Citation
Collaborative governance

Material Information

Title:
Collaborative governance Colorado's foundation for coordinated health care for Medicaid patients
Added title page title:
Colorado's foundation for coordinated health care for Medicaid patients
Creator:
Bolinger, Teri Garland ( author )
Place of Publication:
Denver, Colo.
Publisher:
University of Colorado Denver
Language:
English
Physical Description:
1 electronic file (187 pages) : ;

Thesis/Dissertation Information

Degree:
Doctorate ( Doctor of philosophy)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
School of Public Affairs, CU Denver
Degree Disciplines:
Public affairs

Subjects

Subjects / Keywords:
Medicaid ( lcsh )
Medical policy ( lcsh )
Health insurance ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Review:
This thesis provides an overview of the Medicaid health insurance program to illustrate the significance of finding new approaches to the growing problem of increased enrollment and unsustainable expenditures. It highlights a need for collaboration among state and local resources to better serve Medicaid recipients and draws from collaborative governance literature to identify common characteristics of collaborations. This thesis compares these common characteristics to Colorado's Accountable Care Collaborative (ACC) and the implications of using the ACC to serve the state's Medicaid recipients. By examining actual health outcomes and health care costs for specific groups of Colorado's Medicaid recipients before and after the ACC's implementation, this thesis contributes evidence to the performance of a collaborative governance effort and the impacts on individuals served. This thesis concludes by discussing implications and recommendations for collaborative governance literature and theory, policy and practice, and future areas of research.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: Adobe Reader.
Statement of Responsibility:
by Teri Gaarland Bolinge.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
on10287 ( NOTIS )
1028749523 ( OCLC )
on1028749523
Classification:
LD1193.P86 2017d B66 ( lcc )

Downloads

This item has the following downloads:


Full Text
COLLABORATIVE GOVERNANCE: COLORADOS FOUNDATION
FOR COORDINATED HEALTH CARE FOR MEDICAID RECIPIENTS
by
TERI GARLAND BOLINGER B.A. (Hons), Midwestern State University, 1976 M.S. (Hons), Nova Southeastern University, 1995
A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Doctor of Philosophy Public Affairs Program
2017


2017
TERI GARLAND BOLINGER ALL RIGHTS RESERVED
11


This thesis for the Doctor of Philosophy degree by Teri Garland Bolinger has been approved for the Public Affairs Program by
Danielle Varda, Chair Tanya Heikkila, Advisor Brian Gerber Rene Horton
Date: December 16, 2017


Bolinger, Teri Garland (Ph.D., Public Affairs Program)
Collaborative Governance: Colorados Foundation for Coordinated Health Care for Medicaid Recipients
Thesis directed by Professor Tanya Heikkila
ABSTRACT
This thesis provides an overview of the Medicaid health insurance program to illustrate the significance of finding new approaches to the growing problem of increased enrollment and unsustainable expenditures. It highlights a need for collaboration among state and local resources to better serve Medicaid recipients and draws from collaborative governance literature to identify common characteristics of collaborations. This thesis compares these common characteristics to Colorados Accountable Care Collaborative (ACC) and the implications of using the ACC to serve the states Medicaid recipients. By examining actual health outcomes and health care costs for specific groups of Colorados Medicaid recipients before and after the ACCs implementation, this thesis contributes evidence to the performance of a collaborative governance effort and the impacts on individuals served. This thesis concludes by discussing implications and recommendations for collaborative governance literature and theory, policy and practice, and future areas of research.
The form and content of this abstract are approved. I recommend its publication.
Approved: Tanya Heikkila
IV


DEDICATION
In honor of my beloved daughter, Lauran Bolinger, and my dear mentor, Barbara Bowles, and in loving memory of my parents, Mr. and Mrs. T. K. Garland, my brother, W. T. Garland, my sister, Paula Anne Hatcher, and my inspiring friend, Ruthie Swanson
v


ACKNOWLEDGEMENTS
Something a stranger shared with me several months ago merits acknowledgement. She said she finally discovered the difference between failure and success. Failure: fall down seven times, get up six. Success: fall down seven times, get up eight. I gratefully acknowledge a faith that continues to sustain me at all times and in all circumstances, and I appreciate that I could not have completed this journey alone.
I acknowledge those at the University of Colorado Denver School of Public Affairs who made my experience a positive and productive one. Words inadequately express my appreciation for the unfailing encouragement, guidance, and wisdom my advisor, Tanya Heikkila, extended to me throughout this learning process. I would also like to thank the rest of my dissertation committee, Danielle Varda, Brian Gerber, and Rene Horton, who brought their valued and diverse perspectives and insight to this research. Appreciating the entire faculty and staff who made my experience at the School of Public Affairs a positive and productive one, I extend special thanks to Rob Drouillard, Todd Ely, Mary Guy, Christine Martell, Dawn Savage, and Chris Weible. I am particularly thankful for the support and friendship of my cohort members, Kelley Harp and Lucy Mermagen, whose intelligence, wit, and persistence never faltered.
I am especially grateful to Roger Hartley and Laura Wilson-Gentry, University of Baltimore College of Public Affairs, and Mason Paris, University of Baltimore Office of Technology, for generously providing an academically welcoming and technologically secure environment for me the past two years. I would not have been able to complete my research successfully without their gracious efforts on my behalf.
vi


I particularly recognize the importance of family, friends, and colleagues who daily influence my personal and professional development to a greater extent than they realize. Each of them unconsciously sets an example worth following by invariably saying or doing the perfect thing at the right time. I extend heartfelt love and gratitude to all of my family and specifically acknowledge Bayley Garland, Gloria Garland, Tim Garland, Tom Garland, Truitt Garland, Tina Sanders, Tracee Spore, Sheri Sutton, Julie Yandell, and Keegan Yandell. I am also thankful for priceless gifts of inspiration from Julie Farrar, Francesca Maes, and Jose Torres-Vega. I appreciate the treasures of unconditional friendship and boundless encouragement from Susan Barad, Bill Brewer, Lorri Cluckey, Todd Coffey, Kim Crow, Mary Dwyer, Lois Munson, Suzanne Smith, Drew Strouble, and Sandy Zannino. I remain honored to be part of the Federal Coordinated Health Care Office and the collaborative and ethical environment it cultivates and sustains. I sincerely respect and appreciate my colleagues and am especially grateful for the leadership of Lindsay Barnette, Kerry Branick, Sharon Donovan, Vanessa Duran, Tim Engelhardt, and Sara Vitolo.
Vll


TABLE OF CONTENTS
CHAPTER
I. MEDICAIDS SIGNIFICANCE....................................................... 1
Birth of Medicare and Medicaid.................................................2
Medicaid Design Characteristics................................................5
An Escalating Sense of Urgency................................................10
II. WHY COLLABORATION?.......................................................... 14
From Governance to Collaborative Governance...................................14
Six Defining Criteria.........................................................17
Forum Initiated by Public Agencies or Institutions........................18
Focus on Public Policy or Public Management Issue.........................18
Formally Organized and Collectively Convened Forum........................19
Nonstate Stakeholders as Active Participants..............................19
Forum Participants Involved in Decision Making............................20
Consensus-oriented Decision-making Process................................20
Four Common Elements..........................................................20
Starting Conditions.......................................................21
Institutional Design......................................................22
Facilitative Leadership...................................................24
Collaborative Process.....................................................26
Model Considerations..........................................................28
III. COLORADOS APPROACH.........................................................32
viii
The ACC and Six Defining Criteria
33


The ACC and Four Common Elements................................................34
Starting Conditions..........................................................35
Institutional Design.........................................................39
Facilitative Leadership......................................................44
Collaborative Process........................................................47
Implications for Successful Collaboration.......................................49
Implications for Starting Conditions.........................................50
Implications for Facilitative Leadership.....................................52
Implications for Collaborative Process.......................................52
From Theory to Practice.........................................................54
IV. RESEARCH DESIGN AND METHODS....................................................56
Research Question, Hypothesis, and Variables....................................57
Health Outcomes..............................................................58
Cost of Health Care..........................................................59
Data and Analysis...............................................................60
Data Source and Considerations...............................................60
Data Sample and Research Study Groups........................................61
Research Study Group Characteristics.........................................64
Approach to Analysis.........................................................67
Limitations and Strategies......................................................68
Three Criteria for Causal Inference..........................................69
Four Aspects of Validity.....................................................71
V. RESEARCH STUDY FINDINGS........................................................76
IX


Health Outcomes
76
Hospital Readmissions....................................................77
ER Visits................................................................79
High-cost Imaging........................................................81
Cost of Health Care..........................................................84
Medical Services.........................................................85
Prescription Drugs.......................................................87
PMPM.....................................................................89
Total Cost of Health Care................................................90
Results over Time............................................................92
Control Group Results....................................................92
ACC Group Results........................................................95
Regional Differences.........................................................98
Health Outcomes among RCCOs..............................................99
Cost of Health Care among RCCOs.........................................101
Observations................................................................106
VI. IMPLICATIONS AND RECOMMENDATIONS...........................................108
Literature and Theory.......................................................108
ACC Process Performance.................................................109
ACC Productivity Performance............................................110
Policy and Practice.........................................................112
Approach to Analysis....................................................112
Data Sources............................................................115
x


Prescription Drugs.......................................................116
Contingencies for the Future.............................................117
Future Research.............................................................119
Individual and Multiple RCCOs............................................119
Expanded Timeframe.......................................................121
Medicare-Medicaid Subpopulation..........................................122
Multivariate Analysis....................................................124
Conclusion..................................................................126
REFERENCES......................................................................128
APPENDIX
A. Table of Colorado ACC RCCO Service Regions...............................151
B. Map of Colorado ACC RCCO Service Regions.................................152
C. Organization Of ACC RCCOs................................................153
D. RCCO Relationships and Community Involvement.............................154
E. High-cost Imaging Services...............................................158
F. Behavioral Health Diagnosis Codes........................................159
G. Characteristics of Control and ACC Groups................................161
H. ER Visit Comparisons among RCCOs.........................................163
I. High-cost Imaging Comparisons among RCCOs................................166
J. Medical Service Comparisons among RCCOs..................................169
K. Prescription Drug Comparisons among RCCOs................................170
xi


LIST OF TABLES
TABLE
2.1 Ansell and Gash Elements and Components of Collaborative Governance...........21
3.1 Ansell and Gash Elements and Components and ACC Characteristics...............35
3.2 Ansell and Gash Implications for Success......................................50
4.1 Criteria for Research Study Groups............................................64
4.2 Summarized Characteristics of Control and ACC Groups..........................65
5.1 Potentially Preventable Hospital Readmissions (PPHR)..........................78
5.2 ACC Participation and PPHR....................................................79
5.3 ER Visits.....................................................................80
5.4 ACC Participation and ER Visits...............................................81
5.5 High-cost Imaging.............................................................82
5.6 ACC Participation and High-cost Imaging.......................................83
5.7 Medical Service Claims........................................................86
5.8 ACC Participation and Cost of Medical Service Claims..........................86
5.9 Prescription Drug Claims......................................................87
5.10 ACC Participation and Cost of Prescription Drug Claims........................88
5.11 ACC Group PMPM Calculations...................................................89
5.12 Total Cost of Health Care.....................................................90
5.13 Total Cost of Health Care Differences between Groups........................91
5.14 Control Group Health Outcomes Differences between Periods...................93
5.15 Control Group Cost of Health Care Differences between Periods...............94
5.16 ACC Group Health Outcomes Differences between Periods.......................96
xii


5.17 ACC Group Cost of Health Care Differences between Periods.................97
5.18 ER Visits among RCCOs.........................................................100
5.19 High-cost Imaging among RCCOs...............................................101
5.20 Control Group Medical Service Claims among RCCOs............................103
5.21 ACC Group Medical Service Claims among RCCOs................................103
5.22 Control Group Prescription Drug Claims among RCCOs..........................104
5.23 ACC Group Prescription Drug Claims among RCCOs..............................105
5.24 Summary of Differences........................................................107
xiii


LIST OF FIGURES
FIGURE
1.1 Colorado Medicaid Enrollment from 1995 to 2010 (in thousands).........11
1.2 Colorado Medicaid Annual Expenditures from 1995 to 2010 (in billions).12
3.1 Structure of the ACC...................................................42
xiv


LIST OF ABBREVIATIONS
ABBREVIATION
ACC Accountable Care Collaborative
APCD All Payer Claims Database
BHO Behavioral Health Organization
CIVHC Center for Improving Value in Health Care
CMS Centers for Medicare & Medicaid Services
CPT Current procedural terminology
CT Computed tomography
DSH Disproportionate share hospital (payments)
EPSDT Early and periodic screening, diagnosis, and treatment
ER Emergency room
FMAP Federal medical assistance percentage
FPL Federal poverty level
HMO Health maintenance organization
IGT Intergovernmental transfer
MMIS Medicaid Management Information System
MRI Magnetic resonance imaging
PCMP Primary Care Medical provider
PMPM Per member per month
PPHR Potentially preventable hospital readmissions
RAE Regional Accountable Entity
RCCO Regional Care Collaborative Organization
xv


SDAC
UPL
Statewide Data and Analytics Contractor Upper payment limit (supplements)
xvi


CHAPTER I
MEDICAIDS SIGNIFICANCE
Viewed through the lens of collaborative governance, public management and health policy offer a broad and fruitful field of study. Narrowing the research focus, this thesis examines the development and implementation of Colorados Accountable Care Collaborative (ACC), a model the state designed to serve recipients of its Medicaid health insurance program. The first chapter of this thesis lays the foundation for the object of research by describing Medicaids significance. While the second chapter grounds the object of research in collaborative governance literature, the third chapter determines if Colorados ACC theoretically fits the collaborative governance criteria Ansell and Gash outlined in their 2008 meta-analysis of 137 diverse cases. Satisfied the ACC is a collaborative governance example designed with opportunity to succeed, the fourth chapter of this thesis introduces the research question: What effects will participation in the ACC have on health care for Medicaid recipients in Colorado? The fourth chapter continues by describing the research design and methods used to examine health outcomes and health care costs for specific groups of Colorados Medicaid recipients before and after the ACCs implementation. This thesis presents the research findings in the fifth chapter and concludes with a discussion of implications and recommendations for collaborative governance literature and theory, policy and practice, and future areas of research in the sixth chapter.
This chapter continues by summarizing the significance of the joint federal and state Medicaid health insurance program in three subsections. It describes the birth of Medicare and Medicaid, two of the largest health insurance programs in the United States; outlines
1


Medicaids primary design characteristics; and illustrates the basis for an escalating sense of national and state urgency associated with Medicaid.
Birth of Medicare and Medicaid
In 1965, the United States Congress overwhelmingly approved Social Security Act amendments, Titles XVIII and XIX, to establish Medicare and Medicaid, respectively, as part of President Johnsons set of domestic programs known as the Great Society (Iglehart & Sommers, 2015; Starr, 2015; Stevens, 1996). Medicare and Medicaid sought to provide health insurance to two different subpopulations in the United States: the elderly and the poor (Centers for Medicare & Medicaid Services [CMS], 2015b). Designed to uphold the status quo of private insurance for working individuals, Medicare focused coverage on the retired elderly who could no longer find or afford health insurance (Stevens, 1996). The right to health insurance became a natural extension of the elderlys right to Social Security benefits earned by contributions made during the years they worked (Starr, 2015). Aimed to finance health care services for the poorest individuals in the United States, Medicaid focused coverage on recipients of financial assistance through welfare programs: those who were aged, blind, or disabled or were families with dependent children (CMS, 2015b; Hansen, 2013; Starr, 2015). Medicaids primary purpose was to furnish acute health care services to public assistance recipients based on their income and resources or means (United States Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation [ASPE], 2005). Federal and state programs with eligibility based on personal or family means exist in the United States to provide food, housing, medical care, and social services to those who do not have the ability to get and pay for services on their own (Means-tested, n.d.; Rector & Sheffield, 2016). The difference in Medicare and Medicaids initial
2


focus contributed to a perceived social dichotomy between the deserving elderly and the less deserving indigent (Hudman & Starfield, 1999; Stevens, 1996).
Both programs design and financing also differed. As a national program, Medicare provided the same hospital (Part A) and physician (Part B) benefits to elderly recipients wherever they lived (Starr, 2015). The financing for Medicares Part A came from payroll taxes and Part B came from general federal revenue and premiums paid by recipients (Medicare, n.d.; Starr, 2015). Unlike Medicare, Medicaid was a program chiefly administered by states and included flexibility in program design (ASPE, 2005; CMS, 2017d; Mann & Westmoreland, 2004; Social Security Act, Title XIX, 1965). Medicaid financing came from general state revenues matched in part by general federal revenue (ASPE, 2005; Patrick & Freed, 2012; Starr, 2015). In the first year of operation, Medicare enrollment was 19 million with expenditures of $3.4 billion, and Medicaid enrollment was 4 million with expenditures less than $1 billion (Klemm, 2000; Pear, 1987). However, Medicaids early costs were seriously underestimated. For the first year, projected costs nationwide were $250 million, but actual costs in the state of New York alone reached that amount (Mills, 1987; Starr, 2015).
Although different in focus, design, and financing, both Medicare and Medicaid originally favored hospital-based services over public health and preventive care; promoted medical and surgical specialties instead of primary care; and operated as fee-for-service programs by making a separate payment to a health care provider for every service delivered (Fee-for-service, n.d.; Starr, 2015). As a result, both encouraged a volume-based rather than value-based delivery system (Starr, 2015). Medicare and Medicaid enrollment and costs
3


continued to increase, and both programs expanded after their inception by including other subpopulations and additional services (CMS, 2015b).
Several expansions directly affected Medicaid. An early and periodic screening, diagnosis, and treatment (EPSDT) benefit was created in 1967 for all recipients who were children, and eligibility was linked to the federal Supplemental Security Income program enacted in 1972 for state residents who were elderly or blind or who had a disability (CMS, 2015b). Some federal regulations were waived in 1981 that allowed states to require recipients to get services from a limited set of providers and permitted recipients to get services in home and community-based settings as an alternative to institutional settings (CMS, 2015b; Nathan, 2005; Whitenhill & Shugarman, 2011). In contrast to the original fee-for-service approach, these managed care concepts introduced a health care delivery system aimed at controlling services, physician fees, and a recipients choice of providers with the intention of reducing costs and improving care quality (CMS, 2015b; Managed care, n.d.; Starr, 2015).
An option was created in 1986, which became a mandate in 1988, for states to provide coverage to pregnant women and infants up to 100% of the federal poverty level (FPL) (CMS, 2015b). In 1989, the coverage threshold for pregnant women and children under six years of age was expanded to 133% of the FPL, and EPSDT requirements were established (CMS, 2015b). Phased-in coverage of children between six and 18 years of age and a prescription drug rebate program were established in 1990, and a new low-income group not linked to public assistance was mandated in 1996 (CMS, 2015b). The Balanced Budget Act of 1997 created the Childrens Health Insurance Program and established new managed care options and requirements for states (CMS, 2015b; Starr, 2015). The Ticket to
4


Work and Work Incentives Improvement Act of 1999 expanded the availability of coverage for certain recipients with disabilities who returned to work, established optional eligibility groups, and allowed states to offer a buy-in program for working-age individuals with disabilities.
Medicaid Design Characteristics
In addition to distinct subpopulation and service expansions described in the previous subsection, Medicaids original design differed from Medicares (ASPE, 2005; Iglehart & Sommers, 2015; Starr, 2015; Stevens, 1996). Medicaid operates in a dynamic environment where federal and state government authorities share power (Iglehart & Sommers, 2015). As the social, economic, and political landscapes change, Medicaids inherent design flexibility contributes not only to challenges it may encounter but also to responses it can develop (Iglehart & Sommers, 2015; Nathan, 2005; Weil, Wiener, & Holahan, 1998).
Medicaid eligibilitys initial link to receipt of financial assistance through public assistance programs reinforced a disincentive to work. Low-paying jobs available to public assistance recipients typically lacked benefits such as health insurance coverage, and individuals who accepted such jobs would then lose Medicaid coverage (Iglehart &
Sommers, 2015; Starr, 2015). As a result, those with chronic health conditions or those with a sick child, for example, had a strong incentive not to work to maintain their health insurance coverage through Medicaid (Moffitt & Wolfe, 1992). Although state and federal authorities expanded Medicaid eligibility to other groups of low-income individuals in the 1980s and 1990s, the incentive not to work effectively remained in place until the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 uncoupled Medicaid
5


eligibility from receipt of public assistance (Hudman & Starfield, 1999; Iglehart & Sommers, 2015; Social Security Administration, 1996).
State flexibility concerning Medicaid eligibility, services, and provider rates presented other challenges. For example, a person eligible for Medicaid in one state might not be eligible in another, and services one state provided could differ in scope, quantity, and duration from services another state offered (Klees, Wolfe, & Curtis, 2010). In addition, a state could change its Medicaid eligibility criteria, services, or provider reimbursement rates at any time (Klees et al., 2010). Although a state agreeing to operate a Medicaid program was required to furnish a basic set of health services to all recipients of public assistance, a state could receive additional federal funds if it also covered those in eligible categories with incomes up to 133% of the states public assistance cutoff point (ASPE, 2005; Starr, 2015).
Setting a states threshold for Medicaid eligibility was linked to the configuration of its specific services. Each state varied not only in its public assistance criteria but also in its willingness to cover other individuals among the poor (Starr, 2015). A state also could opt to offer additional services to medically needy individuals who received no public assistance (ASPE, 2005). The type and number of individuals a state included in its Medicaid eligibility parameters influenced a states service categories; the scope, quantity, and duration of services offered; and provider payment rates (Klees et al., 2010; Rosenbaum, 2011; United States Department of Health and Human Services, Health Care Financing Administration, 2000). Because states historically have set Medicaid provider rates lower than Medicare rates, diminished provider participation emerged as another challenge (Ku, 2000;
Rosenbaum, 2011; Starr, 2015).
6


Financing Medicaid proved to be as challenging for states as balancing eligibility, services, and provider rates. The federal government paid states for at least half of the costs of providing Medicaid services to recipients; for every dollar a state spent on Medicaid services, the federal government reimbursed the state no fewer than 50 cents (ASPE, 2005; Iglehart & Sommers, 2015). Each states federal medical assistance percentage (FMAP) is calculated by comparing the states average income per person to the national average; the lower a states average income per person, the higher its FMAP up to a maximum of 83% (ASPE, 2005; Iglehart & Sommers, 2015). Federal funds provided a means by which to increase the value of Medicaid investments, and states regularly adopted strategies to maximize federal support (Klemm, 2000). Based on the premise that demand for social programs increases in periods of economic decline, social programs were designed to compensate for the negative effects of such cycles (Carboni & Milward, 2012). However, as a joint federal and state endeavor, Medicaid was vulnerable to cutbacks during economic downturns when state revenues also decreased (Starr, 2015). During such periods Medicaid enrollment typically increased because unemployed workers lost employer-based health insurance coverage and family income decreased to the Medicaid threshold (Iglehart & Sommers, 2015). As revenues fell and health care costs rose, states increasingly sought to curb Medicaid services, control costs, and leverage financing opportunities (Medicaid and CHIP Payment and Access Commission [MACPAC], 2012b; Patrick & Freed, 2012; Sommers & Epstein, 2011).
Since the early 1980s, states often have used two available financing strategies to maximize federal support for Medicaid: disproportionate hospital share (DSH) payments and upper payment limit (UPL) supplements (Coughlin, Bruen, & King, 2004; MACPAC, 2012a,
7


2012b). The DSH payment policy was crafted to allow supplemental payments to hospitals that served a disproportionate share of low-income patients and was limited to the actual cost of uncompensated services hospitals provided to Medicaid recipients and uninsured individuals (MACPAC, 2012b). Although the purpose was to improve the financial stability of safety net hospitals providing services to low-income patients, broad federal guidelines provided states considerable leeway in determining which hospitals would receive DSH payments and in what amounts (MACPAC, 2012b). The UPL policy allowed states to receive supplemental federal funds for Medicaid fee-for-service payments up to what would have been paid under Medicare; the policy applied to targeted groups of providers such as hospitals, nursing facilities, intermediate care facilities for those with intellectual disabilities, and freestanding non-hospital clinics (Coughlin et al., 2004; Ku, 2000; MACPAC, 2012a). Because Medicaid provider payments traditionally were lower than Medicare payments, considerable potential existed for states to get additional federal dollars with the intention of directing UPL supplemental payments to providers (Coughlin et al., 2004; MACPAC,
2012a).
States soon began to shift the balance in Medicaid spending by using ambiguity in the federal DSH and UPL regulations to transfer funds from local to state government or between state agencies instead of to providers directly (Coughlin et al., 2004; Ku, 2000).
Such intergovernmental transfers (IGTs) utilized legal but convoluted accounting practices that allowed states to increase Medicaid federal matching payments without expending additional state funds. For example, a state would make allowable payments above the states regular Medicaid reimbursement rate to a select group of nursing homes or hospitals, which were usually owned by a county or other local government entity; the facilities would return
8


a large portion of the supplemental payments to the states Medicaid agency through IGTs; and the state would claim a Medicaid federal match for the supplemental payments it made to the facilities (Coughlin et al., 2004; Ku, 2000; MACPAC, 2012b). Federal financing of UPL payments alone rose from $313 million in 1995 to $1.4 billion in 1998 (Ku, 2000). By the time regulations were refined and additional requirements were imposed in the late 1990s and early 2000s, explosive growth in federal Medicaid expenditures for state DSH payments and UPL supplements already had occurred (Ku, 2000; MACPAC, 2012b). In 34 states surveyed, researchers estimated the effective federal match rate in 2001 was an average of three percentage points higher than it would have been without state DSH and UPL practices (Coughlin et al., 2004).
Some safety net hospitals and providers serving Medicaid recipients directly receive and retain DSH payments and UPL supplements as intended, but some do not. Although some states may still legally redirect these federal resources for other purposes such as balancing budgets or cutting taxes, such uses violate the original intent of DSH payments and UPL supplements and distort Medicaid spending (Coughlin et al., 2004; Ku, 2000). These practices prompt the effective federal share of Medicaid spending to rise without a commensurate increase in the state share of funds (Coughlin et al., 2004). Additionally, when states use gains from DSH payments and UPL supplements to finance Medicaid services other than those for which the payments and supplements are intended, spending patterns among Medicaid service categories become clouded (Coughlin et al., 2004; Ku, 2000). These practices can create confusion in policy debates over such topics as Medicaid expansion, which subpopulations of Medicaid recipients to serve, which services to offer, and the amount of provider rates. Coupled with its historical evolution, Medicaids eligibility
9


parameters, service composition, provider rates, and financing opportunities contribute to its increasing significance and distinct role in the nations health care landscape.
An Escalating Sense of Urgency
While Medicaid represented less than five percent of means-tested program spending in 1966, it represented 30% in 1972 and 40% by 1985 (Means-tested, n.d.; Starr, 2015). From 1991 to 2005, the total number of Medicaid recipients increased from 32 to 58 million, and total annual expenditures rose from $110 to $273 billion (Patrick & Freed, 2012). By 2008, Medicaid enrollment included 60 million individuals with expenditures of $330 billion (Jacobson, Neuman, & Damico, 2012). Enrollment and costs continued to climb. Reporting for fiscal year 2010 reflected Medicaid served more than 66 million individuals, which was more than one fifth of the nations population, with total expenditures exceeding $369 billion (Kaiser Family Foundation, 2014; The State Health Care Spending Project, 2014). At Medicaids inception in 1965, health programs received six percent of all federal funds to state and local governments, but by 2010, health-related activities accounted for 58% of all federal funding to state and local governments (Iglehart & Sommers, 2015). Medicaid expenditures represented over 90% of that funding and severely diminished availability of federal support in other areas such as education, transportation and infrastructure, scientific research, and social services (Center on Budget and Policy Priorities, 2016; Iglehart & Sommers, 2015). As Medicaid continued to grow and become one of the largest sources of health coverage in the United States in numbers enrolled and dollars spent, its trajectory appeared unsustainable (CMS, 2008; Iglehart, 2009; Kaiser Family Foundation, 2014; Mann & Westmoreland, 2004).
10


Like the changing national landscape, Colorado was confronted by challenges in
making the growth of its Medicaid program sustainable. Colorado established the states Medicaid program in 1969 (Colorado Center on Law & Policy [CCLP], 2012; Colorado Health Institute [CHI], 2005). Attempting to control costs and streamline services, Colorado Medicaid began expanding managed care in 1981 and implemented an assertive managed care initiative in the 1990s for primary and acute care as well as home and community-based services and long-term care (CCLP, 2012; CHI, 2005). In state fiscal year 1994, Colorado Medicaid served more than 280,000 individuals, which represented slightly more than eight percent of the states population (Colorado Department of Health Care Policy and Financing [HCPF], 1995). The cost of that coverage was more than $1.25 billion (HCPF, 1995). Figure 1.1 illustrates the trend of Colorados Medicaid enrollment from 1995 to 2010 (CMS, 2017f).
Figure 1.1: Colorado Medicaid Enrollment from 1995 to 2010 (in thousands) (CMS, 20171)
As the national growth rate for Medicaid expenditures hovered at almost four percent, Colorados total Medicaid expenditures grew at an average annual rate approaching eight
11


percent from 1995 through 1998, which put additional pressure on the state to control costs (Lutzky, Holahan, & Wiener, 2002).
Second only to spending for education and representing more than 18% of the states budget, Colorado Medicaid served more than 400,000 individuals at a cost of approximately $2 billion in 2005 (CHI, 2005; HCPF, 2013b; The State Health Care Spending Project, 2014). Figure 1.2 shows Colorados Medicaid expenditures from 1995 to 2010 (CMS,
2017f).
Figure 1.2: Colorado Medicaid Annual Expenditures from 1995 to 2010 (in billions) (CMS, 20171)
By 2010, Colorado Medicaid served over 600,000 individuals, which represented 12% of the states population; the cost approached $4 billion (Colorado Department of Local Affairs, 2017; The State Health Care Spending Project, 2014). As Colorados Medicaid enrollment grew 50% between 2005 and 2010, it outpaced the states general population increase of eight percent in the same period (Colorado Department of Local Affairs, 2017). During the national economic crisis from 2007 to 2011, Colorado Medicaid enrollment increased 56%, which resulted in the imposition of cost containment strategies such as
12


reducing provider rates and reimbursements, streamlining enrollment processes, limiting services, establishing prescription drug controls, and increasing recipient responsibility for certain copays (CCLP, 2012; CHI, 2005).
Regardless of political preference, policy makers acknowledged the crisis. They agreed that per person health spending would continue to rise, the number of Medicaid recipients would increase faster than the labor force would grow, additional public benefits would require additional public expenditures, and added taxes or subsidies could offset part of the costs (Aaron, 2007). Even before additional support for Medicaid expansion was included in the Patient Protection and Affordable Care Act (Affordable Care Act) of 2010, federal and state authorities were seeking different approaches and new solutions. Colorado implemented several strategies mentioned in the previous paragraph. Federal and state authorities contemplated additional strategies: engaging states in long-range planning, determining a states FMAP based on program performance, adopting provider pay-for-reporting and pay-for-performance policies, focusing on preventive and primary care services, increasing the use of mandatory managed care, and introducing special initiatives to target high-cost areas such as emergency room (ER) utilization (Aaron, 2007; CMS, 2008; Gifford, Smith, Snipes, & Paradise, 2011; Keckley & Kalkhof, 2007). As seen throughout this chapter, Medicaids multifaceted nature and its increasing significance in the health care landscape warrant innovative and varied responses. The next chapter of this thesis uses collaborative governance literature to lay the foundation for the ensuing discussion of Colorados approach to improving Medicaid.
13


CHAPTER II
WHY COLLABORATION?
As federal and state government partners sought approaches to improve Medicaid, the need for greater collaboration among all invested parties became more apparent. A governance arrangement that fit the problem was required (Ostrom, 2007). It was no longer sufficient for government authorities to operate in isolation to exercise lawful control over or to oversee and make decisions about Medicaid (Governance, n.d.). As the need for greater collaboration was recognized, the use of collaboration expanded.
This chapter begins by defining governance and collaborative governance. It then draws from collaborative governance literature to identify several common characteristics of collaborations and examines the Ansell and Gash (2008) approach to collaborative governance. The chapter concludes with observations regarding the appropriateness of using the Ansell and Gash (2008) approach to assess Colorados ACC as an example of collaborative governance.
From Governance to Collaborative Governance As public management issues continued to expand in scope and complexity in the first decade of this century, the idea of governance evolved to include a broader concept of collaborative governance. By the early 2000s the narrow definition of governance that simply described government activity as exercising lawful control or overseeing and making decisions was inadequate (Governance, n.d.). This section examines several of the leading public management definitions and characteristics of governance that contributed to the more expansive concept of collaborative governance.
14


