce. Through June 20 10, total credits earned by enrollees exceed ed $28.3 million and approximately $14.2 million of that had been redeemed f or products (AHCA 20 10). Table 2 1 lists the top 10 credit earning behaviors in the EBR program from 20 06 2010 (Hall et al. 20 12). A report released by the Florida Foundation for Government
38 Accountability (FGA) in 20 11 took a s napshot of the changes from 20 08 to 20 10 in that can be linked to EBR credit earnings (Bragdon 20 11). Although it shows improvement in most measur es examined in the re port (Figure 2 1 ), no conclusion s can be drawn from this observational study regarding the most recent study conducted by Hall et al. (2012) examined determinants of EBR program awareness, participation, and credit spending. The findings suggested that being a non English speak ing non white, having a lower education level, and being without a personal doctor are all associated with lower odds of program awareness. These factors are also associated with a reduced likelihood o f participating in an EBR program while people with excellent self reported health status are more likely to engage. These are all very interesting findings that shed light on the better understanding of the EBR program and will facilitate further study o n the subject. Similar to the Florida Program (CHIP) who are up to date on their wellness checks will receive a quarterly $30 credit. The s tudy showed a significant impact of the program adherence to wellness visits when compared to children enrolled in Idaho Medicaid but not eligible for the program ( Greene 20 11 ) Considering the current status of EBR program evaluation, the need of a well structured study assessing the impact on Medicaid enrollees health care utilization and expenditures is warranted. Conceptual Model and Hypotheses In his article published in 1972, Michael Grossman explained the demand of health care from a health capital and investment perspective ( Grossman 1972 ) An life course and the magnitude of change
39 depends on the rate of depreciation and individuals investment in the ir stock of health which can b e expressed as the following : H i+1 H i = I i i H i (2 1) where H i represents the initial stock of health or the stock of health in the i th time period ; H i+1 represents the stock of health in the i+1 th time peri od ; i is the depreciation rate during the i th time period ; and I i is the gross investment in health. I i is a function of medical care (M i ), time inputs (TH i ), and human capital stock (E i ). To examine the influence of the EBR program on health care utiliza tion and expenditures, we will focus on the medical care component of the gross health investment (Fig. 2 1). i ) can be seen as a function of an choices (smoking/drinking habits, diet, exercise, substance u se, and the like ), preventive care utilization (wellness check s cancer screening s dental care, vaccination s and so on ), age, and gender, race, and other genetic factors. It is shown in Figure 2 2 health, health depreciation rate, and investment in medical care, which includes both preventive care and more costly acute care like ED visits and hospitalizations. As we discussed earlier, it is presum ed that the financial incentives provided by the EBR program will encourage people to more actively seek preventive services and adopt healthy behaviors (Figure 2 3 ). behaviors and health depreciation rates as w ell as their expenditures on preventive care. Active participation in the EBR program is expected to slow the decline in health status over time while it would also be affected by an
40 along with such other factors as enro investments in preventive care may increase due to the stimulation from financial incentives, if an where much more expensive acute health care services are averted or less demand ed it would offset the increased expenditures on preventive services and generate savings in the long run. With the support of this concept, hypotheses are formulated to answer the following research questions. Question 1: Does participating in an EBR program have any impact on Medicaid s ? Hypothesis 1: Enrollees with a higher level of EBR program participation will be associated with lower total health care expenditures. Question 2a: Does EBR program participation have any impact on Medicaid Hypothesis 2a: Enrollees with a higher level of EBR program participation will be associated with a lower number of total inpatient days within the study period. Question 2b: Does EBR program participation have any impact on Medicaid Hypothesis 2b: Enrollees with higher level s of EBR program participation will have fewer ED visits within the study period. Question 2c: Does EBR program participation have any impact on Medicaid Hypothesis 2c: Enrollees with a higher level of EBR program participation will have fewer avoidable hospitalizations during the study period.
41 Ta ble 2 1. Top 10 credit earning behaviors in the EBR program from 2006 to 20 10 (Hall et al. 2013) Behavior Credits (Unique Count) Dollars Earned (Millions) Recipients (Count) Office Visit Adult/Child 624,970 $9.8 270,963 Childhood Preventive Care 467,5 65 $11.7 198,633 Compliance with Prescribed Maintenance Drug 249,477 $1.9 32,453 Dental Preventive Services Adult/Child 82,299 $2.0 25,807 Vision Exam Adult/Child 42,580 $1.0 34,219 Pap Smear 39,683 $1.0 30,852 Child & Adult Prevent ive Care 27,735 $0.5 20,187 Adult Preventive Care 8,205 $0.1 6,491 Mammography 3,810 $0.1 3,525 Colorectal Screening 2,381 $0.1 2,319
42 Figure 2 1. EBR credit earning s utilizations and outcomes ( Bragdon 20 11)
43 Figure 2 2 Stock of health, health depreciation, and health investment Figure 2 3 EBR program, health behaviors changing, and demand of medical care.
44 CHAPTER 3 METHODS The study examine d the difference s in health care utili zations and expenditures by level of Enhanced Benefits Rewards ( EBR ) participation. The study period was from FY2007 08 to FY2009 10. The study population include d Medicaid Reform participants in urban reform counties (Broward and Duval) who enrolled in a PSN plan. Due to incomplete data, HMO enrollees are excluded from the study. The demonstration was implemented in three rural counties (Baker, Clay, and Nassau) one year after the implementation in Broward and Duval counties. Because the infrastructure and delivery system of Medicaid in these rural counties were different from the ones in urban counties like Broward and Duval, and the size of the enrollee population was much smaller, only beneficiaries from those two urban counties were included in the stud y. For each research question, a base model was used to examine the association between EBR program participation level and health care utilizations/expenditures. In addition to the base model, certain parameters in regression models were modified for sens itivity analysis to evaluate the impact on the outcomes. The effect of EBR program on health care utilizations and expenditures is not assumed to be an instantaneous one. ir health, health care utilizations and health care expenditures. Therefore, the base model used EBR program participation level from the previous year and the year before to estimate Medicaid expenditures and utilizations in the current year with person year level data. It was believed, by adopting this method, we were able to better explore trends in Medicaid utilizations/expenditures following the implementation of the EBR program given different participation levels.
45 Data Source s The primary data sourc es are the Enhanced Benefits Information System (EBIS) data and Medicaid claims and eligibility data. EBIS was funded by AHCA and maintained by Imager Software Inc. to process the Medicaid Reform participating related activities and credits earned. (see Table 3 1 for a complete list of EBR eligible behaviors), dates credits earned, dates credits created, number of credits, dollar amount s of credits earned, and usage of the rewards. EBIS data from FY2006 07 to FY2009 10 were used. Medicaid claims and eligibility data ha d demographic s enrollment history, and health care utilization and expenditures. Eligibility data were s (age, gender, race/ethnicity), eligibility status (Temporary Assistance for Needy Families or TANF, SSI, or other), plan characteristics, and enrollment history. Medicaid claims data were used to including ED visits, total hospital days, and avoidable hospitalizations. Avoidable hospitalizations were identified by using the Preventive Quality Indicators (PQIs) and the co rresponding ICD 9 codes associated with each condition ( http://www.qualityindicators.ahrq.gov/modules/pqi_overview.aspx ) Enrollees total expenditures on health care were also d etermined using Medicaid claims data. Claims and eligibility data from FY2006 07 to FY2009 10 were included in the analysis. Risk scores for Medicaid enrollees were also adopted in the analyses to control for individuals health risk/status.
46 Description o f Dependent and Independent Variables Dependent Variable for Research Q uestion 1 Total health expenditures: Total health expenditures refer to an total expenditures calculated from Medicaid claims data which include d all inpatient, outpa tient, medical, and pharmacy claims generated in a given year. Independent Variables for Research Q uestion 1 level was categorized by using their annual total EBR rewards earning in dollar amoun ts. Based participation level. For all the individuals who had EBR credit earnings, people who earned less than $25 EBR credits in the year were assigned to the lo w group; people who earned between $25 and $50 were assigned to the medium group; and the rest of the participants who earned more than $50 were assigned to the high group. The non earner group was then used as the control group in the analyses and three d ummy variables were created for low, medium, and high groups. For more detail s please refer to the descriptive statistics section in Chapter 4. Risk scores: utilization and expendit ures, as well as his/her EBR program access and usage (Hall et al. 20 12). A variable that represents an was needed for the analysis. Since there was no available health status information in either EBIS data or Medicaid claims d ata, an health status. These are the risk scores that were used to calculate Medicaid Reform HMO risk adjusted premiums by Mercer Health & Benefits LLC, who m AHCA contracted with to calculat e the capitation rates for the Medicaid Reform (Medicaid
47 Managed Care Rate setting). The risk scores were determined by using Medicaid Rx methodology which is a statistical model that uses prescription drug use information and age gender categories for eac h beneficiary enrolled in Medicaid to predict a Eligibility Status: Different eligibility group s in Medicaid could have systematic difference s in their health status and other characteristics, and th ese difference s wo uld in turn have different impact s on their service utilization and expenditures. It i s important be controlled for in this study. This variable indicates on what basis the enrollee was eligible for Medicaid. In Flori da Medicaid Reform, individuals and families who are eligible through SSI or TANF are mandatory participants. Participation of individuals who are dually eligible (for both Medicaid and Medicare) and pregnant women are considered voluntarily and are not re quired to enroll in a reform health plan. The study is limited to beneficiaries who are mandated to enroll in a reform health plan, which includes SSI and TANF eligible enrollees. And eligibility through TANF vs. SSI is controlled for in the analyses. The a forementioned voluntary participants, enrollees participating in a waiver program (e.g. A IDS waiver), and children who received care through a special program for children with special health care needs were excluded from the study population. Enrollment L ength of T ime: Length of time in Medicaid is a continuous variable year to control for exposure time. County: County is a binary variable indicating which county Bro ward or Duval the beneficiary is from
48 Demographics: for in the statistical model. Dependent Variable for Research Q uestion 2a a dummy variable for whether the individual had accrued any inpatient days were used as the dependent variables in the regression model. Independent Variables for R ese arch Q uestion 2a EBR P rogram P articipation L evel: level is the independent variable of interest and is defined in the same way as described earlier. county they are from, and demographics including age, gender, and race/ethnicit y were also controlled for in the statistical models. Dependent Variable for Research Q uestion 2b Total N umber of ED V isits: A dummy variable was created to indicate if there were any ED visits for the individual in a tes of ED visits were also used in a different model. L imited by data availability, this variable only reflect ed the outpatient portion of ED visits, which means it only include d the ED visits that generated claims in Medicaid outpatient claims files (pati ents treated in emergency department then released instead of being admitted to the hospital for inpatient services). ED visits that resulted in inpatient admissions were counted as inpatient treatment and not an ED visit for the purpose of this study.