Lynn, Heinrich, and Hill (2000) defined governance as the regime of laws, administrative rules, judicial rulings and practices that constrain, prescribe, and enable government activity, where such activity is broadly defined as the production and delivery of publicly supported goods and services (p. 235). Incorporating a dimension of flexibility into the relationships in the definition, the authors emphasized a configuration of distinct but interrelated elements comprised of formal and informal structures that involve bargaining and compromise (Lynn, Heinrich, & Hill, 2000). Milward and Provan (2000) highlighted a need to connect networks of actors in public policy domains where power must be shared because missions could not be accomplished by a single institution. Their interpretation of governance included the creation of conditions in which ordered rule and collective action were comprised of agents in the private, public, and nonprofit sectors and focused on mechanisms of government not relying exclusively on government authority and sanctions (Milward & Provan, p. 360, 2000). Complementing earlier definitions that included flexible relationships and networks of shared power, Kettl (2000, 2002) called for a concept of governance that included even more diffused power and more people exercising power in the face of bigger problems. Kettl (2000) described the need for adapting traditional governance models and changing strategies and tactics to enhance governments capacity to govern in a more decentralized and inclusive environment. In their focus on network and collaborative management, Agranoff and McGuire (2001) contributed to broadening the definition of governance by exploring issues related to trust, common purpose, and mutual dependency. Brewer, Neubauer, and Geiselhart (2006) echoed this evolving perspective in their typology of government environments by suggesting conditions of modern times merited new governance architectures. They viewed expanding governance relationships that included
15


public and private sector participants as essential in strengthening capacity and agility in an atmosphere distinguished by rapid and unpredictable changes (Brewer, Neubauer, & Geiselhart, 2006).
As evidenced by the definitions and implications in the previous paragraph, the concept of governance in the early 2000s began to expand beyond the limits of government activity to encompass the collaborative work of public and private sector actors addressing the growing scope and complexity of public management issues. To the extent that some degree of mutual public and private sector participation occurs in service delivery and operates according to established rules, a type of collaborative governance exists (Agbodzakey, 2012). Collaborative governance encourages joint efforts among the public, private, and nonprofit sectors to address complex problems through collective decision making and implementation (Agbodzakey, 2012; Bingham & OLeary, 2008; Farazmand, 2004; Gray, 1989; Huxham & Vangen, 2000). A distinctive feature of collaborative governance is the constructive engagement of people across public agencies, levels of government, and the public, private, and civic spheres to achieve a public purpose that could not otherwise be accomplished (Emerson, Nabatchi, & Balogh, 2012; Williams & Matheny, 1995). Collaborative governance processes share power with stakeholders in decision making to develop recommendations for effective, lasting solutions to public problems (Purdy, 2012, p. 409). Broad stakeholder engagement, inclusive relationships, and collective decision making and action distinguish collaborative governance as a viable vehicle for addressing current public management problems.
Ansell and Gash (2008) distinguished collaborative governance from alternative policy-making approaches such as adversarialism and managerialism (Futrell, 2003;
16


Williams & Matheny, 1995). For example, collaborative governance stakeholders may have adversarial relationships initially, but the collaborative process is intended to transform those relationships into cooperative ones so that mutual objectives can be achieved (Ansell &
Gash, 2008). Unlike managerialism, collaborative governance directly involves stakeholders in decision making rather than merely consulting stakeholders or considering their perspectives (Ansell & Gash, 2008). The authors also differentiated collaborative governance from other types of cooperation or collaboration such as corporatism, associational governance, policy networks, public-private partnerships, participatory management, interactive policy making, stakeholder governance, and collaborative management (Ansell & Gash, 2008). Ansell and Gash (2008) clarified the definition of collaborative governance to describe it specifically as an arrangement in which one or more public agencies directly engage non-state stakeholders in a collective decision-making process that is formal, consensus-oriented, and deliberative and that aims to make or implement public policy or manage public programs or assets (p. 544).
Ansell and Gash (2008, 2012) used their definition for collaborative governance to isolate six criteria, which aided in their meta-analysis of literature and case studies. Their systematic review produced 137 diverse cases using some form of collaborative governance, and their analysis revealed four common elements and related components occurring in each case (Ansell & Gash, 2008, pp. 548-550). Further examination and consideration of the Ansell and Gash approach follows in the next two sections.
Six Defining Criteria
The Ansell and Gash (2008) definition of collaborative governance emphasizes six criteria: a forum initiated by public agencies or institutions; a public policy or public
17


management issue expressed as its focus; a forum organized formally and convened collectively; nonstate stakeholders involved as active participants; decision making included as a role for forum participants; and consensus used as the basis for decision making (pp. 544-545). As seen in the definitions and distinctive characteristics of collaborative governance that appear earlier in this chapter, the six criteria provide a reasonable basis for defining collaborative governance. The six defining criteria also permit the first step in describing, assessing, and comparing examples, which can aid in building theory (Ansell & Gash, 2008). The criteria inform common elements and components essential in collaborative governance, and an example fitting the Ansell and Gash (2008) model could be considered an authentic collaborative governance effort. A summary of each of the six criteria follows.
Forum Initiated by Public Agencies or Institutions
First, public agencies or institutions initiate the forum for collaboration. The initiators role in collaboration is to identify and bring together all legitimate stakeholders (Gray, 1989, p. 71). Because of its right to establish and enforce rules, a public authority convokes the collaborative governance process (Purdy, 2012). Unlike other forms of collaboration, collaborative governance includes a specific convening and leadership role for public agencies to launch a forum to advance their own purposes, to respond to the expressed needs or requests of their constituents, or to fulfill a legislative or judicial mandate (Ansell & Gash, 2008, p. 545).
Focus on Public Policy or Public Management Issue
Second, differing from mediation or dispute resolution, collaborative governance focuses on public policy or public management issues (Ansell & Gash, 2008). The agency or
18


institution that creates the forum does so with an instrumental purpose in mind, usually in response to a challenging area of public policy or management within its purview that warrants a collaborative approach to solutions (Bryson, Crosby, & Stone, 2006; Huxham & Vangen, 2000). According to the Ansell and Gash (2008) definition and criteria, collaborative governance involves public agencies creating a clear strategy to develop or implement public policy or manage public programs or assets.
Formally Organized and Collectively Convened Forum
Third, the forum is organized formally and meets collectively. Compared to less formal interactions cultivated by public agencies and stakeholders, a collaborative governance structure is intentional and explicit (Ansell & Gash, 2008). Along a continuum that moves from cooperation to coordination to collaboration, formality generally moves from low to medium to high (Reilly, 2001). Collaborative governance includes a structure of regular meetings of participants along with communication through other media (Bryson et al., 2006; Huxham & Vangen, 2000).
Nonstate Stakeholders as Active Participants
Fourth, nonstate stakeholders are active participants engaging with public stakeholders and with each other in a deliberative and multilateral process (Ansell & Gash, 2008). Participants work together to build support and stimulate productive and purposeful interaction among other participants (McGuire, 2006). They engage in activities that generate a capacity for joint action that did not previously exist; depending upon the purpose and context of the collaboration, participants may take part in activities such as educating constituents or the public, enacting policy measures, assembling external resources, enacting new practices, overseeing implementation, and ensuring compliance (Emerson et al., 2012).
19


Forum Participants Involved in Decision Making
Fifth, forum participants not only consult and advise but also actively contribute to the decision-making process. A chief characteristic of collaborative governance is participation by interested and affected parties in all phases of the decision-making process (Freeman, 1997). Although the public agency may have ultimate authority, nonstate stakeholders share responsibility for policy outcomes (Ansell & Gash, 2008). Stakeholder participation in decision making promotes collective resolve not only to address the identified problem but also to increase responsibility for outcomes (Agbodzakey, 2012, p. 108).
Consensus-oriented Decision-making Process
Sixth, the decision-making process is oriented toward consensus. Although participants do not always achieve consensus, they strive to find areas where they can reach agreement (Ansell & Gash, 2008; Fung & Wright, 2001). Participants involved in a collaborative governance forum discuss issues until all opinions are expressed and understood, and group agreement on a course of action is required before proceeding to the next topic (Reilly, 2001).
Four Common Elements
Four primary and recurring elements and related components surfaced in the Ansell and Gash (2008) review of case studies fitting their collaborative governance definition and six criteria. The authors stressed common and frequent findings and simplified the representation of elements and components with the intent of making their approach more useful for policy makers and practitioners (Ansell & Gash, 2008). The four elements are starting conditions, institutional design, facilitative leadership, and collaborative process
20


(Ansell & Gash, 2008). Each of the four elements and related components contributing to a collaborative governance effort are discussed in greater detail in the next four subsections of this chapter, and Table 2.1 serves as a guide to the narrative that follows.
Table 2.1: Ansell and Gash Elements and Components of Collaborative Governance
Elements Components
Starting Conditions Prehistory of conflict or cooperation Power and resource imbalances Incentives to participate
Institutional Design Inclusive participation Exclusive forum Clear ground rules Process transparency
Facilitative Leadership Approach Skills Techniques
Collaborative Process Face-to-face dialogue Trust building Commitment to process Shared understanding Intermediate outcomes
Starting Conditions
One element in the Ansell and Gash (2008) collaborative governance examples is starting conditions, which include the prehistory of conflict or cooperation; power and resource imbalances; and incentives to participate. Where gridlock burdens all stakeholders with the costs of a growing problem, it often acts as a catalyst for collaboration. However, sustained conflict predating the collaboration contributes to low levels of trust that can manifest as marginal commitment (Ansell & Gash, 2008; Huxham & Vangen, 2000). In such cases, intentionally creating an atmosphere of mutual reliance among stakeholders and fostering the incremental development of trust aim to increase cooperative participation and encourage greater collaboration among participants (Ansell & Gash, 2008; Purdy, 2012).
21


Initial imbalances in stakeholder power and resources are also integral to starting conditions in collaborative governance case studies. Power issues stem from the ability to exercise influence, authorize action, and control resources (Bryson et al., 2006; Gray, 1989). To sustain commitment to and involvement in the collaboration, stakeholders with less power and fewer resources need assurance that their perspectives are valued and evidence that their interests are included (Agbodzakey, 2012; Ansell & Gash, 2008; Bryson et al., 2006).
Imbalances of power and resources influence stakeholder incentives to participate, the third component of starting conditions in the collaborative governance examples examined (Ansell & Gash, 2008). Financial incentive is not the only and not always the main reason stakeholders participate. An altruistic motive to support the collective good, a rational concern to protect or advance self-interest, a desire to avoid missing an opportunity if absent, or some combination of the three are primary incentives prompting stakeholder participation (Ansell & Gash, 2008; Reilly, 2001; Wood & Gray, 1991). Stakeholders who believe their individual power is ascending or who see alternative vehicles for solutions may have less incentive to collaborate (Ansell & Gash, 2008).
Institutional Design
Another recurring element in the collaborative governance case studies is institutional design, whose components consist of inclusive participation, an exclusive forum, clear ground rules, and process transparency (Ansell & Gash, 2008). Inclusive participation is an intentional design component. Seeking broad participation encourages representation from multiple perspectives and different interests, which stimulates and supports far-reaching consideration of issues and an extensive view of potential benefits and harms (Ansell &
Gash, 2008; Emerson et al., 2012; Reilly, 2001). Deliberate inclusion of stakeholders across a
22


wide array of interests contributes to a public perception of increased transparency and fairness, more informed discussions, and greater legitimacy of decisions (Ansell & Gash, 2008; Freeman, 1997).
An exclusive forum is another design component. In the collaborative governance examples studied, stakeholders are much more likely to participate when the forum is exclusive and alternatives are either absent or unattractive (Ansell & Gash, 2008; Fung & Wright, 2001). Full commitment to the collaborative process is less likely when viable alternatives exist (Reilly, 2001). When the convening authority sends clear signals indicating alternative means of resolution are not under consideration, the legitimacy of the collaborative effort increases and stakeholder uncertainty decreases (Ansell & Gash, 2008; Weber & Khademian, 1997).
A third design component is ground rules. Clear and consistently applied ground rules reassure stakeholders of the collaborations legitimacy (Ansell & Gash, 2008). A set of rules serves to ensure that the collaborative effort cannot be preempted or appealed (Reilly, 2001; Weber & Khademian, 1997). Ground rules also serve as a set of principles to define the collaborations roles and proceedings (Agbodzakey, 2012; Ansell & Gash, 2008; Fung & Wright, 2001; Wood & Gray, 1991).
Process transparency is the fourth component noted in the institutional design of the collaborative governance cases Ansell and Gash (2008) studied. Transparency in collaborative governance encourages an open and documented process (Freeman, 1997; Reilly, 2001) As an instrument of collaborative governance, transparency becomes ingrained in the process and is particularly useful in increasing stakeholder trust and promoting accountability (Ball, 2009).
23


Facilitative Leadership
A third common element in the collaborative governance examples Ansell and Gash (2008) reviewed is facilitative leadership, which distinguishes itself by the approach, skills, and techniques used by those guiding the collaborative effort. Rather than leadership that makes things happen, facilitative leadership helps others to make things happen and supports participants in effectively working with each other (Ansell & Gash, 2012; Chrislip & Larson, 1994; Huxham & Vangen, 2000; Vangen & Huxham, 2003a). Faced with a variety of contexts, goals, and tasks creating different demands, collaborative leaders demonstrate a multitude of skills and techniques to meet the circumstances (Ansell & Gash, 2012). Facilitative leadership is essential to engaging stakeholders, sustaining their interest and action throughout the process, and ensuring they realize the benefits of participating in the collaboration (Ansell & Gash, 2008; Vangen & Huxham, 2003a).
The facilitative leadership approach in the collaborative governance examples studied promotes and safeguards the process to motivate stakeholders to move forward without advocating one point of view or taking unilateral, decisive action (Ansell & Gash, 2008; Bryson et al., 2006; Chrislip & Larson, 1994; Emerson et al., 2012; Huxham & Vangen, 2000). Depending on the situation and circumstances, collaborative governance leaders can find themselves playing multiple roles. Leadership roles include stewards, who convene the collaboration and maintain the integrity of the process; mediators, who facilitate communication and nurture relationships among stakeholders; and catalysts, who stimulate identification and realization of value-creating opportunities (Ansell & Gash, 2012).
The facilitative approach requires a different set of skills than the tactical approach where leaders articulate a clear objective, develop a plan, and lead others in the execution of
24


the plan. Skills needed in facilitative leadership also differ from those in the positional approach where leaders at the top of an organizational structure set goals, organize activities, motivate the completion of tasks, and reward performance (Ansell & Gash, 2008; Chrislip & Larson, 1994). Leadership that employs skills such as engagement, facilitation, negotiation, mediation, and progress evaluation is critical to the success of collaborative governance efforts (Lasker, Weiss, & Miller, 2001; OLeary, Bingham, & Choi, 2010; OLeary, Gerard, & Bingham, 2006). Effective leadership requires skills and abilities diverse and flexible enough to be adapted to fit the circumstances and situations presented in a dynamic collaborative environment (McGuire, 2006; Weber & Khademian, 1997).
Leadership techniques found in collaborative governance examples support a facilitative approach and optimize the critical skills mentioned previously. Enlisting participation from all stakeholders, helping stakeholders develop a common language, inviting all perspectives to be heard, and showing respect to all participants in the collaborations meetings demonstrate the active engagement of facilitative leadership (Agbodzakey, 2012; Ansell & Gash, 2008; Freeman, 1997; Lasker et al., 2001). Employing tact, constructively exploring differences, and building consensus are some of the techniques used in facilitation (Ansell & Gash, 2008; Ball, 2009; Gray, 1989; Lasker et al., 2001; McGuire, 2006).
Leadership skills of negotiation and mediation use techniques such as exhibiting mutual respect for different stakeholder perspectives, finding common ground among diverse interests, and managing conflict to mitigate power struggles and promote mutual gains (Agbodzakey, 2012; Ansell & Gash, 2008; Ball, 2009; Bryson et al., 2006; Emerson et al., 2012). Progress evaluation skills apply leadership techniques such as monitoring progress
25


toward the collaborations goals, acknowledging small victories, and using incremental changes to inspire continued advancement (Ansell & Gash, 2008; Bryson et al., 2006; Emerson et al., 2012).
Collaborative Process
The fourth element in the Ansell and Gash (2008) collaborative governance case studies is the collaborative process itself, which is described as cyclical rather than linear in nature, and its essential components are face-to-face dialogue, trust building, commitment to process, shared understanding, and intermediate outcomes. Communication is the heart of the process, and face-to-face dialogue among stakeholders is particularly advantageous at the beginning (Ansell & Gash, 2008; Emerson et al., 2012; Reilly, 2001). As communication serves as a mechanism in breaking down barriers, it also leads to building relationships and contributes to sustaining engagement over time (Ansell & Gash, 2008; Emerson et al., 2012).
Communication also facilitates building trust among participants. In the case studies examined, Ansell and Gash (2008) noted that trust develops over time when stakeholders work together, establish relationships, and demonstrate they are reasonable, consistent, and reliable. Trust increases as expectations are formed and fulfilled; stakeholder confidence grows along with the belief that all participants will execute their responsibilities without taking advantage of other participants (Emerson et al., 2012; Lasker et al., 2001; Vangen & Huxham, 2003b). Trust enables stakeholders to move beyond their own personal interests and frames of reference toward consideration and understanding of the interests and values others hold (Ansell & Gash, 2008; Emerson et al., 2012).
Communication and trust contribute to the cultivation of commitment to the process (Ansell & Gash, 2008; Purdy, 2012). Stakeholders commit to the collaborative process as a
26


mechanism for addressing their common problems, and their commitment to full participation increases in the absence of viable alternatives (Ansell & Gash, 2008; Lasker et al., 2001; Reilly, 2001). Additionally, in collaborative governance efforts, commitment to the process involves having or developing the belief that the best way to attain desired policy outcomes and achieve mutual gains is through negotiation and compromise (Ansell & Gash, 2008; Lynn et al., 2000; Williams & Matheny, 1995).
Clarifying a common purpose fosters a shared understanding among stakeholders that they can achieve collectively what none can accomplish individually (Ansell & Gash, 2008; Lasker et al., 2001; Tett, Crowther, & OHara, 2003; Vangen & Huxham, 2003b). Shared understanding shifts the perception of ownership from the convening public agency or institution to the collective body of stakeholders (Ansell & Gash, 2008). Shared understanding is an integral part of each stage of the collaborations collective learning experience and implies agreement from the problems definition to the solutions goals and objectives (Ansell & Gash, 2008).
Shared understanding promotes another component in the collaborative process: the recognition that intermediate outcomes are important to long-term success (Ansell & Gash, 2008; Chrislip & Larson, 1994). Articulating those outcomes, producing a shared plan of action, and creating assessment criteria give participants an incremental path forward (Emerson et al., 2012). Openly recognized, short-term accomplishments and progress fuel momentum that encourages stakeholders to remain committed and involved (Ansell & Gash, 2008; Reilly, 2001).
27


Model Considerations
Earlier sections in this chapter outlined six defining criteria and four common elements and related components in the Ansell and Gash (2008) model for collaborative governance. An approach was described in which public and private actors work together in distinct ways using particular processes to establish laws and rules for providing public goods and services. Ansell and Gash (2008) hold a view of collaborative governance that pertains specifically to public affairs. Unlike the broader integrative framework for collaborative governance introduced by Emerson, Nabatchi, and Balogh (2012), for example, the Ansell and Gash (2008) criteria and elements align more closely with examples where a public agency is the catalyst for convening a public forum focused on addressing a public policy issue.
In addition, the Ansell and Gash (2008) approach draws fundamental characteristics of collaborative governance from not only existing literature but also a meta-analysis of case studies across multiple disciplines. From the case studies reviewed, Ansell and Gash (2008) provide a realistic approach for use by practitioners and identify factors crucial to the functioning of the collaborative process. Several of the case studies involved community health partnerships or interdisciplinary health initiatives (Alexander, Comfort, & Weiner, 1998; el Ansari, 2003; Fawcett et al., 1995; Gilliam et al., 2002; Mitchell & Shortell, 2000; Mizrahi & Abramson, 2000; Roussos & Fawcett, 2000; Weech-Maldonado & Merrill, 2000). Comparing features of those partnerships and initiatives to the Ansell and Gash (2008) criteria and elements outlined earlier in this chapter indicates their approach would be appropriate for determining whether or not a complex public health care delivery system is an example of collaborative governance.
28


Several articles referencing the Ansell and Gash (2008) model of collaborative governance were read in conjunction with this research study. Review was conducted to determine if there were additional considerations for or modifications to the Ansell and Gash (2008) approach relevant to this studys collaborative governance lens. Some authors suggest broadening collaborative governance beyond a public affairs focus (Binz-Scharf, Lazer, & Mergel, 2012; Clarke, 2017; Emerson & Nabatchi, 2015; Emerson et al., 2012; Mosley, 2014). Another proposes enriching the collaborative governance approach with additional topics such as social capital and its ability to enable and encourage mutually advantageous cooperation (Oh & Bush, 2016; Social capital, n.d.). One recommends expanding the existing element of facilitative leadership (Page, 2010). Another advocates incorporating additional details about bureaucratic features and routines (Michel, 2017). Sorensen and Torfing (2011) suggest the benefit of incorporating innovation. These examples offer valuable perspectives, but several others amplify two issues in the Ansell and Gash (2008) approach that are particularly relevant to this research study.
Several authors elaborated on the effects of local influences and time in collaborative governance efforts. Authors stressed the importance of considering potentially different effects of a collaborative dispersed across multiple jurisdictions, initiated in local and regional settings, or underrepresented in certain geographic areas (Andres & Chapain, 2013; Gibson, 2011; Herranz, 2008; Koebele, 2015; Siddiki, Carboni, Koski, & Sadiq, 2015).
Ansell and Gash (2008) also emphasized the importance of collaborative governance in local environments where previous attempts at downstream implementation of policy had failed. They caution that leadership at the local level could be limited by circumstances and resource
29


availability, and effective collaboration would be constrained when there was a lack of leadership (Ansell & Gash, 2008).
Several authors addressed the importance of considering the effect of time on collaborative governance efforts (Berardo, Heikkila, & Gerlak, 2014; de Loe, Murray, & Simpson, 2015; Gerlak & Heikkila, 2011; Gollagher & Hartz-Karp, 2013; Heikkila &
Gerlak, 2016). From a pragmatic perspective, stakeholders considered spending time on collaboration as taking time from normal business operations (de Loe et al., 2015). Authors also advised that how a collaborative process unfolds and evolves over time may be different than how it initially appears when designed and that one indicator of a collaboratives success is sustainability over time (Gollagher & Hartz-Karp, 2013; Heikkila & Gerlak,
2016). Ansell and Gash (2008) recognized collaborative governance as a time-consuming process requiring a long-term commitment to achieve outcomes. The authors also commented on the positive effect the investment of time at the beginning of the collaborative effort has on implementation and later operations (Ansell & Gash, 2008). Without a view over time, the ability to determine the collaboratives successful functioning would be limited (Heikkila & Gerlak, 2016). Working together over time provides an additional benefit. Stakeholders learn how to collaborate successfully, which may serve as a foundation to support collaboration on additional issues or in new areas (Ansell & Gash, 2017). Failure to consider the effects of local influences and time on a collaborative governance effort could impact outputs and outcomes (Koebele, 2015).
As a prerequisite for selecting a model by which to determine whether or not Colorados ACC is an example of a collaborative governance effort, this research study sought to understand the criteria and elements in the Ansell and Gash (2008) model, consider
30


its public affairs constraints, review health-related case studies used in their research, and read a subset of related articles published after their research findings. As a result, the Ansell and Gash (2008) model was selected as an appropriate choice for examining the collaborative governance features of the ACC. The next chapter of this thesis explores the ACCs features through the lens of the Ansell and Gash (2008) collaborative governance model.
31


CHAPTER III
COLORADOS APPROACH
The first chapter of this thesis summarized Medicaids significance and the need for innovative solutions to address its escalating growth. The second chapter identified collaborative governance as one means of approaching Medicaid improvement and used the collaborative governance definition, criteria, and elements from the Ansell and Gash (2008) research to ground the object of this research study. This chapter examines Colorados move to reform its predominantly fee-for-service Medicaid program by developing the ACC and uses the Ansell and Gash (2008) approach to determine whether or not the ACC is a collaborative governance example.
References and examples in this chapter related specifically to Colorados Medicaid program and the ACC are resources currently available or previously found on publicly available websites. Main sources for the reports and documentation cited are the United States Department of Health and Human Services Centers for Medicare & Medicaid Services (CMS), the federal agency responsible for administering the Medicare and Medicaid programs; Colorados Department of Health Care Policy and Financing, the agency responsible for administering the states Medicaid program; national and state nonprofit organizations; and the seven Regional Care Collaborative Organizations (RCCOs) participating in the ACC. Sources include such documentation as requests for information, requests for proposals, annual reports, and committee and subcommittee minutes. All sources are cited appropriately in the text in this chapter and included in the References section at the end of this thesis.
32


The ACC and Six Defining Criteria
As seen in the previous chapter of this thesis, the Ansell and Gash (2008) definition of collaborative governance includes six criteria: a forum initiated by public agencies or institutions; a public policy or public management issue expressed as its focus; a forum organized formally and convened collectively; nonstate stakeholders involved as active participants; decision making included as a role for forum participants; and consensus used as the basis for decision making. All are present in the ACC. The designation of the Colorado Department of Health Care Policy and Financing (the Department) as the public agency responsible for developing, implementing, maintaining, and advancing the ACC satisfies the first criterion, and the identification of Colorado Medicaid reform as the public policy and management issue the ACC addresses fulfills the second (CMS, 2012; HCPF, 2009b; Rodin & Silow-Carroll, 2013; State Policy Options, 2012).
The ACC is the formally organized and collectively convened forum, which satisfies the third criterion (HCPF, 2009a; Karabatsos, 2011; Rodin & Silow-Carroll, 2013; State Policy Options, 2012). ACC participants such as the Regional Care Collaborative Organizations (RCCOs), the Statewide Data and Analytics Contractor (SDAC), Primary Care Medical Providers (PCMPs), specialists and other health care providers, community-based organizations, and Medicaid recipients and advocates fulfill the fourth criterion in their roles as active nonstate stakeholders (CMS, 2012; HCPF, 2009a; Karabatsos, 2011; Rodin & Silow-Carroll, 2013; State Policy Options, 2012).
Satisfying the fifth criterion, stakeholders play an active role in ACC decision making through participation as RCCOs, the SDAC, PCMPs, and members of the central and regional advisory committees and subcommittees (Colorado Access, 2017; Colorado
33


Community Health Alliance, 2017; Community Care Central Colorado, 2017; HCPF, 2017c; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017). Additional Medicaid recipients and advocates as well as other community-based organizations influence decision making through participation in open meetings and workgroups, public comment periods, and survey completion (CMS, 2012; HCPF, 2017c).
Finally, as seen in the Departments reports, advisory committee by-laws, and committee and subcommittee minutes, the ACCs consensus-based approach to decision making satisfies the sixth criterion (Colorado Access, 2017; Colorado Community Health Alliance, 2017; Community Care Central Colorado, 2017; HCPF, 2012c, 2017c; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017). As Ansell and Gash (2008) stated, Although public agencies may have the ultimate authority to make a decision, the goal of collaboration is typically to achieve some degree of consensus among stakeholders (pp. 546-547). The Department is ultimately responsible for decisions affecting the ACC; however, the ACCs consensus-oriented collaboration offers stakeholders a forum in which to discover areas of agreement, make formal recommendations, and participate in the decision-making process (HCPF, 2017c; McGinnis & Small, 2012).
The ACC and Four Common Elements
This section compares the four common elements and primary components that surfaced in the Ansell and Gash (2008) meta-analysis to characteristics of the ACC. This section is arranged in subsections containing information about starting conditions, institutional design, facilitative leadership, and collaborative process, which were discussed during the examination of the Ansell and Gash (2008) model in the previous chapter of this thesis. Table 3.1 serves as a guide to the narrative that follows.
34


Table 3.1: Ansell and Gash Elements and Components and ACC Characteristics
Ansell and Gash Elements ACC Characteristics
Starting Conditions Prehistory of conflict or cooperation Power and resource imbalances Incentives to participate Starting Conditions Historical dissatisfaction and mistrust Multiple interests and diverse perspectives Focal representation and accountability
Institutional Design Inclusive participation Exclusive forum Clear ground rules Process transparency Institutional Design Focal, regional, statewide engagement Vision and interdependent structure Roles and expectations Emphasis on openness and documentation
Facilitative Leadership Approach Skills Techniques Facilitative Feadership Guidance Promoting participation and managing process Active listening and leveraging common ground
Collaborative Process Face-to-face dialogue Trust building Commitment to process Shared understanding Intermediate outcomes Collaborative Process Monthly and quarterly meetings Mediators and local leaders Shared ownership and mutual gains Clear mission and common goals Incremental progress and future strategy
Starting Conditions
Components found in the starting conditions of the collaborative governance examples Ansell and Gash (2008) analyzed included a prehistory of conflict or cooperation, power and resource imbalances, and incentives to participate. These components also existed in Colorado before the ACCs inception. Literature and case studies indicate that conditions existing before collaboration begins can encourage either conflict or cooperation among public agencies and stakeholders (Ansell & Gash, 2008; Emerson et al., 2012; Gray, 1989). Discord and tension were apparent in Colorado due to previous unsuccessful attempts to implement managed care in the state. Like many other states, Colorado attempted to curb Medicaid fee-for-service spending by migrating to managed care in the mid-1990s (Holahan, Zuckerman, Evans, & Rangarajan, 1998; Hurley & Somers, 2003; Weissert, 2002). The states chiefly hierarchical approach proved unsuccessful as evidenced by issues such as
35


limited health care access for Medicaid recipients, inadequate provider networks, insufficient rate analysis and reimbursement, and dissatisfaction among individuals, providers, and health plans (Hill, 2004; Lutzky et al., 2002; Thomson, Rhodes, & Cowie, P. C., 2000; Weissert, 2002). In 2000 a Colorado judge awarded Rocky Mountain Health Maintenance Organization (HMO) $18 million in back Medicaid payments, and both the state and the HMO agreed that independent actuaries would analyze the rate-setting mechanism the Department used (Lutzky et al., 2002). Later that year Kaiser Foundation Health Plan of Colorado filed suit to recover years of Medicaid payments the Department allegedly withheld (Lutzky et al., 2002). In less than one year, the Department lost four of its participating managed care health plans and suspended enrollment in its only remaining plan, Colorado Access (Hill, 2004). These issues contributed to increasing distrust and conflict among the Department, health plans and providers, Medicaid recipients, and other stakeholders. Reverting to a fee-for-service approach, Colorado then became one of only two states with voluntary managed care enrollment for all eligible Medicaid beneficiary groups; 85% of its Medicaid recipients elected a fee-for-service approach (Gifford et al., 2011; Hill, 2004; Karabatsos, 2011). The distrustful environment precipitated by the failure of managed care coupled with escalating costs fueled by the fee-for-service resurgence prompted Colorado to regroup (Kaiser Family Foundation, 2012, 2013; Rodin & Silow-Carroll, 2013).
In his 2006 campaign for governor, Bill Ritter outlined policy positions in The Colorado Promise, including a plan for health care. Ritter (2006) stated, Quite simply, our health care system is broken, and this crisis will not be fixed by tinkering around the edges and making only small incremental changes (p. 11). He included collaborative development in his eight fundamental principles for health care reform (Ritter, 2006, p. 11). In Colorados
36