49 Ind ependent Variables for Research Q uestion 2b EBR P rogram P articipation L evel: Using the same methods described earlier, EBR participation level was determined based on an s evels from each year. home county, and demographics including age, gender, and race/ethnicity were controlled for in the models. Dependent Variable for Research Q uestion 2c Total N umber of A voidable H ospitalizations: Annual rates of avoidable hospitalizations of an individual enrollee were identified by using the Agency for Healthcare Research and Quality (AHRQ) PQI guide and associated ICD 9 diagnosis codes ( http://www.qualityindicators.ahrq.gov/Modules/PQI_TechSpec.aspx ). Table 3 1 lists all the conditions that were included. Also, a dummy variable was created to indicate if there were any avoi dable hospitalizations identified in each year during the study period. Independent Variables for Research Q uestion 2c E BR P rogram P articipation L evel : This was categorized in the same pattern as bility status, enrollment length of time, home county, and demographics including age, gender, and race/ethnicity were also controlled for in the statistical models. Statistical Analysis The purpose of the study was to examine the influence of the EBR pr ogram participation on Florida Medicaid expenditures and utilization. It is well known that health expenditure data are likely to be heavily skewed, as there may be a large portion
50 of people reporting zero expenditure s while a small portion of the populati on have the highest expenditures. Simple linear regression models would create biased and inefficient estimates. One of the most widely used approaches to address this issue is the t wo p art model (2PM). The 2PM was developed for the Rand Health Insurance S tudy (HIE) that ran through the 1970s and 80s and was widely adopted because of its ability to deal with the issue s stated above in health care expenditure studies ( Manning et al. 1987 ) The first part of the model was used to estimate the probability of expenditure value being zero versus not being zero. Then for the sec ond part of the model, depending on the nature of the data and distribution of the outcome variables, a suitable regression with necessary transformation was used to estimate the expenditures conditional on having any spending during a given period of time In short, it can be expressed as the following equation: E(Y i ) = P(Y)*E(Y|Y>0) (3 1) where Y represents the outcome measure of interest, in this case health care expendit ures ; P(Y) is the probability of having any expenditures ; and E(Y|Y>0) is the predicted outcome given its value greater than zero. Research Q uestion 1 A logit model can be used in the first part of the 2PM, when necessary, to estimate the probability of ha ving any expenditures. However, with only 19,860 out of 359,099 observations (5.5%) reporting zero expenditure a single generalized estimating equation (GEE) using a gamma family with a log link was used in the analyses. Based on a recent study using Flor ida Medicaid Demonstration expenditures
51 data, this one part model is able to handle data distribution without giving biased estimators ( Harman et al. 20 11 ) To be more specific, the estimation equation can be expressed as: 1) + (3 2) where th year within the study period. On the right hand side of the whether the value of EBR program participation level was different from the previous year and the year before when available. This means that the EBR participation level in the first year was used to estimate expenditures in the second year and participation level in the first year and the second year were used to estimate expenditures in the third year and variables including length of time, home following a gamma distri bution. The rationale behind this model is that, when studying the relationship between EBR program participation and expenditures, health care expenditures in the current year should be a function of participation level in previous years following the imp lementation of the program. In other words, there will be a lag effect of EBR program participation on health care expenditures. This model allows us to better understand the relationship and observe the trend of chang e in expenditures under the influence of the EBR program. Research Q uestion s 2a, b, and c First, a logistic model was used to examine the association between EBR program participation level and the odds of having any ED visits, inpatient days, or
52 avoidable hospitalizations. And then, consideri ng the dependent variables being count data, a negative binomial model was adopted to analyze the relationship between EBR participation level and annual total ED visits, avoidable hospitalizations, or total hospital days for individuals with positive util ization. The general form of the estimation equations is: log(# of ED visits i /avoidable hospitalizations i /hospital days i ) = 0 + 1 *Year + 2 *(EBR_participation_level i 1 ) + 3 *(EBR_participation_level i 2 ) + 4 *Controls + (3 3) where outcome variables refer to ; from the previous year and the year before ; nts a vector of covariates eligibility status, enrollment length, county, age, gender, and race/ethnicity ; a the error term for all the other unobserved factors that influence d the estimation of dependent variables. With the design of the study and the longitudinal nature of the data, GEEs were used for both models to address potential correlation s betw een outcomes. Compared to the likelihood based generalized linear mixed model which is sensitive to variance structure specification, GEEs use quasi likelihood estimation to obtain rather consistent parameter estimates and standard errors even when covari ance structure is not correctly specified (Liang and Zeger 1986; Diggle, Liang and Zeger, 1994). All statistical analyses were performed using SAS 9.3
53 Table 3 1. PQI indicators for avoidable hospitalizations PQI Number Description ADULT (18 years and Older) 1 Diabetes s hort term c omplications (ketoacidosis, hyperosmolarity, coma) 2 Perforated a ppendix 3 Diabetes l ong t erm c omplications (renal, eye, neurological, circulatory) 5 Chronic o bstructive p ulmonary d isease 7 Hypertension 8 Conges tive h eart f ailure 10 Dehydration 11 Bacterial p neumonia 12 Urinary t ract i nfection 13 Angina without p rocedure 14 Uncontrolled d iabetes (not short term or long term) 15 Adult a sthma 16 Lower extremity amputation for diabetes CHILD 14 Asthma (ag es 2 17) 15 Diabetes (ages 6 17 years) 16 Gastroenteritis (3 months 17 years) 17 Perforated a ppendix ( 1 year 17 years) 18 Urinary tract infection (3 months to 17 years)
54 CHAPTER 4 RESULTS Descriptive Statistics Through the first three years of the E BR program, a total of $ 8,811,227.82 of EBR credits was earned by the study population Overall, the most credits were earned by using preventive care for children and adults followed by o ffice visits by children and adults (Table 4 1) All c ancer screenin gs, disease management programs ( diabete s, congested heart failure asthma, HIV/AIDS, hypertension, and others ) alcohol and narcotics /drugs program s smoking cessation, weight management, and exercise programs combined together had fewer counts of occurre nce than any other type s of behaviors listed in Table 4 1 (all preventive care, office visits, dental services, or prescription drug maintenance) in any given year From FY2006 07 to FY2008 09, the first three years of the EBR program, the study population program had been increasing. Total EBR credits earn ed increased from over $1.5 million in FY2006 07 to over $3.8 million in FY2008 09. During the first two years of the program, primary care physician (PCP) office visits was the highest credit earning category. As the overall earning and earning by different behavior types consistently increased over this three year period, earning by having physician visits dropped significantly in FY2008 09 due to the policy change regardin g EBR credits earning by attending physician visits. Enrollees participated in the EBR program were assigned to three groups based on their annual total EBR credits earn ed to represent their high, medium, or low level of EBR program participation (also see Chapter 3, p. 46) The average percentages of earning from different behavior types by EBR program participation level are reported in
55 Table s 4 2, 4 3, and 4 4. For each year, significant differences in percentages of earning from different eligible behav ior types exist across participation level groups. A summary of outcome variables is given in Table 4 5. Average annual health expenditures for the study population decreased over time. Acute care utilization decreased over time as well. The characteristic s of the study population were compared by year based on their assigned EBR program participation level group. It appears that to non participants People partici pating in the EBR program had longer average enrollment length in the year than non earners. Additionally, the average risk scores of earners were higher than the non earner group. A detailed comparison for all the other characteristics can be found in Tab le s 4 6, 4 7, and 4 8. Model Statistics Research Question 1 The base model for research question 1 used the prior year EBR program participation level to estimate annual total health expenditures for Medicaid Reform enrollees. The parameter estimates of th e model were estimated by a GEE model clustered on the person level with multi year observations potentially correlated. Two based standard d the specified covarianc e structure in the model was correct. If the covariance structure is not correctly specified, the standard error may not be accurate although the GEE e stimates of model parameters can ed this iss estimator to obtain consistent estimators even when covariance structure was not
56 correctly assigned in the model (Liang and Zeger 1986 ). In the following sections, only sed. The base model examined the influence of EBR program participation levels in the that, when compared to FY2007 08, Medicaid enrollees had significantly less annual expen diture in FY2008 09 and FY2009 10 (Table 4 9). For EBR program participation level in previous year, people who participated in the EBR program, regardless of their participation level, spent less than non participants. However, among EBR program earners, the relationship is reversed as people with the lowest level of earnings had the least health expenditures compared to medium and high level earners. People in the medium group also spent less than people in the high participation group. All the above effe cts are statistical significant ( p <0. 001). Also, male beneficiaries spent significantly less than females, and African Americans and Hispanics had significantly lower expenditures than white s while people of other race s or ethnicit ies spent more. The sec annual the previous year and the year before. S imilar to the base model, beneficiaries had significantly lower expenditures in FY2 008 09 and FY2009 10 compared to their spending in FY2007 08 ( Table 4 10 ). For EBR program participation levels, all earners had significantly lower expenditures than non earners, with the exception of the high level earners from the prior year who were no t significantly different from the non earners. Participation levels from two years prior appear to have a stronger effect on current health expenditures than EBR participation levels from the prior year. And similar to the trend showed in the
57 base model, among participants, enrollees in lower participation level group actually had significantly lower health expenditures than their peers with higher participation level. Since EBR program participation level was determined rather arbitrarily based on cutoff points, annual total EBR credit earnings in dollar amount were used as a continuous variable in sensitivity analysis to further examine the association between Table 4 11). When o nly EBR earnings from the previous year were used in the model, it shows that for each $10 increase in reward dollar earning s the health expenditure for the next year will have a 0.6% decrease ( p = 0.005). After adding EBR earnings from two years prior t he model suggests that EBR earning from one year ago was no longer significantly associated with an Table 4 12 ). However, EBR earnings from two year s current health expenditures. For each $10 increase in EBR credit earning from two year p <0. 0001). Given the year period, for every $10 increase in EBR earnings from two year prior, there was a $38.80 For all previous models, the study sample contains all beneficiaries enrolled in Medicaid for at least one month for the gi ven year. To determine if the direction and magnitude of the relationship between EBR program participation and Medicaid expenditures would change when the study was limited to people who ha d been more consistently enrolled in Medicaid, EBR program partici pation levels from the past two years were used to estimate health expenditures of enrollees who had been enrolled in
58 Medicaid for more than 6 months during the year. Among this population, EBR participation from the previous year was associated with highe r Medicaid expenditures while EBR participation levels from two years ago was still significantly associated with decreasing health expenditures ( Table 4 13 ). Also, holding other variables constant, the expenditures in FY2008 09 and FY2009 10 were both hi gher than Medicaid expenditures in FY2007 08. When using EBR credit earnings as a continuous variable, the results show a similar trend ( Table 4 14 ). Research Question 2 Inpatient days L ogistic model s were used to examine the odds of having any hospitaliza tions given different levels of EBR program participation. Participation in the EBR program in the previous year or two years ago reduce d the odds of having any inpatient services for Medicaid enrollees for people with both low and high levels of participa tion, but the effect on medium level participation was only significant for EBR credits earned two years prior. The EBR participation level from two years ago appeared to be associate d with greater decreased odds of having any hospitalizations than EBR par ticipation from the previous year. And enrollees in the low participation category from two years ago ha d lower odds of having any inpatient services than any other group. When EBR credit earnings in dollar amount s were used as an indication of participati on level, the same trend was discovered as EBR earnings from two years ago showed greater effect than EBR earnings from one year ago (Table s 4 15 4 16). After limiting the sample population to people enrolled more than 6 months in Medicaid for the given y ear, the models using either EBR program participation level groups or annual EBR credits earning showed that participating in EBR program in the previous year became
59 insignificant while EBR program participation from two years ago was still significantly associated with lower odds of having any inpatient services (Table s 4 17, 4 18 ). Compared to the full model (Table s 4 15, 4 16), participating in the EBR program two years prior appear ed to have a slightly smaller effect on lowering the odds of having any inpatient services when the analyses were limited to people who had more than 6 months of coverage. For all the models discussed above, results also suggest that there was an upward trend in the odds of being hospitalized in FY2008 09 and FY2009 10 compar ed to FY2007 08. To estimate the effect of EBR program participation on annual total inpatient days for enrollees who had had any inpatient services in a given year, a negative binomial regression model was used for both the whole study population and a sa mple of people who enrolled in Medicaid for more than 6 months in a given year. Without the limitation on enrollment length, the parameter estimates suggest ed that EBR program participation was associated with a lower number of total inpatient days, althou gh only the previous year showed an effect on a statistical ly significant level ( Table 4 19 ). The model that use d total annual EBR credit earning s showed similar results (Table 4 20). EBR credit earning s from two years ago did days. However, for each $10 increase in the was a 1.6% decrease in the number of annual total inpatient day s ( p <0. 001). Results were similar when the analysis only included individuals with more than 6 months of enrollment in Medicaid in a given year. Only EBR participation from the previous year ys ( Table s 4 21, 4 22).