2006 legislative session, Senate Bill 06-208 established the Blue Ribbon Commission on Health Care Reform (the Commission) (Access to Affordable Health Care Act, 2006; Burger & Stapleton, 2009). After months of deliberation and discussions with stakeholders, constituents, legislators, and executive officials, the Commission presented a comprehensive report in 2007 (Blue Ribbon Commission for Health Care Reform, 2008a, 2008b). The report provided a set of recommendations for health care reform in Colorado. The recommendations became a series of legislative initiatives, the Building Blocks to Health Care Reform, which were passed during the 2008 legislative session (Blue Ribbon Commission for Health Care Reform, 2008a, 2008b; Hill, Courtot, Bovbjerg, & Adams, 2012). The Medicaid Value-Based Care Coordination Initiative was one of the Commissions recommendations, an approach subsequently known as the ACC (HCPF, 2009b; Rodin & Silow-Carroll, 2013).
The starting conditions of dissatisfaction and mistrust that originated in the 1990s over Medicaid managed care negatively influenced Colorados stakeholder environment at the outset of the ACCs design and development (Kaiser Family Foundation, 2012; Thomson et al., 2000; Weissert, 2002). As a result, the Department sought to balance disparities among power and resources by including multiple interests and diverse perspectives in creating the ACC. The Department approached a less hierarchical approach and attempted to engage a broad spectrum of stakeholders. As discussed in the second chapter of this thesis, fostering trust and respect, employing an inclusive strategy to gain wide representation, and offsetting power imbalances among stakeholder groups aid in ensuring meaningful participation, which is vital to collaborations success (Ansell & Gash, 2008). The Department conducted multiple public forums with both in-person and electronic opportunities for participation (Kaiser Family Foundation, 2012; Karabatsos, 2011). In addition, the Department issued a
37


formal request for information that solicited feedback from client groups, community and statewide social service organizations, local governments, independent physician associations, physicians, hospital systems, providers, provider collaboratives, managed care organizations, health plans, clients, quality organizations, health foundations, and other interested parties (HCPF, 2009b, p. 1). The request for information was 54 pages in length and contained more than 200 questions categorized by regional entities; Medicaid recipients and advocates; statewide data and health information technology entities; Medicaid providers such as primary care physicians, specialists, hospitals, pharmacies, home health agencies, and nursing facilities; and all other interested parties willing to participate (HCPF, 2009b). Stakeholders who engaged in shaping the ACCs design represented diverse interests and divergent opinions (Karabatsos, 2011; Rodin & Silow-Carroll, 2013; State Policy Options,
2012).
Considering the states geography, the demographics of its Medicaid population, and existing provider networks, the scope of Colorados Medicaid delivery system redesign was comprehensive. As incentives to participate, the Department emphasized local representation and accountability (HCPF, 2009b; Karabatsos, 2011). The Department envisioned regional areas of accountability with existing and new support services for Medicaid recipients and providers, health information technology services, data sharing and analysis capabilities, contracts focused on outcomes, and gain-sharing provisions (HCPF, 2009b; Rodin & Silow-Carroll, 2013; State Policy Options, 2012). The request for information presented a compelling incentive to comment and participate. Stakeholders were motivated to enter the dialogue and voice their perspectives as one means of upholding and advancing their interests (Ansell & Gash, 2008; Reilly, 2001). Some of the same vendors and providers who
38


had supported or opposed the Departments previous position on managed care participated in public forums and contributed to the ACCs design (Kaiser Family Foundation, 2012; Karabatsos, 2011; Thomson et al., 2000). The Department built upon the negative historical context of dissatisfaction and mistrust inherent in the starting conditions for health care reform in Colorado and motivated stakeholders to work collaboratively in designing the ACC as a regional-based, accountable Medicaid delivery system.
Institutional Design
Aligned with the Ansell and Gash (2008) element of inclusive participation, the Department intentionally engaged local, regional, and statewide resources, which aided in establishing the legitimacy of the collaboration itself. Initial influences were the Accountable Care Organizations emphasis on coherent local physician and hospital delivery systems that seek to increase the quality and decrease the cost of health care (Fisher, Staiger, Bynum, & Gottlieb, 2007; HCPF, 2009b); a regional approach to health care in which health, business, and political leaders work together to improve health care in their communities (HCPF, 2009b; Wagner, Austin, & Coleman, 2006); and the Triple Aims focus on improving the experience of care, improving the health of populations, and reducing per person health care costs statewide (Berwick, Nolan, & Whittington, 2008; HCPF, 2009b). The ACCs design evolved from a series of iterations. A delivery system for Colorado Medicaid recipients emerged that synthesized input from numerous perspectives. Seeking a middle ground between a fee-for-service system and managed care, Colorado chose a managed fee-for-service hybrid. It combined characteristics of a regional Accountable Care Organization with a Primary Care Case Management system (Fisher et al., 2007; HCPF, 2010a; Verdier, Byrd, & Stone, 2009). A managed fee-for-service hybrid assigns responsibility for the delivery and
39


monitoring of a recipients Medicaid services to a primary care provider or coordinating entity; as payment, the provider or other designated entity receives a small monthly case management fee in addition to customary fees for services delivered and billed (HCPF, 2010a; Verdier et al., 2009). As illustrated in the table in Appendix A and the map in Appendix B, the Department configured the state into seven geographic areas based on an algorithm that considered Medicaid recipient referral and access patterns, population density, provider capacity, public health districts, and the presence of Federally Qualified Health Centers and safety net providers (State Policy Options, 2012). Stakeholders acknowledged the uniqueness of and supported the need for each region. As a result, localized resources became essential to the ACCs development, and the Department encouraged each region to develop approaches suitable for and sustainable by its population and local communities (Karabatsos, 2011; State Policy Options, 2012). Appendix C contains information about each regions organization, and Appendix D includes details about regional relationships and community involvement.
The Ansell and Gash (2008) research also includes an exclusive forum as an element in the institutional design of a collaborative governance effort. Researchers observed that stakeholders were much more likely to participate when the forum was exclusive and alternatives were absent or unattractive (Ansell & Gash, 2008; Fung & Wright, 2001). Early in the process, the Department messaged the ACC as the vision of the future and promoted the appeal of its exclusive design. Envisioning the ACC as an evolving model, the Department began conversations with all interested parties about its long-term plan to integrate federal, state, and local health care programs into a more cohesive and efficient delivery system (HCPF, 2009b, p. 8). The Department signaled the ACC as the foundation
40


for Medicaids future in Colorado, a foundation that would ultimately support all Medicaid recipients in the state (HCPF, 2009b, p. 8). The ACC was created as a regional model of accountability for improving the health, functioning and self-sufficiency of Medicaid clients (HCPF, 2009a, p. 14). It also embodied a distinct and interdependent structure that contributed to its exclusiveness.
Three design elements central to the ACCs structure were intended to work in concert with each other and with the Department: Regional Care Collaborative Organizations (RCCOs), responsible for achieving health and cost outcomes for Medicaid recipients in the states seven regions; Primary Care Medical Providers (PCMPs), responsible for providing comprehensive primary care and serving as a Medicaid recipients medical home; and the Statewide Data and Analytics Contractor (the SDAC), responsible for serving as a data repository and providing information to RCCOs, PCMPs, and the Department (HCPF, 2010a, 2010b; Kaiser Family Foundation, 2013; Karabatsos, 2011, p. 11). In addition, the Department divided responsibilities among the RCCOs, PCMPs, and the SDAC. All ACC activities were critical and contributed to the overall functioning of the ACC, but none were duplicated by another design element (HCPF, 2010a, 2010b, 2011; Karabatsos, 2011). Distinct contractual and data exchange relationships existed among the elements as seen in Figure 3.1.
41


Solid lines indicate contractual relationships. Dotted lines indicate data relationships.
Figure 3.1: Structure of the ACC
Ansell and Gash (2008) identified clear ground rules as a third component in the institutional design element of the collaborative governance case studies they analyzed. During the ACCs development and implementation phases, the Department used clear ground rules to establish roles and expectations for key participants. In its request for information concerning the development of the ACC, the Department posed questions concerning the composition of governing boards, the development and maintenance of stakeholder relationships, and mechanisms for meeting, monitoring, reporting, and reviewing performance (HCPF, 2009b). Later the Department formally solicited proposals for RCCOs and the SDAC in accordance with the states competitive procurement process (HCPF, 2010a, 2010b). The Departments requests for proposals for RCCOs and the SDAC also contained clear definition of roles, structure, and expectations (HCPF, 2010a, 2010b). For example, in addition to describing the general, strategic, provider support, and accountability
42


requirements for the RCCOs and the SDAC, the Department included details in its requests for proposals related to maintaining the collaborative aspects of the ACC (HCPF, 2010a, 2010b). The Department stated its intention to chair a committee to guide ACC performance improvement that would include representation from each RCCO, the SDAC, and Medicaid recipients and providers (HCPF 2010a, 2010b). The Department also set the expectation for each RCCO to guide regional performance improvement through a committee directed and chaired by the RCCO; the committee would have a formal membership and governance structure, include broad Medicaid recipient and provider representation, and meet no less frequently than quarterly (HCPF, 2010a, p. 51). Additionally, the Department stipulated various committees the SDAC would facilitate or attend (HCPF, 2010b, pp. 62-63). The Department ultimately executed contracts with the RCCOs and the SDAC. Because PCMPs were required to be Medicaid providers, they had existing contracts with the Department (HCPF, 2010a, 2017e; Karabatsos, 2011, p. 12). As a result, the Department established and maintains a contractual relationship with all seven RCCOs, the SDAC, and all PCMPs; in addition, each RCCO has a contractual relationship with each PCMP in its service area (HCPF, 2010a; Karabatsos, 2011).
From engaging stakeholders and obtaining input to creating the ACCs structure and executing contractual relationships, the Department facilitated an open and documented process. As the ACCs implementation began, ground rules and responsibilities prescribed in the contracts continued to promote process transparency, the fourth component Ansell and Gash (2008) describe in institutional design. The ACC has a diverse program improvement advisory committee designed to provide guidance and make written recommendations to aid in improving health outcomes, access, cost, and client and provider experience (HCPF,
43


2017c). The advisory committees by-laws specify its membership, structure, and process, and the committee meets in person with Department staff at least quarterly with all meetings open to the public (HCPF, 2017c). The advisory committees three current standing subcommittees meet no fewer than eight times per year and currently focus on improving health, systems, and provider and community issues (HCPF, 2017c). The Department posts meeting agendas, handouts, and minutes on its website and makes recordings of the proceedings available for sixty days after each meeting (HCPF, 2017c). In addition, each RCCO hosts regular member, stakeholder, or performance advisory committee meetings in its area communities (Colorado Access, 2017; Colorado Community Health Alliance, 2017; Community Care Central Colorado, 2017; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017). Meetings are open, meeting schedules are communicated in advance, and transcripts or minutes along with other reports and materials are publicly available (Colorado Access, 2017; Colorado Community Health Alliance, 2017; Community Care Central Colorado, 2017; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017).
Facilitative Leadership
Facilitative leadership, the fourth component of collaborative governance examples Ansell and Gash (2008, 2012) studied, is characterized by its approach, skills, and techniques. Continuity and diversity in leadership provide a stable and broad foundation for further shaping and guiding policy and process (Ansell & Gash, 2008; Huxham & Vangen, 2000; Ostrom, 2002). Laurel Karabatsos, a member of the Departments executive team and current Delivery System and Payment Innovation Division Director and Deputy Medicaid Director, is one of the ACCs original architects and continues to oversee the program and its
44


supporting staff (HCPF, 2017d). A combination of shared and distinct representation exists throughout ACC leadership not only in the Departments executive team and supporting staff but also among leaders of the RCCOs, the SDAC, PCMPs, Medicaid recipients and advocates, and other stakeholders (HCPF, 2017c, 2017d). Similarities are apparent when ACC leadership roles are compared to those Ansell and Gash (2008, 2012) described as stewards, mediators, and catalysts. Since initially convening the ACC, the Department continues to serve as its steward to guide and oversee the overall effort and maintain the integrity of the process (CMS, 2012; HCPF 2009a, 2012c, 2016b, 2017b).
Facilitating communication and nurturing relationships among stakeholders in accordance with the Ansell and Gash (2008, 2012) description, ACC mediators exist not only in the Department but also among other ACC stakeholders. During the ACCs development and implementation, the Department used its own staff or professional consultants to facilitate meetings and workgroups and develop relationships among stakeholders (CMS, 2012; Karabatsos, 2011). The Department also facilitated two monthly operations meetings to ensure communication and coordination among RCCOs, the SDAC, and the Department (CMS, 2012). To promote further collaboration, RCCOs met independently to reinforce their relationships and work together on specific issues (CMS, 2012). The ACCs program improvement advisory committee maintains its membership roster and by-laws on the Departments website; the by-laws illustrate the roles of co-chairs in facilitating communication among committee and subcommittee members and other ACC stakeholders (HCPF, 2017c). In addition, each RCCO facilitates its own ACC advisory committee and retains committee membership, participation, meetings, and minutes on its website (Colorado Access, 2017; Colorado Community Health Alliance, 2017; Community Care Central
45


Colorado, 2017; HCPF, 2017f; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017).
Consistent with the description from the Ansell and Gash (2008, 2012) research, ACC leadership also includes catalysts that stimulate identification of opportunities to create value. For example, the SDACs focus on data revealed opportunities to improve health care and outcomes, including peer-to-peer learning in and among RCCOs (HCPF, 2010b). The SDAC used national and state data to provide demographic and service utilization information to ACC stakeholders and to develop baseline performance measures for the ACC to gauge progress in improving health outcomes and reducing costs (HCPF, 2010b). The SDACs role as a catalyst stimulated RCCO and PCMP activity to enhance service delivery to Medicaid recipients. RCCOs established formal data-sharing agreements or arrangements with providers in their regions in accordance with security and privacy guidelines contained in the Health Insurance Portability and Accountability Act (1996) to identify Medicaid recipients for outreach and care coordination purposes (Health Services Advisory Group, 2013a). RCCOs explored using data in a variety of innovative ways in the ACC: to pinpoint Medicaid recipients with multiple ER visits; to distinguish those needing post-hospitalization assistance; to determine specialists Medicaid recipients used; to funnel specialty care to PCMPs; to distinguish high-volume Medicaid practices where care coordinators could be assigned; and to populate web-based portals that providers could access (Health Services Advisory Group, 2013a, 2013b, 2014, 2015).
As stewards, mediators, and catalysts, ACC leadership includes multiple perspectives and involves individuals in diverse ways to empower collaboration and increase the potential for success (Ansell & Gash, 2008, 2012; Bryson & Crosby, 2005; Emerson et al., 2012;
46


Ostrom, 2002). ACC leadership examples in the first three paragraphs of this subsection illustrate the use of engagement, facilitation, mediation, and progress evaluation skills described in the second chapter of this thesis, which have been identified as necessary to move a collaborative effort forward (Ansell & Gash, 2008; Lasker et al., 2001; OLeary et al., 2010; OLeary et al., 2006). Like examples from the collaborative governance literature referenced in the second chapter of this thesis, instances of ACC facilitative leadership techniques such as enlisting participation from all stakeholders, helping stakeholders develop a common language, inviting all stakeholder perspectives to be heard, and showing respect to all stakeholders appear earlier in this subsection (Agbodzakey, 2012; Ansell & Gash, 2008; Freeman, 1997; Lasker et al., 2001). Constructively exploring differences, building consensus, finding common ground among varied interests, acknowledging small victories, and using incremental changes to inspire continued advancement are additional techniques reflected in this subsections ACC leadership examples, which are similar to those Ansell and Gash (2008, 2012) described.
Collaborative Process
In addition to the starting conditions, institutional design, and facilitative leadership components common to collaborative governance case studies examined by Ansell and Gash (2008), the ACC includes a process conducive to collaboration. Face-to-face dialogue is one of the identified components in a collaborative process (Ansell & Gash, 2008), and the ACC demonstrates evidence of such dialogue in the documentation of its monthly and quarterly meetings (Colorado Access, 2017; Colorado Community Health Alliance, 2017; Community Care Central Colorado, 2017; HCPF, 2017c; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017).
47


Trust building is another collaborative process component emphasized by Ansell and Gash (2008) and substantiated in the ACC through documented accounts of Department staff, consultants, facilitators, RCCO leadership, and committee members serving as mediators and local leaders to develop relationships and increase cooperative participation within the regions and across the state (CMS, 2012; Colorado Access, 2017; Colorado Community Health Alliance, 2017; Community Care Central Colorado, 2017; HCPF, 2010a, 2010b, 2017c; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017).
The third component in the collaborative process Ansell and Gash (2008) defined is commitment to the process. The ACCs interdependent structure, described in the discussion of its institutional design earlier in this chapter, reinforces commitment to the collaborative effort through shared ownership of the process and outcomes. The Department, RCCO leadership, the SDAC, PCMPs, Medicaid recipients and advocates, and other stakeholders have distinct roles in the ACC, and each role is a vehicle for distributing ownership of and responsibility for the collaborations success (HCPF, 2017c, 2017d).
Ansell and Gash (2008) described shared understanding as the fourth component of a collaborative process. To promote a shared understanding among ACC participants, the Department explicitly states its mission as well as the ACCs two central goals and four primary objectives. The Department pursues a mission not only to improve health care access and outcomes for Medicaid recipients but also to be a responsible steward of financial resources (HCPF, 2017a). Concurrently, the ACC progresses toward its goals to improve health outcomes and control costs by expanding access to comprehensive primary care; providing a focal point of care for all members including coordinated and integrated access to
48


other services; ensuring a positive member and provider experience and promoting member and provider engagement; and effectively applying an unprecedented level of statewide data and analytics functionality to support transparent, secure data sharing and enable the near-real-time monitoring and measurement of health care outcomes and costs (CMS, 2012; HCPF, 2010a, 2010b).
Intermediate outcomes included as the fifth component in a collaborative process appear in the ACC as a focus on incremental progress with an attentive eye on future strategy (Ansell & Gash, 2008; CMS, 2012; Karabatsos, 2011; State Policy Options, 2012). To move toward improved health outcomes and reduced costs, the ACCs collaborative process provides a basis for systematic actions. Combined and executed properly, seven features embedded in the ACCs process contribute to measured progress: a regional approach to managing, providing, and coordinating care; principles of a client-centered medical home model; an integrated network of providers; provision of high-quality care coordination and medical management services; a focus on accountability to improve outcomes and control costs; analysis and application of informatics and benchmarking to review, measure, and compare utilization, outcomes, and costs; and a focus on continuous improvement and innovation, constant learning, and the sharing of best practices (CMS, 2012).
Implications for Successful Collaboration
In the common elements of starting conditions, facilitative leadership, and collaborative process in the Ansell and Gash (2008) model, the authors suggest 10 implications to facilitate success in collaborative governance. Although a collaborative governance effort still could be considered an authentic example without these implications, Ansell and Gash (2008) perceive it would be less likely to succeed. Assessment of the ACC
49


based on the Ansell and Gash (2008, 2012) definition, criteria, and components as described in the second chapter and earlier in this chapter indicates the ACC is an example of a collaborative governance effort. Before moving forward with a research study designed to study health outcomes and cost of health care for individuals served in the ACC, this thesis examines the ACC in relation to the 10 implications for success (Ansell & Gash, 2008). Table 3.2 summarizes the implications to guide the narrative.
Table 3.2: Ansell and Gash Implications for Success
Ansell and Gash Implications for Success
Starting Conditions
Prehistory of conflict or cooperation
- High degree of interdependence among stakeholders or steps taken to remediate low levels of trust and social capital
Power and resource imbalances
- Commitment to positive strategy of empowerment and representation of weaker or disadvantaged stakeholders
Incentives to participate
- Stakeholder perception of high interdependence
______- Respect from courts, legislators, and executives to honor collaborative process outcomes
Facilitative Leadership
Approach
- Honest broker accepted and trusted by stakeholders
- Organic leaders from within the stakeholder community respected and trusted at the
__________outset_____________________________________________________________________________
Collaborative Process
Trust building
- Time allowed if remedial efforts are needed
Commitment to process
- Achievement of buy in
- Suitability for situations requiring ongoing cooperation
Intermediate outcomes
- Production of small wins
Implications for Starting Conditions
As described in the earlier discussion of Colorados Medicaid environment before the ACCs creation, a prehistory of conflict and distrust existed (Hill, 2004; Lutzky et al., 2002; Thomson et al., 2000; Weissert, 2002). In designing the ACC, the Department deliberately
50


took steps to remediate low levels of trust and social capital by investing time in multiple opportunities to obtain diverse stakeholder participation through varied feedback channels (HCPF, 2009b; Kaiser Family Foundation, 2012; Karabatsos, 2011). Throughout the design process, the Department engaged stakeholders who represented diverse interests and divergent opinions (Karabatsos, 2011; Rodin & Silow-Carroll, 2013; State Policy Options, 2012). In addition, the ACC was intentionally created as a regional health care delivery system for Medicaid recipients as seen by the interdependent structure and roles and responsibilities of the RCCOs (HCPF, 2009a).
Intentional engagement of a broad spectrum of stakeholders signaled an inclusive strategy designed to offset power imbalances among groups and ensure meaningful participation (Karabatsos, 2011; Rodin & Silow-Carroll, 2013; State Policy Options, 2012). Additionally, a strategy of empowerment was employed in the creation of the program improvement advisory committee and subcommittees at the overarching ACC level and in area advisory committee meetings hosted by RCCOs in their communities (Colorado Access, 2017; Colorado Community Health Alliance, 2017; Community Care Central Colorado, 2017; HCPF, 2017c; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017).
Providing incentives to participate, Colorados 2008 legislative session passed a series of initiatives, which included what would come to be known as the ACC (Blue Ribbon Commission for Health Care Reform, 2008a, 2008b; HCPF, 2009b; Hill et al., 2012; Rodin & Silow-Carroll, 2013). Contributing further legitimacy and respect for the effort, the Department was required to submit an annual report to the legislatures Joint Budget Committee providing information about the ACCs implementation (HCPF, 2012a, 2013a).
51


High interdependence was embedded in the ACCs structure of distinct roles for the SDAC, RCCOs, and PCMPs, and stakeholders were encouraged to participate too (HCPF, 2010a, 2010b; Kaiser Family Foundation, 2013; Karabatsos, 2011).
Implications for Facilitative Leadership
The ACC satisfied facilitative leadership implications for success in utilizing both honest brokers and respected members of the stakeholder community throughout the design and implementation process. In some cases, consultants and facilitators were used, and in other situations, RCCO leadership, committee members, or other stakeholders served as mediators and facilitators to develop relationships and increase participation (CMS, 2012; Colorado Access, 2017; Colorado Community Health Alliance, 2017; Community Care Central Colorado, 2017; HCPF, 2010a, 2010b, 2017c; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017).
Implications for Collaborative Process
As documented earlier in this chapters Implications for Starting Conditions subsection, time was consciously included in the collaborative process to foster building trust (HCPF, 2009b; Kaiser Family Foundation, 2012; Karabatsos, 2011; Rodin & Silow-Carroll, 2013; State Policy Options, 2012). The additional investment in time was particularly appropriate considering the atmosphere of mistrust and conflict present in the environment before the ACC was developed (Hill, 2004; Lutzky et al., 2002; Thomson et al., 2000; Weissert, 2002).
When the Department signaled the ACC as the foundation for Medicaids future in Colorado, which would ultimately support all Medicaid recipients in the state, it provided stakeholders with knowledge that the ACC was a sanctioned initiative that would require
52


long-term participation and cooperation (HCPF, 2009b). As seen in the Starting Conditions subsection earlier in this chapter, some of the same stakeholders who opposed the Departments previous attempts at Medicaid reform in Colorado attended public forums and contributed to the ACCs design by voicing divergent opinions (Kaiser Family Foundation, 2012; Karabatsos, 2011; Thomson et al., 2000). Participation over time achieved a level of buy in not experienced in earlier efforts (Kaiser Family Foundation, 2012; Karabatsos, 2011; Thomson et al., 2000).
Intermediate outcomes were addressed in the ACCs design, and the importance of small wins was acknowledged early in the process. The collaborative process included the use of SDAC data and analysis to review, measure, and compare ACC enrollment, utilization, outcomes, and costs at macro and micro levels (CMS, 2012; HCPF, 2010b; Karabatsos, 2011; State Policy Options, 2012). From the outset, focus on these aspects of the ACC contributed to an atmosphere of incremental improvement, constant learning, and ongoing innovation (CMS, 2012).
Satisfying the six defining criteria, the four common elements and related components, and the 10 implications for success established by Ansell and Gash (2008), the ACC appears to be an example of collaborative governance designed with opportunity to succeed. Yet the question remains: is the ACC a governance arrangement matched to the specific problems it was created to address (Ostrom, 2007)? To determine if the ACC fulfills its purpose of transforming service delivery for Medicaid recipients in Colorado to produce positive effects in health outcomes and cost of health care, this thesis transitions from theory to practice and approaches the analysis of health outcomes and health care costs before and after the ACCs implementation.
53


From Theory to Practice
As Ansell and Gash (2008) noted, the need exists to examine the strength and influence, if any, of collaborative governance outcomes. The authors indicated the purpose of their research and meta-analysis was to draw positive and negative findings together in a common model that could begin to identify conditions under which collaborative governance could be said to work or not work in terms of process outcomes (Ansell & Gash, 2008).
As McGuire (2006) wrote, From comparative case studies to large-N quantitative research, there is a growing realization that collaboration is not an end in itself and that only by examining its impact will general management theory be advanced (p. 40). The purpose of this research study is to take the next step and examine ACC performance outcomes: changes in health outcomes and health care costs. In so doing, this thesis contributes evidence to the performance of a collaborative governance effort and the impacts on individuals served.
Enrollment of the first Medicaid recipients in Colorados ACC began in May 2011, and the Department began to report favorable outcomes in 2012 (CMS, 2012; HCPF, 2012c; Karabatsos, 2011; State Policy Options, 2012). In the annual report submitted to the Colorado General Assemblys Joint Budget Committee on November 1, 2012, Susan E.
Birch, the Departments Executive Director, stated the ACCs first-year performance resulted in reduced utilization rates for ER visits, hospital readmissions, and high-cost imaging services; lower rates of exacerbated chronic health conditions such as asthma and diabetes; and reduced total cost of care for enrolled Medicaid recipients (HCPF, 2012c, p. 2). Additionally, the report indicated the program produced positive results during the enrollment ramp-up phase in which access to claims data, provider network infrastructure, and care coordination activities were still under development (HCPF, 2012c, p. 4). The
54


Departments reported results ensured the ACCs continuation. The Department projected the ACC would demonstrate positive results in future years and concentrated its efforts to make the ACC the primary Medicaid delivery system in Colorado (HCPF, 2012c).
The next chapter of this thesis explains the research design and methods used to study ACC performance outcomes. This research examines actual health outcomes and health care costs for specific groups of Colorado Medicaid recipients before and after implementation of the ACC. In so doing, findings in the fifth chapter of this thesis contribute evidence to the performance of a collaborative governance effort and the impacts on individuals when served through such a process. While collaboration presents a different way for the public, private, and nonprofit sectors to work together, it is important to understand if collaborative governance affects outcomes in measurable ways for the individuals it intends to benefit (Padgett, Bekemeier, & Berkowitz, 2004; Varda, 2011; Varda & Retrum, 2012; Varda,
Shoup, & Miller, 2012). As Varda and Retrum (2012) indicated, benefits of collaboration have become widely accepted and the practice of collaboration is growing, but evaluation and analysis are needed because collaboration has the potential to improve the process of healthcare which can create better outcomes, but also reduce the cost of delivering services by eliminating waste, unnecessary work, and rework (p. 170).
55


CHAPTER IV
RESEARCH DESIGN AND METHODS
The first chapter of this thesis highlighted the significance of finding different approaches to the escalating problems associated with increasing Medicaid enrollment and rising costs. The second and third chapters then examined collaborative governance literature and compared the collaborative governance definition, criteria, and elements distilled from the Ansell and Gash (2008) meta-analysis of 137 diverse case studies to Colorados ACC. Satisfied that the ACC can be categorized as a collaborative governance effort, this chapter specifies a research design to examine implications of the ACC for Colorados Medicaid recipients. In addition to contributing another example to collaborative governance literature, this thesis examines actual health outcomes and health care costs associated with Colorados approach, thereby contributing evidence about the performance of a collaborative governance process and the impacts on individuals served.
While collaboration may present a different way for the public, private, and nonprofit sectors to work together, it is important to understand if collaborative governance affects outcomes in measurable ways for the individuals it intends to benefit (Padgett et al., 2004; Varda, 2011; Varda & Retrum, 2012; Varda et al., 2012). As Varda and Retrum (2012) indicated, the benefits of collaboration have become widely accepted and the practice of collaboration is growing, but the ability to measure, document, and strategize to affect practice has been weak (p. 170). To that end, this chapter frames the research question, hypothesis, and variables; discusses data and analysis; and considers limitations and strategies to this approach.
56


Research Question, Hypothesis, and Variables
Examination of the escalating problems associated with increasing Medicaid enrollment and rising costs coupled with Colorados collaborative approach led to formulation of the research question: What effects will participation in the ACC have on health care for Medicaid recipients in Colorado? The unit of observation is the individual, the Medicaid recipient participating or not participating in the ACC. The unit of analysis is the ACC. The null hypothesis and its alternative follow:
Ho: Participation in the ACC is expected to have no effect on health outcomes and cost of health care for Medicaid recipients in Colorado.
Hi: Participation in the ACC is expected to (a) improve health outcomes and (b) reduce cost of health care for Medicaid recipients in Colorado.
To examine effects, this study used a quasi-experimental nonequivalent control group research design (Singleton & Straits, 2010, pp. 250-252). Two study groups were configured. Those who do not participate in the ACC are the control group, and those who do participate are the ACC group. Three periods were constructed: the 12 months before ACC implementation, May 1, 2010 through April 30, 2011 (Ti); the first 12 months after ACC implementation, May 1, 2011 through April 30, 2012 (T2); and the second 12 months after ACC implementation, May 1, 2012 through April 30, 2013 (T3). The categorical independent variable is participation in the ACC. Participation in the ACC is operationalized as enrolled Medicaid recipients residing in a focus community county and having an existing relationship with a PCMP. Additional details about the research study sample appear in this chapters Data and Analysis section. This research studys dependent variables are health
57


outcomes and cost of health care, which align with the ACCs two central goals of improving health outcomes and controlling costs (HCPF, 2012a, 2012c).
Health Outcomes
Health outcomes are operationalized using the three utilization measures related to health outcomes the ACC initially established as key performance indicators: (a) 30-day allcause hospital readmissions, (b) emergency room (ER) visits, and (c) high-cost imaging services (HCPF, 2010b). Hospital readmissions, ER visits, and high-cost imaging services have gained national and state attention as areas in which excessive or inappropriate use can be reduced to improve health outcomes (Berwick et al., 2008; CHI, 2012; CMS, 2012;
Patient Protection and Affordable Care Act, 2010; Stem, 2013; Zhang, Wan, Rossiter, Murawski, & Patel, 2008). In a coordinated system of health care delivery such as the ACC, reductions in these health outcome indicators would represent improved performance. Hospital readmissions, ER visits, and high-cost imaging services appear in the limited data set at the Medicaid claim level, are linked to the appropriate individual by a unique deidentified member number, and are measured as the number of occurrences. Each measure is defined and explained further in the following three paragraphs.
Hospital readmissions. The ACC defined 30-day all-cause hospital readmissions as any inpatient case that occurred within 30 days of an inpatient discharge of an individual along with specific exclusionary criteria related to eligibility and discharge status, transfers, and interim bill and continued stay (Treo Solutions, 2012, pp. 5-6). The limited data set obtained for this research study contained an indicator for potentially preventable hospital readmissions within 30 days of discharge, which was used as a proxy for all-cause hospital readmissions. This indicator is derived from the individuals inpatient discharge and
58


readmission dates, diagnosis, prior admission procedures, and readmission reason (3M Health Information Systems, 2015).
ER visits. ER visits were considered any outpatient emergency department claim that did not have an inpatient stay on the same date of service with the same client identification number (Treo Solutions, 2012, p. 6). Any ER visit that resulted in an inpatient admission was excluded (Treo Solutions, 2012, p. 6). This research used the indicator for ER visits in the acquired limited data set.
High-cost imaging. The ACC classified high-cost imaging services as any claim in an enhanced ambulatory payment group related to computed tomography (CT) scans, a method using x-rays to create cross-sectional pictures of bones and soft tissue in the body, or magnetic resonance imaging (MRI), a technique using a magnetic field and radio waves to create detailed images of organs and tissues in the body (Mayo Clinic, 2017a, 2017b; Treo Solutions, 2012, p. 7). The list of specific current procedural terminology (CPT) codes included as high-cost imaging services in the ACC appears in Appendix E. The codes in Appendix E were used to identify CT scans and MRI services in the limited data set obtained for this research, and an indicator was created to distinguish them from other diagnostic services.
Cost of Health Care
Cost of health care is operationalized as (a) the cost of claims paid by Medicaid for medical services, (b) the cost of claims paid by Medicaid for prescription drugs, and (c) the per member per month (PMPM) primary care case management fees paid to RCCOs and PCMPs, which was described in the previous chapter of this thesis (HCPF, 2010a, 2012a; Verdier et al., 2009). Amounts Medicaid paid for medical services and prescription drugs are
59


distinct fields in the limited data set. The PMPM fees for ACC members paid to RCCOs and PCMPs were calculated based on amounts published in the first two ACC annual reports as $0 for Ti, which was before the ACCs implementation; $13 and $4, respectively, during T2; and $10.50 and $3, respectively, during T3 (HCPF, 2012a, 2013a; Kaiser Family Foundation,
2013). Published amounts for the appropriate period were multiplied by the number of months a participant was in the ACC for that period, and all per participant calculations were summed. Additional information is included in the fifth chapter in the PMPM subsection and Table 5.11.
Data and Analysis
This section of the chapter provides details about the research studys data and analysis by describing the data source and related considerations, the data sample and configuration of the research study groups, characteristics of the study groups, and the approach selected for data analysis.
Data Source and Considerations
To examine health outcomes and cost of health care for this research study, it was necessary to obtain historical health record and claims information for Colorado Medicaid recipients. The repository for such data in Colorado is the all payer claims database (APCD) (All Payer Claims Database Council, 2017). In 2010 Colorado revised state statutes to create the APCD and authorized the Department to oversee it (Colorado All Payer Claims Database, 2010). The Department named a nonprofit organization, the Center for Improving Value in Health Care (CIVHC), as the APCD administrator and holds CIVHC accountable for the APCDs regulatory compliance through monitoring, auditing, and reporting activities (All Payer Claims Database Council, 2017; Center for Improving Value in Health Care [CIVHC],
60