60 ED visits The odds of having any emergency department visits in a given year were estimated using logistic regression models. While holding all other variables constant, when compared to FY2007 08, the odds of having any ED visits were higher in both FY2008 09 and FY2009 10. Although EBR program participation in the previous year was associated with higher odds of having any ED visits, EBR participation from two years prior was associated with lower odds of having any ED visits in t he current year ( Table s 4 23, 4 24). When the analyses were limited to people who had been enrolled in Medicaid for more than 6 months in a given year, similar results were observed ( Table s 4 25, 4 26 ). While participat ion in the EBR program in the previou s year was associated with an increased likelihood of having any ED visits, EBR program participation from two years ago was significantly associated with lower odds of having any outpatient ED visits. Also, from FY2007 08 to FY2009 10, there was a signifi cant and consistent increase in the odds of having any ED visits for Medicaid enrollees. Negative binomial regression was used to examine the effect of EBR program had o utpatient ED visits, EBR program participation in both the previous year and two years prior was significantly association with fewer ED visits ( Table s 4 27, 4 28 ). When analyzing only individuals who had more than 6 months of Medicaid enrollment in a give n year, the effect of EBR participation in the previous year became borderline significant; participation in EBR two years ago was still associated with fewer number s of ED visits on a statistically significant level ( Table s 4 29, 4 30 ).
61 Avoidable hospital izations Overall, EBR program participation in both the previous year and two years ago was significantly associated with lower odds of having any avoidable hospitalizations in the given year (Table s 4 31, 4 32). EBR program participation in the previous y ear appears to have an overall greater effect on decreasing the likelihood of having an avoidable hospitalization for Medicaid enrollees when compared to participation in the EBR program from two years ago. This association was strongest when total EBR cre dits earned was used to represent participation level in the model instead of being categorized as high, medium, or low levels of credit earning s When the analyses were limited to beneficiaries enrolled more than 6 months in Medicaid for the year, the rel ationship between EBR participation level and the odds of having any avoidable hospitalizations becomes less clear ( Table s 4 33, 4 34). Among enrollees who had any avoidable hospitalizations, no significant effects of EBR program participation on the rates of avoidable hospitalization s were observed from the analyses with or without enrollment length limitation (Table s 4 35 to 4 38).
62 Table 4 1. Annual total EBR credits earning and counts of occurrence by behavior type FY 20 06 20 07 FY 20 07 20 08 FY 20 08 20 09 Behavior Type Total EBR Credit Earning s ($) Counts of Occurrence Total EBR Credit Earning ($) Counts of Occurrence Total Credit Earning s ($) Counts of Occurrence All Preventive Services 413,740.00 16,815 1,172,280.00 47,826 2,185,052.50 89,474 Office Vi sits Adult/Child 780,182.50 35,034 1,550,187.50 68,325 563,550.00 75,141 Dental Preventive Services 165,130.00 6,608 225,482.50 9,043 444,365.00 17,832 Prescribed Drug Maintenance 80,975.00 10,764 215,577.50 28,979 301,367.50 40,475 All Others 113,610.3 2 4,559 208,475.00 8,434 391,252.50 15,713 Total 1,553,637.82 73,780 3,372,002.50 162,607 3,885,587.50 238,635
63 Table 4 2. Composition of EBR credits earning by service type and participation level for FY2006 07 High (N=10362) Medium (N=9740) Low (N=8571) Prevent Services 26.75% 15.04% 39.75% Office Visits 43.91% 68.69% 49.60% Dental Services 12.02% 10.15% 7.39% Rx Maintenance 7.43% 2.34% 15.33% All Others 9.89% 3.78% 26.48% All percentages are significantly different across groups Table 4 3. Composition of EBR credits earning by service type and participation level for FY2007 08 High (N=25657) Medium (N=14914) Low (N=12642) Prevent Services 33.68% 23.14% 52.96% Office Visits 43.80% 65.44% 37.94% Dental Services 7.12% 5.72% 0.94% R x Maintenance 7.71% 2.00% 2.09% All Others 7.69% 3.69% 6.07% *All percentages are significantly different across groups Table 4 4. Composition of EBR credits earning by service type and participation level for FY2008 09 High (N=27949) Medium (N=30702) Low (N=40412) Prevent Services 55.80% 58.60% 32.42% Office Visits 9.18% 18.24% 60.53% Dental Services 16.35% 6.43% 0.62% Rx Maintenance 9.33% 3.51% 2.89% All Others 9.35% 13.22% 3.54% All percentages are significantly different across groups Tab le 4 5. Average personal annual total expenditures and acute care utilizations by FY FY2007 08 FY2008 09 FY2009 Total Expenditure ($) 2997.98 1692.51 1087.33 Total ED Visits 0.486 0.377 0.352 Total Inpatient days 0.517 0.296 0.193 Total Avoidable Ho spitalization 0.013 0.013 0.008 All reported means are significantly different across years
64 Table 4 6. Sample characteristics by EBR program participation level for FY2007 08 High (N=10362) Medium (N=9740) Low (N=8571) Non earner (N=35983) Age 15.7 3 17.49 12.60 13.08 Gender Female 56.48% 53.06% 52.47% 51.87% Male 43.52% 46.94% 47.53% 48.13% Race/Ethnicity White 22.13% 22.29% 17.38% 20.13% Black 44.31% 47.70% 56.39% 53.24% Hispanic 18.11% 15.35% 14.54% 14.78% Other 15.45% 14.66% 11.69 % 11.84% County Duval 44.81% 44.28% 50.16% 50.91% Broward 55.19% 55.72% 49.84% 49.09% Eligibility Status SSI 27.96% 29.75% 18.17% 15.22% TANF 72.04% 70.25% 81.83% 84.78% Enrollment l ength (months) 9.35 9.32 9.05 7.08 Risk Score 1.29 1.30 0.89 0.50 Annual Expenditures ($) 5447.21 5181.76 2496.87 1820.92 ED Visits 0.68 0.63 0.51 0.39 Inpatients Days 0.75 0.86 0.50 0.36 Avoidable h ospitalizations (every 100 enrollees) 1.40 1.79 0.96 1.19 All characteristics percentage or means are significantly different across groups
65 Table 4 7. Sample characteristics by EBR program participation level for FY2008 09 High (N=25657) Medium (N=14914) Low (N=12642) Non earner (N=57181) Age 13.71 15.26 12.06 13.63 Gender Female 55. 75% 54.24% 52.67% 53.45% Male 44.25% 45.76% 47.33% 46.55% Race/Ethnicity White 20.95% 20.67% 15.83% 22.37% Black 46.77% 51.84% 62.19% 52.42% Hispanic 18.95% 16.42% 13.52% 15.09% Other 13.33% 11.07% 8.46% 10.13% County Duval 45.50% 46.44% 47 .34% 52.50% Broward 54.50% 53.56% 52.66% 47.50% Eligibility Status SSI 18.29% 17.08% 9.98% 8.69% TANF 81.71% 82.92% 90.02% 91.31% Enrollment l ength (months) 6.79 6.47 6.18 4.79 Risk Score 1.12 1.00 0.64 0.26 Annual Expenditures ($) 2912.6 6 2580.69 1177.08 1027.33 ED Visits 0.51 0.46 0.34 0.31 Inpatients Days 0.43 0.48 0.20 0.21 Avoidable h ospitalizations (every 100 enrollees) 1.49 1.31 0.44 1.32 All characteristics percentage or means are significantly different across gro ups
66 Table 4 8. Sample characteristics by EBR program participation level for FY2009 10 High (N=27949) Medium (N=30702) Low (N=40412) Non earner (N=84986) Age 12.19 12.15 14.73 14.26 Gender Female 55.02% 56.31% 53.87% 52.65% Male 44.98% 43.69% 4 6.13% 47.35% Race/Ethnicity White 20.22% 19.52% 20.87% 20.97% Black 49.13% 51.96% 53.56% 53.39% Hispanic 18.59% 17.64% 15.53% 15.95% Other 12.06% 10.88% 10.04% 9.70% County Duval 48.33% 49.26% 50.09% 46.99% Broward 51.67% 50.74% 49.91% 53.0 1% Eligibility Status SSI 15.06% 9.86% 12.08% 7.92% TANF 84.94% 90.14% 87.92% 92.08% Enrollment l ength (months) 6.00 5.33 5.08 4.90 Risk Score 0.90 0.86 0.80 0.30 Annual Expenditures ($) 1910.65 1146.94 1173.96 753.84 ED Visits 0.50 0.38 0.35 0.29 Inpatients Days 0.28 0.18 0.24 0.15 Avoidable h ospitalizations (every 100 enrollees) 0.98 0.47 0.68 0.81 All characteristics percentage or means are significantly different across groups
67 Table 4 9. GEE base model for Medicaid expen ditures Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.0817 0.0238 0.1283 0.0351 0.0006 FY0910 0.1572 0.0228 0.2020 0.1125 <0 .0001 EBR Level (Non earner) High 0.0681 0.0177 0.1027 0.0334 0.0001 Medium 0.1873 0.0182 0.2230 0.1515 <0 .0001 Low 0.2865 0.0183 0.3223 0.2507 <0 .0001 Age 0.0119 0.0007 0.0106 0.0132 <0 .0001 Gender (Female) Male 0.0741 0.0167 0.1069 0.0414 <0 .0001 Race/Ethnicity (White) Black 0.1269 0.0183 0.1628 0.0911 <0 .000 1 Hispanic 0.0796 0.0256 0.1298 0.0293 0.0019 Other 0.2908 0.0439 0.2048 0.3768 <0 .0001 County (Broward) Duval 0.0812 0.0166 0.1138 0.0486 <0 .0001 Eligibility (SSI) TANF 1.4494 0.0269 1.5021 1.3968 <0. 0001 Enrollment Length 0.1805 0.0015 0.1776 0.1835 <0. 0001 Risk Score 0.2143 0.0047 0.2051 0.2236 <0. 0001
68 Table 4 10. GEE base model with two years of EBR data for Medicaid expenditures Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.0165 0.0259 0.067 3 0.0344 0.5251 FY0910 0.0723 0.0264 0.1240 0.0206 0.0061 EBR Level Previous Year (Non earner) High 0.0157 0.0174 0.0499 0.0185 0.3683 Medium 0.1375 0.0180 0.1728 0.1021 <0. 0001 Low 0.2406 0.0176 0.2751 0.2061 <0. 0001 EBR Level Two ye ars ago (Non earner) High 0.1783 0.0159 0.2094 0.1471 <0. 0001 Medium 0.1768 0.0215 0.2189 0.1347 <0. 0001 Low 0.3539 0.0189 0.3911 0.3168 <0. 0001 Age 0.0118 0.0006 0.0105 0.0131 <0. 0001 Gender (Female) Male 0.0741 0.0166 0.1065 0 .0416 <0. 0001 Race/Ethnicity (White) Black 0.1143 0.0180 0.1497 0.0790 <0. 0001 Hispanic 0.0738 0.0254 0.1236 0.0239 0.0037 Other 0.2909 0.0436 0.2054 0.3764 <0. 0001 County (Broward) Duval 0.0786 0.0164 0.1108 0.0464 <0. 0001 Eligib ility (SSI) TANF 1.4561 0.0270 1.5089 1.4033 <0. 0001 Enrollment Length 0.1808 0.0015 0.1779 0.1838 <0. 0001 Risk Score 0.2333 0.0046 0.2243 0.2423 <0. 0001
69 Table 4 11. GEE model using EBR credits earning for Medicaid expenditures Estimate S.E 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.0775 0.0222 0.1210 0.0339 0.0005 FY0910 0.1980 0.0212 0.2396 0.1565 <0. 0001 Annual EBR credits 0.0006 0.0002 0.0011 0.0001 0.0160 Age 0.0117 0.0007 0.0104 0.0130 <0. 0001 Gender (Fe male) Male 0.0749 0.0164 0.1070 0.0428 <0. 0001 Race/Ethnicity (White) Black 0.1325 0.0209 0.1734 0.0915 <0. 0001 Hispanic 0.0831 0.0277 0.1374 0.0288 0.0027 Other 0.2978 0.0308 0.2374 0.3583 <0. 0001 County (Broward) Duval 0.0 836 0.0167 0.1165 0.0508 <0. 0001 Eligibility (SSI) TANF 1.4527 0.0303 1.5120 1.3934 <0. 0001 Enrollment Length 0.1791 0.0019 0.1754 0.1827 <0. 0001 Risk Score 0.1952 0.0078 0.1800 0.2104 <0. 0001