2017a). The APCD contains individual Medicaid, Medicare, Medicare Advantage, and commercial health plan insurance claims for millions of Colorado health care recipients, and in 2013 CIVHC began releasing approved custom data requests through a process compliant with the Privacy Act of 1974, the Health Insurance Portability and Accountability Act (HIPAA) of 1996, the American Recovery and Reinvestment Acts Health Information Technology for Economic and Clinical Health (HITECH) Act (2009), and other applicable state and federal rules and regulations (All Payer Claims Database Council, 2017; CIVHC, 2017a). The APCD protects the privacy and security of Colorados health claims data by limiting use to purposes permitted by law and by restricting information to data elements reasonably necessary to accomplish an approved purpose (CIVHC, 2017c).
The APCD and its Data Release Review Committee approved this research studys application for a limited data set. Although this limited data set included ZIP Codes and Medicaid eligibility and service dates, which are protected health information data elements, it did not include direct individual identifiers such as name, telephone number, Social Security number, or medical record number (CIVHC, 2017b). In addition to safeguarding data through the APCDs privacy and security measures, this research study preserved data integrity by following University of Colorado Denver information technology security program policies and all Colorado Multiple Institutional Review Board policies and procedures, which include data privacy and security and investigator responsibilities (Colorado Multiple Institutional Review Board, 2017; University of Colorado Denver, 2017). Data Sample and Research Study Groups
The data request application for this research study included Colorado Medicaid medical and prescription drug claims for calendar years 2010, 2011, 2012, and 2013 for those
61


individuals who were between the ages of 21 and 60 in 2010, who had fee-for-service Medicaid as their only source of health insurance coverage for the years in question, and whose records included physical health diagnoses or behavioral health diagnoses for anxiety or depression. Behavioral health diagnoses of interest for anxiety or depression were limited to those covered under the Colorado Community Behavioral Health Services Program, and the list appears in Appendix F (HCPF & Colorado Department of Human Services, 2014).
The limited data set provided by the APCD consisted of 17 files containing more than 28 million records representing over 37 thousand unduplicated individuals (Colorado All Payer Claims Database, personal communication, June 20, 2015). Data provided were based upon Colorado Medicaid eligibility for calendar years 2010 through 2013. All records received were then examined and compared to records needed for this research study. For example, some records received contained Medicare instead of Colorado Medicaid health insurance designations, invalid or non-Colorado ZIP Codes, invalid Medicaid eligibility or service dates, or behavioral health diagnoses other than those for anxiety or depression. Records received that were not needed were then filtered from the limited data set to narrow the focus to only those Colorado Medicaid recipients of interest for this research study.
The Departments Medicaid eligibility requirements for initial ACC enrollment were considered next. To be eligible for enrollment in the ACC and included in the analysis for its key performance indicators, which were described in the discussion about dependent variables in this chapters Health Outcomes subsection, an individual could not have fewer than three months of Medicaid eligibility in a 12-month period (Colorado Access, 2013; M. Ly and H. Schum, personal communication, November 16, 2015; Treo Solutions, 2012). For purposes of this research, records associated with individuals having no fewer than 45 of 48
62


possible months of eligibility during the four calendar years from 2010 through 2013 were selected to ensure continuity of Medicaid eligibility.
The remaining pool of records provided the basis for configuring the control and ACC groups. Before configuring the groups, two remaining parameters were considered: if an individual resided in one of the focus community counties in which the ACC was first implemented and if an individuals Medicaid claims history indicated a relationship with a PCMP. The ACC was initially implemented in a limited number of focus community counties and ensured that every RCCO had some participants in at least one county in its service area (Colorado Access, 2013; Karabatsos, 2011). Records for each unduplicated Medicaid recipient in the limited data set contained a ZIP Code of residence, which then was associated with the appropriate county and the corresponding RCCO in the ACC. Also, for an eligible Medicaid recipient in a focus community county to be enrolled in the ACC, the Department required the recipient to have a previous relationship with a PCMP (Lindrooth, Tung, Santos, Hardy, & OLeary, 2016). The limited data set the APCD provided for this research did not contain a designation for those Medicaid providers who were ACC PCMPs; however, the APCD was able to provide an additional data file in November 2016 that allowed ACC PCMPs to be identified and matched to Medicaid providers in the original limited data set. This enabled configuration of the research study groups. Criteria for both groups are summarized in Table 4.1.
63


Table 4.1: Criteria for Research Study Groups
Criteria Control ACC
Group Group
Between the ages of 21 and 60 in 2010 Yes Yes
Fee-for-service Medicaid as only health insurance coverage from 2010 through 2013 Yes Yes
Claims records containing physical health diagnoses and a behavioral health diagnosis of anxiety or depression Yes Yes
No fewer than 45 out of 48 possible months of Medicaid eligibility during the calendar years from 2010 through 2013 Yes Yes
Resident of an ACC focus community county No Yes
Established relationship with a PCMP No Yes
Number of research study group members 380 2,683
Research Study Group Characteristics
Characteristics of the control and ACC groups are examined in this subsection. Table 4.2 summarizes gender, generation, and RCCO distribution for each group. A more detailed view appears in Appendix G. Race and ethnicity distributions were not included since individuals are not required to report race and ethnicity during their Medicaid application process. Race and ethnicity information that appeared in the limited data set was incomplete and considered unreliable (HCPF, 2015c; Kaiser Family Foundation, 2011c; Wilmot, 2006).
64


Table 4.2: Summarized Characteristics of Control and ACC Groups
Characteristic Tota Number (%) Control Group Number (%) ACC Group Number (%) Difference between Groups
Gender
Female 2,669 (87) 303 (80) 2,366 (88) 8
Male 394 (13) 77 (20) 317 (12) -8
Generation
Gen Y 841 (27) 66 (17) 775 (29) 12
Gen X 1,343 (44) 167 (44) 1,176 (44) 0a
Baby Boomers 879 (29) 147 (39) 732 (27) -12
RCCO
1 328 (11) 129 (34) 199 (7) -21
2 259 (8) 27 (7) 232 (9) 2a
3 852 (28) 22 (6) 830 (31) 25
4 542 (18) 173 (45) 369 (14) -32
5 457 (15) 0 (0) 457 (17) 17
6 355 (11) 11 (3) 344 (13) 10
7 270 (9) 18 (5) 252 (9) 4
Total 3,063 (100) 380 (100) 2,683 (100)
aWhen p > .05, the difference between percentages in the control and ACC groups is not significant.
Gender. Gender distribution of Colorados Medicaid population aligns closely with
national percentages: 58% female and 42% male (CHI, 2017; Kaiser Family Foundation,
2011b). However, both the control and ACC groups reflect a sample with a higher percentage
of females and lower percentage of males. As mentioned earlier in this chapter, records in the
limited data set used for this research study only included Colorado Medicaid recipients
between the ages of 21 and 60 in 2010. When correlated to this research studys age range of
interest, approximately 35% of the entire Medicaid population is represented, which has a
higher concentration of females than the total Medicaid population (Kaiser Family
Foundation, 2011a).
65


The control group has a lower percentage of female participants than the ACC group. Since females have a higher rate of health care utilization than males and incur greater health care costs, this difference could relate to higher health outcomes and lower cost of health care for the control group compared to the ACC group (Cylus, Hartman, Washington, Andrews,
& Catlin, 2010).
Generation. For purposes of this research study, the age range was categorized as three generations: Gen Y, whose members were born between 1982 and 2000; Gen X, whose members were born between 1965 and 1981; and Baby Boomers, whose members were born between 1946 and 1964 (United States Census Bureau, 2015). Also, because this research study focused on adults between the ages of 21 and 60 in 2010, the limited data set contained Gen Y members with birth years only between 1982 and 1989. Taking those factors into consideration, gender and generation distributions were more consistent with what percentages would be in a similar national sample that excluded individuals under the age of 21 and over the age of 60 (Kaiser Family Foundation, 2011a, 2011b).
The control group also has a lower percentage of Gen Y participants with a corresponding higher percentage of Baby Boomer participants than the ACC group. Because older Medicaid recipients typically have more complex health conditions than younger recipients, their overall service utilization and use of higher cost services tend to be greater (Cylus et al., 2010). As a result, this difference could relate to lower health outcomes and higher cost of health care for the control group compared to the ACC group.
RCCO. This chapters Data Sample and Research Study Groups subsection explained that records for each unduplicated Medicaid recipient in the limited data set contained a ZIP Code of residence, which then was associated with its appropriate county to
66


determine the designated RCCO assignment. The table of RCCOs and counties and the state map in Appendix A and Appendix B, respectively, indicate the distribution of RCCOs across Colorado: RCCO 1 serves the west; RCCO 2 covers the northeast; RCCO 4 operates in the southeast; RCCO 5 is the Denver metropolitan area; and RCCOs 3, 4, and 6 cluster around RCCO 5 and cover the center of the state.
Table 4.2 indicates a significant difference in RCCO distribution between the groups in six of the seven regions, and the control group contained no participants in RCCO 5.
Based on analysis of each participants county of residence in the limited data set, the control groups distribution is heavily weighted in more rural areas of the state while the ACC groups distribution is concentrated in more urban areas. As described in the Institutional Design subsection of the third chapter of this thesis, RCCOs were configured to normalize influences, such as rural and urban geography, which otherwise might negatively impact access to and equality of care. However, the differences and potential impact on health outcomes and cost of health care are worth noting before the discussion of findings in the next chapter.
Approach to Analysis
This research study utilized SPSS 23.0 software for both descriptive and inferential statistical analysis of the data. Table 4.2 summarizes the sample and provides information about the characteristics of the research study groups (Healey, 1999). For example, crosstabulations were used to illustrate frequency distributions of the variables of interest in both research study groups. Inferential statistics examined relationships among variables between the two research study groups, within each research study group over time, and among RCCOs. Health outcome variables are categorical and indicate whether or not, based
67


on definitions provided at the beginning of this chapter, a hospital readmission, ER visit, or high-cost imaging service occurred. Chi-square tests were utilized for categorical variables (Hays, 1994; Healey, 1999). For continuous variables related to the cost of health care, such as claims paid by Medicaid for medical services and for prescription drugs, independent t-tests were employed (Hays, 1994; Healey, 1999). The limited data set did not contain PMPM fees paid to RCCOs and PCMPs for each ACC member; therefore, amounts calculated for ACC group participants for T2 and T3 were based on published amounts specified earlier in the Research Question, Hypothesis, and Variables section of this chapter. As noted, the PMPM amount was not the same in T3 as in T2. Results of statistical tests and calculations appear in the next chapter of this thesis in the discussion of this research studys findings.
Limitations and Strategies
After considering the research question and hypothesis along with data and analytical approaches, this section of the thesis addresses possible limitations of the research study and mitigating strategies. In research, validity refers to the relationship between premises and conclusions (Singleton & Straits, 2010). In assessing research results, it is necessary to determine if the independent variable influences the dependent variable and if a causal inference can be made (Singleton & Straits, 2010). Over time scholars developed a typology for assessing validity in experimental or quasi-experimental research studies, and the criteria are discussed at greater length in the Four Aspects of Validity subsection later in this chapter. This research design supports valid causal inference in its examination of the relationship between the ACC and health outcomes in two ways. It satisfies the three criteria for causal inference, and it addresses the mitigation of threats to internal validity and ensures statistical,
68


construct, and external validity. The next two subsections elaborate on the three criteria for causal inference and the four aspects of validity as they relate to this research study.
Three Criteria for Causal Inference
Three elements of causality confirm validity: direction of influence, association, and nonspuriousness (Singleton & Straits, 2010). First, direction of influence reflects a temporal element: cause precedes effect (Singleton & Straits, 2010; Van de Ven, 2007). Assessing health outcomes and cost of health care for Colorado Medicaid recipients during the 12 months before the ACCs implementation, the first 12 months after implementation, and the second 12 months after implementation, as described in the Research Question, Hypothesis, and Variables section of this chapter, establishes a temporal sequence that supports direction of influence.
Second, association illustrates a relationship between the cause or independent variable and its effect or dependent variable (Parker, 1993; Singleton & Straits, 2010; Van de Ven, 2007). Although the effects on Medicare are more pronounced, an aging population and increased life expectancy also influence Medicaid: personal income typically declines, enrollment grows, health conditions worsen, cost of health care increases, and health outcomes decline (Joffe, 2015). Some effects may also be disproportionate. For example, before development and implementation of the ACC, Colorados elderly and individuals with disabilities accounted for one-quarter of Medicaids recipients but almost two-thirds of its expenditures (CHI, 2005). Aging Medicaid recipients or those with a disability typically need more expensive services such as long-term care, prescription drugs, and specialty services (Burger & Stapleton, 2009; CHI, 2005; Rosenbaum, 2011). Colorado anticipated that implementing the ACCs collaborative governance approach to provide a coordinated
69


delivery system for the states Medicaid recipients would have positive impacts (HCPF, 2010a, 2010b; Kaiser Family Foundation, 2012). For example, it was estimated that participation in the ACC would improve health outcomes by increasing follow-up procedures after hospital discharges, thereby decreasing hospital readmissions; by increasing preventive and primary care, consequently decreasing ER visits; and by increasing use of X-rays and low-cost diagnostic alternatives, thus decreasing unnecessary high-cost CT scans and MRI services (Karabatsos, 2011; State Policy Options, 2012; Treo Solutions, 2012).
Third, nonspuriousness indicates the maintenance of the relationship when all extraneous variables are held constant and rival hypotheses are eliminated (Singleton & Straits, 2010; Van de Ven, 2007). Identifying and isolating relevant extraneous antecedent and intervening variables assist in eliminating confounding effects and substantiating nonspuriousness (Parker, 1993; Van de Ven, 2007). Typically, health outcomes decline and health care costs escalate for Medicaid recipients over time while other factors vary little (Joffe, 2015). For example, the overall distribution of Colorado Medicaids recipients by age, gender, ethnicity, geography, and income fluctuated little during the years examined in this research study (CHI, 2017). As described in the Data Sample and Research Study Groups subsection earlier in this chapter, the control and ACC groups were selected from a limited data set of unduplicated Colorado Medicaid recipients similar except for ACC participation, which was determined by residence in a focus community county and a relationship with a PCMP. Similarity in all other aspects of the two research study groups would tend to support nonspuriousness. Yet, some variation is seen between the groups in gender, generation, and RCCO distribution illustrated in Table 4.2. Differences other than ACC participation could
70


indicate the presence of extraneous factors that may confound statistical test results. This is considered in the next chapter during the discussion of research findings.
Four Aspects of Validity
Different aspects of validity are typically classified in four categories: internal, statistical conclusion, construct, and external (Singleton & Straits, 2010; Van de Ven, 2007). Internal validity examines if covariance is the result of a causal relationship (Parker, 1993; Van de Ven, 2007). To a large degree, research design aspects discussed in the previous subsection, which aim to satisfy the three criteria for causal inference, address internal validity. Statistical conclusion validity is concerned with appropriate use of statistics to determine covariance between the independent and dependent variables (Shadish, Cook, & Campbell, 2002; Van de Ven, 2007). Construct validity is concerned with the generalizability of the studys findings to its theory (Shadish, 2011; Van de Ven, 2007). External validity deals with generalizability of the studys findings to its population (Singleton & Straits,
2010; Van de Ven, 2007). To highlight potential pitfalls, the following paragraphs provide examples of potential threats to validity in the four general categories as they relate to this research study.
Internal. Threats to internal validity considered for this research study included history, maturation, testing, regression, instrumentation, selection and interactions with selection, and mortality. This research studys quasi-experimental nonequivalent control group design naturally mitigated the four threats of history, maturation, testing, and regression, and the use of pre-ACC implementation and post-ACC implementation periods further controlled the threat of history and strengthened the research design (Singleton & Straits, 2010; Van de Ven, 2007). However, three remaining threats were assessed.
71


The threat of instrumentation primarily relates to the process of data collection (Singleton & Straits, 2010). As explained in the Data Source and Considerations subsection earlier in this chapter, this research study did not collect data but obtained a limited data set of historical health record and claims information for Colorado Medicaid recipients from the APCD, Colorados official repository of such data (All Payer Claims Database Council, 2017). As described in this chapters Data Sample and Research Study Groups subsection, the limited data set, which was based upon Colorado Medicaid eligibility for calendar years 2010 through 2013, was examined upon receipt to determine if files contained the information requested. Some records received contained Medicare instead of Colorado Medicaid health insurance designations, invalid or non-Colorado ZIP Codes, invalid Medicaid eligibility or service dates, and behavioral health diagnoses other than those for anxiety or depression. Records received that were not needed were then filtered from the limited data set to narrow the focus to only those Colorado Medicaid recipients of interest for this research study. Contents of some data fields were incomplete and could not be used. For example, as mentioned earlier in this chapters Research Study Group Characteristics subsection, race and ethnicity distributions were not included because individuals are not required to report race and ethnicity during their Medicaid application process. Where possible, data demographic, enrollment, and service distributions were verified against published state reports to conclude that the data were reasonable.
Next, selection problems can occur when participants self-select or are assigned to research study groups based on their preferences (Parker, 1993). Although neither selfselection nor preferential assignment occurred in this study, concern about differences in gender, generation, and RCCO distribution between the two research study groups was
72


expressed in the earlier discussion of nonspuriousness in the Three Criteria for Causal Inference subsection in this chapter. Further consideration is given to these differences during the discussion of research findings in the next chapter.
Finally, the threat of mortality refers to the loss of participants during the study and the effects of attrition on results (Parker, 1993; Singleton & Straits, 2010). However, as explained in the Data Sample and Research Study Groups subsection earlier in this chapter, records for unduplicated individuals included in the two research study groups were Medicaid recipients with no fewer than 45 of 48 possible months of eligibility during the four calendar years from 2010 through 2013; the same individuals were in the control and ACC groups at the beginning of this research study and at the end.
Statistical conclusion. Threats to statistical conclusion relate to the appropriate use of statistics to arrive at accurate decisions about accepting or rejecting research study hypotheses (Parker, 1993). These threats were lessened in this research study by careful examination of statistical procedures and the assumptions used (Shadish et al., 2002).
Mindful that statistical power is a function of the level of significance, sample size, and effect size, this research study was designed with reliable measures and sufficient power (Parker, 1993; Van de Ven, 2007). Also, participants in the two research study groups were similar, which minimized the threat of random heterogeneity (Parker, 1993).
Construct. Potential threats to construct validity in this research study included inadequate explication, mono-operation bias, and researcher expectancies (Parker, 1993; Van de Ven, 2007). Specifically defining and accurately measuring study variables minimized the first threat (Parker, 1993; Van de Ven, 2007). Measures in this research study were sufficiently defined and operationalized before data receipt and analysis so that results could
73


be accurately attributed to the constructs of interest. As mentioned previously in the Research Question, Hypothesis, and Variables section of this thesis, health outcomes and cost of health care were operationalized using multiple factors to mitigate the possibility of mono-operation bias (Parker, 1993). Also, a rigorous approach to quantitative analysis and review lessened any negative effect that researcher expectancies might have had on the interpretation of results.
External. Concern for external validity relates to generalizability of a research studys results and meaning outside its specific context (Singleton & Straits, 2010). The ability to generalize research findings across persons, settings, and time requires research study samples representative of the population of interest (Parker, 1993). This research study minimized threats to external validity by obtaining a sample representative of Colorados Medicaid population in the age group of interest, in a range of geographic settings across the state, and across time before, during, and after implementation of the ACC (Parker, 1993; Shadish et al., 2002; Van de Ven, 2007). While it is unrealistic to expect that all threats to internal, statistical conclusion, construct, and external validity could be eliminated in any research study, awareness of threats likely to occur aid in the examination and selection of strategies to improve the research design and interpretation of its results (Parker, 1993; Singleton & Straits, 2010).
This chapters explanation of this research studys design and methods sets the stage for discussion of its findings, which appear in the next chapter of this thesis. Examination of health outcomes and health care costs for specific groups of Colorado Medicaid recipients before and after the ACCs implementation contributes evidence to the performance of a
74


collaborative governance effort and the impacts on individuals when served through such a
process.
75


CHAPTER V
RESEARCH STUDY FINDINGS
This chapters discussion of findings is organized in four primary sections, one related to health outcomes, another associated with cost of health care, the third concerning results over time, and the fourth exploring regional differences. Results from statistical analysis of each dependent variable are summarized. Differences between groups in each of the three periods as well as differences within each group between the first and third periods are considered. Outcomes and explanatory tables are included, and issues related to practical interpretation of test results are raised. This chapter concludes with observations about the research studys overall findings and implications for the hypothesis.
Health Outcomes
In the Research Question, Hypothesis, and Variables section in the previous chapter, this thesis noted that health outcomes for this research study are operationalized as (a) potentially preventable hospital readmissions within 30 days of discharge, (b) ER visits, and (c) high-cost imaging services. These three utilization measures relate to health outcomes the ACC initially established as key performance indicators (HCPF, 2010b). As noted in the Health Outcomes subsection in the fourth chapter of this thesis, these measures would be expected to show a reduction over time in a coordinated health care delivery system like the ACC. Reductions would correspond with improved health outcomes.
Each of the dependent variables operationalizing health outcomes is categorical and indicates whether or not, based on the definitions provided at the beginning of the previous chapter, a hospital readmission, ER visit, or high-cost imaging service occurred. These services appear in the limited data set at the Medicaid claim level, are linked to the
76


appropriate individual by a unique deidentified member number, and are measured as the number of occurrences. It is possible that one Medicaid recipient could have more than one of these services in any period. Independence of observations exists in each group.
Frequency tables were generated. Chi-square goodness of fit results were reported based on unequal proportions, which were determined before frequency tables were produced and supported by reasonable assumptions (Hays, 1994; Healey, 1999). Pearsons chi-square test for association was used to determine if a relationship existed between participation in the ACC and potentially preventable hospital readmissions within 30 days of discharge, ER visits, and high-cost imaging (Hays, 1994; Healey, 1999). The strength of associations was assessed using the phi coefficient (Healey, 1999). For each of the three dependent variables representing health outcomes, supporting tables along with anticipated and actual results appear in the following subsections.
Hospital Readmissions
Reducing hospital readmissions that occur within 30 days of discharge is considered an appropriate measure for a collaborative health care system seeking to improve coordination across inpatient, outpatient, and follow-up services (HCPF, 2012a, 2016a). As explained in the fourth chapter of this thesis, this research study used the indicator for potentially preventable hospital readmissions within 30 days of discharge as a proxy for allcause hospital readmissions. This indicator is derived from an individuals inpatient discharge and readmission dates, diagnosis, prior admission procedures, and readmission reason (3M Health Information Systems, 2015). Potentially preventable hospital readmission rates vary due to factors such as age, health status, and number and type of chronic conditions. Relying on national research studies that have calculated reasonable averages for
77


various populations and subpopulations, the assumption was made in this research study that 20% of all hospital readmissions within 30 days of discharge would be potentially preventable for the Medicaid recipients included in this study (Agency for Healthcare Research and Quality, 2015; Donze, Aujesky, Williams, & Schnipper, 2013; Goldfield et al., 2008). As a result, 20% was used for weighting unequal proportions during statistical analysis. Additionally, the indicator only applies to inpatient claims. Outpatient and professional claims are not applicable for analysis of this variable, and they were excluded. Table 5.1 illustrates the frequency of potentially preventable hospital readmissions within 30 days of discharge compared to other hospital readmissions within the same amount of time in both groups for each of the three periods.
Table 5.1: Potentially Preventable Hospital Readmissions (PPHR)
Ti........................Ti......................Ts
pp HR Control Group Number (%) ACC Group Number (%) Total Number (%) Control Group Number (%) ACC Group Number (%) Total Number (%) Control Group Number (%) ACC Group Number (%) Total Number (%)
Yes 2 151 153 8 136 144 5 122 127
(2) (12) (11) (15) (10) (10) (6) (9) (9)
No 79 1,152 1,231 54 1,280 1,334 82 1,178 1,260
(98) (88) (89) (87) (90) (90) (94) (91) (91)
Total 81 1,303 1,384 62 1,416 1,478 87 1,300 1,387
Q0Q) Q0Q) QOQ) (Ml) QOQ) QOQ) QOQ) QOQ) QOQ)
If the control and ACC groups were similar except for ACC participation, no significant difference would be expected between the groups in Ti, but the ACC groups potentially preventable hospital readmissions would be expected to begin decreasing during T2 and continue decreasing during T3 due to improved RCCO and provider coordination across inpatient, outpatient, and follow-up services (HCPF, 2012a, 2016a). A similar decrease
78


would not be expected in the control group since its participants would not be subject to care coordination provided by the ACC. Table 5.2 summarizes the results for each period.
Table 5.2: ACC Participation and PPHR
Measure Ti Ti Ts
Chi-square Value 69.213 97.186 101.929
Chi-square Significanceab .000 .000 .000
Pearson Value 6.450 .735 1.297
Pearson Significance13 .011 .391 .255
Phi Value .068 -.022 .031
Phi Significance13 .011 .391 .255
aWhen p < .05, observed frequencies in the sample are not the same as expected frequencies in the population; minimum PPHR observed were fewer than expected. bWhen p < .05, significant difference exists.
However, as Table 5.2 illustrates, a significant difference in potentially preventable hospital readmissions existed between the control and ACC groups in Ti before the ACCs implementation, but no significant difference appeared to exist between the groups during the first two years of the ACCs operation. As a result, no inference can be drawn that the change was attributable to ACC participation. Additionally, the strength of the relationship was negligible in all three periods although it was significant in Ti.
ER Visits
The second key performance indicator the ACC selected to measure health outcomes was ER visits not resulting in a hospital admission (HCPF, 2010a, 2010b, 2012a). Before the ACCs implementation, the majority of Colorados Medicaid recipients accessed health care through a fee-for-service system that supported episodic rather than coordinated service delivery, which led many recipients to ER rather than primary care provider visits (HCPF, 2010a, 2010b, 2012a). For reporting based on unequal proportions, it is reasonable to
79


presume that 40% of all medical service claims would be ER visits (Centers for Disease Control and Prevention, 2016; CHI, 2012).
The limited data set utilized for this research study contains service dates and an indicator for ER visits. The categorical indicator appears at the claim level and specifies whether or not the claim was an ER visit fitting the criteria in the previous paragraph. Claims are linked to the appropriate individual by a unique deidentified member number and are measured as the number of occurrences. It is possible that one Medicaid recipient could have more than one ER visit in any period. Table 5.3 reflects the number of ER visits compared to other inpatient, outpatient, and professional medical services within the same period for both groups in Ti, T2, and T3.
Table 5.3: ER Visits
ER Visits Control Group Number (%) Ti ACC Group Number (%) Total Number (%) Control Group Number (%) t2 ACC Group Number (%) Total Number (%) Control Group Number (%) t3 ACC Group Number (%) Total Number (%)
Yes 502 5,924 6,426 476 6,029 6,505 490 5,888 6,378
(7) (8) (8) (6) (8) (7) (6) (7) (7)
No 6,922 65,744 72,666 7,278 73,150 80,428 7,817 73,932 81,749
(93) (92) (92) (94) (92) (93) (94) (93) (93)
Total 7,424 71,668 79,092 7,754 79,179 86,933 8,307 79,820 88,127
Q0Q) QOO) QOO) QOQ) (Ml) QOO) Q00) QOQ) Q00)
No significant difference between the groups would be expected in Ti if the groups
were similar except for ACC participation. Compared to ER visits in the control group, ER
visits in the ACC group would be expected to decrease during T2 and continue decreasing during T3 due to strengthened PCMP relationships and extended evening and weekend PCMP
hours of operation (HCPF, 2010a, 2012a). Statistical test results in Table 5.4 indicate a
significant difference in ER visits between the groups in Ti, T2, and T3. Because a significant
80


difference was present in Ti, which was prior to the ACCs implementation, significant differences in T2 and T3 cannot be attributed solely to the ACC. Also, Table 5.4 indicates the relationships strength was negligible but significant in all three periods.
Table 5.4: ACC Participation and ER Visits
Measure Ti Ti T3
Chi-square Value 33,483.39 38,300.14 39,414.41
Chi-square Significanceab .000 .000 .000
Pearson Value 20.386 22.214 24.481
Pearson Significance13 .000 .000 .000
Phi Value .016 .016 .017
Phi Significance13 .000 .000 .000
aWhen p < .05, observed frequencies in the sample are not the same as expected frequencies in the population; minimum ER visits observed were fewer than expected. bWhen p < .05, significant difference exists.
High-cost Imaging
The third key performance indicator measuring health outcomes in the ACC was high-cost imaging, which was defined as CT scans and MRI services (HCPF, 2010a, 2010b, 2012a). Used appropriately, alternatives such as X-rays and ultrasound prove to be as efficacious and are more cost effective than CT scans and MRI services (HCPF, 2012a; Mayo Clinic, 2017c; Treo Solutions, 2012). Reporting based on unequal proportions used a reasonable assumption that 35% of all diagnostic imaging services would be CT scans or MRI services (Regents Health Resources, 2012; Smith-Bindman, Miglioretti, & Larson, 2008).
Each claim record in the limited data set used for this research study contains a field with the CPT code for the medical service procedure performed. CPT codes for diagnostic imaging procedures range from 70010 through 76499 (American Medical Association,
81


2017). Claims for services with those specific codes were filtered from the remainder of the medical services in the data. Appendix E contains the list of specific CT scan and MRI service CPT codes used to identify high-cost imaging services in the ACC. An indicator was created in the data set for this research study to distinguish CT scans and MRI services from other diagnostic imaging procedures. Table 5.5 includes the frequency of high-cost imaging services compared to other diagnostic services for the control and ACC groups for each of the three periods. The categorical indicator for high-cost imaging services appears at the claim level and specifies whether or not the service fit the CPT code criteria outlined above. Claims are linked to the appropriate individual by a unique deidentified member number and are measured as the number of occurrences. It is possible that one Medicaid recipient could have more than one high-cost imaging service in any period.
Table 5.5: High-cost Imaging
CT and MRI Control Group Number (%) Ti ACC Group Number (%) Total Number (%) Control Group Number (%) T2 ACC Group Number (%) Total Number (%) Control Group Number (%) t3 ACC Group Number (%) Total Number (%)
Yes 388 4,060 4,448 255 2,732 2,987 201 2,545 2,746
(39) (32) (33) (26) (22) (22) (19) (20) (20)
No 619 8,438 9,058 713 9,875 10,589 879 9,940 10,819
(61) (68) (67) (74) (78) (78) (81) (80) (80)
Total 1,007 12,498 13,505 968 12,607 13,575 1,080 12,485 13,565
(Ml) (Ml) QM) Q0Q) Q0Q) (100) Q0Q) Q0Q) (MM
If the control and ACC groups were similar except for ACC participation, no
significant difference between the groups would be anticipated in high-cost imaging services
in Ti. The ACC groups utilization would be expected to decrease during T2 with continued
decrease during T3 due to an increase in the alternative use of X-rays and ultrasound and
82


possibly a decrease in diagnostic testing due to overall health improvement (HCPF, 2012a;
Mayo Clinic, 2017c; Treo Solutions, 2012). Table 5.6 summarizes the results. Table 5.6: ACC Participation and High-cost Imaging
Measure Ti Ti T3
Chi-square Value 25.290 1,007.857 1,298.430
Chi-square Significanceab .000 .000 .000
Pearson Value 15.418 11.436 1.936
Pearson Significance13 .000 .001 .164
Phi Value .034 -.029 .012
Phi Significance13 .000 .001 .164
aWhen p < .05, observed frequencies in the sample are not the same as expected frequencies in the population; minimum high-cost imaging services observed were fewer than expected. bWhen p < .05, significant difference exists.
Statistical tests indicated a significant difference in high-cost imaging utilization between both groups in Ti and T2 but not in T3. Because a significant difference was present in Ti before the ACCs implementation, the differences significance in T2 could not be attributed exclusively to the ACC. Although the strength of the relationship was negligible in all three periods, it was significant in Ti and T2.
In reviewing health outcome results between the two groups in each of the three periods, insufficient evidence exists to reject the research studys null hypothesis and attribute changes in outcome indicators solely to participation in the ACC. Results between groups were significantly different in potentially preventable hospitalization readmissions only in the period before the ACCs implementation. ER visits were significantly different in all three periods, and high-cost imaging was significantly different in the first two periods. The magnitude of the differences was negligible. Outcomes suggest other influences may be responsible. For example, dissimilarities may exist between the groups in demographic
83


characteristics or health status, or a statewide focus on health improvement and health initiatives in Colorado may have influenced not only ACC participants but also other Medicaid recipients. Before addressing results over time and regional differences, the next section examines differences in cost of health care between the control and ACC groups in each period.
Cost of Health Care
As discussed in the Research Question, Hypothesis, and Variables section in the previous chapter of this thesis, cost of health care is operationalized in this research study as (a) the cost of claims paid by Medicaid for medical services, (b) the cost of claims paid by Medicaid for prescription drugs, and (c) PMPM fees paid to RCCOs and PCMPs in T2 and T3.
Each of the dependent variables operationalizing cost of health care is continuous, and independence of observations exists in each group. Frequency tables were generated. The independent-samples t-test was used to determine if a difference existed between the mean or average of each continuous dependent variable of interest in the control and ACC groups (Hays, 1994). The mean difference was examined for significance, and 95% confidence intervals were generated to confirm the reasonableness of the mean differences for each variable in each period (Hays, 1994; Healey, 1999). Equality or homogeneity of variances in the variables for the two groups was tested. Results from Levenes test were used when homogeneity existed, and results from Satterthwaites method were used in the absence of homogeneity (Hays, 1994; Institute for Digital Research and Education, 2017). Quantile-quantile and box plots were generated to conduct a visual inspection for outliers, and statistical normality of the distributions was determined using the Kolmogorov-Smirnov test
84


Full Text

PAGE 1

FOR COORDINATED HEALTH CARE FOR MEDICAID RECIPIENTS by TERI GARLAND BOLINGER B.A. (Hons), Midwestern State University, 1976 M.S. (Hons), Nova Southeastern University, 1995 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Doctor of Philosophy Public Affairs Program 2017

PAGE 2

ii 2017 TERI GARLAND BOLINGER ALL RIGHTS RESERVED

PAGE 3

iii This thesis for the Doctor of Philosophy degree by Teri Garland Bolinger has been approved for the Public Affairs Program by Danielle Varda Chair Tanya Heikkila, Advisor Brian Gerber Rene Horton Date: December 16, 2017

PAGE 4

iv Bolinger, Teri Garland (Ph.D., Public Affairs Program ) Recipients Thesis directed by Professor Tanya Heikkila ABSTRACT This thesis provides an overview of the Medicaid health insurance program to illustrate the significance of finding new approaches to the growing problem of increased enrollment and unsustainable expenditures It hig hlights a need for collaboration among s tate and local resources to better serve Medicaid recipients and draws from col laborative go vernance literature to identify common characteristics of collaboration s This thesis tive recipient s. By examining actual health outcomes and health care cost Medicaid recipient s before and after the ACC this thesi s contributes evidence to the performance of a collaborative governance effort a nd the impacts on individuals served This thesis concludes by discussing implications and recommendations for collaborative governance literature and theory, policy and practi ce, and future areas of research The form and content of this abstract are approved. I recommend its publication. Approved: Tanya Heikkila

PAGE 5

v DEDICATION In honor of my beloved daughter, Lauran Bolinger, and my dear mentor, Barbara Bowles, a nd in loving memory of my parents, Mr. and Mrs. T. K. Garland, my brother, W. T. Garland, my sister, Paula Anne Hatcher and my inspir ing friend, Ruthie Swanson

PAGE 6

vi ACKNOWLEDGEMENTS Something a stranger shared with me s everal months ago merits acknowledgement. She said she finally discovered the difference between failure and success. Failure: fall down seven times, get up six. Success: fall down seven times, get up eight. I gratefully acknowledge a faith that continues to sus tain me at all times and in all circumstances and I appreciate that I could not have completed th is journey alone. I acknowledge those at the University of Colorado Denver School of Public Affairs who made my experience a positive and productive one. Wor ds inadequately express my appreciation for the unfailing encouragement guidance, and wisdom my advisor Tanya Heikkila, extended to me throughout this learning process. I would also like to thank the rest of my dissertation committee, Danielle Varda, Bri an Gerber, and Rene Horton, who brought their valued and diverse perspectives and insight to this research. Appreciating the entire faculty and staff who m ade my experience at the School of Public Affairs a positive and productive one I extend special tha nks to Rob Drouillard, Todd Ely, Mary Guy, Christine Martell, Dawn Savage, and Chris Weible I am particularly thankful for the support and friendship of my cohort members, Kelley Harp and Lucy Mermagen whose intelligence, wit, and persistence never fa lte red I am especially grateful to Roger Hartley and Laura Wilson Gentry, University of Baltimore College of Public Affairs, and Mason Paris, University of Baltimore Office of Technology, for generously providing an academically welcoming and technologicall y secure environment for me the past two years. I would not have been able to complete my research successfully without their gracious efforts on my behalf.