70 Table 4 12. GEE model with two years of EBR c redits earning for Medicaid expenditures Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.0418 0.0255 0.0917 0.0081 0.1004 FY0910 0.1309 0.0261 0.1819 0.0798 <0. 0001 Annual EBR credits (previous year) 0.0001 0.0002 0.000 5 0.0004 0.8071 Annual EBR credits (two years ago) 0.0024 0.0002 0.0028 0.0021 <0. 0001 Age 0.0114 0.0007 0.0101 0.0127 <0. 0001 Gender (Female) Male 0.0743 0.0168 0.1073 0.0412 <0. 0001 Race/Ethnicity (White) Black 0.1313 0.0184 0.167 4 0.0953 <0. 0001 Hispanic 0.0772 0.0259 0.1280 0.0264 0.0029 Other 0.2961 0.0447 0.2085 0.3837 <0. 0001 County (Broward) Duval 0.0818 0.0168 0.1146 0.0489 <0 .0001 Eligibility (SSI) TANF 1.4675 0.0269 1.5202 1.4149 <0 .0001 Enrollme nt Length 0.1793 0.0015 0.1762 0.1823 <0 .0001 Risk Score 0.2119 0.0047 0.2026 0.2211 <0 .0001
71 Table 4 13. GEE model for Medicaid expenditures (more than 6 months enrollment) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.08 64 0.0195 0.0482 0.1245 <0 .0001 FY0910 0.0446 0.0209 0.0036 0.0855 0.0328 EBR Level Previous Year (Non earner) High 0.2903 0.0196 0.2519 0.3288 <0 .0001 Medium 0.1551 0.0199 0.1162 0.1940 <0 .0001 Low 0.0133 0.0181 0.0222 0.0487 0.4632 EBR Level Two years ago (Non earner) High 0.1582 0.0168 0.1912 0.1253 <0 .0001 Medium 0.1521 0.0193 0.1900 0.1143 <0 .0001 Low 0.3201 0.0202 0.3597 0.2804 <0 .0001 Age 0.0099 0.0006 0.0086 0.0112 <0 .0001 Gender (Female) Male 0.0447 0.0152 0.0 744 0.0150 0.0032 Race/Ethnicity (White) Black 0.1149 0.0174 0.1490 0.0808 <0 .0001 Hispanic 0.0378 0.0246 0.0860 0.0105 0.1251 Other 0.1349 0.0341 0.0680 0.2018 <0 .0001 County (Broward) Duval 0.0686 0.0160 0.0999 0.0372 <0 .0001 El igibility (SSI) TANF 1.5406 0.0290 1.5975 1.4837 <0 .0001 Enrollment Length 0.0424 0.0043 0.0341 0.0508 <0 .0001 Risk Score 0.2764 0.0050 0.2666 0.2863 <0 .0001
72 Table 4 14. GEE model with two years of EBR credits earning for Medicaid expenditure s (more than 6 months enrollment) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.0424 0.0186 0.0059 0.0789 0.0229 FY0910 0.0230 0.0199 0.0159 0.0619 0.2458 Annual EBR credits (previous year) 0.0031 0.0002 0.0026 0.0035 <0 .00 01 Annual EBR credits (two years ago) 0.0018 0.0002 0.0021 0.0014 <0 .0001 Age 0.0099 0.0006 0.0086 0.0111 <0 .0001 Gender (Female) Male 0.0446 0.0151 0.0742 0.0150 0.0032 Race/Ethnicity (White) Black 0.1265 0.0170 0.1598 0.0933 <0 .0 001 Hispanic 0.0377 0.0242 0.0852 0.0098 0.1197 Other 0.1341 0.0342 0.0670 0.2012 <0 .0001 County (Broward) Duval 0.0702 0.0159 0.1014 0.0390 <0 .0001 Eligibility (SSI) TANF 1.5438 0.0290 1.6006 1.4871 <0 .0001 Enrollment Length 0.039 9 0.0043 0.0314 0.0484 <0 .0001 Risk Score 0.2752 0.0050 0.2655 0.2850 <0 .0001
73 Table 4 15. GEE model for the likelihood of having any inpatient services Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.1039 0.0264 0.0523 0.15 56 <0 .0001 FY0910 0.0797 0.0278 0.0252 0.1342 0.0042 EBR Level Previous Year (Non earner) High 0.1331 0.0296 0.1911 0.0751 <0 .0001 Medium 0.0364 0.0283 0.0920 0.0191 0.1986 Low 0.0767 0.0281 0.1317 0.0217 0.0063 EBR Level Two years ago ( Non earner) High 0.2882 0.0323 0.3515 0.2249 <0 .0001 Medium 0.2460 0.0350 0.3147 0.1773 <0 .0001 Low 0.4212 0.0419 0.5033 0.3392 <0 .0001 Age 0.0269 0.0007 0.0256 0.0283 <0 .0001 Gender (Female) Male 0.3930 0.0217 0.4354 0.3506 <0 0001 Race/Ethnicity (White) Black 0.0610 0.0258 0.0104 0.1116 0.0180 Hispanic 0.3025 0.0401 0.3811 0.2240 <0 .0001 Other 0.0874 0.0369 0.1596 0.0151 0.0178 County (Broward) Duval 0.0449 0.0214 0.0869 0.0028 0.0365 Eligibility (SSI) TANF 1.1381 0.0314 1.1996 1.0767 <0 .0001 Enrollment Length 0.1353 0.0023 0.1308 0.1398 <0 .0001 Risk Score 0.1854 0.0081 0.1696 0.2013 <0 .0001
74 Table 4 16. GEE model for the likelihood of having any inpatient services (using EBR credits earned ) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.0495 0.0252 0.0000 0.0990 0.0498 FY0910 0.0183 0.0270 0.0346 0.0712 0.4971 Annual EBR credits (previous year) 0.0016 0.0003 0.0022 0.0010 <0 .0001 Annual EBR credits (two y ears ago) 0.0031 0.0004 0.0038 0.0024 <0 .0001 Age 0.0270 0.0007 0.0257 0.0283 <0 .0001 Gender (Female) Male 0.3951 0.0217 0.4375 0.3526 <0 .0001 Race/Ethnicity (White) Black 0.0512 0.0258 0.0007 0.1018 0.0471 Hispanic 0.3027 0.0401 0. 3812 0.2241 <0 .0001 Other 0.0871 0.0369 0.1594 0.0148 0.0182 County (Broward) Duval 0.0399 0.0214 0.0819 0.0021 0.0624 Eligibility (SSI) TANF 1.1309 0.0312 1.1920 1.0698 <0 .0001 Enrollment Length 0.1343 0.0023 0.1298 0.1388 <0 .0001 Risk Score 0.1832 0.0080 0.1676 0.1988 <0 .0001
75 Table 4 17. GEE model for the likelihood of having any inpatient services (more than 6 months of enrollment) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.1271 0.0305 0.0672 0.1870 <0 .0001 FY0910 0.1991 0.0317 0.1369 0.2613 <0 .0001 EBR Level Previous Year (Non earner) High 0.0256 0.0344 0.0419 0.0931 0.4571 Medium 0.1182 0.0333 0.0528 0.1835 0.0004 Low 0.0670 0.0336 0.0011 0.1329 0.0464 EBR Level Two years ago (Non earner) High 0.2525 0.0374 0.3258 0.1792 <0 .0001 Medium 0.2242 0.0406 0.3038 0.1447 <0 .0001 Low 0.4089 0.0496 0.5060 0.3118 <0 .0001 Age 0.0254 0.0008 0.0238 0.0270 <0 .0001 Gender (Female) Male 0.4042 0.0254 0.4541 0.3544 <0 .000 1 Race/Ethnicity (White) Black 0.1079 0.0307 0.0477 0.1682 0.0004 Hispanic 0.2972 0.0486 0.3925 0.2018 <0 .0001 Other 0.0673 0.0429 0.1514 0.0168 0.1169 County (Broward) Duval 0.0537 0.0254 0.1035 0.0039 0.0345 Eligibility (SSI) TANF 1.2002 0.0371 1.2729 1.1275 <0 .0001 Enrollment Length 0.0065 0.0069 0.0201 0.0071 0.3494 Risk Score 0.1996 0.0096 0.1807 0.2185 <0 .0001
76 Table 4 18. GEE model for the likelihood of having any inpatient services (using EBR credits earned and more than 6 months of enrollment) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.0574 0.0289 0.0009 0.1140 0.0466 FY0910 0.1426 0.0307 0.0825 0.2028 <0 .0001 Annual EBR credits (previous year) 0.0002 0.0004 0.0009 0.0006 0.6 751 Annual EBR credits (two years ago) 0.0025 0.0004 0.0033 0.0017 <0 .0001 Age 0.0255 0.0008 0.0239 0.0271 <0 .0001 Gender (Female) Male 0.4063 0.0254 0.4561 0.3564 <0 .0001 Race/Ethnicity (White) Black 0.0984 0.0307 0.0381 0.1586 0.001 4 Hispanic 0.2969 0.0487 0.3923 0.2016 <0 .0001 Other 0.0682 0.0430 0.1524 0.0160 0.1124 County (Broward) Duval 0.0515 0.0254 0.1013 0.0017 0.0426 Eligibility (SSI) TANF 1.1981 0.0370 1.2706 1.1257 <0 .0001 Enrollment Length 0.00 58 0.0069 0.0193 0.0077 0.3984 Risk Score 0.2009 0.0095 0.1822 0.2196 <0 .0001
77 Table 4 19. GEE model for annual total inpatient days Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.0440 0.0403 0.0350 0.1230 0.2750 FY0910 0.0124 0.0415 0.0689 0.0938 0.7642 EBR Level Previous Year (Non earner) High 0.1841 0.0379 0.2584 0.1097 <0 .0001 Medium 0.1608 0.0418 0.2428 0.0788 0.0001 Low 0.0551 0.0445 0.1423 0.0322 0.2163 EBR Level Two years ago (Non earner) High 0.0446 0.0437 0.1303 0.0411 0.3077 Medium 0.0613 0.0490 0.1573 0.0346 0.2103 Low 0.1045 0.0604 0.2228 0.0139 0.0836 Age 0.0017 0.0009 0.0002 0.0035 0.0771 Gender (Female) Male 0.1366 0.0320 0.0739 0.1994 <0 .0001 Race/Ethnicity (White ) Black 0.0083 0.0321 0.0713 0.0547 0.7967 Hispanic 0.1616 0.0577 0.2746 0.0485 0.0051 Other 0.2573 0.0577 0.1441 0.3704 <0 .0001 County (Broward) Duval 0.1276 0.0293 0.1849 0.0702 <0 .0001 Eligibility (SSI) TANF 0.7727 0.0421 0.8552 0.6901 <0 .0001 Enrollment Length 0.0188 0.0043 0.0105 0.0272 <0 .0001 Risk Score 0.0681 0.0078 0.0528 0.0833 <0 .0001
78 Table 4 20. GEE model for annual total inpatient days (using annual EBR earnings) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.0199 0.0385 0.0555 0.0954 0.6049 FY0910 0.0166 0.0402 0.0954 0.0623 0.6809 Annual EBR credits (previous year) 0.0016 0.0004 0.0024 0.0007 0.0002 Annual EBR credits (two years ago) 0.0003 0.0005 0.0012 0.0006 0.5572 Age 0.0014 0.0009 0.0004 0.0033 0.1258 Gender (Female) Male 0.1405 0.0320 0.0778 0.2033 <0 .0001 Race/Ethnicity (White) Black 0.0126 0.0322 0.0758 0.0506 0.6961 Hispanic 0.1657 0.0572 0.2779 0.0536 0.0038 Other 0.2538 0.0580 0.1402 0 .3675 <0 .0001 County (Broward) Duval 0.1241 0.0295 0.1819 0.0663 <0 .0001 Eligibility (SSI) TANF 0.7707 0.0421 0.8533 0.6881 <0 .0001 Enrollment Length 0.0154 0.0042 0.0072 0.0237 0.0002 Risk Score 0.0630 0.0078 0.0477 0.0783 <0 .0001
79 Table 4 21. GEE model for annual total inpatient days (more than 6 months of enrollment) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.0436 0.0450 0.0446 0.1318 0.3325 FY0910 0.0530 0.0482 0.0415 0.1474 0.2720 EBR Level Previous Year (Non earner) High 0.1325 0.0452 0.2211 0.0439 0.0034 Medium 0.1210 0.0487 0.2164 0.0255 0.0130 Low 0.0212 0.0512 0.1216 0.0791 0.6784 EBR Level Two years ago (Non earner) High 0.0445 0.0475 0.1375 0.0486 0.3488 Mediu m 0.0291 0.0535 0.1339 0.0757 0.5864 Low 0.1050 0.0703 0.2428 0.0329 0.1355 Age 0.0031 0.0011 0.0009 0.0052 0.0047 Gender (Female) Male 0.1372 0.0375 0.0636 0.2107 0.0003 Race/Ethnicity (White) Black 0.0101 0.0378 0.0640 0.0842 0.7897 Hispanic 0.1388 0.0761 0.2879 0.0103 0.0681 Other 0.1190 0.0619 0.0024 0.2404 0.0548 County (Broward) Duval 0.1443 0.0346 0.2121 0.0764 <0 .0001 Eligibility (SSI) TANF 0.8811 0.0531 0.9852 0.7771 <0 .0001 Enrollment Length 0.0277 0 .0112 0.0497 0.0056 0.0138 Risk Score 0.0814 0.0078 0.0661 0.0966 <0 .0001
80 Table 4 22. GEE model for annual total inpatient days (using annual EBR earnings and more than 6 months of enrollment) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708 ) FY0809 0.0181 0.0428 0.0659 0.1020 0.6730 FY0910 0.0317 0.0466 0.0596 0.1230 0.4963 Annual EBR credits (previous year) 0.0010 0.0005 0.0020 0.0000 0.0483 Annual EBR credits (two years ago) 0.0002 0.0005 0.0012 0.0008 0.7107 Age 0.0030 0. 0011 0.0008 0.0052 0.0065 Gender (Female) Male 0.1426 0.0375 0.0691 0.2160 0.0001 Race/Ethnicity (White) Black 0.0070 0.0377 0.0669 0.0809 0.8528 Hispanic 0.1454 0.0750 0.2925 0.0016 0.0526 Other 0.1169 0.0620 0.0047 0.2385 0.0595 Coun ty (Broward) Duval 0.1433 0.0350 0.2119 0.0747 <0 .0001 Eligibility (SSI) TANF 0.8761 0.0532 0.9803 0.7719 <0 .0001 Enrollment Length 0.0318 0.0115 0.0543 0.0094 0.0055 Risk Score 0.0777 0.0078 0.0624 0.0930 <0 .0001
81 Table 4 23. GE E model for the likelihood of having any ED visits Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.1491 0.0122 0.1252 0.1730 <0 .0001 FY0910 0.1611 0.0124 0.1368 0.1854 <0 .0001 EBR Level Previous Year (Non earner) High 0. 1215 0.0126 0.0968 0.1463 <0 .0001 Medium 0.0326 0.0129 0.0073 0.0579 0.0115 Low 0.0775 0.0125 0.1021 0.0529 <0 .0001 EBR Level Two years ago (Non earner) High 0.0440 0.0143 0.0721 0.0158 0.0022 Medium 0.0814 0.0161 0.1130 0.0497 <0 .0001 Low 0.1927 0.0166 0.2253 0.1601 <0 .0001 Age 0.0009 0.0004 0.0017 0.0001 0.0251 Gender (Female) Male 0.1576 0.0095 0.1763 0.1389 <0 .0001 Race/Ethnicity (White) Black 0.0713 0.0123 0.0473 0.0953 <0 .0001 Hispanic 0.0206 0.0163 0.0113 0.0526 0.2056 Other 0.1784 0.0185 0.2147 0.1421 <0 .0001 County (Broward) Duval 0.2115 0.0098 0.1923 0.2307 <0 .0001 Eligibility (SSI) TANF 0.1774 0.0176 0.2118 0.1430 <0 .0001 Enrollment Length 0.1763 0.0010 0.1743 0.1783 <0 .0001 Risk Score 0.0995 0.0052 0.0892 0.1098 <0 .0001
82 Table 4 24. GEE model for the likelihood of having any ED visits (using annual EBR earnings) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.1263 0.0119 0.1030 0.1497 <0 .0001 FY0910 0.1338 0.0121 0.1101 0.1574 <0 .0001 Annual EBR credits (previous year) 0.0012 0.0001 0.0009 0.0015 <0 .0001 Annual EBR credits (two years ago) 0.0007 0.0001 0.0010 0.0004 <0 .0001 Age 0.0008 0.0004 0.0016 0.0000 0.0485 Gender (Female) Male 0 .1588 0.0095 0.1775 0.1401 <0 .0001 Race/Ethnicity (White) Black 0.0621 0.0122 0.0381 0.0860 <0 .0001 Hispanic 0.0203 0.0163 0.0116 0.0522 0.2129 Other 0.1791 0.0185 0.2154 0.1428 <0 .0001 County (Broward) Duval 0.2100 0.0098 0.1908 0.22 92 <0 .0001 Eligibility (SSI) TANF 0.1760 0.0175 0.2103 0.1416 <0 .0001 Enrollment Length 0.1755 0.0010 0.1735 0.1776 <0 .0001 Risk Score 0.0962 0.0051 0.0862 0.1062 <0 .0001
83 Table 4 25. GEE model for the likelihood of having any ED visits (mor e than 6 months of enrollment) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.1150 0.0147 0.0861 0.1439 <0 .0001 FY0910 0.2548 0.0148 0.2257 0.2838 <0 .0001 EBR Level Previous Year (Non earner) High 0.2247 0.0156 0.1942 0 .2552 <0 .0001 Medium 0.1441 0.0161 0.1127 0.1756 <0 .0001 Low 0.0345 0.0158 0.0035 0.0654 0.0290 EBR Level Two years ago (Non earner) High 0.0369 0.0180 0.0721 0.0018 0.0396 Medium 0.0712 0.0201 0.1106 0.0319 0.0004 Low 0.1673 0.0207 0.20 79 0.1267 <0 .0001 Age 0.0037 0.0005 0.0047 0.0027 <0 .0001 Gender (Female) Male 0.1295 0.0119 0.1527 0.1062 <0 .0001 Race/Ethnicity (White) Black 0.1121 0.0155 0.0819 0.1424 <0 .0001 Hispanic 0.0612 0.0209 0.0203 0.1022 0.0034 Other 0 .1554 0.0229 0.2003 0.1105 <0 .0001 County (Broward) Duval 0.2137 0.0123 0.1896 0.2378 <0 .0001 Eligibility (SSI) TANF 0.2843 0.0210 0.3256 0.2431 <0 .0001 Enrollment Length 0.0761 0.0033 0.0697 0.0825 <0 .0001 Risk Score 0.1110 0.0072 0.0 969 0.1250 <0 .0001
84 Table 4 26. GEE model for the likelihood of having any ED visits (using annual EBR earnings and more than 6 months of enrollment) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.0863 0.0142 0.0584 0.1142 <0 .0 001 FY0910 0.2386 0.0144 0.2104 0.2668 <0 .0001 Annual EBR credits (previous year) 0.0021 0.0002 0.0018 0.0025 <0 .0001 Annual EBR credits (two years ago) 0.0004 0.0002 0.0008 0.0000 0.0411 Age 0.0036 0.0005 0.0046 0.0026 <0 .0001 Gender (Female) Male 0.1310 0.0119 0.1543 0.1078 <0 .0001 Race/Ethnicity (White) Black 0.1050 0.0154 0.0748 0.1353 <0 .0001 Hispanic 0.0625 0.0209 0.0216 0.1034 0.0028 Other 0.1565 0.0229 0.2014 0.1116 <0 .0001 County (Broward) Duval 0.2122 0.0123 0.1881 0.2364 <0 .0001 Eligibility (SSI) TANF 0.2874 0.0210 0.3285 0.2462 <0 .0001 Enrollment Length 0.0760 0.0032 0.0697 0.0824 <0 .0001 Risk Score 0.1130 0.0071 0.0990 0.1270 <0 .0001
85 Table 4 27. GEE model for rates of ED visits Estimate S.E 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.0477 0.0159 0.0167 0.0788 0.0026 FY0910 0.0376 0.0198 0.0011 0.0763 0.0570 EBR Level Previous Year (Non earner) High 0.0487 0.0185 0.0849 0.0125 0.0084 Medium 0.0706 0.0219 0.113 5 0.0277 0.0012 Low 0.0158 0.0213 0.0575 0.0259 0.4574 EBR Level Two years ago (Non earner) High 0.0892 0.0215 0.1313 0.0471 <0 .0001 Medium 0.0502 0.0277 0.1044 0.0041 0.0698 Low 0.0522 0.0232 0.0976 0.0067 0.0244 Age 0.0061 0.0006 0. 0050 0.0073 <0 .0001 Gender (Female) Male 0.0444 0.0138 0.0714 0.0174 0.0013 Race/Ethnicity (White) Black 0.0907 0.0193 0.1285 0.0529 <0 .0001 Hispanic 0.0910 0.0244 0.1389 0.0432 0.0002 Other 0.1251 0.0328 0.1894 0.0609 0.0001 C ounty (Broward) Duval 0.0148 0.0128 0.0399 0.0103 0.2480 Eligibility (SSI) TANF 0.1098 0.0333 0.1751 0.0444 0.0010 Enrollment Length 0.0745 0.0026 0.0694 0.0796 <0 .0001 Risk Score 0.0013 0.0077 0.0138 0.0165 0.8627
86 Table 4 28. GEE m odel for rates of ED visits (using annual EBR earnings) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.0463 0.0164 0.0141 0.0784 0.0048 FY0910 0.0321 0.0206 0.0082 0.0724 0.1180 Annual EBR credits (previous year) 0.0006 0.0 002 0.0010 0.0002 0.0033 Annual EBR credits (two years ago) 0.0010 0.0002 0.0015 0.0006 <0 .0001 Age 0.0059 0.0006 0.0048 0.0071 <0 .0001 Gender (Female) Male 0.0440 0.0136 0.0707 0.0172 0.0013 Race/Ethnicity (White) Black 0.0911 0.0 190 0.1284 0.0539 <0 .0001 Hispanic 0.0913 0.0241 0.1385 0.0441 0.0001 Other 0.1251 0.0324 0.1887 0.0616 0.0001 County (Broward) Duval 0.0145 0.0127 0.0394 0.0104 0.2525 Eligibility (SSI) TANF 0.1095 0.0332 0.1746 0.0443 0.0010 Enrollment Length 0.0740 0.0026 0.0689 0.0792 <0 .0001 Risk Score 0.0001 0.0077 0.0150 0.0151 0.9924
87 Table 4 29. GEE model for rates of ED visits (more than 6 months of enrollment) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY 0809 0.0474 0.0145 0.0189 0.0758 0.0011 FY0910 0.0719 0.0147 0.0432 0.1007 <0 .0001 EBR Level Previous Year (Non earner) High 0.0166 0.0154 0.0468 0.0135 0.2785 Medium 0.0415 0.0164 0.0735 0.0094 0.0112 Low 0.0004 0.0172 0.0340 0.0332 0.981 4 EBR Level Two years ago (Non earner) High 0.0907 0.0182 0.1263 0.0551 <0 .0001 Medium 0.0481 0.0218 0.0909 0.0053 0.0275 Low 0.0728 0.0192 0.1104 0.0353 0.0001 Age 0.0038 0.0004 0.0030 0.0047 <0 .0001 Gender (Female) Male 0.0601 0.0100 0.0798 0.0404 <0 .0001 Race/Ethnicity (White) Black 0.0584 0.0139 0.0857 0.0310 <0 .0001 Hispanic 0.0739 0.0164 0.1062 0.0417 <0 .0001 Other 0.1047 0.0211 0.1460 0.0635 <0 .0001 County (Broward) Duval 0.0178 0.0098 0.0015 0.0 371 0.0705 Eligibility (SSI) TANF 0.1575 0.0206 0.1978 0.1172 <0 .0001 Enrollment Length 0.0365 0.0037 0.0292 0.0438 <0 .0001 Risk Score 0.0240 0.0049 0.0143 0.0337 <0 .0001
88 Table 4 30. GEE model for rates of ED visits (using annual EBR earnings and more than 6 months of enrollment) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.0450 0.0145 0.0166 0.0734 0.0019 FY0910 0.0706 0.0153 0.0407 0.1005 <0 .0001 Annual EBR credits (previous year) 0.0003 0.0002 0.0006 0.000 0 0.0919 Annual EBR credits (two years ago) 0.0011 0.0002 0.0015 0.0007 <0 .0001 Age 0.0037 0.0004 0.0029 0.0046 <0 .0001 Gender (Female) Male 0.0598 0.0101 0.0796 0.0401 <0 .0001 Race/Ethnicity (White) Black 0.0602 0.0140 0.0875 0.03 28 <0 .0001 Hispanic 0.0738 0.0165 0.1061 0.0414 <0 .0001 Other 0.1044 0.0212 0.1459 0.0628 <0 .0001 County (Broward) Duval 0.0180 0.0098 0.0013 0.0373 0.0671 Eligibility (SSI) TANF 0.1583 0.0209 0.1992 0.1174 <0 .0001 Enrollment Len gth 0.0366 0.0038 0.0292 0.0441 <0 .0001 Risk Score 0.0229 0.0051 0.0130 0.0328 <0 .0001
89 Table 4 31. GEE model for the likelihood of having any avoidable hospitalizations Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.2178 0 .0537 0.1125 0.3231 <0 .0001 FY0910 0.1001 0.0599 0.2176 0.0173 0.0947 EBR Level Previous Year (Non earner) High 0.6740 0.0530 0.7778 0.5702 <0 .0001 Medium 0.7044 0.0600 0.8220 0.5868 <0 .0001 Low 0.8941 0.0695 1.0302 0.7579 <0 .0001 EBR Level Two years ago (Non earner) High 0.0914 0.0642 0.2172 0.0344 0.1546 Medium 0.2567 0.0769 0.4075 0.1060 0.0008 Low 0.8999 0.1151 1.1255 0.6743 <0 .0001 Age 0.0269 0.0014 0.0242 0.0295 <0 .0001 Gender (Female) Male 0.1672 0.0445 0.2545 0.0799 0.0002 Race/Ethnicity (White) Black 0.1309 0.0521 0.0287 0.2330 0.0120 Hispanic 0.4552 0.0875 0.6267 0.2837 <0 .0001 Other 0.1247 0.0741 0.2699 0.0204 0.0922 County (Broward) Duval 0.1220 0.0433 0.0371 0.2068 0.0048 Eli gibility (SSI) TANF 1.5370 0.0556 1.6459 1.4281 <0 .0001 Enrollment Length 0.0047 0.0044 0.0039 0.0134 0.2810 Risk Score 0.1401 0.0096 0.1213 0.1589 <0 .0001
90 Table 4 32. GEE model for the likelihood of having any avoidable hospitalizations (usi ng annual EBR earnings) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.1744 0.0523 0.0718 0.2770 0.0009 FY0910 0.2382 0.0602 0.3562 0.1202 <0 .0001 Annual EBR credits (previous year) 0.0061 0.0007 0.0074 0.0048 <0 .0001 Annual EBR credits (two years ago) 0.0024 0.0007 0.0039 0.0010 0.0010 Age 0.0275 0.0013 0.0249 0.0301 <0 .0001 Gender (Female) Male 0.1669 0.0444 0.2539 0.0798 0.0002 Race/Ethnicity (White) Black 0.1124 0.0521 0.0103 0.2145 0.0309 Hisp anic 0.4612 0.0873 0.6324 0.2900 <0 .0001 Other 0.1225 0.0740 0.2677 0.0226 0.0979 County (Broward) Duval 0.1489 0.0431 0.0644 0.2333 0.0006 Eligibility (SSI) TANF 1.4897 0.0552 1.5980 1.3815 <0 .0001 Enrollment Length 0.0018 0.0044 0.0104 0.0067 0.6738 Risk Score 0.1255 0.0089 0.1081 0.1429 <0 .0001
91 Table 4 33. GEE model for the likelihood of having any avoidable hospitalizations (more than 6 months of enrollment) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.2409 0.0778 0.0884 0.3935 0.0020 FY0910 0.2207 0.0857 0.0526 0.3887 0.0101 EBR Level Previous Year (Non earner) High 0.1794 0.0763 0.3290 0.0298 0.0188 Medium 0.0752 0.0810 0.2340 0.0836 0.3535 Low 0.2981 0.0950 0.4843 0.1120 0. 0017 EBR Level Two years ago (Non earner) High 0.0606 0.0874 0.2319 0.1107 0.4880 Medium 0.2428 0.1036 0.4458 0.0397 0.0191 Low 0.6202 0.1513 0.9167 0.3237 <0 .0001 Age 0.0432 0.0019 0.0393 0.0470 <0 .0001 Gender (Female) Male 0.242 2 0.0668 0.3730 0.1113 0.0003 Race/Ethnicity (White) Black 0.1528 0.0735 0.0087 0.2970 0.0377 Hispanic 0.7373 0.1475 1.0264 0.4481 <0 .0001 Other 0.3767 0.1093 0.5909 0.1625 0.0006 County (Broward) Duval 0.2619 0.0665 0.1315 0.3923 < 0 .0001 Eligibility (SSI) TANF 1.6585 0.0951 1.8450 1.4721 <0 .0001 Enrollment Length 0.1741 0.0158 0.2051 0.1431 <0 .0001 Risk Score 0.1753 0.0142 0.1475 0.2031 <0 .0001