PAGE 7

vii I particularly recognize the importance of family, friends, and colleagues who daily influence my personal and professional development to a greater extent than they realize. Each of them unconsciously set s an example worth following by invariably saying or doing the perfect thing at the right time I extend heartfelt love and gratitude t o all of my family and specifically acknowledge Bayley Garland Gloria Garland Tim Garland, Tom Garland, Truitt Garland, Tina Sanders, Tracee Spore, Sheri Sutton, Julie Yandell, and Keegan Yandell I am also thankful for priceless gifts of inspiration fro m Julie Farrar, Francesca Maes, and Jose Torres Vega. I appreciat e the treasures of unconditional friendship and boundless encouragement from Susan Barad, Bill Brewer, Lorri Cluckey, Todd Coffey, Kim Crow, Mary Dwyer, Lois Munson, Suzanne Smith, Drew Strou ble, and Sandy Zannino. I remain honored to be part of the Federal Coordinated Health Care Office and the collaborative and ethical environment it cultivate s and sustain s I sincerely respect and a ppreciate my colleagues and am especially grateful for the leadership of Lindsay Barnette, Kerry Branick, Sharon Donovan, Vanessa Duran, Tim Engelhardt, and Sara Vitolo.

PAGE 8

viii TABLE OF CONTENTS CHAPTER I. ................................ ................................ ........................ 1 Birth of Medicare and Medicaid ................................ ................................ ......................... 2 Medicaid Design Characteristics ................................ ................................ ........................ 5 An Escalating Sense of Urgency ................................ ................................ ....................... 10 II. WHY COLLABORATION? ................................ ................................ ............................ 14 From Governance to Collaborative Governance ................................ .............................. 14 Six Defining Criteria ................................ ................................ ................................ ......... 17 Forum Initiated by Public Agencies or Institutions ................................ .................... 18 Focus on Public Policy or Public Management Issue ................................ ................. 18 Formally Organized and Collectively Convened Forum ................................ ............ 19 Nonstate Stakeholders as Active Participants ................................ ............................. 19 Forum Participants Involved in Decision Making ................................ ...................... 20 Consensus oriented Decision making Process ................................ ........................... 20 Four Common Elements ................................ ................................ ................................ ... 20 Starting Conditions ................................ ................................ ................................ ..... 21 Institut ional Design ................................ ................................ ................................ ..... 22 Facilitative Leadership ................................ ................................ ................................ 24 Collaborative Process ................................ ................................ ................................ .. 26 Model Considerations ................................ ................................ ................................ ....... 28 III. COL ................................ ................................ .......................... 32 The ACC and Six Defining Criteria ................................ ................................ .................. 33

PAGE 9

ix The ACC and Four Common Elements ................................ ................................ ............ 34 Starting Conditions ................................ ................................ ................................ ..... 35 Institutional Design ................................ ................................ ................................ ..... 39 Facilitative Leadership ................................ ................................ ................................ 44 Collaborative Process ................................ ................................ ................................ .. 47 Implic ations for Successful Collaboration ................................ ................................ ........ 49 Implications for Starting Conditions ................................ ................................ ........... 50 Implications for Facilitative Leadership ................................ ................................ ..... 52 Implications for Collaborative Process ................................ ................................ ....... 52 From Theory to Practice ................................ ................................ ................................ ... 54 IV. RESEARCH DESIGN AND METHODS ................................ ................................ ........ 56 Research Question, Hypothesis, and Variables ................................ ................................ 57 Health Outcomes ................................ ................................ ................................ ......... 58 Cost of Health Care ................................ ................................ ................................ ..... 59 Data and Analysis ................................ ................................ ................................ ............. 60 Data Source and Considerations ................................ ................................ ................. 60 Data Sample and Research Study Groups ................................ ................................ .. 61 Research Study Group Characteristics ................................ ................................ ........ 64 Approach to Analysis ................................ ................................ ................................ .. 67 Limitations and Strategies ................................ ................................ ................................ 68 Th ree Criteria for Causal Inference ................................ ................................ ............ 69 Four Aspects of Validity ................................ ................................ ............................. 71 V. RESEARCH STUDY FINDINGS ................................ ................................ .................... 76

PAGE 10

x Health Outcomes ................................ ................................ ................................ ............... 76 Hospital Rea dmissions ................................ ................................ ................................ 77 ER Visits ................................ ................................ ................................ ..................... 79 High cost Imaging ................................ ................................ ................................ ...... 81 Cost of He alth Care ................................ ................................ ................................ ........... 84 Medical Services ................................ ................................ ................................ ......... 85 Prescription Drugs ................................ ................................ ................................ ...... 87 PMPM ................................ ................................ ................................ ......................... 89 Total Cost of Health Care ................................ ................................ ........................... 90 Re sults over Time ................................ ................................ ................................ ............. 92 Control Group Results ................................ ................................ ................................ 92 ACC Group Results ................................ ................................ ................................ .... 95 Regional Differences ................................ ................................ ................................ ........ 98 Health O utcomes among RCCOs ................................ ................................ ............... 99 Cost of Health Care among RCCOs ................................ ................................ ......... 101 Observations ................................ ................................ ................................ ................... 106 VI. IMPLICATIONS AND RECOMMENDATIONS ................................ ......................... 108 Literature and Theory ................................ ................................ ................................ ..... 108 ACC Process Performance ................................ ................................ ........................ 109 ACC Productivity Performance ................................ ................................ ................ 110 Policy and Practice ................................ ................................ ................................ .......... 112 Approach to Analysis ................................ ................................ ................................ 112 Data Sources ................................ ................................ ................................ ............. 115

PAGE 11

xi Prescription Drugs ................................ ................................ ................................ .... 116 Contingencies for the Future ................................ ................................ ..................... 117 Future Research ................................ ................................ ................................ .............. 119 Individual and Multiple RCCOs ................................ ................................ ............... 119 Expanded Timeframe ................................ ................................ ................................ 121 Medicare Medicaid Subpopulation ................................ ................................ ........... 122 Multivariate Analysis ................................ ................................ ................................ 124 Conclusion ................................ ................................ ................................ ...................... 126 REFERENCES ................................ ................................ ................................ ..................... 128 APPENDIX A. Table of Colorado ACC RCCO Service Regions ................................ ..................... 151 B. Map of Colorado ACC RCCO Service Regions ................................ ....................... 152 C. Organization Of ACC RCCOs ................................ ................................ .................. 153 D. RCCO Relationships and Community Involvement ................................ ................. 154 E. High cost Imaging Services ................................ ................................ ...................... 158 F. Behavioral Health Diagnosis Codes ................................ ................................ ......... 159 G. Characteristics of Control and ACC Groups ................................ ............................ 161 H. ER Visit Comparisons among RCCOs ................................ ................................ ..... 163 I. High cost Imaging Comparisons among RCCOs ................................ ..................... 166 J. Medical Service Comparisons among RCCOs ................................ ..................... 169 K. Prescription Drug Comparisons among RCCOs ................................ ................... 170

PAGE 12

xii LIST OF TABLES TABLE 2.1 Ansell and Gash Elements and Co mponents of Collaborative Governance .......... 21 3.1 Ansell and Gash Elements and Co mponents and ACC C haracteristics ................ 35 3.2 Ansell and Gash Implications for Success ................................ ............................. 50 4.1 Criteria for Research S tudy Groups ................................ ................................ ....... 64 4.2 Summarized Characteristics of Control and ACC Groups ................................ .... 65 5.1 Potentially Preventable Hospital Readmissions (PPHR) ................................ ....... 7 8 5.2 ACC Participation and PPHR ................................ ................................ ................ 7 9 5.3 ER Visits ................................ ................................ ................................ ................ 80 5.4 ACC Participation and ER Visits ................................ ................................ ........... 8 1 5.5 High cost Imaging ................................ ................................ ................................ 8 2 5.6 ACC Participation and High cost Imaging ................................ ............................ 8 3 5.7 Medical Service Claims ................................ ................................ ......................... 8 6 5.8 ACC Participation and Cost of Medical Service Claims ................................ ....... 8 6 5.9 Prescription Drug Claims ................................ ................................ ....................... 8 7 5.10 ACC Participation and Cost of Prescription Drug Claims ................................ ..... 8 8 5.11 ACC Group PMPM Calculations ................................ ................................ .......... 8 9 5.12 Total Cost of Health Care ................................ ................................ ...................... 90 5.13 Total Cost of Health Care Differences between Groups ................................ ..... 9 1 5.14 Control Group Health Outcomes Differences between Periods .......................... 9 3 5.15 Control Group Cost of Health Care Differences between Periods ...................... 9 4 5.16 ACC Group Health Outcomes Differences between Periods .............................. 9 6

PAGE 13

xiii 5.17 ACC Group Cost of Health Care Differences between Periods ......................... 9 7 5.18 ER Visits among RCCOs ................................ ................................ ..................... 100 5.19 High cost Imaging among RCCOs ................................ ................................ ...... 10 1 5.20 Control Group Medical Service Claims among RCCOs ................................ ..... 10 3 5.21 ACC Group Medical Service Claims among RCCOs ................................ ......... 10 3 5.22 Control Group Prescription Drug Claims among RCCOs ................................ ... 10 4 5.23 ACC Group Prescription Drug Claims among RCCOs ................................ ....... 10 5 5.24 Summary of Differences ................................ ................................ ...................... 10 7

PAGE 14

xiv LIST OF FIGURES FIGURE 1.1 Colorado Medicaid Enrollment from 1995 to 2010 (in thousands) ....................... 11 1.2 Colorado Medicaid Annual Expenditures from 199 5 to 2010 (in billions) ........... 1 2 3.1 Structure of the ACC ................................ ................................ ............................. 42

PAGE 15

xv LIST OF ABBREVIATIONS ABBREVIATION ACC Accountable Care Collaborative APCD All Payer Claims Database BHO Behavioral Health Organization CIVHC Center for Improving Value in Health Care CMS Centers for Medicare & Medicaid Services CPT Current procedural terminology CT Computed tomography DSH Disproportionate share hospital (payments) EPSDT Early and periodic screening, diagnosis, and treatment ER Em ergency room FMAP Federal medical assistance percentage FPL Federal poverty level HMO Health maintenance organization IGT Intergovernmental transfer MMIS Medicaid Management Information System MRI Magnetic resonance imaging PCMP Primary Care Medical p rovider PMPM Per member per month PPHR Potentially preventable hospital readmissions RAE Regional Accountable Entity RCCO Regional Care Collaborative Organization

PAGE 16

xvi SDAC Statewide Data and Analytics Contractor UPL Upper payment limit (supplements)

PAGE 17

1 CHAPTER I Viewed through the lens of collaborative governance, p ublic management and health policy offer a broad and fruitful field of study Narrowing the research focus, this thesis Collaborative (ACC) a model the state designed to serve recipients of its Medicaid health insurance program The first chapter of this t hesis lays the foundation for the obj ect of research by describing Medicaid While the second chapter grounds the object ACC theoretically fits the collabor ative governance criteria A nsell and Gash outlined in their 2008 meta analysis of 137 diverse cases. Satisfied the ACC is a collaborative governance example designed with opportunity to succeed, t he fourth chapter of this thesis introduces the research question: What effects will pa rticipation in the ACC have on health care for Medicaid recipients in Colorado? The fourth chapter continues by describ ing the research design and methods used to examine health outcomes and health care costs for specific pients This thesis presents the research findings in the fifth chapter and concludes with a discussion of implications and recommendations for collaborative governance literature and theory, policy and practice, a nd future areas of research in the sixth chapter This chapter continues by summarizing the significance of the joint federal and state Medicaid health insurance program in three subsections. It describes the birth of Medicare and Medicaid, two of the largest health insurance program s in the United States; outlines

PAGE 18

2 esign characteristics; and illustrates the basis for an escalating sense of national and state urgency associated with Medicaid Birth of Medicare and Medicaid In 1965, t he United States Congress overwhelmingly approved Social Security Act amend ments Titles XVIII and XIX, to establish Medicare and Medicaid, respectively, as part ( Iglehart & Sommers 2015; Starr, 2015; Stevens, 1996) Medicare and Medicaid sought to provide health insurance to two different subpopulations in the United States : the elderly and the poor (Centers for Medicare & Medicaid Services [CMS], 2015b ) Designed to uphold the status quo of private insurance for working individuals, Medicare focused coverage on the retired elderly who could no longer find or afford health insurance (Stevens, 1996). The right to health insurance became a natural extension earned by contributions made during the years they worked (Starr, 2015). Aim ed to finance health care services for the poorest individuals in the United States, Medicaid focused coverage on recipients of f inancial assis tance through welfare programs : those who were aged, blind, or disabled or were families with dependent children ( CMS, 2015b ; Hansen, 2013 ; Starr, 2015 ). Medicaid to furnish acute health care services to public assistan ce recipients based on their income and resources or means ( U nited States Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation [ASPE], 2005). Federal and state programs with eligibility based on personal or family means exist in the United States to provide food, housing, medical care, and social services to those who d o not have the ability to get and pay for services on their own (Means tested, n.d.; Rector & Sheffield, 2016). The difference in initial

PAGE 19

3 focus contributed to a perceived social dichotomy between the deserving elderly and the less d eserving indigent ( Hudman & Starfield, 1999; Stevens, 1996). d esign and financing also differed. As a national program, Medicare provided the same hospital ( Part A) and physician ( Part B) benefits to elderly recipients wherever they lived (Starr, 2015). The f inancing for Medicare Part A came from payroll taxes and Part B came from general federal revenue and premiums paid by recipients ( Medicare, n.d.; Starr, 2015). Unlike Medicare, Medicaid was a program chiefly administered by states an d included flexibility in program design ( ASPE, 2005; CMS, 2017 d ; Mann & Westmoreland, 2004 ; Social Security Act, Title XIX, 1965 ). Medicaid financing came from general state reve nues matched in part by general federal revenue ( ASPE, 2005; Patrick & Freed, 2012 ; Starr, 2015 ). In the first year of operation, Medicare enrollment was 19 million with expenditures of $3.4 billion and Medicaid enrollment was 4 million with expenditures less than $1 billion (Klemm, 2000; Pear, 1987). arly cos ts were seriously underestimated. For the first year projected costs nationwide were $250 million, but actual costs in the state of New York alone reached that amount (Mills, 1987; Starr, 2015). Although different in focus, design, and financing, b oth Medicare and Medicaid originally favored hospital based services over public health and preventive care; promoted medical and surgical specialties instead of primary care; and operated as fee for service programs by making a separate payment to a healt h care provider for every service delivered (Fee for service, n.d.; Starr, 2015). As a result, both encouraged a volume based rather than value based delivery system (Starr, 2015). Medicare and Medicaid enrollment and costs

PAGE 20

4 continued to increase, and both programs expanded after their inception by including other subpopulations and additional services (CMS, 2015b ) Several expansions directly affected Medicaid A n early and periodic screening, diagnosis, and treatment (EPSDT) benefit was created in 1967 for all recipients who were children and eligibility was linked to the federal Supplemental Security I ncome program enacted in 1972 for state residents who were elderly or blind or who had a disability (CMS, 2015b ). Some f ederal regulations were waived in 1981 that allowed states to require recipients to get services fr om a limited set of providers and permitted recipients to get services in home and community based settings as an alternative to institutional settings (CMS, 2015b ; Nathan, 2005; Wh itenhill & Shugarman, 2011). In contrast to the original fee for service approach, these managed care concepts introduced a health care delivery system intention of reducing costs and improving care quality (CMS, 2015b ; Managed care, n.d.; Starr, 2015). A n option was created in 1986, which became a mandate in 1988, for states to provide coverage to pregnant women and infants up to 100% of the federal poverty level (FPL) (CMS, 2015b ). In 1989, the coverage threshold for pregnant women and children under six years of age was expanded to 133% of the FPL, and EPSDT requirements were established (CMS, 2015b ). Phased in coverage of children between six and 18 years of age and a prescription drug rebate program were established in 1990 and a new low income group not linked to public assistance was mandated in 1996 (CMS, 2015b ). T he Balanced ed new managed care options and requirements for states (CMS, 2015b ; Starr, 2015 ) T he Ticket to

PAGE 21

5 Work and Work Incentives Improvement Act of 1999 expanded the availability of coverage for certain recipients with disabilities who returned to work established optional eligibility groups and allowed states to offer a buy in program for working age individuals with disabilities. Medicaid Design Ch aracteristics In addition to distinct subpopulation and service expansion s describ ed in the previous sub section, original design ( ASPE, 2005; Iglehart & Sommers, 2015; Starr, 2015; Stevens, 1996) Medicaid operates in a dynamic environment where federal and state government authorities share power (Iglehart & Sommers, 2015). As esign f lexibilit y contributes not only to challenges it may encounter but also to responses it can develop ( Iglehart & Sommers, 2015; Nathan, 2005; Weil, Wiener, & Holahan, 1998). Medicaid eligibility link to receipt of financial assistance th rough public assistance programs reinforc ed a disincentive to work. Low paying jobs available to public assistance recipients typically lacked benefit s such as health insuranc e coverage, and individuals who accepted such jobs would then lose Medicaid coverage ( Iglehart & Sommers, 2015; Starr, 2015) As a result, t hose with chronic health conditions or those with a sick child, for example, had a strong incentive not to work to maintain their health insurance coverage through Medicaid (Moffitt & Wolfe, 1992) Although state and federal authorities expanded Medicaid eligibility to other groups of low income individuals in the 1980s and 1990s, the incentive not to work effectively remained in place until the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 uncoupled Medicaid

PAGE 22

6 eligibility from receipt of public assistance (Hudman & Starfield, 1999; Iglehart & Sommers, 2015; Social Securi ty Administration, 1996 ). State flexibility conc erning Medicaid eligibility, services and provider rates presented o ther challenge s For example, a person eligible for Medicaid in one state might not be eligible in another, and services one state provided could differ in scope, quantity, and duration from services another state offered (Klees, Wolfe, & Curtis, 2010) In addition, a state could change its Medicaid eligibility criteria, services, or provider reimbursement rates at any time (Klees et al., 2010). Although a state agreeing to operate a Medicaid program was requir ed to furnish a basic set of health services to all recipients of public assistance a state could receive additional federal funds if it also covered those in eligible categorie s with incomes up to 133% cutoff point ( ASPE, 2005; Starr, 2015 ). thre shold for Medicaid eligibility was linked to the configuration of its sp ecific services. Each state varied not only in its public assistance criteria but also in its willingness to cover other individuals among the poor (Starr, 2015) A state also could opt to offer additional services to medically needy individuals who received no public assistance (ASPE, 2005). The type and number of individuals a state included in its Medicaid eligibility parameters influenced service categories ; the sc ope, quantity and duration of services offered ; and provider payment rates ( Klees et al. 2010; Rosenbaum, 2011; U nited States Department of Health and Human Services, Health Care Financing Administration, 2000). B ecause states historically have set Medic aid provider rates lower than Medicare rates, diminished provider participation emerged as another challenge (Ku, 2000; Rosenbaum, 2011; Starr, 2015).

PAGE 23

7 Financing Medicaid proved to be as challenging for sta tes as balancing eligibility, services and provi der rates The federal government paid states for at least half of the costs of providing Medicaid services to recipients ; for every dollar a state spen t on Medicaid services, t he federal government reimbursed the state no fewer than 50 cents (ASPE, 2005; Iglehart & Sommers, 2015) Each state medical assistance percentage (FMAP) is average income per person to the national average ; the FMAP up to a maximu m of 83% ( ASPE, 2005; Iglehart & Sommers, 2015). F ederal funds provided a means by which to increase the value of Medicaid investments, and states regularly adopted strategies to maximize federal support (Klemm, 2000). Based on the premise that demand for social programs increases in periods of economic decline, social programs were designed to compensate for the negative effects of such cycles (Carboni & Milward, 2012). However, as a joint federal and state endeavor, Medicaid was vulnerable to cutbacks dur ing economic downturns when state revenues also decreased (Starr, 2015). During such periods Medicaid enrollment typically increase d because unemployed workers lost employer based health insurance cove rage and family income decreased to the Medicaid threshold (Iglehart & Sommers, 2015). As revenues fell and health care costs rose, states increasingly sought to curb Medicaid services, control costs, and leverage financing opportunities ( Medicaid and CHIP Payment and Access Commission [M ACPAC ], 2012b; Patrick & Freed, 2012; Sommers & Epstein, 2011). S ince the early 1980s, s tates often have use d two available financing strategies to maximize federal support for Medicaid : disproportionate hospital share (DSH) payments and upper payment li mit (UPL) supplements (Coughlin, Bruen, & King, 2004; MACPAC 2012 a

PAGE 24

8 2012b ). T he DSH payment policy was crafted to allow s upplemental payments to hospitals that served a disproportionate share o f low income patients and was limited to the actual cost of uncompensated services hospitals provided to Medicaid recipients and uninsured i ndividuals (MACPAC, 2012b) Although the purpose was to improve th e financial stability of safety net hospitals providing services to low income patients, broad federal guideli nes provided states considerable leeway in determining which hospitals would receive DSH payments and in what amounts (MACPAC, 2012b). The UPL policy allowed states to receive supplemental federal funds for Medicaid fee f or service payments up to what woul d have been paid under Medicare; th e policy applied to targeted groups of providers such as hospitals, nursing facilities, intermediate care facilities for those with intellectual disabilities, and freestanding non hospital clinics ( Coughlin et al., 2004; Ku, 2000; MACPAC, 2012a). Because Medicaid provider payments traditionally were lo wer than Medicare payments, considerable potential existed for states t o get additional federal dollars with the intention of directing UPL supplemental payments to providers (Coughlin et al., 2004 ; MACPAC, 2012a ). S tates soon began to shift the balance in Medicaid spending by us ing ambiguity in the federal DSH and UPL regulations to transfer funds from local to state government or between state agencies instead of to provide rs directly (Coughlin et al., 2004; Ku, 2000). Such intergovernmental transfers (IGTs) utilized legal but convoluted accounting practices that allowed states to increase Medicaid federal matching payments without expending additional state funds For example, a state would make allowable payment s regular Medicaid reimbursement rate to a select group of nursing homes or hospitals which were usually owned by a county or other local government en tity; the facilities would return

PAGE 25

9 a d agency through IGTs; and the state would claim a Medicaid federal match for the supplemental payments it made to the facilities ( Coughlin et al., 2004; Ku, 2000; MACPAC, 2012b). Federal fi nancing of UPL payments alone rose from $313 million in 1995 to $1.4 billion in 1998 (Ku, 2000). By the time r egulations were refined and additional requirements were imposed in the late 1990s and early 2000s, explosive growth in federal Medicaid expenditu res f or state DSH payments and UPL supplements already had occurred (Ku, 2000 ; MACPAC, 2012b ). In 34 states surveyed, r esearchers estimated the effective federal match rate in 200 1 was an average of three percentage points higher than it would have been wi thout state DSH and UPL practices (Coughlin et al., 2004). Some safety net hospitals and providers serving Medicaid recipients directly receive and retain DSH payments and UPL supplements as intended, but some do not Although some states may still legally redirect these federal resources for other purposes such as balancing budgets or cutting taxes, such uses violate the original intent of DSH payments and UPL supplements and distort Medicaid s pending ( Coughlin et al., 2004; Ku, 2000). These practic es prompt t he effective federal share of Medicaid spending to rise without a commensurate increase in the state share of funds (Coughlin et al., 2004). Additionally, w hen states use gains from DSH payments and UPL supplements to finance Medicaid services o ther than those for which the payments and supplements are intended, spending patterns among Medicaid service categories become clouded (Coughlin et al., 2004; Ku, 2000). These practices can create confusion in policy debates over such topics as Medicaid e xpansion, which subpopulations of Medicaid recipients to serv e, which services to offer, and the amount of provider rates. Coupled with its historical evolution eligibility

PAGE 26

10 parameters, service composition, provider rates, and financing opportun ities contribute to its increasing significance health care landscape A n Escalating Sense of Urgency While Medicaid represented less than five percent of means tested program spending in 1966, it represented 30% in 1972 and 40% by 1985 ( Means tested n.d.; Starr, 2015). From 1991 to 2005, the total number of Medicaid recipient s incre ased from 32 to 58 million, and total annual expenditures rose from $110 to $273 billion (Patrick & Freed, 2012). By 2008, Medicaid enrollment included 60 million individuals with expenditures of $330 billion (Jacobson, Neuman, & Damico, 2012 ). Enrollment and costs continued to climb. Reporting for fiscal year 2010 reflected Medicaid served more than 66 million individuals, which was more than one fifth (Kaiser Family Foundation, 2014; The State Health Care Spending Project, 2014 ). At inception in 1965, health programs received six percent of all federal funds to state and local governments, but by 2010, health rela ted act ivities accounted for 58% of all federal funding to state and local governments (Iglehart & Sommers, 2015). Medicaid expenditures r epresent ed over 90% of that funding and severely diminish ed availability of federal support in other areas such as education, transportation and infrastructure, scientific research, and social services (Center on Budget and Policy Priorities, 2016; Iglehart & Sommers, 2015). As Medic aid continued to grow and become one of the largest source s of health coverage in the United States in numbers enrolled and dollars spent its trajectory appeared unsustainable ( CMS, 2008; Iglehart, 2009; Kaiser Family Foundation, 2014; Mann & Westmoreland, 2004).

PAGE 27

11 Like the changing national landscape Colora do was confronted by challenges in making the growth o f its Medicaid program sustainable Medicaid program in 1969 (Colorado Center on Law & Policy [CCLP] 20 12; Colorado Health Institute [CHI] 2005). Attempting to control costs and streamline services, Colorado Medicaid began expanding managed care in 1981 and implemented an assertive managed care initiative in the 1990s for primary and acute care as well as home and community based services and long term care (CCLP, 2012; CHI, 2005). In state fiscal year 1994, Colorado Medicaid served more than 280,000 individuals, which represented slightly more than eight nt of Health Care Policy and Financing [HCPF], 1995). The cost of that coverage was more than $1.25 billion (HCPF, 1995). Figure Figure 1.1: Colorado Medicaid Enrollment from 1995 to 2010 (in thousands) (CMS, 2017f ) As the national growth rate for Medicaid expenditures hovered at almost four percent, eight

PAGE 28

12 percent from 1995 through 1998, which put additional pressure on the state to control costs ( Lutzky, Holahan, & Wiener, 2002 ). budget, Colorado Medicaid served more than 400,000 individuals at a cost of approximately $2 billion in 2005 (CHI, 2005; HCPF, 2013b; The State Health Care Spending Project, 2014). Figure 1.2 es from 1995 to 2010 (CMS, 2017f ). Figure 1.2: Colorado Medicai d Annual Expenditures from 199 5 to 2010 (in billions) (CMS, 2017f ) By 2010, Colorado Medicaid served over 600,000 individuals which represented $4 billion ( Colorado Department of Local Affairs, 2017; The State Health Care Spending Project, 2014). As enrollment grew 50% between 2005 and 2010 it outpaced general population increase of eight percent in the same period (Colorado Department of Local Affairs, 2017). During th e national economic crisis from 2007 to 2011 Colorado M edicaid enrollment increased 56% which resulted in the imposition of cost containment strategies such as

PAGE 29

13 reducing provider rates and reimbursements, streamlining enrollment processes, limiting servic es, establishing prescription drug controls, and increasing recipient responsibility for certain copays (CCLP, 2012 ; CHI, 2005 ). Regardless of political preference, policy makers acknowledged the crisis They agreed that per person health spending would co ntinue to rise, the number of Medicaid recipients would increase faster than the labor force would grow, additional public benefits would require additional public expenditures, and added taxes or subsidies could offset part of the costs (Aaron, 2007). Eve n before additional support for Medicaid expansion was included in the Patient Protection and Affordable Care Act (Affordable Care Act) of 2010, federal and state authorities were seeking different approaches and new solutions. Colorado implemented several strategies mentioned in the previous paragraph. F ederal and state authorities contemplated additional strategies: e ngaging states in long range planning, program performance, adopting provider pay for reporting and pay for performance policies focusing on preventive and primary care services increasing the use of mandatory managed care, and introducing special initiatives to target high cost areas such as emergency room (ER) utilization ( Aaron, 2007; CMS, 2008; Gif ford Smith, Snipes, & Paradise, 2011 ; Keckley & Kalkhof, 2007 ). As seen throughout landscape warrant innovative and varied responses. The next chapter of this thesis uses collaborative governance literature to lay the foundation for the ensuing discussion of

PAGE 30

14 CHAPTER II WHY COLLABORATION? As federal an d state government partners sought approaches to improve Medicaid the need for greater collaboration among all invested parties became more apparent. A governance arrangement that fit the problem was required (Ostrom, 2007). It was no longer sufficient for government authorities Medicaid (Governance, n.d.). As the need for greater collaboration wa s recogniz ed, the use of collaboration expanded This chapter begins by defining governance and collaborative governance. It then draws from collaborative governance literature to identify several common characteristics of collaboration s and examines the Ansell and Gash (2008) approach to collaborative governance. The chapter concludes with observations regarding the appropriate ness of using the Ansell and Gash (2008) approach to assess an example of collaborative governance. From Governance to Collaborative Governance As public management issues continued to expand in scope and complexity in the first decade of this century, the idea of governance evolved to include a broader concept of collaborative governance. By the early 2000s the narrow definition of governance that simply described government activity as exercising lawful c ontrol or overseeing and making decisions was inadequate (Governance, n.d.). This section examines several of the leading public management definitions and characteristics of governance that contributed to the more expansive concept of collaborative governance.