92 Table 4 34. GEE model for the likelihood of having any avoidable hospital izations (using annual EBR earnings and more than 6 months of enrollment) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.1436 0.0719 0.0027 0.2845 0.0458 FY0910 0.1169 0.0838 0.0473 0.2811 0.1630 Annual EBR credits (previous year) 0.0007 0.0008 0.0023 0.0009 0.3626 Annual EBR credits (two years ago) 0.0006 0.0010 0.0025 0.0012 0.5111 Age 0.0435 0.0019 0.0396 0.0473 <0 .0001 Gender (Female) Male 0.2432 0.0667 0.3738 0.1125 0.0003 Race/Ethnicity (White) Bl ack 0.1439 0.0735 0.0003 0.2880 0.0504 Hispanic 0.7427 0.1473 1.0315 0.4540 <0 .0001 Other 0.3768 0.1092 0.5909 0.1628 0.0006 County (Broward) Duval 0.2807 0.0663 0.1507 0.4107 <0 .0001 Eligibility (SSI) TANF 1.6391 0.0950 1.8253 1. 4529 <0 .0001 Enrollment Length 0.1828 0.0158 0.2137 0.1519 <0 .0001 Risk Score 0.1716 0.0139 0.1444 0.1988 <0 .0001
93 Table 4 35. GEE model for annual rates of avoidable hospitalizations Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.1163 0.0428 0.0325 0.2001 0.0065 FY0910 0.0494 0.0455 0.0399 0.1387 0.2780 EBR Level Previous Year (Non earner) High 0.0616 0.0754 0.2093 0.0862 0.4143 Medium 0.0154 0.0726 0.1577 0.1269 0.8321 Low 0.1587 0.0898 0.0174 0.3348 0 .0773 EBR Level Two years ago (Non earner) High 0.0235 0.0547 0.0837 0.1307 0.6676 Medium 0.0911 0.0809 0.0674 0.2496 0.2601 Low 0.3206 0.2137 0.0983 0.7396 0.1336 Age 0.0015 0.0015 0.0015 0.0045 0.3182 Gender (Female) Male 0.0196 0.04 73 0.0731 0.1123 0.6786 Race/Ethnicity (White) Black 0.0157 0.0519 0.0860 0.1173 0.7625 Hispanic 0.1433 0.0502 0.2416 0.0449 0.0043 Other 0.0099 0.0641 0.1158 0.1356 0.8771 County (Broward) Duval 0.0570 0.0520 0.0450 0.1590 0.2732 E ligibility (SSI) TANF 0.2076 0.0796 0.3635 0.0516 0.0091 Enrollment Length 0.0065 0.0055 0.0042 0.0172 0.2326 Risk Score 0.0222 0.0108 0.0011 0.0433 0.0392
94 Table 4 36. GEE model for annual rates of avoidable hospitalizations (using annual EB R earnings) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.1475 0.0425 0.0641 0.2308 0.0005 FY0910 0.0974 0.0390 0.0210 0.1737 0.0125 Annual EBR credits (previous year) 0.0010 0.0007 0.0024 0.0004 0.1646 Annual EBR credits (two years ago) 0.0003 0.0006 0.0008 0.0014 0.6424 Age 0.0019 0.0014 0.0008 0.0047 0.1691 Gender (Female) Male 0.0186 0.0474 0.0743 0.1114 0.6952 Race/Ethnicity (White) Black 0.0256 0.0564 0.0850 0.1363 0.6495 Hispanic 0.1456 0.0500 0.2436 0.0477 0.0036 Other 0.0062 0.0639 0.1190 0.1314 0.9226 County (Broward) Duval 0.0545 0.0515 0.0464 0.1554 0.2899 Eligibility (SSI) TANF 0.2148 0.0824 0.3764 0.0533 0.0091 Enrollment Length 0.0084 0.0058 0.0031 0.0198 0.1517 R isk Score 0.0300 0.0094 0.0115 0.0485 0.0015
95 Table 4 37. GEE model for annual rates of avoidable hospitalizations (more than 6 months of enrollment) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.1293 0.0715 0.0109 0.2695 0.0706 FY0910 0.0318 0.0802 0.1254 0.1890 0.6916 EBR Level Previous Year (Non earner) High 0.0831 0.0911 0.2617 0.0955 0.3617 Medium 0.0043 0.0896 0.1713 0.1798 0.9619 Low 0.2358 0.1138 0.0128 0.4589 0.0383 EBR Level Two years ago (Non earne r) High 0.0216 0.0704 0.1164 0.1597 0.7588 Medium 0.1084 0.0972 0.0821 0.2988 0.2647 Low 0.4435 0.2839 0.1130 1.0000 0.1183 Age 0.0002 0.0021 0.0039 0.0043 0.9216 Gender (Female) Male 0.0008 0.0772 0.1520 0.1504 0.9920 Race/Ethnicity (White) Black 0.0192 0.0762 0.1302 0.1685 0.8014 Hispanic 0.2829 0.0785 0.4366 0.1291 0.0003 Other 0.0165 0.1062 0.2246 0.1917 0.8767 County (Broward) Duval 0.1004 0.0914 0.0787 0.2795 0.2720 Eligibility (SSI) TANF 0.2834 0.1 287 0.5356 0.0312 0.0276 Enrollment Length 0.0098 0.0149 0.0391 0.0195 0.5117 Risk Score 0.0304 0.0135 0.0040 0.0568 0.0241
96 Table 4 38. GEE model for annual rates of avoidable hospitalizations (using annual EBR earnings and more than 6 months of enr ollment) Estimate S.E. 95% Confidence Limits Pr > |Z| Year (FY0708) FY0809 0.1865 0.0670 0.0552 0.3178 0.0054 FY0910 0.1119 0.0608 0.0073 0.2311 0.0658 Annual EBR credits (previous year) 0.0015 0.0010 0.0034 0.0004 0.1314 Annual EBR credits ( two years ago) 0.0002 0.0006 0.0011 0.0014 0.8020 Age 0.0003 0.0021 0.0039 0.0045 0.8957 Gender (Female) Male 0.0032 0.0800 0.1537 0.1601 0.9679 Race/Ethnicity (White) Black 0.0380 0.0847 0.1280 0.2040 0.6536 Hispanic 0.2686 0.0799 0. 4251 0.1120 0.0008 Other 0.0238 0.1059 0.2315 0.1838 0.8219 County (Broward) Duval 0.0956 0.0915 0.0837 0.2750 0.2960 Eligibility (SSI) TANF 0.3045 0.1347 0.5684 0.0406 0.0237 Enrollment Length 0.0006 0.0149 0.0286 0.0297 0.9681 Ri sk Score 0.0369 0.0124 0.0127 0.0612 0.0029
97 CHAPTER 5 DISCUSSION AND CONCLUSIONS Discussion Descriptive Results Due to the fact that nearly 60% of EBR credits earned during the first eighteen months had been through keeping primary care appointments by a dults and children, the state decided in March 20 08 to decrease the number o f allowed occurrences for each year from t wo occurrences to one and the amount of credits rewarded for each visit reduced by half to $7.5 0 per visit. The change was implemented at the beginning of FY 20 08 09, the third year of the EBR program. As a result, the earning by keeping office visits peaked in FY 20 07 08 an d dropped significantly in the following year (Table 4 1 ). Compared to office visits, the numbers of occurrence s and cred its earned in dollar amount s for other EBR credit earning behaviors have been con sistently increasing over the years. The trend reflects the changes in regulation regarding EBR program and the growing awareness of the program within the Medicaid enrollee p opulation. Although significant differences were found in all three years in terms of the average percentages of credits earned for each behavior type, it is difficult to identify a clear pattern for the differences and changes over time. However, when th e average personal annual total expenditures and acute care utilizations were compared by year, significant decrease s over time in all outcomes were observed. Two factors can potentially contribute to these differences. Florida Medicaid Reform was implemen ted at the beginning of FY2006 07 and the decrease in expenditures and utilizations can be attributed to the overall effect of Medicaid Reform as a whole, the EBR program included. Due to the economic downturn from the year of 20 07 to 20 09, considerably
98 more people enrolled in Medicaid during that period of time. As these new enrollees became eligible for Medicaid, they may not have as much health care utilizations compared to old enrollees and as a result, lowered average personal annual expenditures and acute care usages. Demographics of the study population were compared among groups with different EBR participation levels. Results suggest that, for each year, earner groups generally have a longer enrollment length than non earners. It was interesting t o notice that average annual health care expenditures are also higher for earner groups than non earner group. More importantly, participant groups had higher average risk scores and more enrollees eligible through SSI. This could potentially lead to biase s in analyses caused by self selection and once again prove the necessity to control for these factors in the analyses. More detailed discussion on selection bias can be found in the following sections. Medicaid Expenditure s The effect of prior year EBR p rogram participation on annual Medicaid expenditures was examined and it was found that, as hypothesized, all earners groups (high/medium/low) were associated with lower annual Medicaid expenditures compared to enrollees who did not participate in EBR prog ram after controlling for other factors. However, contrary to the original hypothesis, lower level participation was actually associated with lower health expenditures when compared among all earners groups. The same pattern could still be observed after p articipation level from two years ago was added into the model. Participating in EBR program in the past was associated with lower expenditures compared to non participants (Table 4 6). However, as explained in the conceptual framework, participating in EB R program would have directly contributed
99 to an other healthy behaviors used to earn EBR credits. This could be the reason for the reversed pattern among EBR program participa nts described above. It is also interesting to see that for the same participation level group, participation from two years prior had a greater effect on lowering Medicaid expenditures than participation in the previous year. The effect of a program like EBR will not be instantaneous and its influence on health expenditures and utilizations could take time to have an effect. The results from the model are fully consistent with this supposition. When the analyses were limited to Medicaid enrollees who had m ore than 6 months of enrollment in the given year, participation in the EBR program in the previous year was found to be associated with more expenditures than non participants while participation from two years ago was still linked to significantly lower Medicaid spending. Enrollees with longer enrollment length were likely to be the group of people that were sicker and/or poorer or the type of people who seek health care more actively. The effect of the EBR program on lowering health care expenditures co uld take an even longer time to show in this population. It is also possible that the risk scores used in the unobserved differences between people who had more than 6 months enrollment and the general enrollee population. Sensitivity analyses using annual EBR credits as the indicator of participation level rather than categorical participation level groups showed similar patterns and very similar results in all models which suggests that the way EBR program participation level was defined did not significantly impact the results.