PAGE 31

15 Lyn n, Heinrich, and Hill (2000) administrative rules, judicial rulings and practices that constrain, prescribe, and enable government activity, where such activity is broadly defined as the production and delivery of publicly suppo p. 235). Incorporating a dimension of flexibility into the re lationships in the definition, the authors emphasize d a configuration of distinct but interrelated elements comprised of formal and informal structures that involve bargaining and compromise (Lynn, Heinrich, & Hill, 2000). Milward and Provan (2000) highlight ed a need to connect networks of actors in public policy domains where power must be shared becau se missions c ould not be accomplished by a single institution Their interpretation of governance include d the creation of conditions in which orde red rule and collective action were comprised of agents in the private public, and nonprofit sectors and foc used on mechanisms of government not relying exclusively on government authority and sanctions ( Milward & Provan, p. 360, 2000). Complementing earlier definitions that included flexible relationships and networks of shared power, Kettl (2000, 2002) called for a concept of governance that included even more diffused power and more people exercising power in the face of bigger problems Kettl (2000) described the need for adapting traditional governance models and changing strategies and tactics to enhance more decentralized and inclusive environmen t In their focus on network and collaborative management, Agranoff and McGuire (2001) contributed to broadening the definition of governance by explor ing issues related to tru st, common purpose, and mutual dependenc y. Brewer, Neubauer, and Geiselhart (2006) echoed this evolving perspective in their typology of government environments by suggesting conditions of modern times merit ed new governance architectures They viewed expa nd ing governance relationships that include d

PAGE 32

16 public and private sector participants as essential in strengthening capacity and agility in an atmosphere distinguished by rapid and unpredictable changes ( Brewer, Neubauer, & Geiselhart, 2006) As evidenced b y the definitions and implications in the previous paragraph, the concept of governance in the early 2000s began to expand beyond the limits of government activity to encompass the collaborative work of publ ic and p rivate sector actors addressing the growing scope and complexity of public management issues To the extent that some degree of mutual public and private sector participation occurs in service delivery and operates according to established rules, a type of collaborative governance exists (Ag bodzakey, 2012). C ollaborative governance encourages joint efforts among the public, private, and nonprofit sectors to address complex problems through collective decision making and implementation 200 4; Gray, 1989; Huxham & Vangen, 2000). A distinctive feature of collaborative governance is the c onstructive engag ement of people across public agencies, levels of government, and the public, private, and civic spheres to achieve a public purpose that coul d not otherwise be accomplished ( Emerson, Nabatchi, & Balogh 2012 ; Williams & Matheny, 1995 ). Collaborative governance processes share power with stakeholders in decision making to develop recommendations for effective, lasting solutions to public problem s (Purdy, 2012, p. 409). Broad stakeholder engagement, inclusive relationships, and collective decision making and action distinguis h collaborative governance as a viable vehicle for addressing current public management problems. Ansell and Gash (2008) distinguish ed collaborative governance from alternative policy making approaches such as adversarialism and managerialism (Futrell 2003;

PAGE 33

17 Williams & Matheny 1995). For example, c ollaborative governance stakeholders may have adversarial relationships initi ally, but the collaborative process is intended to transform those relationships into cooperative ones so that mutual objectives can be achieved (Ansell & Gash, 2008). Unlike managerialism, collaborative governance directly involves stakeholders in decisio n making rather than merely consulting stakeholders or considering their perspectives (Ansell & Gash, 2008). The authors also differentiate d collaborative governance from other types of cooperation or collaboration such as corporatism, associational govern ance, policy networks, public private partnerships, participatory management, i nteractive policy making, stake holder governanc e, and collaborative management (Ansell & Gash, 2008). Ansell and Gash (2008) clarified the definition of collaborative governance to engage non state stakeholders in a collective decision making process that is formal, consensus oriented, and deliberative and that aims to make or implement publ ic policy or Ansell and Gash (2008, 2012) used the ir definition for collaborative governance to isolate six criteria, which aided in their meta analysis of literatu re and case studies Their systematic re view produced 137 diverse cases using some f orm of collaborative governance and their analysis revealed four common elements and related components occurring in each case (Ansell & Gash, 2008, pp. 548 550). Further examination and consideration of the Ansell a nd Gash approach follows in the next two sections Six Defining Criteria The Ansell and Gash (2008) definition of collaborative g overnance emphasizes six criteria : a forum initiated by public agencies or institutions; a public policy or public

PAGE 34

18 m anagement issue expressed as its focus; a forum organized formally and convened collectively; nonstate stakeholders involved as active participants ; decision making included as a role for forum participants; and consensus used as the basis for decision making ( pp. 5 44 545). As seen in the definitions and distinctive characteristics of collaborative governance that appear earlier in this chapter, t he six criteria provide a reasonable basis for defining collaborative governance. The six defining criteria also permit th e first step in describing, assessing, and comparing examples which can aid in building theory (Ansell & Gash, 2008). T he criteria inform common elements and components essential in collaborative governance and a n example fitting the Ansell and Gash (2008) model could be considered an authentic collaborative governance effort A summary of each of the six criteria follows. Forum Initiated by Public Agencies or Institutions First, public agencies or institutions initiate the forum for collaboration. Th e initiator (Gray, 1989, p. 71). Because of its right to establish and enforce rules, a public authority convokes the collaborative governance process (Purdy, 2012). Unli ke other forms of collaboration, collaborative governance includes a specific convening and leadership role for public agencies to launch a forum to advance their own purposes, to respond to the expressed needs or requests of their constituents, or to fulf ill a legislative or judicial mandate (Ansell & Gash, 2008, p. 545) Focus on Public Policy or Public Management Issue Second, differing from mediation or dispute resolution, collaborative governance focuses on public policy or public management issues (Ansell & Gash, 2008) The agency or

PAGE 35

19 institution that creat es the forum does so with an instrumental purpose in mind, usually in response to a challenging area of public policy or management within its purview that warrants a collaborative approach to sol utions (Bryson, Crosby, & Stone, 2006 ; Huxham & Vangen, 2000 ). According to the Ansell and Gash (2008) definition and criteria, collaborative governance involves public agencies creating a clear strategy to develop or implement public policy or manage publ ic programs or assets Formally Organized and Collectively Convened Forum Third, the forum is organized formally and meets collectively. Compared to less formal interactions cultivated by public agencies and stakeholders, a collaborative governance structure is intentional and explicit (Ansell & Gash, 2008) Along a continuum that moves from cooperation to coordination to collaboration, formality generally moves from low to medium to high ( Reilly, 2001). Collaborative governance includes a structure of regular meetings of participants along with communication through other media ( Bryson et al. 2006; Huxham & Vangen, 2000). Nonstate Stakeholders as Active Participants Fourth, nonstate stakeholder s are ac tive participants engaging with public stakehol ders and with each other in a deliberative and multilateral process (Ansell & Gash, 2008) Participants work together to build support and stimulate productive and purposefu l interaction among other participants (McGuire, 2006). They engage in activities t hat generate a capacity for joint action that did not previously exist; depending upon the purpos e and context of the collaboration, participants may take part in activities such as educating constituents or the public, enacting policy measures, assembling external resources, enacting new practices, overseeing implementation, and ensuring compliance (Emerson et al., 2012)

PAGE 36

20 Forum Participants Involved in Decision Making Fifth forum participants not only consult and advise but also actively contribute to the decision making process. A chief characteristic of collaborative governance is participation by interested and affected parties in all phases of the decision making process (Freeman, 1997). Although the public agency may have ultimate authority nonst ate stakeholders share responsibility for policy outcomes (Ansell & Gash, 2008) Stakeholder participation in decision making promotes collective resolve not only to address the identified problem but also to increase responsibility for outcomes (Agbodzake y, 201 2 p. 108). Consensus oriented Decision making Process Sixth the dec ision making process is oriented toward consensus Although participants do not always achieve consensus, they strive to find areas w here they can reach agreement (Ansell & Gash, 2 008 ; Fung & Wright, 2001 ). P articipants involved in a collaborative governance forum discuss issues until all opinions are expressed and understood, and group agreement on a course o f action is required before proceeding to the next topic (Reilly, 2001). Four Common Elements Four primary and recurring elements and related components surfaced in the Ansell and Gash (2008) review of case studies fitting their collaborative governanc e definition and six criteria The authors stressed common and fr equent findi ngs and simplified the representation of elements and components with the intent of making their approach more useful for policy makers and practitioners (Ansell & Gash, 2008). The four elements are starting conditions, institutional design, facilitative leadership, and collaborative process

PAGE 37

21 (Ansell & Gash, 2008) Each of the four elements and related components contributing to a collaborative governance effort are discussed in greater detail in the next four subsections of this chapter and Table 2.1 serves as a guide to the narrative that follows. Table 2.1: Ansell and Gash Elements and Co mponents of Collaborative Governance Elements Components Starting Conditions Prehistory of conflict or cooperation Power and resource imbalances Incentives to participate Institutional Design Inclusive participation Exclusive forum Clear ground rules Process transparency Facilitative Leadership Approach Skills Techniques Collaborative Process Face to face dialogue Trust building Commitment to process Shared understanding Intermediate outcomes Startin g C onditions One element in the Ansell and Gash (2008) collaborative governance examples i s starting conditions, which include the pre histo ry of conflict or cooperation ; p ower and resource imbalances ; and i ncentives to participate Where gridlock burden s all stakeholders with the cos ts of a growing problem, it often act s as a catalyst for collaboration However, sustained conflict predating t he collaboration contribute s to low levels of trust that can manifest as marginal commitment (Ansell & Gash, 2008 ; Huxham & Vangen, 2000 ). In such cases, intentionally creating an atmosphere of mutual reliance among stakeholders and fostering the incre m ent al development of trust aim to increase cooperative participation and encourage greater collaboration among participants (Ans ell & Gash, 2008; Purdy, 2012).

PAGE 38

22 Initial imbalances in stakeholder power and resources are also integral to starting conditions in collaborative governance case studies. Power issues stem from the ability to exercise influence, authorize action, and control resources ( Bryson et al. 2006; Gray, 19 89 ). To sustain commitment to and involvement in the collaboration, stakeholders with less power and fewer resources need assurance that their perspectives are valued and ev idence that their interests are included ( Agbodzakey, 2012; Ansell & Gash, 2008; Bryson et al. 2006) I mbalance s of power and reso urces influence stakeholder incentives to participate, the third component of starting conditions in the collaborative governance examples examined (Ansell & Gash, 2008). Financial incentive is not the only and not always the main reason stakeholders parti cipate. An altruistic motive to support the collective good, a rational concern to protect or advance self interest, a desire to avoid missing an opportunity if absent or some combination of the three are primary incentives prompting stakeholder participa tion (Ansell & Gash, 2008; Reilly, 2001; Wood & Gray, 1991). Stakeholders who believe their individual power is ascending or who see alternative vehicles for solutions may have less incentive to collaborate (Ansell & Gash, 2008). Institutional D esign Another recurring element in the collaborative governance case studies i s institutional design, whose components consist of inclusive participation an exclusive forum clear ground rules, and process transparency (Ansell & Gash, 2008) Inclusive participa tion i s an intentional design component. S eeking b road participation encourages representation from multiple perspectives and different interests, which stimulates and supports far reaching consideration of issues and an extensive view of potential benefit s and harms ( Ansell & Gash, 2008; Emerson et al., 2012; Reilly, 2001) Deliberate inclusion of stakeholders across a

PAGE 39

23 wid e array of interests contributes to a public perception of increased transparency and fairness more informed discussions, and greater legitimacy of decisions (Ansell & Gash, 2008; Freeman, 1997). An exclusive forum i s another design component. In the collaborative governance examples studie d s takeholders are much more likely to participate when the forum i s exclusive and alternatives a re either absent or unattractive (Ansell & Gash, 2008; Fung & Wright, 2001). Full commitment to the collaborative process i s less likely whe n viable alternatives exist (Reilly, 2001). Wh en the convening authority sends clear signals indicating alternative means of resolution are not under consideration, the legitimacy of the collaborati ve effort increases and stakeholder uncertainty decreases (Ansell & Gash, 2008; Weber & Khademian, 1997). A third design component i s ground rules. Clear and consistently applied ground rules rea ssure stakeholders of the legitim acy (Ansell & Gash, 2008). A set of rules serves to ensur e that the collaborative effort can not be preempted or appealed (Reilly, 2001; Weber & Khademian, 1997). Ground rules also ser ve as a set of principles to define the ( Agbodzakey, 2012; Ansell & Gash, 2008 ; Fung & Wright, 2001 ; Wood & Gray, 1991 ). Process transparency i s the fourth component noted in the institutional design of the collaborat ive governance cases Ansell and Gash (2008) studied T ransparency in col laborative governance encourages an open and documented process ( Freeman, 1997; Reilly, 2001) As an instrument of collaborative governance transparency becomes ingrained in the process and is particularly useful in increasing stakeholder trust and promoting accountability (Ball, 2009).

PAGE 40

24 Facilit ative L eadership A third common element in the colla borative governance examples Ansell and Gash (2008) reviewe d i s facilitative leadership, which distinguishes itself by the approach, skills, and techniques used by those guiding the collaborative effort Rather than leadership that makes things happen, facilitative leadership helps others to make things happen and support s participants in effectively working with each other (Ansell & Gash, 2012; Chrislip & Larson, 1994; Huxham & Vangen, 2000 ; Vangen & Huxham, 2003a ). Faced with a variety of contexts, goals, and tasks creating different demands, collaborative le aders demonstrate a multitude of skills and techniques to meet the circumstances (Ansell & Gash, 2012). Facilitative l eadership i s e ssential to engaging stake holders sustaining their interest and action throughout the process and ensuring they realize th e benefits of participating in the collaboration (Ansell & Gash, 2008 ; Vangen & Huxham, 2003a ). The facilitative leadership approach in the collaborative governance example s studied promotes a nd s afeguards the process to motivate stakeholders to move forward without advocating one point of view or tak ing unilateral, decisive action (Ansell & Gash, 2008; Bryson et al., 2006; Chrislip & Larson, 1994 ; Emerson et al., 2012 ; Huxham & Vangen, 2000 ). Depending on the situation and circum stances, collabor ative governance leaders can find themselves playing multiple roles. Leadership roles include stewards, who convene the collaboration and maintain the integr ity of the process ; mediators, who facilitate communication and nurture relationsh ips among stakeholders ; and catalysts, who stimulate identification and realization of value creating opportunities (Ansell & Gash, 2012). T h e facilitative approach requires a different set of skills than the tactical ap proach where leaders articulate a clear objective, develop a plan, and le a d others in the execution of

PAGE 41

25 the plan Skills needed in facilitative leadership also differ from those in the positional approach where leaders at the top of an organizational structure set goals, organize activities motivate the c ompletion of tasks, and reward performance ( Ansell & Gash, 2008; Chrislip & Larson, 1994 ). Leadership that employs skills such as engagement, facilitation, negotiation, mediation and progress evaluation is critical to the success of collaborative governance efforts ( Lasker, Weiss, & Miller, 2001; Gerard, & Bingham 2006) Effective leadership requires skills and abilities diverse and flexible enough to be adapted to fit the circumstances and s ituations presented in a dynamic collaborative environment (McGuire, 2006; Weber & Khademian, 1997 ). Leadership t echniques found in collaborative governance examples support a fac ilitative approach and optimize the critical skills mentioned previously. E n listing participation from all stakeholders, helping stakeholders develop a common language, inviting all perspectives to be heard, and showing respect to all participants in the meetings demonstrate the active engagement of facilitative le adership (Agbodzakey, 2012 ; Ansell & Gash, 2008; Freeman, 1997 ; Lasker et al. 2001 ) E mploy ing tact, constructively explor ing differences, and buil ding consensus are some of the techniques used in facilitation ( Ansell & Gash, 2008; Ball, 2009; Gray, 1989 ; Lasker et al., 2001; McGuire, 2006 ) Leadership skills of neg otiation and mediation use techniques such as exhibiting mutual respect for different stakeholder perspectives, finding common ground among diverse interests, and managing conflict to mitigate power struggles and promote mutual gains (Agbodzakey, 2012 ; Ansell & Gash, 2008; Ball, 2009; Bryson et al., 2006; Emerson et al., 2012 ) Pr ogress evaluation skills apply leadership t echniques such as monitoring progress

PAGE 42

26 cknowledging small victories, and using incremental changes to inspire continued advancement ( Ansell & Gash, 2008; Bryson et al., 2006 ; Emerson et al., 2012 ) C ollaborative P rocess The fourth element in the Ansell and Gash (2008) c ollaborative governance case studies i s the collaborative process itself, which is described as cyclical rather than linear in nature, and its essential componen ts are f ace to face dialogue, trust building, commitment to process, shared understan ding, and intermediate outcomes Communication i s the heart of the process, and face to fac e dialogue among stakeholders i s particularly advantageous at the beginning (Ansell & Gash, 2008; Emerson et al., 2012; Reilly, 2001 ). As communication serves as a mechanism in breaking down barrier s, it also le a d s to buildi ng relationships and contributes to sustaining engagement over time (Ansell & Gash, 2008; Emerson et al., 2012) Communication also facilitates building trust among partici pants. In the case studies examined, Ansell and Gash (2008 ) noted that trust develops over time when stakeholders work together, establish relationsh ips, and demonstrate they are reasonable consistent, and reliable Trust increases a s expectations are formed and fulfilled ; stakeholder confidence grows along with the be lief that all participants will execute their responsibilities without taking advantage of other participants ( Emerson et al., 2012; Lasker et al., 2001 ; Vangen & Huxham, 2003b ) Trust enables stakeholders to move beyond their own personal interests and frames of reference toward consideration and understanding of the interests and values others hold (Ansell & Gash, 2008; Emerson et al., 2012). Communication and trust contribute to the cultiva t ion of c ommitment to the process ( Ansell & Gash, 2008; Purdy, 2012 ). Stakeholders commit to the c ollaborative process as a

PAGE 43

27 mechanism for addressing their common problems, and their commitment to full participation increases in the absence of viable altern atives ( Ansell & Gash, 2008; Lasker et al., 2001 ; Reilly, 2001 ) Additionally, in collaborative governance efforts, commitment to the process involves having or developing the belief that the best way to attain desired policy outcomes and achieve mutual ga ins is through negotiation and compromise (Ansell & Gash, 2008; Lynn et al., 2000 ; Williams & Matheny, 1995 ). Clar ifying a common purpose fosters a shared understanding among stakeholders that they can achi eve collectively what none can accomplish individually (Ansell & Gash, 2008; ; Vangen & Huxham, 2003b ). Shared understanding shifts the perception of ownership from the convening public agency or institution to the collective body of s takeholders (Ansell & Gash, 2008). Shared understanding is an integral part of each stage of the learning experience and implie s agreement from the problem objectives (Ansell & Gash, 2008) Shared understanding promotes another component in the collaborative process: the recognition that i ntermediate outcomes are important to long term success (Ansell & Gash, 2008; Chrislip & Larson, 1994) Articulating those outcomes, producing a shared p lan of action, and creating assessment criteria gi ve participants an incremental path forward (Emerson et al., 2012). Openly recognized, short term accomplishments and progress fuel momentum that encourages stakeholders to remain committed and involved (Ansell & Gash, 2008; Reilly, 2001).

PAGE 44

28 Model Considerations Earlier sections in this chapter outlined six defining criteria and four common elements and related compo n ents in the Ansell and Gash (2008) model for collaborative governance A n approach was described in which public and private actors work together in distinct ways using particular processes to establish laws and rules for providing public goods and services. Ansell and Gash (2008) hold a view of collaborative governance that pertains spe cifically to publ ic affairs Unlike the broader integrative framework for collaborative governance introduced by Emerson, Nabatchi, and Balogh (2012), for example, the Ansell and Gash (2008) criteria and elements align more closely with examples where a pu blic agency is the catalyst for convening a public forum focused on addressing a public po licy issue In addition, the Ansell and Gash (2008) approach draws fundamental characteristics of collaborative governance from not only existing literature but also a meta analysis of case studies across multiple disciplines. From the case studies reviewed, Ansell and Gash (2008) provide a realistic approach for use by practitioners and identify factors crucial to the functioning of the collaborative process. Several of the case studies involved community health partnerships or interdisciplinary health initiatives (Alexander, Comfort, & Weiner, 1998; el Ansari, 2003; Fawcett et al., 1995; Gilliam et al., 2002; Mitchell & Shortell, 2000; Mizrahi & Abramson, 2000; Rousso s & Fawcett, 2000; Weech Maldonado & Merrill, 2000). Comparing features of those partnerships and initiatives to the Ansell and Gash (2008) criteria and elements outlined earlier in this chapter indicat es their approach would be appropriate for determin ing whether or not a complex public health care delivery system is an example of collaborative governance.

PAGE 45

29 Several a rticles referencing the Ansell and Gash (2008) model of collaborative governance were read in conjunction with this research study Review was conducted to determine if there were additional considerations for or modifications to the Ansell and Gash (2008) approach relevant to Some authors suggest broadening collaborative governance beyond a public aff airs focus (Binz Scharf, Lazer, & Mergel, 2012; Clarke, 2017; Emerson & Nabatchi, 2015; Emerson et al., 2012 ; Mosley, 2014 ). Another propose s enriching the collaborative governance approach with additional topics such as social capital and its ability to e nable and encourage mutually advantageous cooperation (Oh & Bush 2016; Social capital, n.d.). One recommends expanding the existing element of facilitative leadership (Page, 2010). A nother advocates incorporating additional details about bureaucratic features and routines (Michel, 2017). Sorensen and Torfing (2011) suggest the benefit of incorporating innovation. T hese examples offer valuable perspectives, but several others amplify t wo is sues in the Ansell and Gash (2008) approach that are particular ly relevant to this research study Several authors elaborated on the effects of local influences and time in collaborative governance efforts Authors stressed the importance of considering pot entially different effects of a collaborative dispersed across multiple jurisdictions, initiated in local and regional settings, or underrepresented in certain geographic areas (Andres & Chapain, 2013; Gibson, 201 1; Herranz, 2008; Koebele, 2015 ; Siddiki, Carboni, Koski, & Sadiq, 2015 ) Ansell and Gash (2008) also emphasize d the importance of collaborative governance in local environments where previous attempts at downstream implementation of policy had failed. The y caution that leadership at the local level could be limited by circumstances and resource

PAGE 46

30 availability, and effective collaboration would be constrained when there was a lack of leadership (Ansell & Gash, 2008). Several authors address ed the importance o f considering the effect of time on collaborative governance efforts ( Berar d o, Heikkila, & Gerlak, 2014; de Loe, Murray, & Simpson, 2015; Gerlak & Heikkila, 2011; Gollagher & Hartz Karp, 2013; Heikkila & Gerlak, 2016). From a pragmatic perspective, stakeholders considered sp e nding time on collaboration as taking time from normal business operations (de Loe et al., 2015). Authors also advise d that h ow a collaborative process unfolds and evolves over time may be different than how it initially appears when designed and that success is sustainability over time ( Gollagher & Hartz Karp, 2013; Heikkila & Gerlak, 2016). Ansell and Gash (2008) recognized collaborative governance as a time consum ing process requiring a long term commitment to achieve outcomes. The authors also comment ed on the positive effect the investment of time at the beginning of the collaborative effort has on implementation and later operation s (Ansell & Gash, 2008) Withou t a view s successful functioning would be limited ( Heikkila & Gerlak, 2016). Working together over time provides an additional benefit. S takeholders learn how to collaborate successfully, which may ser ve as a foundation to support collaboration on additional issues or in new areas (Ansell & Gash, 2017). F ailure to consider the effects of local influences and time on a collaborative governance effort could impact outputs and outcomes (Koebele, 2015). As a prerequisite for selecting a model by which to determine whether or not an example of a collaborative governance effort, t his research study sought to understand the criteria and elements in the Ansell and Gash (2008) model, consider

PAGE 47

31 its public affairs constraints, review health related case studies used in their research, and read a subset of related articles published after their research findings. As a result, the Ansell and Gash (2008) model was selected as an appropriate choice fo r examining the collaborative governance features of the ACC. The next chapter of this thesis through the lens of the Ansell and Gash (2008) collaborative governance model.

PAGE 48

32 CHAPTER III The first chapter of this thesis summarized innovative solutions to address its escalating growth. The second chapter identified collaborative governance as one means of approaching Medicaid improvement and used the collaborative governance definition, criteria, and elements from the Ansell and Gash (2008) research to ground the object of this research study This chapter examines to reform its predominantly fee for service Medicaid program by developing the ACC and uses the Ansell an d Gash (2008) approach to determine whether or not the ACC is a collaborative governance example. program and the ACC are resource s currently available or previously found on publicly available websites. Main sources for the reports and d ocumentation cited are the United Centers for Medicare & Medicaid Services (CMS), the federal agency responsible for administering the Medicare and Medicaid programs; Department of Health Care Policy and Financing, the agency responsible for administering Medicaid program; national and state nonprofit organizations; and the seven R egional Care Collaborative Organizations (RCCOs) participating in the ACC. Sources include such documentation as requests for information, requests for proposals, annual reports, and committee and subcommittee minutes. All sources are cited appropriately i n the text in this chapter and included in the References section at the end of this thesis.

PAGE 49

33 The ACC and Six Defining Criteria As seen in the previous chapter of this thesis, t he Ansell and Gash (2008) definition of collaborative governance includes six criteria : a forum initiated by public agencies or institutions; a public policy or public management issue expressed as its focus; a forum organized formally and convened collectively; nonstate stakeholders involved as active participants; de cision making included as a role for forum participants; and consensus used as the basis for decision making. All are present in the ACC. The designation of the Colorado Department of Health Care Policy and Financing (the Department) as the public agency responsible for developing implementing, maintaining, and advancing the ACC satisfies the first criterion, and the identification of Colorado Medicaid reform as the public policy and management issue the ACC addresses fulfills the second ( CMS, 2012; HCPF, 2009b; Rodin & Silow Carroll, 2013; State Policy Options, 2012) The ACC is the formally organized and collectively convened forum which satisfies the third criterion (HCPF, 2009a; Karabatsos, 2011; Rodin & Silow Carroll, 2013; State Policy Options, 2012). ACC participants such as the Regional Care Collaborative Organizations (RCCOs), the Statewide Data and Analytics Contractor (SDAC), Primary Care Medical Providers (PC MPs), specialists and other health care providers, community based organizations, and Medicaid recipients and advocates fulfill the fourth criterion in their roles as active nonstate stakeholders ( CMS, 2012; HCPF, 2009a; Karabatsos, 2011; Rodin & Silow Car roll, 2013; State Policy Options, 2012) Satisfying the fifth criterion, s takeholders play an active role in ACC decision making through participation as RCCOs, the SDAC, PCMPs, and members of the central and regional advisory committees and subcommittees (Colorado Access, 2017; Colorado

PAGE 50

34 Community Health Alliance, 2017; Community Care Central Colorado, 2017; HCPF, 2017c ; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017) Additional Medicaid recipients and advocates as well as o ther community based organizations influence decision making through participation in open meetings and workgroups public comment periods and survey completion ( CMS, 2012; HCPF, 2017c ). Finally, as seen in the advisory committee by laws and based approach to decision making satisfies the sixth criterion (Colorado Access, 2017; Colorado Community Health Alliance, 2017; Community Care Central Colorado, 2017; HCPF, 2012 c 2017c ; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017). As Ansell and Gash (2008) stated, decision, the goal of collaboration is typically to achieve some degree of conse nsus among pp. 546 547). The Department is ultimately responsible for decisions affecting the ACC; however, the consensus oriented collaboration offers stakeholders a forum in which to discover areas of agreement, make formal r ecommendations, and participate in the decision making process ( HCPF, 2017c ; McGinnis & Small, 2012). The ACC and Four Common Elements This section compares the four common elements and primary components that surfaced in the Ansell and Gash (2008) meta analysis to characteristics of the ACC. This section is arranged in subsections containing information about starting conditions, institutional design, facilitative leadership, and collaborative process, which were discussed during the examination of the Ansell and Gash (2008) model in the previous chapter of this thesis. T able 3.1 serves as a guide to the narrative that follows.