100 Acute Care Utilizations Inpatient days Both prior year EBR program participation and participation from two years ago were associated with lo wer odds of having any inpatient services for the year. And participation from two years ago showed a much greater effect than prior year participation, regardless of how EBR credits earned was operationalized. With the limitation on enrollment length appl ied, the effect of EBR program participation level from previous year becomes less apparent while the effect of participation from two years ago still holds (Table s 4 17, 4 18). Overall, the influence of EBR program participation on the likelihood of havi ng any inpatient services mirrors its effect on total Medicaid expenditures. For enrollees who have used some inpatient services, prior year EBR participation appears to be associated with lower number of annual total inpatient days while the effect of pa rticipation from two years ago is not conclusive. With categorical EBR program participation level, prior year low level participation does not have a significant effect on the number of inpatient days in any model analyzed, yet both high and medium groups showed significant results. More importantly, in both models with or without enrollment length limitation, high level participation ha d a greater effect on lowering annual inpatient days than medium level. This means, for people who were sick enough to u se any inpatient services, using more services eligible for EBR credit earning s could reduce their inpatient length of stay in coming years. ED visits Prior year EBR program participation is associated with greater odds of having ED visits although partic ipation from two years ago was significantly associated with
101 lower likelihood of having any ED visits. Considering the absolute value, EBR program participation from prior year appear ed to have a greater effect on the likelihood of having any ED visits tha n participation from two years ago. This finding was still consistent with the assumption that a program like EBR would need a certain period of time for its influence to take effect. Also, regardless of the limitation on enrollment length, models that use d categorical participation level showed an interesting trend. For participation levels in the previous year, results suggest ed that a higher level of participation led to greater odds of having one or more ED visits. However, for EBR program participation levels from two years ago, it was shown that although all participants would have lower odds of having ED visits compared to non participants, the lower level of participation had a greater effect on lowering the odds of having ED visits than higher leve l of participation. It is possible that people in the high participation level group would also more actively seek health care in general. When ED visits become their choice of care, people in the high level group would also be more likely to have ED visit s than other groups. When the lag effect of the EBR program was finally strong enough to take over, the trend was reversed but due to the same reason discussed above, lower level of participation was associated with lower odds of having ED visits than hi gher level of participation. Among enrollees who had ED visits, EBR program participation during both the previous year and two years prior was associated with a lower number of ED visits although the effect of prior year participation became insignifican t when the analyses were limited to individuals who had had more than 6 months of enrollment in Medicaid in a given year. This result further confirmed the impact of the EBR program when
102 for their health status, and other factors have been adjusted for in the models. The effect of EBR program participation from two years ago still had a greater effect on lowering annual total ED visits than prior year participation, once again demonstrating that it takes time for an Avoidable hospitalizations Avoidable hospitalizations are generally sensitive to the changes in access/utilization of primary care. The results from analyzing the effect of EBR program participation on the likelihood of having any avoidable hospitalizations agree with previous findings (Rosano et al. 20 13). The models that examined the whole study population showed that EBR program participation in both the previou s year and two years ago was significantly associated with lower odds of having any avoidable hospitalizations. Prior year EBR program participation appears to have a greater effect on reducing likelihood of having avoidable hospitalization compared to EBR program participation from two years ago (Table s 4 31, 4 32). However, the relationship between EBR participation level and the odds of having any avoidable hospitalizations becomes less clear when enrollment length limitation was applied to the model. Al so, no significant effects of EBR program participation on the rates of avoidable hospitalization s were observed among enrollees who had had avoidable hospitalizations in a given year. A longer study period and more data are needed to more accurately measu re the hospitalizations.
103 Policy Implications As one of the first state s to incorporate financial incentives for enrollees in its state Medicaid program individual health outc omes will improve as people take an active role in managing and understanding their health Waiver 20 05). Financial incentives were given to Medicaid enrollees to promote the u se of preventive services and adopting healthy behaviors, such as exercise, weight control, and the like It was anticipated that, with the EBR program, people will be incentivized to be more responsible for their own health and health care needs and in tu rn achieve better health outcomes. Eventually, the effect of the program will translate into less utilization of expensive and preventable health services and cost savings for Medicaid. The EBR program in Florida is one of the first public programs that us e s financial incentives to promote healthy behaviors. Previously, most financial incentive programs were implemented in private settings on a much smaller scale. Also, most of these programs only focused on one or a few behaviors. For that reason, Florida Medicaid Reform EBR program has generated great interest nationwide Although only data from FY2007 08 to FY2009 10 was available for this study, with a rigorous study design and sensitivity analyses, the study was able to show that participating in EBR pr ogram is generally associated with lower Medicaid expenditures and acute care utilization. During the first three years of EBR program, there were multiple regulation changes related to eligible healthy behaviors and previously documented concerns regardin g the awareness of the program (Hall et al. 20 12). Despite those issues that may have negatively affect ed the effectiveness in the early stage of implementation, the findings from this study are quite
104 encouraging. It has been shown that, com pared to non participants in the reform counties, EBR program participants were more likely to have lower health expenditures and lower odds of having acute care utilization than non participants. By using EBR program participation in previous years to est imate changes in Medicaid expenditures and utilizations, the lag effect of this financial incentive program was observed. Interesting trends were also found for Medicaid expenditures and ED visits as lower level participation was associated with lower expe nditures and lower odds of having ED visits compared to higher level of EBR program participation. It could be attributed to the possibility of people in the high level group being more likely to actively seek care than people in other groups. It may also be an indication that the current incentive structure of the program is not optimal and should be modified in the future. As illustrated by descriptive statistics (Table s 4 1 to 4 4), most EBR credits were earned by keeping appointments and uti lizing preventive services yet only very small portion of credits were earned by participating in disease management programs and other healthy behaviors. It may be necessary for policy makers to consider lowering the obstacles that prevent people from pa rticipating in these programs and earning credits from them. At the same time, the potential lack of access to primary care for certain population s cannot be overlooked. Nonetheless, the findings from this study suggest that the effectiveness of the EBR pr early stage s Limitations The results of this study demonstrate that it takes time for the EBR program to fully impact health care expenditures and utilization. Only three years of data were available for thi s study. Although the influence of the EBR program has already been
105 identified in the study, more data and a longer study period will certainly benefit future studies. Given the adjustment of the program in the early stage of implementation, the effect of the EBR program would eventually stabilize after a certain period of time. If more recent data were available, the effectiveness of EBR program on health care expenditures and utilizations could be examined with better accuracy. Enrollee s s w ere controlled for in the analyses as a proxy for health status. For the enrollees without assigned risk scores, they were assumed to be risk neutral in the study. If there were unknown reasons regarding unassigned risk scores that were not captured in the st udy, it could potentially lead to selection bias. The statistical models also included eligibility status in the analyses. Generally, individuals enrolled through SSI have worse health conditions than people enrolled through TANF (Ref) This can be viewed In the base models, the EBR program participation level was arbitrarily categorized based on distribution of the data. However, in reality, it is difficult to predict the threshold at which EB R program participation will show its effect on utilizations and expenditures. Although significant finding s were observed with the current definition s of high, medium, and low level s of EBR program participation, it is possible that it was not optimized. Sensitivity analyses were conducted by using annual EBR earnings as a continuous variable and the results were consistent with what was observed using categorical participation levels. Th is study limits the study population to Medicaid Reform participants from Broward and Duval counties who enrolled in PSN plans. HMO enrollees were not included in the study because of their incomplete utilization data. PSN plans have a
106 larger proportion of beneficiaries eligible through SSI compared to HMO plans and people enrolled through SSI are generally sicker than TANF eligible enrollees. Therefore, it is not advised to generalize the findings from this study to the general Medicaid Reform population (including the eligibility status variable in the model that helped ad dress this issue to some extent) without further validation. Also, the large variation in enrollment length of time among enrollees and discontinued eligibility for some enrollees created certain difficulties for the analyses, although enrollment length of time was used in the statistical models to control for exposure time. Sensitivity analyses were also conducted to examine the effect of the program for individuals who enrolled more than 6 months in Medicaid in the given year. Future Study This is the fir st time that the Florida Medicaid EBR program has been evaluated for its influence on health expenditures and acute care utilization. Given limitations on data availability, the study was still able to present significant findings regarding the effectivene ss of the program with a rigorous study design. It also points to some future research opportunities that would investigate the program in more depth or help address some limitations mentioned in the previous section. A study over a longer period of time c an be carried out when more recent Medicaid data become available. The data from FY2006 07 and FY2007 08 may contain some unwanted variations that related to regulation changes in the program and other factors. A future study can start from FY2008 09 and e xamine the effect of the program after it has stabilized over time. Also, more than two years of lag value of EBR program participation can be added into the analysis to monitor any long term relationship between EBR program participation and health care
107 e xpenditures/utilization. Currently, only a small proportion of EBR credits were earned through participation in disease management programs or other more complex behavior changes. However, active participation in these programs could substantially lower he alth risk for the target population and reduce their preventable acute care utilization. It would be intriguing to design a study that specifically examines enrollees who participated in these programs and compare them to other participants and non partici pants for their health expenditures and utilization. Similar studies can be designed for different type of EBR eligible behaviors and the results from these studies can potentially help policy makers adjust the incentive structure of the EBR program. One o f the end goals of the EBR program is to achieve cost savings for Florida Medicaid. Therefore, we should not only consider its effect on lowering health expenditures, but also take the costs of the program into account. The s ervices center and data managem ent system were built specifically for the EBR program and administrative costs should also be considered as sunk costs in the evaluation of the program. Future study can compare the accrued and projected savings in Medicaid expenditures and the costs of t he program to determine whether the EBR program will be economically favorable to the state government over the longer term The findings from this study are encouraging and inspiring. It is the first study that systematically examined the effect of the EB R program on Medicaid expenditures and utilization. Future study is needed to investigate the program more closely to provide additional information to the evidence base on using a financial incentives program in the public sector.
108 LIST OF REFERENCES Agenc y for Health Care Administration. Florida's Medicaid Reform Choice Counseling and Enhanced Benefits. Technical Advisory Panel Meeting Tallahassee Florida 20 10. American Cancer Society 20 12 Cancer Facts and Figures. at http://www.cancer.org/acs/groups/content/@epidemiologysurveilance/documents /document/acspc 031941.pdf Atkinson, W. Charles Wolfe; and J. Hamborsky, eds. 20 12. Epidemiology and Prevention of Vacci ne Preventable Diseases, 12 th Edition. Atlanta: Centers for Disease and Prevention. Berns, G. S., D. Laibson, and G. Loewenstein. 20 toward an Trends Cogn Sci 11(11): 482 8. Birkhead, G. S., C. W. LeBaron, P. Parsons, J. C. Grabau, E. Maes, L. Barr Gale, J. of children enrolled in the Special Supplemental Food Program for Women, Infants, and Children (WIC). The JAMA 274(4): 312 6. Bloch, M. J., D. S. Armstrong, L. Dettling, A. Hardy, K. Caterino, and S. Barrie. 20 06. program, monetary incentives, or usual care in the treatment of dyslipidemia J Occup Environ Med 48(7): 675 81. Blue, C. L. AAOHN J 43(2): 76 86. Bra gdon, T. 2011. Naples, FL: Foundation for Government Accountability. Burr, M. L., J. Trembeth, K. B. Jones, J. Geen, L. A. Lynch, and Z. E. Roberts. 20 07. e intake of fruit and fruit juice by Public Health Nutr 10(6): 559 65. Cahill, K. and R. Perera. 20 Cochrane Database Syst Rev (3): CD004307. Cahill, K. and R. Perera. 20 Cochrane Database Syst Rev (4): CD004986. Center s for Disease Control and Prevention (CDC). Data and Statistics for Overweight and Obesity. http://www.cdc.gov/obesity/data/index.html
109 Chapter 20 05 133 2005. Tallahassee, FL: Laws of Florida. Inquiry 27(1): 24 38. The Commonwealth Fund. 2007. "Public Programs Are Using Incent ives to Promote Healthy Behavior." New York: The Commonwealth Fund Cutler, D. M. and A. Lleras Muney. 20 J Health Econ 29(1): 1 28. Diggle, P K.Y. Liang and S.L. Zeger 1994 "Analysis of Longitudinal Data ". Oxford: Oxford University Press. Florida Medicaid Reform Application for 1115 Re search Demonstration and Waiver 2005. Tallahassee, FL: Florida Agency for Healthcare Administration Georgetown University Healthy Policy Institute 20 08. Policy Briefing #6" The Enhanced Benefits Rewards Program: Is it changing the way Medicaid beneficia ries approach their health?" Washington, DC: Georgetown University. Goodman, C., & Anise, A. 20 06. "What is known about the effectiveness of economic instruments to reduce consumption of foods high in saturated fats and other energy dense foods for preven ting and treating obesity?" Copenhagen, Denmark: WHO Regional Office for Europe (Health Evidence Network report). Greene, J. 20 child visits among low Med Care Res Rev 68(5): 579 93. Grossma h ealth c apital and d emand for h Journal of Political Economy 80(2): 223 25. Hall, A.G., C.H. Lemak, A.Y. Landry, and R.P. Duncan. 20 1 3 h ealthy b ehaviors: Experience f rom Florida Medicaid's Enhanced Benefit Re wards Program Journal of Primary Care & Community Health 4(2); 112 18 Newcastle exercise project: a randomised controlled trial of methods to promote physical activity in BMJ 319(7213): 828 32. Harman, J. S., C. H. Lemak, M. Al Amin, A. G. Hall, and R. P. Duncan. 20 in per member per month expenditures after implementation of Florida's Medicaid Health Serv Res 46(3): 787 80 4. Herman, C. W., S. Musich, C. Lu, S. Sill, J. M. Young, and D. W. Edington. 20 06. based online physical activity intervention on J Occup Environ Med 48(9): 889 95.
110 Henry J. Kaiser Family Foundation 2012. A t http://www.statehealthfacts.org/profileind.jsp?cat=4&sub=47&rgn=11&cmprgn=1 Henry J. Kaiser Family Foundation, Medicaid/CHIP A t http://kff.org/state category/medicaid chip/ Higgins, S. T. and H. D. Chilcoat. 20 Drug Alcohol Depend 104 Suppl 1: S1 5. Higgins, S. T., K. Silverman, S. C. Sigmon, and N A. Naito. 20 Prev Med 55 Suppl: S2 6. Hutchins, S. S., J. Rosenthal, P. Eason, E. Swint, H. Guerrero, and S. Hadler. 1999. effectiveness of linking the special supplemental program for J Public Health Policy 20 (4): 408 26. Isaacs, S. L. and S. A. Schroeder. 20 the ignored determinant of the N Engl J Med 351(11): 1137 42. Jeffery, R. W. 20 12. Prev Med 55 Suppl: S61 7. Jepson, R., A. Clegg, C. Forbes, R. Lewis, A. Sowden, and J. Kleijnen. 20 determinants of screening uptake and interventions for increasing uptake: a Health Technol Assess 4(14): i vii, 1 133. Kane, R. L., P. E. Johnson, R. J. Town, and M. Butler. 20 Am J Prev Med 27(4): 327 52. Kaplan, C. P., R. Bastani, T. R. Be lin, A. Marcus, K. Nasseri, and M. Y. Hu. 20 00. up after an abnormal pap smear: results from a quasi J Womens Health Gend Based Med 9(7): 779 90. Kenkel, D. S. 1994. "The demand for preventative medical care." Appl Econ 26:313 25 Kerpelman, L. C., D. B. Connell, and W. J. Gunn. 20 JAMA 284(1): 53 9. Laken, M. P. centives to increase participation in prenatal Obstet Gynecol 85(3): 326 9. Am J Health Promot 9(2): 85 7, 124.
111 Liang K. Y., a nd S Zeger. 1986. "Longitudinal data analysis using generalized linear models". Biometrika 73 (1): 13 22. Lillard L W. Manning C. E. Peterson, N. Lurie, G. A. Goldberg, and C. E. Phelps. 1986. "Preventive medical care: standards, usage, and efficacy." Santa Monica CA: Rand Corporation. Loewenstein, G., T. Brennan, and K. G. Volpp. 20 JAMA 298(20): 2415 7. Manning, W. G., J. P. Newhouse, N. Duan, E. B. Keeler, A. Leibowitz, and M. S. Marquis. 198 Am Econ Rev 77(3): 251 77. Marcus, A. C., L. A. Crane, C. P. Kaplan, A. E. Reading, E. Savage, J. Gunning, G. ce to screening follow up among women with abnormal Pap smears: results from a large clinic based trial Med Care 30(3): 216 30. Marcus, A. C., C. P. Kaplan, L. A. Crane, J. S. Berek, G. Bernstein, J. E. Gunning, and M. W. to follow up among women with abnormal Pap smears. Results from a randomized trial testing an intensive follow Med Care 36(3): 397 410. Matson Koffman, D. M., J. N. Brownstein, J. A. N einer, and M. L. Greaney. 20 site specific literature review of policy and environmental interventions that Am J Health Promot 19(3): 167 93. Melnikow, J., M. Paliesches Obstet Gynecol 89(6): 1023 7. Morone, J., Litman, T. & Robins, L. 20 08. Health Politics and Policy 4 th Edition Cl ifton Park, NY : Delmar Cengage Learning influenza immunization among high risk patients: education or financial Am J Med 101(6): 612 20. Nexoe, J., J. Kragstru on influenza vaccination rates among the elderly. A randomized controlled trial in Scand J Prim Health Care 15(2): 109 12. Palmer, J. D., and C. T. Pettit. 20 12. "Medicaid r isk based managed care: a nalysis of financial results for 20 11 http://www.milliman.com/insight/research/health/Medicaid risk based managed care Analysis of financial results for 2011/
112 Perlino, C. M. 20 Rosano A C. A. Loha R. Falvo J. van der Zee W. Ricciardi G. Guasticchi and A. G. de Belvis 20 13. "The relationship between avoidable hospitalization and accessibility to primary care : a systematic review Eur J Public Health 23(3):356 60. Rosenthal, M. B., Z. Li, A. D. Robertson, and A. Milstein. 20 incentives for prenatal care on birth outcomes and Health Serv Res 44(5 Pt 1): 1465 79. Rula, E. Y., J. E. Pope, and J. C. Hoffman. 20 Popul Health Manag 14 Suppl 1: S35 44. g the growth of medical spending." Washington, DC: National Coalition on Health Care, October 20 07. N Z Med J 110(1038): 58 60. Schroeder, S. A. 20 improving the health of N Engl J Med 357(12): 1221 2 8. Slater, J. S., G. A. Henly, C. N. Ha, M. E. Malone, J. A. Nyman, S. Diaz, and P. G. McGover n. 20 based strategy to increase mammography use among low income underinsured women ages 40 to Cancer Epidemiol Biomarkers Prev 14(10): 2346 52. Stone, E. G., S. C. Morton, M. E. Hulscher, M. A. Maglio ne, E. A. Roth, J. M. Grimshaw, B. S. Mittman, L. V. Rubenstein, L. Z. Rubenstein, and P. G. Shekelle. 20 02. services: a meta Ann Intern Med 136(9): 641 51. Sutherland, K., J. B. Christianson, and S. Leatherman. 20 Med Care Res Rev 65(6 Suppl): 36S 78S. U.S. Department of Health and Human Services. Healthy People 20 10 2nd Ed. 2000. "With Understanding and Improving Health and Objectives for Improving Health." Washington, DC: U.S. Govern ment Printing Office. Wang, Y. C., K. McPherson, T. Marsh, S. L. Gortmaker, and M. Brown. 20 Lancet 378(9793): 815 25.
113 Wing, R. R., R. W. Jeffery, N. Pronk, and W. L. He trainer and financial incentives on exercise adherence in overweight women in a Obes Res 4(5): 457 62.
114 BIOGRAPHICAL SKETCH Shuo Yang joined the Department of Health Services Resea rch, Management, and Policy at the University of Florida in January 20 09. Before enroll ing in the PhD program in h ealth s ervices research, he received a M aster of Science in Pharmac y (Dec ember 20 05 ) with a major in pharmaceutical sciences medicinal chemistry and a M aster of Science (A ug ust 2008) with a major in chemistry both also at the Univ ersity of Florida. For the past four years, he has been working on multiple research projects and the results of the studies have been presented at professional conferences and published in pe e r reviewed journals. He specialize s in design and conducting ri gorous evidence based health services research using large healthcare databases. He is also skilled at econometric modeling and data intensive analyses. His dissertation Evaluating the Effect of the Enhanced Benefits Rewards Program Participation on Flor ida Medicaid Utilizations and Expenditures EBR program within Florida Medicaid Reform. It provides meaningful insight for policymakers on both the federal and state level.