PAGE 51

35 Table 3.1 : Ansell and Gash Elements and Components and ACC Characteristics Ansell and Gash Elements ACC Characteristics S tarting Conditions Prehistory of conflict or cooperation Power and resource imbalances Incentives to participate Starting Conditions Historical dissatisfaction and mistrust Multiple interests and diverse perspectives Local representation and accountability Institutional Design Inclusive participation Exclusive forum Clear ground rules Process transparency Institutional Design Local, regional, statewide e ngagement Vision and interdependent structure Roles and expectations Emphasis on o penness and documentation Facilitative Leadership Approach Skills Techniques Facilitative Leadership Guidance Promoting participation and managing process Active listening and leveraging common ground Collaborative Process Face to face dialogue Trust building Commitment to process Shared understanding Intermediate outcomes Collaborative Process Monthly and quarterly meetings Mediator s and local leaders Shared ownership and mutual gains Clear mission and common goals Incremental progress and future strategy Starting Conditions Components found in the starting conditions of the collaborative governance examples Ansell and Gash (2008) analyzed included a prehistory of conflict or cooperation, power and resource imbalances, and incentives to participate. These c omponents also existed in C olorado before the ACC Literature and case studies indicate that conditions existing before collaboration begins can encourage either conflict or cooperation among public agencies and stakeholders (Ansell & Gash, 2008; Emerson et al., 2012; Gray, 1989). Discord and tension were apparent in Colorado due to previous unsuccessful attempts to implement managed care in the state Like many other states, Colorado att empted to curb Medicaid fee for service spending by migrating to managed care in the mid 1990s (Holahan, Zuckerman, Evans, & Rangarajan, 1998 ; Hurley & Somers, 2003; Weissert, 2002). The chiefly hierarchical approach proved unsuccessful as evidenced by issues such as

PAGE 52

36 limited health care access for Medicaid recipient s, inadequate provider networks, insufficient rate analys is and reimbursement, and dissatisfaction among individuals, provider s, and health plans ( Hill, 200 4; Lutzky et al. 2002; Thomson, Rhodes, & Cowie, P. C., 2000; Weissert, 2002). In 2000 a Colorado judge awarded Rocky Mountain Health Maintenance Organization (HMO) $18 million in back Medic aid payments, and both the state and the HMO agreed that independ ent actuaries would analyze the rate setting mechanism the Department used ( Lutzky et al., 2002). Later that year Kaiser Foundation Health Plan of Colorado filed suit to recover years of Medicaid payments the Department allegedly withheld (Lutzky et al., 2002). In less than one year the Department lost four of its participating managed care he alth plans and suspended enrollment in its only remaining plan, Colorado Access (Hill, 2004). These issues contributed to increasing distrust and conflict among the Department health plans and providers, Medicaid recipients, and other stakeholders. Revert ing to a fee for service approach, Colorado then became one of only two states with voluntary managed care enrollment for all eligible Medicaid beneficiary groups ; 85% of i ts Medicaid recipient s elect ed a fee for service approach (Gifford et al. 2011 ; Hill, 2004; Karabatsos, 2011). The distrustful environment precipitated by the failure of managed care coupled with escalating costs fueled by the fee for service resurgence prompted Colorado to regroup (Kaiser Family Foundation, 2012 2013 ; Rodin & Silow Carroll, 2013 ). In his 2006 campaign for governor, Bill Ritter outlined policy positions in The Colorado Promise including a plan for health care. Ritter (2006) health care system is broken, and this crisis will not be fixed by tinkering around the edges ( p. 11). He included collaborative development in his eight fundamental principles for health care reform ( Ritter,

PAGE 53

37 2006 legislative session, Senate Bill 06 20 8 established the Blue Ribbon Commission on Health Care Reform (the Commission) (Access to Affordable Health Care Act, 2006 ; Burger & Stapleton, 2009 ). After months of deliberation and discussions with stakeholders, constituents, legislators, and executive officials, the Commission presented a comprehensive report in 2007 (Blue Ribbon Commission for Health Care Reform, 2008 a, 2008b ). The report provided a set of recommendations for health care reform in Colorado. The recommendations became a series of legis lative initiatives, the Building Blocks to Health Care Reform, which were passed during the 2008 legislative session (Blue Ribbon Commission for Health Care Reform, 2008 a, 2008b ; Hill, Courtot, Bovbjerg, & Adams, 2012). The Medicaid Value Based Care Coordi nation Initiative was one recommendations, an approach subsequently known as the ACC ( HCPF 2009 b ; Rodin & Silow Carroll, 2013 ). The starting conditions of d issatisfaction and mistrust that originated in the 1990s over Medicaid managed care negatively influenced stakeholder environment at design and development ( Kaiser Family Foundation, 2012; Thomson et al., 2000; Weissert, 2002). As a result, the Department sought to balance disparities among power and resources by including multiple interests and diverse per spectives in creating the ACC T he D epartment approached a less hierarchical approach and attempted to engage a broad spectrum of stakeholders. As discussed in the second chapter of this thesis fostering trust and respect, employing an inclusive strategy to gain wide representation, and offsetting power imbalances among stakeholder groups aid in ensuring meaningful participation, which is ( Ansell & Gash, 2008). The Department conducted multiple public forums with both in person and electronic opportunities for participation (Kaiser Family Foundation, 2012; Karabatsos, 2011). In addition, the Department issued a

PAGE 54

38 formal request for information that solicited feedback from client groups, community and statewide social service organizations, local governments, independent physician associations, physicians, hospital systems, providers, provider collaboratives, managed care organizations, health plans, clients, quality or ganizations, health foundations, and other interested parties (HCPF, 2009 b p. 1). The request for information was 54 pages in length and contained more than 200 questions categorized by regional entities; Medicaid recipients and advocates; statewide data and health information technology entities; Medicaid providers such as primary care physicians specialist s hospitals, pharmacies, home health agen ci es and nursing facilities; and all other interested parties willing to particip ate (HCPF, 2009 b ). Stakeholders who divergent opinions (Karabatsos, 2011; Rodin & Silow Carroll, 2013 ; State Policy Options, 2012). of its Medicaid population, and existing provider networks, t he scope of comprehensive As incentives to participate, the Department emphasi zed local representation and accountability (HCPF, 2009 b ; Karabatso s, 2011) The Department envisioned regional areas of accountability with existing and new support services for Medicaid recipients and providers, health information technology services, data sharing and analysis capabilities, contracts focused on outcomes and gain sharing provisions (HCPF, 2009 b ; Rodin & Silow Carroll, 2013; State Policy Options, 2012 ). T he request for information presented a compelling incentive to comment and participate Stakeholders were motivated to enter the dialogue and voice their perspectives as one means of upholding and advancing their interests (Ansell & Gash, 2008; Reilly, 2001). Some of the same vendors a nd providers who

PAGE 55

39 had previous position on manage d care participated in pub Kaiser Family Foundation, 2012; Karabatsos, 2011; Thomson et al., 2000). The Department built upon the negative historical context of dissatisfaction and mistrust inherent in the starting cond itions for healt h care reform in Colorado and motivate d stakeholders to work collaboratively in designing the ACC as a regional based, accountable Medicaid delivery system Institutional Design Aligned with the Ansell and Gash (2008) element of inclusive participation the Department intentional ly engaged local, regional, and statewide resources which aided in establishing the legitim acy of the collaboration itself. I nitial influences were the Accountable coherent local physician and hospital delivery systems that seek to increase the quality and decrease the cost of health care (Fisher, Staiger, Bynum, & Gottlieb, 2007; HCPF, 2009b); a regional approach to health care in which health, business, and political leaders work together to i mprove health care in their communities (HCPF, 2009b; Wagner, Austin, & Coleman, 2006 ) ; and the improving the experience of care, improving the health of populations, and reducing per person health care costs statewide (Berwick, Nolan & Whittington, 2008; HCPF, 2009b ). evolved from a series of iterations A delivery system for Colorado Medicaid recipients emerged that synthesized input from numerous perspectives. Seeking a middle ground between a fee for service syste m and managed care, Colorado chose a managed fee for service hybrid. It combined characteristics of a regional Accountable Care Organization with a Primary Care Case Management system ( Fisher et al., 2007; HCPF, 2010a ; Verdier, Byrd, & Stone, 2009). A managed fee for service hybrid assigns responsibility for the delivery and

PAGE 56

40 entity ; as payment, the provider or other designated entity receives a small monthly case m anagement fee in addition to customary fee s for services delivered and billed (HCPF, 2010a; Verdier et al., 2009 ) As illustrated in the table in Appendix A and the map in Appendix B t he Department configured the state into s even geographic areas based on an algorithm that considered Medicaid recipient referral and access patterns, population density, provider capacity, public health districts, and the presence of Federally Qualified Health Centers and safety net providers (State Policy Options, 2012). Stakeholders acknowledged the uniqueness of and sup p orted the need for each region As a result, localized resources became essential to the ACC and the Department encouraged each region to develop approaches suitable for and sustai nable by its population and local communities (Karabatsos, 2 011; State Policy Options, 2012 ). Appendix C contains information about each community involvement. The Ansell and Gash (2008) research also includes an e xclusive forum as an element in the institutional design of a collaborative governance effort. Researchers observed that stakeholders wer e much more likely to participate when the forum was exclusive and alternatives were absent or unattractive (Ansell & Gash, 2008; Fun g & Wright, 2001). Early in the process, the Department messaged the ACC as the vision of the future and promoted the appeal of its exclusive design. E nvisioning the ACC as an evolving model, t he Department began conversation s with all interested parties a bout its long term plan to integrate federal, state, and local health care programs into a more cohesive and efficient delivery system ( HCPF, 2009b p. 8) The Department signaled the ACC as the foundation

PAGE 57

41 for Medicaid in Colorado a foundation th at would ultimately support all Medicaid recipients in the state ( HCPF, 2009b, p. 8 ). The ACC was created as a regional model of accountability for improving the health, function ing and self sufficiency of Medicaid clients (HCPF, 2009a, p. 14) It also embodied a distinct and interdependent structure that contributed to its exclusiveness Three design elements cen were intended to work in concert with each other and with the Department: Regional Care Collaborative Organizations (RCCOs), responsible for achieving health and cost outcomes for Medicaid recipients in the comprehensive primary care and serving as a Medicaid recipient me; and the Statewide Data and Analytics Contractor (the SDAC), responsible for serving as a data repository and providing information to RCCOs, PCMPs, and the Departme nt (HCPF, 2010a, 2010b ; Kaiser Family Foundation, 2013; Karabatsos, 2011, p. 11). In add ition, the Department divided responsibilities among the RCCOs, PCMPs, and the SDAC. All ACC activities were critical and contributed to the overall functioning of the ACC but none were duplicated by another design element (HCPF, 2010a, 2010b, 2011; Karab atsos, 2011 ). Distinct contractual and data exchange relationships existed among the elements as seen in Figure 3.1.

PAGE 58

42 Figure 3.1: Structure of the ACC Ansell and Gash (2008) identified clear ground rules as a third component in the institutional design element of the collaborative governance case studies they analyzed During the development and implementation phases, the Department used clear ground rules to establish roles and expectations for key participants In its request for information concerning the development of the ACC, the Department posed questions concerning the composition of governing boards, the development and maintenance of stakeholder relationships, and mechanisms for meeting, monitoring, reporting, and reviewing performance (HCPF, 2009b). Later t he Department formally solicited proposals for RCCOs an d the SDAC competitive procurement process (HCPF, 2010a, 2010b ). The D contained clear definition of roles, structure, and expectations (HCPF, 2010a, 2010b). For example, in addition to describing the general, strategic, provider support, and accountability

PAGE 59

43 requir ements for t he RCCOs and the SDAC, the Department included details in its requests for proposals related to maintaining the collaborative aspects of the ACC (HCPF, 2010a, 2010b). The Department stated its intention to chair a committee to guide ACC performance improvement that would include representation from each RCCO, the SDAC, and Medicaid recipients and providers (HCPF 2010a, 2010b). The Department also set the expectation for each RCCO to g uide regional performance improvement through a committee directed and chaired by the RCCO; the committee would have a formal membership and governance structure, include broad Medicaid recipient and provider representation, and meet no less frequently tha n quarterly (HCPF, 2010a, p. 51). Additionally, the Department stipulated various committees the SDAC would facilitate or attend (HCPF, 2010b, pp. 62 63). The Department ultimately executed contracts with the RCCOs and the SDAC. Because PCMPs were required to be Medicaid providers, they had existing contracts with the Department (HCPF, 2010a, 2017 e ; Karabatsos, 2011, p. 12 ). As a result, the Department established and maintains a contractual relationship with all seven RCCOs, the SDAC and all PCMPs ; in add ition, each RCCO has a contractual relationship with each PCMP in its service area (HCPF, 2010a; Karabatsos, 2011 ). executing contractual relationships, the Department facilitated an open and documented process As the ACC began, ground rules and responsibilities prescribed in the contracts continued to promote process transparency the fourth component Ansell and Gash (2008) describe in institutional de sign. The ACC has a diverse program improveme nt advisory committee designed to provide guidance and make written recommendations to aid in improving health outcomes, access, cost, and client and provider experience ( HCPF,

PAGE 60

44 2017c ). The advisory committee by laws specify its membership, structure, and process, and the committee meets in person with Department staff at least quarterly with all meetings open to the public ( HCPF, 2017c ). The a three current standing subcommittees meet no few er than eight times per year and currently focus on improving health s ystems, and provider and community issues ( HCPF, 2017c ). The Department posts meeting agendas, handouts, and minutes on its website and makes recordings of the proceedings available for sixty days after each meeting ( HCPF, 2017c ). In addition, e ach RCCO hosts regular member, stakeholder, or performance advisory committee meetings in its area co mmunities (Colorado Access, 2017 ; Colorado Community Health Alliance, 2017; Community Care Cent ral Colorado, 2017; Integrated Community Health Partners, 2017 ; R ocky Mountain Health Plans, 2017 ). Meetings are open, meeting schedules are communicated in advance, and transcripts or minutes along with other reports and materials are publicly available ( Colorado Access, 2017 ; Colorado Community Health Al liance, 2017 ; Community Care Central Colorado, 2017 ; Integrated Community Health Partners, 2017 ; R ocky Mountain Health Plans, 2017 ). Facilitative Leadership Facilitative l eadership, the fourth component of collaborative governance examples Ansell and Gash (2008, 2012) studied, is characterized by its approach, skills, and techniques Continuity and diversity in leadership provide a stable and broad foundation for further shaping and guid ing policy and process (Ansell & Gash, 2008; Huxham & Vangen, 2000; Ostrom, 2002). Laurel Karabatsos, a executive team and current Delivery System and Payment Innovation Division Director and Deputy Medicaid Director is one of t

PAGE 61

45 support ing staff (HCPF, 201 7d ). A combination of shared and distinct representation e xists throughout ACC leadership ing staff but also among leaders of the RCCOs, the SDAC, PCMPs, Medicaid recipients and advocates, and other stakeholders ( HCPF, 2017c, 2017d ). Similarities are apparent w hen ACC leadership roles are compared to those Ansell and Gash (2008, 2012) described as stewards, mediators, and catalysts. Since initially convening the ACC, the Department continues to serve as its steward to guide and oversee the overall effort and maintain the integrity of the process (CMS, 2012; HCPF 2009a 2012 c 2016 b 201 7b ) Facilitating communication and nurturing relationships among stakeholders in accordance with the Ansell and Gash (2008, 2012) description ACC mediators exist not on ly in the Department but also among other ACC stakeholders. D uring the development and implementation, the Department used its own staff or professional consultants to facilitate meetings and workgroups and develop relationships among stakeholders ( CMS, 2012 ; Karabatsos, 2011). The Department also facilitated two monthly oper ations meetings to ensure communication and coordination among RCCOs, the SDAC, and the Department (CMS, 2012). T o promote further collaboration, RCCOs met independently to reinfor ce their relationships and work together on specific issues (CMS, 2012). improvement advisory committee maintains its membership roster and by laws on the website; the by laws illustrate the roles of co chairs in f acilitating communication among committee and subcommittee members and other ACC sta keholders ( HCPF, 2017c ). In addition, each RCCO facilitates its own ACC advisory committee and retains committee membership, participation, meetings, and minutes on its website ( Colorado Access, 2017; Colorado Community Health Alliance, 2017; Community Care Central

PAGE 62

46 Colorado, 2017; HCPF, 2017f ; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017 ). Co nsistent with the description from the Ansell and Gash ( 2008, 2012) research, ACC leadership also includes c atalysts t hat stimulate identification of opportunities to create value. For example, the SDAC reveal ed opportunities to improve health care and outcomes, including peer to peer learning i n and among RCCOs (HCPF, 2010b). The SDAC used national and state data to provide demographic and service utilization information to ACC stakeholders and to develop baseline performance measures for the ACC to gauge as a catalyst stimulated RCCO and PCMP activity to enhance service delivery to Medicaid recipients. RCCOs est ablished formal data sharing agreements or arrangements with providers in their regions in accordance with security and privacy guidelines contained in the Health Insurance Portability and Accountability Act (1996) to identify Medicaid recipients for outre ach and care coordination purposes (Health Services Advisory Group, 2013a). RCCOs explored using data in a variety of innovative ways in the ACC : to pinpoint Medicaid recipients with multiple ER visits ; to distinguish those needing post hospitalization assistance; to determine specialists Medicaid recipients used ; to funnel specialty care to PCMPs ; to distinguish high volume Medicaid practices where care coordinator s could be assigned; and to p opulate web ba sed portals that providers could access (Health Services Advisory Group, 2013a, 2013b, 2014, 2015). As stewards, mediators, and catalysts, ACC leadership includes multiple perspectives and involves individuals in diverse ways to empower collaboration and increase the potential for success (Ansell & Gash, 2008 2012 ; Bryson & Crosby, 2005; Emerson et al., 2012;

PAGE 63

47 Ostrom, 2002). ACC leadership examples in the first three paragraphs of this subsection illustrate the use of engagement, facilitation, mediation, and progress evaluation skills described in the second chapter of this thesis which have been identified as necessary to move a collaborative effort forward ( Ansell & Gash, 2008; Lasker et al. al. Like examp les from the collaborative governance literature referenced in the second chapter of this thesis, instances of ACC facilitative leadership techniques such as e nlisting participation from all stakeholders, helping stakeholders develop a common language, inv iting all stakeholder perspectives to be heard, and sho wing respect to all stakeholders a ppear earlier in this subsection (Agbodzakey, 2012; Ansell & Gash, 2008; Freeman, 1997; Lasker et al., 2001 ). C onstructi vely exploring differences, building consensus finding common ground among varied interests, acknowledging small victories, and using incremental changes to inspire continued advancement are additional techniques reflected in ACC leadership examples, which are similar to those Ansell and Gash (2008, 2012) described Collaborative Process In additio n to the starting conditions, institutional design and facilitative leadership components common to collaborative governance case studies examined by Ansell and Gash (2008) the ACC includes a process conducive to collaboration. Face to face dialogue is one of the identified components in a collaborative process ( Ansell & Gash, 2008), and the ACC demonstrates evidence of such dialogue in the documentation of its monthly and quarterly meetings (Colorado Access, 2017; Colorado Community Health Alliance, 2017; Community Care Cen tral Colorado, 2017; HCPF, 2017c ; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017)

PAGE 64

48 Trust building is another collaborative process component emphasized by Ansell and Gash (2008) and substantiated in the ACC through documented accounts of Department staff, consultants, facilitators, RCCO leadership, and committee members serving as mediators and local lea ders to develop relationships and increase cooperative participation within the regions and across the state (CMS, 2012; Colorado Access, 2017; Colorado Community Health Alliance, 2017; Community Care Central Colorado, 2017; HCPF, 2010a, 2010b, 2017c ; Inte grated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017 ) The third component in the collaborative process Ansell and Gash (2008) defined is commitment to the process. of its institutional design earlier in this chapter, reinforce s commitment to the collaborative effort through shared ownership of the process and outcomes The Departm ent RCCO leadership, the SDAC, PCMPs, Medicaid recipients and advocates, and other stakeholde rs have distinct roles in the ACC, and each role is a vehicle for distributing ownership of and resp onsibility for the collaboration HCPF, 2017c 2017d ). Ansell and Gash (2008) described shared understanding as the fourth component of a collaborative process. T o promote a shared understanding among ACC participants, t he Department explicitly states nd four primary objectives. T h e Department pursues a mission not only to improve health care access and outcomes for Medicaid recipients but also to be a responsible ste ward of financial resources (HCPF, 2017 a ) Concurrently, the ACC progresses toward its goals to improve health outcomes and control costs by e xpand ing acces s to comprehensive primary care; providing a focal point of car e for all m embers including coordinated and inte grated access to

PAGE 65

49 other services; ensuring a positive m ember and p rovider experience a nd promoting member and provider engagement; and e ffectively apply ing an unprecedented level of statewide data and analytics functionality to support transparent, secure data sharing and enable the near real time monitoring and measurement of health care o utcomes and costs (CMS, 2012; HCPF, 2010a, 2010b). I ntermediate outcomes included as the fifth component in a collaborative process appear in the ACC as a focus on incremental progress with an attentive eye on future strategy (Ansell & Gash, 2008; CMS, 20 12; Karabatsos, 2011; State Policy Options, 2012). To move toward improved health outcomes and reduced costs, the ACC collaborative process provides a basis for systematic actions. Combined and executed properly, seven features embedded in the ACC contribute to measured progress : a regional approach to managing, providing, and coordinating care; principles of a client centered medical home model; an integrated network of providers; provision of high quality care coordination and medical m anagement services; a focus on accountability to improve outcomes and control costs; analysis and application of informatics and benchmarking to review, measure, and compare utilization, outcomes, and costs; and a focus on continuous improvement and innova tion, constant learning, and the sharing of best practices ( CMS, 2012). Implications for Success ful Collaboration In the common elements of starting conditions, facilitative leadership, and collaborative process in the Ansell and Gash (2008) model, the au thors suggest 10 implications to facilitate success in collaborative governance. Although a collaborative governance effort still could be considered an authentic example without these implications, Ansell and Gash (2008) perceive it w ould be less likely t o succeed. Assessment of the ACC

PAGE 66

50 based on the Ansell and Gash (2008, 2012) definition, criteria, and components as described in the second chapter and earlier in this chapter indicates the ACC is an example of a collaborative governance effort Before movi ng forward with a research study designed to study health outcomes and cost of health care for individuals served in the ACC, this thesis examines the ACC in relation to the 10 implications for success (Ansell & Gash, 2008). Table 3.2 summarizes the implic ations to guide the narrative. Table 3.2: Ansell and Gash Implications for Success Ansell and Gash Implications for Success Starting Conditions Prehistory of conflict or cooperation High degree of interdependence among stakeholders or steps taken to remediate low levels of trust and social capital Power and resource imbalances Commitment to positive strategy of empowerment and representation of weaker or disadvantaged stakeholders Incentives to participate Stakeholder perception of high interdependenc e Respect from courts, legislators, and executives to honor collaborative process outcomes Facilitative Leadership Approach Honest broker accepted and trusted by stakeholders Organic leaders from with in the stakeholder community respected and trusted at the outset Collaborative Process Trust building Time allowed if remedial efforts are needed Commitment to process Achievement of buy in Suitability for situations requiring ongoing cooperation Intermediate outcomes Production of small wins Implications for Starting Conditions (Hill, 2004; Lutzky et al., 2002; Th omson et al., 2000; Weissert, 2002). In designing the ACC, the Department deliberate ly

PAGE 67

51 took steps to remediate low levels of trust and social capital by investing time in multiple opportunities to obtain diverse stakeholder participation through varied fee dback channels (HCPF, 2009b; Kaiser Family Foundation, 2012; Karabatsos, 2011). Throughout the design process, the Department engaged stakeholders who represented diverse interests and divergent opinions (Karabatsos, 2011; Rodin & Silow Carroll, 2013; Stat e Policy Options, 2012). In addition, the ACC was intentionally created as a regional health care delivery system for Medicaid recipients as seen by the interdependent structure and roles and responsibilities of the RCCOs (HCPF, 2009a ). Intentional engage ment of a broad spectrum of stakeholders signaled an inclusive strategy designed to offset power imbalances among groups and ensure meaningful participation (Karabatsos, 2011; Rodin & Silow Carroll, 2013; State Policy Options, 2012). Additionally, a strate gy of empowerment was employed in the creation of the program improvement advisory committee and subcommittees at the overarching ACC level and in area advisory committee meeting s hosted by RCCOs in their communities (Colorado Access, 2017; Colorado Commun ity Health Alliance, 2017; Community Care Central Colorado, 2017; HCPF, 2017c; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017). series of initiatives, which included what would come to be known as the ACC (Blue Ribbon Commission for Health Care Reform, 2008a, 2008b; HCPF, 2009b; Hill et al., 2012; Rod in & Silow Carroll, 2013). Contributing further legitimacy and respect for the effort, the Committee providing information implementation (HCPF, 2012a, 201 3a).

PAGE 68

52 H igh interdependence was RCCOs, and PCMPs and stakeholders were encouraged to participate too (HCPF, 2010a, 2010b; Kaiser Family Foundation, 2013; Karabatsos, 2011). Implications for Fa cilitative Leadership The ACC satisfied facilitative leadership implications for success in utilizing both honest brokers and respected members of the stakeholder community throughout the design and implementation process. In some cases, consultants and fa cilitators were used, and in other situations, RCCO leadership, committee members, or other stakeholders served as mediators and facilitators to develop relationships and increase participation (CMS, 2012; Colorado Access, 2017; Colorado Community Health A lliance, 2017; Community Care Central Colorado, 2017; HCPF, 2010a, 2010b, 2017c; Integrated Community Health Partners, 2017; Rocky Mountain Health Plans, 2017) Implications for Collaborative Process As documented e arlier Implications fo r Starting Conditions subsection, time was consciously included in the collaborative process to foster building trust (HCPF, 2009b; Kaiser Family Foundation, 2012; Karabatsos, 2011; Rodin & Silow Carroll, 2013; State Policy Options, 2012). The additional i nvestment in time was particularly appropriate considering the atmosphere of mistrust and conflict present in the environment before the ACC was developed (Hill, 2004; Lutzky et al., 2002; Thomson et al., 2000; Weissert, 2002). W hen the Department signaled Colorado which would ultimately support all Medicaid recipients in the state, it provided stakeholders with knowledge that the ACC was a sanctioned initiative that would require

PAGE 69

53 long term participation a nd cooperation (HCPF, 2009b). As seen in the Starting Conditions subsection earlier in this chapter, s ome of the same stakeholders who opposed the attempts at Medicaid reform in Colorado attended publ ic forums and contributed to the A by voicing divergent opinions (Kaiser Family Foundation, 2012; Karabatsos, 2011; Thomson et al., 2000). Participation over time achieved a level of buy in not experienced in earlier efforts (Kaiser Family Foundation, 2012; Karabatsos, 2011; Tho mson et al., 2000). small wins was acknowledged early in the process The collaborative process included the use of SDAC data and analysis to review, measure, and compare ACC e nrollment, utilization, outcomes, and costs at macro and micro levels (CMS, 2012; HCPF, 2010b; Karabatsos, 2011; State Policy Options, 2012) From the outset, focus on these aspects of the ACC contributed to an atmosphere of incremental improvement constant learning, and ongoing innovation (CMS, 2012). Satisfying the six defining criteria, the four common elements and related components, and the 10 implications for success established by Ansell and Gash (2008), the ACC appears to be an example of co llaborative governance designed with opportunity to succeed. Yet t he question remains : i s the ACC a governance arrangement matched to the specific problems it was created to address (Ostrom, 2007) ? To determine if the ACC fulfills its purpose of transformi ng service delivery for Medicaid recipients in Colorado to produce positive effects in health outcomes and cost of health care this thesis transitions from theory to practice and approaches the analysis of health outcomes and health care costs before and after implementation.

PAGE 70

54 From Theory to Practice As Ansell and Gash (2008) noted, the need exists to examine the strength and influence, if any, of collaborative governance outcomes. The authors indicated the purpose of their research and meta ana lysis was to draw positive and negative findings together in a common model that could begin to identify conditions under which collaborative governance could be said to work or not work in terms of process outcomes (Ansell & Gash, 2008). As McGuire (2006) N quantitative research, there is a growing realization that collaboration is not an end in itself and that only by thi s research study is to take the next step and examine ACC performance outcomes: changes in health outcomes and health care costs. In so doing, this thesis contributes evidence to the performance of a collaborative governance effort and the impacts on indiv iduals served. Enrollment of the first Medicaid recipients in began in May 2011, and the Department began to report favora ble outcomes in 2012 (CMS, 2012; HCPF, 2012 c ; Karabatsos, 2011; State Policy Options, 2012 ). In the annual report submitted to the Director, stated year performance resulted in reduced utilization rates for ER visits, hospital readmissions and high cost imaging services; lower rates of exacerbated chronic health conditions such as asthma and diabetes; and reduced total cost of care for enrolled Medicaid recipients ( HCPF, 2012 c p. 2). Additionally, the report indicated th e program produced positive results during the enrollment ramp up phase in which access to claims data, provider network infrastructure, and care coordination activities were sti ll under development (HCPF, 2012 c p. 4 ). The

PAGE 71

55 report ed results continuation The Department projected the ACC would demonstrate positive results in future years and concentrated its efforts to make the ACC the primary Medicaid delivery system in Colorado (HCPF, 2012 c ) T he next chapter of this thesis explains the research design and methods used to study ACC performance outcomes. This research examines actual health outcomes and health care costs for specific groups of Colorado Medicaid recipients before and after implementation of the ACC. In so doing, findin gs in the fifth chapter of this thesis contribute evidence to the performance of a collaborative governance effort and the impacts on individuals when served through such a process While collaboration presents a different way for the public, private, and nonprofit sectors to work together, it is important to understand if collaborative governance affects outcomes in measurable ways for the individuals it intends to benefit (Padgett, Bekemeier, & Berkowitz, 2004; Varda, 2011; Varda & Retrum, 2012; Varda, Sh oup, & Miller, 2012). As Varda and Retrum (2012) indicated, benefits of collaboration have become widely accepted and the practice of collaboration is growing, but e valuation rocess of healthcare which can create better outcomes, but also reduce the cost of delivering services by eliminating waste,

PAGE 72

56 CHAPTER IV RESEARCH DESIGN AND METHODS The first chapter of this thesis highlighted the significance of finding different approaches to the escalating problems associated with increasing Medicaid enrollment and rising costs. The second and third chapters then examined collaborative governance literature and compared the collaborative governa nce definition, criteria, a nd elements distilled from the Ansell and Gash (2008) meta analysis of 137 diverse case studies Satisfied that the ACC can be categorized as a collaborative governance effort, t his chapter specifies a research design to examine implications of the ACC for Medicaid recipients. In addition to contributing another example to collaborative governance literature, this thesis examines actual health outcomes and health care c approach, thereby contributing evidence about the performance of a collaborative governance process and the impacts on individuals served. While collaboration may present a different way for the public, private, and nonprof it sectors to work together, it is important to understand if collaborative governance affects outcomes in measurable ways for the individuals it intends to benefit (Padgett et al. 2004; Varda, 2011; Varda & Retrum, 2012; Varda et al. 2012). As Varda and Retrum (2012) indicated, the benefits of collaboration have become widely accepted and the practice of the ability to measure, document, and strategize to affect p. 170). To that end, this chapter fr ames the research question, hypothesis, and variables; discusses data and analysis ; and considers limitations and strategies to this approach

PAGE 73

57 Research Question, Hypothesis, and Variables Examination of the escalating problems associated with increasing Medicaid enrollment and rising costs coupled with collaborative approach led to formulation of the research question: What effects will participation in the ACC have on health care for Medicaid recipient s in Colorado? The unit of observation is the individual, the Medicaid recipient participating or not participating in the ACC T he unit of analysis is the ACC. The null hypothesis and its alternati ve follow : H 0 : Participation in the ACC is expected to have no effect o n health outcomes and cost of health care for Medicaid recipient s in Colorado. H 1 : Participation in the ACC is expected to (a) improve health outcomes and (b) reduce cost of health care for Medicaid recipient s in Colorado. To examine effects, this study used a quasi experimental nonequivalent control group research design (Singleton & Straits, 2010, p p 250 252). Two study groups were configured T hose who do not participate in the ACC are the control group, and those who do participate are the ACC group. T hree periods were construct ed : the 12 months before ACC implementation, May 1, 2010 through April 30, 2011 (T 1 ) ; the first 12 months after ACC implementation, May 1, 2011 through April 30, 2012 (T 2 ) ; and the second 12 months after ACC implementation May 1, 2012 through April 30, 2013 (T 3 ) The categorical independent variable is participation in the ACC Participation in the ACC is operationalized as enrolled Medicaid recipient s residing in a focus community county and having a n existing relationship wit h a PCMP Additional details about the research study sample appear in this Data and Analysis section This research dependent variables are health

PAGE 74

58 outcomes and cost of health care goals of improving health outcomes and controlling costs (HCPF, 2012a, 2012c ). Health Outcomes Health outcomes are operationalized using the three utilization measures related to health outcomes the ACC initially established as key performance indicators: (a) 3 0 day all cause hospital readmissions, (b) emergency room ( ER ) visits, and (c) high cost imaging servi ces (HCPF, 2010b) Hospital readmissions, ER visits, and high cost imaging services have gained national and state attention as areas in which excessive o r inappropriate use can be reduced to improve health outcomes ( Berwick et al., 2008; CHI, 2012; CMS, 2012; Patient Protection and Affordable Care Act, 2010; Stern, 2013; Zhang, Wan, Rossiter, Murawski, & Patel, 2008). In a coordinated system of health care delivery such as the ACC, reductions in these health outcome indicators would represent improved performance. Hospital readmissions, ER visits, and high cost imaging services appear in the limited data set at the Medicaid claim level, are linked to the ap propriate individual by a unique deidentified member number, and are measured as the number of occurrences. Each measure is defined and explained further in the following three paragraphs. Hospital r eadmissions The ACC defined 30 day all cause hospital readmissions a s any inpatient case that occurred within 30 days of an inpati ent discharge of an individual along with specific exclusionary criteria related to eligibility and discharge status, transfers, and interim bill and continued stay (Treo Solutions 2012, pp. 5 6). The limited data set obtained for this research study contained an indicator for potentially preventable hospital readmissions within 30 days of discharge, which was used as a proxy for all cause hospital readmissions. This indicator is d

PAGE 75

59 readmission dates, diagnosis, prior admission procedures, and readmission reason (3M Health Information Systems, 2015). ER visits. ER visits were considered any outpatient emergency department claim tha t did not have an inpatient stay on the same date of service with the sa me client identification number (Treo Solutions, 2012, p. 6). Any ER visit that resulted in an inpatient admission was excluded (Treo Solutions, 2012, p. 6). This research used the ind icator for ER visits in the acquired limited data set. High cost i maging. The ACC classified high cost imaging services as any claim in an enhanced ambulatory payment group related to computed tomography (CT) scans, a method using x rays to create cross se ctional pictures of bones and soft tissue in the body, or magnetic resonance imaging (MRI), a technique using a magnetic field and radio waves to create detailed images of organs and tissues in the body (Mayo Clinic, 2017a, 2017b; Treo Solutions, 2012, p. 7). The list of specific current procedural terminology (CPT) codes included as high cost imaging services in the ACC appears in Appendix E. The codes in Appendix E were used to identify CT scans and MRI services in the limited data set obtained for this r esearch, and an indicator was created to distinguish them from other diagnostic services. Cost of Health Care Cost of health care is o perationalized as (a ) the cost of claims paid by Medicaid for medical services, (b) the cost of claims paid by Medicaid for prescription drugs and (c ) t he per member per month (PMPM) primary care case management fee s paid to RCCOs and PCMPs, which was described in the previous chapter of this thesis (HCPF, 2010a 2012a ; Verdier et al., 2009). A mounts Medicaid paid for med ical services and prescription drugs are

PAGE 76

60 distinct fields in the limited data set. The PMPM fees for ACC members paid to RCCO s and PCMP s were calculated based on amounts published in the first two ACC annual reports as $0 for T 1 which was before the ACC implementation; $13 and $4, respectively, during T 2 ; and $10.50 and $3, respectively, during T 3 (HCPF 2012a, 2013a; Kaiser Family Foundation, 2013). P ublished amounts for the appropriate period were multiplied by the number of months a participant was in the ACC for that period, and all per participant calculations were summed. Additional information is included in the fifth chapter in the PMPM subsecti on and Table 5.11. Data and Analysis This section of the chapter provides details about t he research analysis by describing the data source and related considerations, the data sample and configuration of the research study groups characteristics of the study groups, and the approach selected for data analysis. Data Source and Considerations To examine health outcomes and cost of health ca re for this research study it was necessary to obtain historical health record and claims information for Colorado Medicaid recipien ts The repository for such data in Colorado is the all payer claims database (APCD) (All Payer Claims Database Council, 2017) In 2010 Colorado revised state statutes to create the APCD and authorized the Departme nt to oversee it (Colorado All Payer Claims Database, 2010) The Department named a nonprofit organization, the Center for Improving Value in Health Care (CIVHC), as the APCD administrator and holds CIVHC accountable for the rough monitor ing audit ing and reporting activities (All Payer Claims Database Council, 2017 ; Center for Improving Value in Health Care [CIVHC],

PAGE 77

61 2017a ) The APCD contains individual Medicaid, Medicare, Medicare Advantage, and commercial health plan insura nce claims for millions of Colorado health care recipients, and in 2013 CIVHC began releasing approved custom data requests through a process compliant with the Privacy Act of 1974, the Health Insurance Portability and Accountability Act (HIPAA) of 1996, t Health Information Technology for Economic and Clinical Health (HITECH) Act (2009), and other applicable state and federal rules and regulations (All Payer Claims Database Council, 2017; CIVHC, 2017a) The APCD p rotects by limiting use to purposes permitted by law and by restricting information to data elements reasonably necessary to accomplish an approved purpose (CIVHC, 2017c ). T he APCD and its Data Relea se Review Committee application for a li mited data set A lthough this limited data set included ZIP Codes and Medicaid eligibility and service dates, which are protected health information data elements, it did not inclu de direct individual identifiers such as name, telephone number, Social Security number, or medical record number (CIVHC, 2017b). In addition to safeguarding data through the APCD privacy and security measures, this research study preserved dat a integrity by following University of Colorado Denver information technology security program policies and all Colorado Multiple Institutional Review Board policies and procedures, which include data privacy and security and investigator responsibilities ( Colorado Multiple Institutional Review Board, 2017; University of Colorado Denver, 2017). Data Sample and Research Study Groups The data request application for this research study included Colorado Medicaid medical and prescription drug cl aims for calend ar years 2010, 2011, 2012, and 2013 for those

PAGE 78

62 individuals who were between the ages of 21 and 60 in 2010 who had fee for service Medicaid as their only source of health insurance coverage for the years in question, and whose records included physical health diagnoses or behavioral health diagnose s for anxiety or depression Behavioral health diagnoses of interest for anxiety or depression were limited to those covered under the Colorado Community Behavioral Health Services Program, a nd the list appears in Appendix F (HCPF & Colorado Department of Human Services, 2014). T he limited data set provided by the APCD consisted of 17 files containing more than 28 million records represent ing over 37 thousand unduplicated individuals (Colorad o All Payer Claims Database, personal communication, June 20, 2015). Data provided were based upon Colorado Medicaid eligibility for calendar years 2010 through 2013. All r ecords received were then examined and compared to records needed for this research study For example, some records received contained Medicare instead of Colorado Medicaid health insurance designations invalid or non Colorado ZIP C odes invalid Medicaid eligibility or service dates or behavioral health diagnoses other than those for anxiety or depression R ecords received that were not needed were then filtered from the limited data set to narrow the focus to only those Colorado Medicaid recipients of interest for this research study T he r equirements for initial ACC enrollment were considered next To be eligible for enrollment in the ACC and included in the analysis for its key performance indicators which were described in the discussion about dependent Health Outcomes subsection an individual could not have fewer than three months of Medicaid eligibility in a 12 month period (Colorad o Access, 2013 ; M. Ly and H. Schum, personal commu nication, November 16, 2015; Treo Solutions, 2012) For purposes of this research, records associated with individuals having no fewer than 45 of 48

PAGE 79

63 possible months of eligibility during the four calendar years from 2010 thr ough 2013 were selected to ensur e continuity of Medicaid eligibility The remaining pool of records provided the basis for configuring the control and ACC groups Before configuring the groups, two remaining parameters were considered: if an individual resided in one of the focus communit y counties in which the ACC was first implemented and if Medicaid claims history indicated a relationship with a PCMP The ACC was initially implemented in a limited number of focus community counties and ensured that every RCCO had some participants in at least one county in its service area (Colorado Access, 2013; Karabatsos 2011). Records for each unduplicated Medicaid recipient in the limited data set contained a ZIP Code of residence which then was associated with the appropriate county and the corresponding RCCO in the ACC Also, for an eligible Medicaid recipient in a focus community county to be enrolled in the ACC, the Department required the recipient to have a previous relationship with a PCMP ( Lindrooth, Tung, Sa The limited data set the APCD provided for this research did not contain a designation for those Medicaid providers who were ACC PCMPs; however, the APCD was able to pro vide an additional data file in November 2016 that allow ed ACC PCMPs to be identified and matched to Medicaid providers in the original limited data set This enabled configuration of the research study groups. C riteria for both groups a re summarized in Table 4.1.

PAGE 80

64 Table 4.1: Criteria for Research Study Group s Criteria C ontrol Group ACC Group Between the ages of 21 and 60 in 2010 Yes Yes Fee for service Medicaid as only health insurance coverage from 2010 through 2013 Yes Yes Claims records containing physical health diagnoses and a behavioral health diagnosis of anxiety or depression Yes Yes No fewer than 45 out of 48 possible months of Medicaid eligibility during the calendar years from 2010 through 2013 Yes Yes Resident of an ACC focus community county No Yes Established relationship with a PCMP No Yes Number of research study group members 380 2,683 Research Study Group Characteristics Characteristics of the control and ACC groups are examined in this subsection. Table 4.2 summariz es gender, generation, and RCCO distribution for each group A more detailed view appears in Appendix G. Race and ethnicity distributions were not included since i ndividuals are not required to report race and ethnicity during the ir Medicaid application process R ace and ethnicity i nformation that appeared in the limited data set was incomplete and considered unreliable (HCPF, 2015 c ; Kaiser Family Foundation, 2011 c ; Wilmot, 2006).

PAGE 81

65 Table 4.2: Summarized Characteristics of Control and ACC Groups a When p > .05, the difference between percentages in the control and ACC groups is not significant. Gender. Medicaid population aligns closely with national percentages : 58% female and 42% male (CHI 2017; Kaiser Family Foundation, 2011b). However, both the control and ACC groups reflect a sample with a higher percentage of females and lower percentage of males. As mentioned earlier in this chapter, records in the limited data set used for this research study only included Colorado Medicaid recipients between the ages of 21 and 60 in 2010 When correlated interest, approximately 35% of the entire Medicaid population is represented, which has a higher concentration of females than the total Medicaid population (Kaiser Family Foundation, 2011a). Characteristic Total Control Group ACC Group Differen c e between Groups Number (%) Number (%) Number (%) Gender Female 2,669 (87) 303 (80) 2,366 (88) 8 Male 394 (13) 77 (20) 317 (12) 8 Generation Gen Y 841 (27) 66 (17) 775 (29) 12 Gen X 1,343 (44) 167 (44) 1,176 (44) 0 a Baby Boomers 879 (29) 147 (39) 732 (27) 12 RCCO 1 328 (11) 129 (34) 199 (7) 21 2 259 (8) 27 (7) 232 (9) 2 a 3 852 (28) 22 (6) 830 (31) 25 4 542 (18) 173 (45) 369 (14) 32 5 457 (15) 0 (0) 457 (17) 17 6 355 (11) 11 (3) 344 (13) 10 7 270 (9) 18 (5) 252 (9) 4 Total 3,063 (100) 380 (100) 2,683 (100)

PAGE 82

66 The control group has a lower percentage of female participants than the ACC group. Since females have a higher rate of health care utilization than males and incur greater health care costs, this difference could relate to higher health outcomes and lower cost of health care for the control group compared to the ACC group (Cylus, Hartman, Washington, Andrews, & Catlin, 2010). Generation. For purposes of t his research study, the age range was categorized as three generations: Gen Y, whose members were born between 1982 and 2000; Gen X, whose members were born between 1965 and 1981; and Baby Boomers, whose members were born between 1946 and 1964 (United States Census Bureau, 2015). Also, be cause this research study focused on adults between the ages of 21 and 60 in 2010 the limited data set contained Gen Y members with birth years only between 1982 and 1989. Taking those factors into consideration, gender and generation distributions were more consistent with what percentages would be in a similar national sample that excluded individuals under the age of 21 and over the age of 60 (Kaiser Family Foundation, 2011a 2011b ). The control group also has a lower percentage of Gen Y participants with a corresponding higher percentage of Baby Boomer participants than the ACC group. Because older Medicaid recipients typically have more complex health conditions than younger recipients, their overall service utilization and use of higher cost service s tend to be greater (Cylus et al., 2010). As a result, this difference could relate to lower health outcomes and higher cost of health care for the control group compared to the ACC group. RCCO. Data Sample and Research Study Groups subsec tion explain ed that records for each unduplicated Medicaid recipient in the limited data set contained a ZIP Code of residence, which then was associated with its appropriate county to

PAGE 83

67 determine the designated RCCO assignment. T he table of RCCOs and counties and the state map in Appendix A and Appendix B, respectively, indicate the distribution of RCCOs across Colorado: RCCO 1 serves the west; RCCO 2 covers the northeast; RCCO 4 operates in the southeast; RCCO 5 is the Denver met ropolitan area; and RCCOs 3, 4, and 6 cluster around RCCO 5 and cover the center of the state. Table 4.2 indicates a significant difference in RCCO distribution between the groups in six of the seven regions and the control group contained no participan ts in RCCO 5 the control distribution is heavily weighted in more rural areas of the state while the ACC distribution is concentrated in more urban areas. As described in the Institutional Design subsection of the third chapter of this thesis, RCCOs were configured to normalize influences, such as rural and urban geography, which otherwise might negatively impact access to and equality of care. However, the differences and potential impact on health outcomes and cost of health care are worth noting before the discussion of findings in the next chapter Approach to Analysis This research study utilized SPSS 23.0 software for both descrip tive and inferential statistical analysis of the data. Table 4.2 summarize s the sample and provide s information about the characteristics of the research study groups (Healey, 1999) For example, c rosstabulations were used to illustrate frequency distribut ions of the variables of interest in both research study groups. Inf erential statistics examine d relationships among variables between the two research study groups within each research study group over time and among RCCOs Health outcome variables are categorical and indicate whether or not, based

PAGE 84

68 on definitions provided at the beginning of this chapter, a hospital readmission, ER visit, or high cost imaging service occurred. Chi square tests were utilized f or categorical variables (Hays, 1994; Healey, 1999). For continuous variables related to the cost of health care, such as claims paid by Medicaid for medical services and for prescription drugs, independent t tests were employed (Hays, 1994; Heal e y, 1999). The limited data set did not contain PMPM f ees paid to RCCOs and PCMPs for each ACC member ; therefore, amount s calculated for ACC group participants for T 2 and T 3 were based on published amounts specified earlier in the Research Question, Hypothesis, and Variables section of this chapter. As noted, the PMPM amount was not the same in T 3 as in T 2 Results of statistical tests and calculations appear in the next chapter of this thesis in the discussion of this Limitations and Strategies After considering the research question and hypothes i s along with data and analytical approaches, this section of the thesis addresses possible limitations of the research study and mitigating strategies. In research, validity refers to the relationship between premises and conclusions (Singleton & Straits, 2010). In assessing research results, it is necessary to determine if the independent variable influences the dependent variable and if a causal inference can be made (Singleton & Straits, 2010). Over time scholars developed a typolog y for assessing validity in experimental or quasi experimental research studies, and the criteria are discussed at greater length in the Four Aspects of Validity subsecti on later in this chapter. This research design supports valid causal inference in its examination of the relationship betwe en the ACC and health outcomes in two ways. It satisfies the three criteria for causal inference, and it addresses the mitigation of threats to internal validity and ensures statistical,

PAGE 85

69 construct, and external validity The next two subsections elaborate on the three criteria for causal inference and the four aspects of validity as they relate to this research study Three Criteria for Causal Inference Three elements of causality confirm validity: direction of influe nce, association, and nonspuriousness (Singleton & Straits, 2010). First, direction of influence reflects a temporal element: cause precedes effect (Singleton & Straits, 2010; Van de Ven, 2007 ). Assessing health outcomes and cost of health care for Colorad o Medicaid recipients durin g the 12 the first 12 months after implementation, and the second 12 months after implementation as described in the Research Question, Hypothesis, and Variables section of this chapter, establish es a temporal sequence that supports direction of influence. Second, association illustrates a relationship between the cause or indep endent variable and its effect or dependent variable (Parker, 1993; Singleton & Straits, 2010; Van de Ven, 2007) Although the effects on Medicare are more pronounced, an aging population and increased life expectancy also influence Medicaid: personal income typically declines, enrollment grows, health conditions worsen, cost of health care increases, and health out comes decline (Joffe, 2015). Some effects may also be disproportionate. For example, b efore development and implementation of the ACC elderly and individuals with disabilities account ed for one quarter of Medicaid recipients but almost two th irds of its expendi tures (CHI 2005 ). Aging Me dicaid recipients or those with a disability typically need more expensive services such as long term care, prescription drugs, and specialty ser vices ( Burger & Stapleton, 2009; CHI 2005 ; Rosenbaum, 2011 ). Colorado anticipated that

PAGE 86

70 delivery system for Medicaid recipients would have positive impacts ( HCPF, 2010a, 2010b; Kaiser Family Foundation, 2012 ). For example, i t was estimated that participation in the ACC would improve health outcomes by increasing follow up procedures after hospital discharges, thereby decreasing hospital readmissions; by increas ing preventive and primary care, consequently decreas ing ER visits ; and by increasing use of X rays and low cost diagnostic alternatives, thus decreasing unnecessary high cost CT scans and MRI services ( Karabatsos, 2011; State Policy Options, 2012; Treo Solutions, 2012 ). Third, nonspuriousness indicates the maintenance o f the relationship when all extraneous variables are held constant and rival hypotheses are eliminated (Singleton & Straits, 2010; Van de Ven, 2007). Identifying and isolating relevant extraneous antecedent and intervening variables assist in eliminating c onfounding effects and substantiating nonspuriousness (Parker, 1993; Van de Ven, 2007). Typically, health outcomes decline and health care costs escalate for Medicaid recipients over time while other factors vary little ( Joffe, 2015 ). For example, the over gender, ethnicity, geography, and income fluctuated little during the years examined in this research study ( CHI 2017 ). As described in the Data Sample and Research Study Groups subsection earlier in this chapter, the control and ACC groups were selected from a limited data set of unduplicate d Colorado Medicaid recipients similar except for ACC participation which was determined by residence in a focus community county and a rel ationship with a PCMP. Similarity in all other aspects of the two research study groups would tend to support nonspuriousness. Yet, some variation is seen between the groups in gender, generation, and RCCO distribution illustrated in Table 4.2 D ifferences other than ACC participation could

PAGE 87

71 indicate the presence of extraneous fa ctors that may confound statistical test results This is considered in the next chapter during the discussion of research findings Four Aspects of Validity Different aspects of v alidity are typically classified in four categories: internal, statistical conclusion, construct, and external (Singleton & Straits, 2010; Van de Ven, 2007). Internal validity examines if covariance is the result of a causal relationship (Parker, 1993; Van de Ven, 2007). To a large degree, research design aspects discussed in the previous subsection, which aim to satisfy the three criteria for causal inference, address internal validity. Statistical conclusion validity is concerned with appropriate use of s tatistics to determine covariance between the independent and dependent variables ( Shadish, Cook, & Campbell, 2002; Van de Ven, 2007). Construct validity is concerned with the generalizability n, 2007). External validity 2010; Van de Ven, 2007). To highlight potential pitfalls, the following paragraphs provide examples of potential threats to validity in the four general categories as they relate to this research study Internal. T h reats to internal validity considered for this research study included history, maturation, testing, regression, instrumentation, selection and interactions with selection and mortality experimental nonequivalent control group design naturally mitigate d the four threats of history, maturation, testing, and regression, and the use of pre ACC implementation a nd post ACC implementation periods furth er contro lled the threat of history and strengthen ed the research design (Singleton & Straits, 2010; Van de Ven, 2007). However, three remaining threats were assessed.

PAGE 88

72 The threat of i nstrumentation primarily relates to the process of data collection (Singleton & Straits, 2010) As explained in the Data Source and Considerations subsection earlier in this chapter t his research study did not collect data but obtained a limited data set of historical health record and claims information for Colorado Me dicaid recipients from the APCD of such data (All Payer Claims Database Council, 2017). As described in th Data Sample and Research Study Groups subsection, the limited data set which was based upon Colorado Medicaid eligibility for calendar years 2010 through 2013, was examined u pon receipt to determine if files contained the information requested. Some records received contained Medicare instead of Colorado Medicaid health insurance designations, invalid or non Colorado ZIP Codes, invalid Medicaid eligibility or service dates, and behavioral health diagnoses other than those for anxiety or depression. R ecords received that were not needed were then filtered from the limited data set to narrow the focus to onl y those Colorado Medicaid recipients of interest for this research study Contents of some data fields were incomplete and could not be used. For example, as mentioned earlier in th Research Study Group Characteristics subsection race and ethn icity distributions were not included because individuals are not required to report race and ethnicity during their Medicaid application process. Where possible, data demographic, enrollment, and service distributions were verified against published state reports to conclude that the data were reasonable Next, s election problems can occur when participants self select or are assigned to research study groups based on their preferences (Parker, 1993). Although neither self selection nor preferential assignment occurred in this study, concern about differences in gender, generation, and RCCO distribution between the two research study groups was

PAGE 89

73 expressed in the earlier discussion of nons puriousness in the Three Criteria for Causal Inference subsection in this chapter. F urther consideration is given to these differences during the discussion of research findings in the next chapter. Finally, t he threat of mortality refers to the loss of p articipants during the study and the effects of attrition on results (Parker, 1993; Singleton & Straits, 2010). However, as explained in the Data Sample and Research Study Groups subsection earlier in this chapter, records for unduplicated individuals incl uded in the two research study groups were Medicaid recipients with no fewer than 45 of 48 possible months of eligibility during the four calendar years from 2010 through 2013 ; the same individuals were in the control and ACC groups at the beginning of thi s research study and at the end. S tatistical conclusion Threats to statistical conclusion relate to the appropriate use of statistics to arrive at accurate decisions about accepting or rejecting research study hypotheses (Parker, 1993). These threats were lessened in this research study by careful examination of statistical procedures and the assumptions used (Shadish et al., 2002). Mindful that statistical power is a function of the level of significance, sample size, and effect size this research st udy was designed with reliable measures and sufficient power (Parker, 1993; Van de Ven, 2007). Also, participants in the two research study groups were similar, which minimized the threat of random heterogeneity ( P arker, 1993). Construct. Potential threat s to construct validity in this research study include d inadequate explication, mono operation bias, and researcher expectancies (Parker, 1993; Van de Ven, 2007). Specifically defining and accurately measuring study variables minimize d the first threat (Pa rker, 1993; Van de Ven, 2007). Measures in this research study were sufficiently defined and operationalized before data receipt and analysis so that results could

PAGE 90

74 be accurately attributed to the constructs of interest. As mentioned previously in the Resea rch Question Hypothes i s and Variables section of thi s thesis health outcomes and cost of health care were operationalized using multiple factors to mitigate the possibility of mono operation bias (Parker, 1993). Also, a rigorous approach to quantitative analysis and review lessen ed any negative effect that research er expectancies might have had on the interpretation of results. External. Concern for external validity relates to generalizability of a research study (Singleton & Straits, 2010). The ability to generalize research findings across persons, settings, and time requires research study samples representative of the population of interest (Parker, 1993). Thi s research study minimized threats to external validity by obtaining a sample Medicaid population in the age group of interest, in a range of geographic settings across the state, and across time before, during, and after imple mentation of the ACC (Parker, 1993; Shadish et al., 2002; Van de Ven, 2007). While it is un realistic to expect that all threats to internal, statistical conclusion, cons truct, and external validity could be eliminated in a ny research study, awareness of t h reats likely to occur aid in the examination and selection of strategies to impr ove the research design and interpretation of its results ( Parker, 1993; Singlet on & Straits, 2010 ). xplanation of th design and methods set s the stage for discussion of its f indings which appear in the next chapter of this thesis. Examination of health outcomes and health care costs for specific groups of Colorado Medicaid recipients before and after implementation contributes evide nce to the performance of a

PAGE 91

75 collaborative governance effort and the impacts on individuals when served through such a process.

PAGE 92

76 CHAPTER V RESEARCH STUDY FINDINGS is organized in four primary sections, one related to health outcomes another associated with cost of health care the third concerning results over time and the fourth exploring regional differences Results from s tatistical analysis of each dependent variable are summarized Differences b etween groups in each of the three periods as well as differences within each group between the first and third periods are considered. O utcomes and explanatory tables are included and issues related to practical interpretation of test results are raised. This c hapter concludes with observations about the Health Outcomes In the Research Question, Hypothesis, and Variables section in the previous chapter, this thesis noted that health outcomes for this research study are operationalized as (a) potentially preventable hospital readmissions wi thin 30 days of discharge (b) ER visits, and (c) high cost imaging services. These three utilization mea sures relate to health outcomes the ACC initially e stablished as key performance indicators (HCPF, 2010b). As noted in the Health Outcomes subsection in the fourth chapter of this thesis, these measures would be expected to show a reduction over time in a coordinated health care delivery system like the AC C. Reductions would correspond with improved health outcomes. E ach of the dependent variables operationalizing health outcomes is categorical and indicates whether or not, based on the definitions provided at the beginning of the previous chapter a hospi tal readmission, ER visit, or high cost imaging service occurred. These services appear in the limited data set at the Medicaid claim level, are linked to the

PAGE 93

77 appropriate individual by a unique deidentified member number, and are measured as the number of occurrences. It is possible that one Medicaid recipient could have more than one of these services in any period. I ndependence of observations exists in each group F requency tables were generated. C hi square g oodness of fit results were reported based on unequal proportions which were determined before frequency tables were produced and supported by reasonable assumptions ( Hays, 1994; Healey, 1999). chi square test for association was used to determine if a relationship existed between participa tion in the ACC and potentially preventable hospital readmissions within 30 days of discharge, ER visits, and high cost imaging (Hays, 1994; Healey, 1999) The strength of associations was assesse d using the p hi coefficient (Healey, 1999) For each of the three dependent variables representing health outcomes, supporting tables along with anticipated and actual results appear in the following subsections. Hospital R eadmissions Reducing hospital readmissions that occur within 30 days of discharge is considered an appropriate measure for a collaborative health care system seeking to improve coordination across inpatient, outpatient, and follow up services (HCPF, 2012a 2016a ). As explained in the fourth chapter of this thesis, this research study used the indicator for potentially preventable hospital readmissions within 30 days of discharge as a proxy for all cause hospital rea dmissions discharge and readmission dates, diagnosis, prior admission procedures, and readmission reason (3M Health Information Systems, 2015). Potentially preventable hospital readmission rates vary due to factors such as age, health status, and number and type of chronic conditions. Relying on national research studies th at have calculated reasonable averages for

PAGE 94

78 various populations and subpopulations, the assumption was made in this research study that 20% of all hospital readmissions within 30 days of discharge would be potentially preventable for the Med icaid recipients included in this study ( Agency for Healthcare Research and Quality, 2015; Donz, Aujesky, Williams, & Schnipper, 2013; Goldfield et al., 2008). As a result, 20% was used for weighting unequal proportions during statistical analysis. Additionally, the indi cator only applies to inpatient claims O utpatient and professional claims are not applicable for analysis of this variable and they were excluded Table 5.1 illustrates the frequency of potentially preventable hospital readmissions within 30 days of disc harge compared to other hospital readmissions within the s ame amount of time in both groups for each of the three periods. Table 5.1: Potentially Preventable Hospital Readmissions (PPHR) -------------T 1 --------------------------T 2 -------------------------T 3 -------------PP HR Control Group ACC Group Total Control Group ACC Group Total Control Group ACC Group Total Number (%) Number (%) Number (%) Number (%) Number (%) Number (%) Number (%) Number (%) Number (%) Yes 2 (2) 151 (12) 153 (11) 8 (15) 136 (10) 144 (10) 5 (6) 122 (9) 127 (9) No 79 (98) 1,152 (88) 1,231 (89) 54 (87) 1,280 (90) 1,334 (90) 82 (94) 1,178 (91) 1,260 (91) Total 81 (100) 1,303 (100) 1,384 (100) 62 (100) 1,416 (100) 1,478 (100) 87 (100) 1,300 (100) 1,387 (100) If the control and ACC groups were similar except for ACC participation, no significant difference would be expected between the groups in T 1 but the ACC potentially preventable hospital readmissions would be expected to begin decreasing during T 2 and continue decreasing during T 3 due to improved RCCO and provider coordination across inpatient, outpatient, and follow up services (HCPF, 2012a, 2016a) A similar decrease

PAGE 95

79 would not be expected in the control group since its participants would not be subject to care coordination provided by the ACC Table 5.2 summarizes t he results for each period Table 5.2: ACC Participation and PP HR Measure T 1 T 2 T 3 Chi s quare Value 69.213 97.186 101.929 Chi s quare Significance a b .000 .000 .000 Pearson Value 6.450 .735 1.297 Pearson Significance b .011 .391 .255 Phi Value .068 .022 .031 Phi Significance b .011 .391 .255 a When p < .05, observed frequencies in the sample are not the same as expected frequencies in the population; minimum PPHR observed were fewer than expected b When p < .05, significant difference exis ts. However, as Table 5.2 illustrates, a significant difference in potentially preventable hospital readmissions existed between the control and ACC groups in T 1 implementation but no significant difference appeared to exist between the groups during the first two years of the ACC As a result, no inference can be drawn that the change was attributable to ACC participation. Additionally, the strength of the relationship was negligible in all three periods although it was significant in T 1. ER V isits The second key performance indic ator the ACC selected to measure health outcomes was ER visits not result ing in a hospital admission (HCPF, 2010a, 2010b, 2012a ). Before the through a fee for service system that supported episodic rather than coordinated service delivery, which led many recipients to ER rather than primary care provider visits (HCPF, 2010a, 2010b, 2012a). For reporting based on unequal proportions, it is reasonable to

PAGE 96

80 presume that 40% of all medical service claims would be ER visits (Centers for Disease Control and Prevention, 2016; CHI 2012). The limited data set utilized for this research study contains service dates and an indicator for ER visits The categorical indicator appears at the claim level and specifies whether or not the claim was an ER visit fitting the criteria in the previous paragraph. Claims are linked to the appropriate individual by a unique deidentified member number and are measu red as the number of occurrences. It is possible that one Medicaid recipient could have more than one ER visit in any period. T able 5.3 reflects the number of ER visits compared to other inpatient, outpatient, and professional medical services within the s ame period for both groups in T 1 T 2, and T 3 Table 5.3: ER Visits -------------T 1 --------------------------T 2 -------------------------T 3 -------------ER Visits Control Group ACC Group Total Control Group ACC Group Total Control Group ACC Group Total Number (%) Number (%) Number (%) Number (%) Number (%) Number (%) Number (%) Number (%) Number (%) Yes 502 (7) 5,924 (8) 6,426 (8) 476 ( 6) 6,029 (8) 6,505 (7) 490 (6) 5,888 (7) 6,378 (7) No 6,922 (93) 65,744 (92) 72,666 (92) 7,278 (94) 73,150 (92) 80,428 (93) 7,817 (94) 73,932 (93) 81,749 (93) Total 7,424 (100) 71,668 (100) 79,092 (100) 7,754 (100) 79,179 (100) 86,933 (100) 8,307 (100) 79,820 (100) 88,127 (100) No significant difference between the groups would be expected in T 1 if the groups were similar except for ACC part icipation. Compared to ER visits in the control group, ER visits in the ACC group would be expected to decrease during T 2 and continue decreasing during T 3 due to strengthened PCMP relationships and extended evening and weekend PCMP hours of operation (HCPF, 2010a, 2012a) St atistical tes t results in Table 5.4 indicate a significant difference in ER visits between the groups in T 1 T 2 an d T 3 Because a significan t

PAGE 97

81 difference was present in T 1 differences in T 2 and T 3 cannot be attributed solely to the ACC. Also, Table 5.4 indicates the nificant in all three periods. Table 5.4 : ACC Participation and ER Visits Measure T 1 T 2 T 3 Chi square Value 33,483.39 38,300.14 39,414.41 Chi square Significance ab .000 .000 .000 Pearson Value 20.386 22.214 24.481 Pearson Significance b .000 .000 .000 Phi Value .016 .016 .017 Phi Significance b .000 .000 .000 a When p < .05, observed frequencies in the sample are not the same as expected frequencies in the population; min imum ER visits observed were few er than expected. b When p < .05, significant difference exists. High cost I maging The third key performance indicator measur ing health outcomes in the ACC was high cost imaging which was defined as CT scans and MRI services (HCPF, 2010a, 2010b, 2012a). Used appropriately, a lternatives such as X rays and ultrasound prove to be as efficacious and are more cost effective than CT scans and MRI s ervices (HCPF, 2012a; Mayo Clinic, 2017c; Treo Solutions, 2012 ). Reporting based on unequal proportions used a reasonable assumption tha t 35% of all diagnostic imaging services would be CT scans or MRI services ( Regents Health Resources, 2012; Smith Bindman, Miglioretti, & Larson, 2008). Each claim record in t he limited data set used for this research study contains a field with the CPT code for the medical service procedure performed CPT codes for diagnostic imaging procedures range from 70010 through 76499 (American Medical Association,

PAGE 98

82 2017). Claims for s ervices with those specific codes were filtered from the remainder of the medical services in the data. Appendix E contain s the list of specific CT scan and MRI service CPT codes used to identify high cost imaging services in the ACC A n i ndicator was created in the data set for this research study to distinguish CT scans and MRI s ervi ces from other diagnostic imaging procedures Table 5.5 includes the frequency of high cost imaging services compared to other diagnostic services for the control and ACC groups for each of the three periods The categorical indicator for high cost imaging services appears at the claim level and specifies whether or not the service fit the CPT code criteria outlined above. Claims are linked to the appropriate individual by a unique deidentified member number and are measured as the number of occurrences. It is possible that one Medicaid recipient could have more than one high cost imaging service in any period. Table 5.5: High cost Imaging -------------T 1 --------------------------T 2 --------------------------T 3 -------------CT and MRI Control Group ACC Group Total Control Group ACC Group Total Control Group ACC Group Total Number (%) Number (%) Number (%) Number (%) Number (%) Number (%) Number (%) Number (%) Number (%) Yes 388 (39 ) 4,060 ( 3 2) 4,448 (33 ) 255 (26 ) 2,732 (22 ) 2,987 (22 ) 201 (1 9 ) 2,545 (20 ) 2,746 (20 ) No 619 (61 ) 8,438 (68 ) 9,058 (67 ) 713 (74 ) 9,875 (78 ) 10,589 (78 ) 879 (81 ) 9,940 (80 ) 10,819 (80 ) Total 1,007 (100) 12,498 (100) 13,50 5 (100) 968 (100) 12,607 (100) 13,575 (100) 1,080 (100) 12,485 (100) 13,565 (100) If the control and ACC groups were similar except for ACC participation, no significant difference between the groups would be anticipated in high cost imaging services in T 1 T utilization would be expected to decrease during T 2 with continued decrease during T 3 due to an increase in the alternative use of X rays and ultrasound and

PAGE 99

83 possibly a decrease in diagnostic testing due to overall health improvement (HCPF, 2 012a; Mayo Clinic, 2017c; Treo Solutions, 2012). T able 5.6 summarizes the results. Table 5.6: ACC Participation and High cost Imaging Measure T 1 T 2 T 3 Chi square Value 25.290 1,007.857 1,298.430 Chi square Significance ab .000 .000 .000 Pearson Value 15.418 11.436 1.936 Pearson Significance b .000 .001 .164 Phi Value .034 .029 .012 Phi Significance b .000 .001 .164 a W hen p < .05, observed frequencies in the sample are not the same as expected frequencies in the population; minimum high cost im aging services observed were few er than expected. b When p < .05, significant difference exists. Statistical tests indicated a s ignificant difference in high cost imaging utilization between both groups in T 1 and T 2 but not in T 3 Because a significant difference was present in T 1 2 could not be attributed exclusively to the ACC. Although the strength of the r elationship was negligible in all three pe riods, it was significant in T 1 and T 2 In r eviewing health outcome results between the two groups in each of the three periods insufficient evidence exists to attribute changes in outcome indicators solely to participation in the ACC. R esults between groups were significantly different in potentially preventable hospitalization readmissions ER visits were significant ly different in all three periods, and high cost imaging was significant ly different in the first two periods. The magnitude of the differences was negligible. O utcomes suggest other influences may be responsible. For example, dissimilarities may exist between the gro ups in demographic

PAGE 100

84 characteristics or health status or a statewide focus on health improvement and health initiatives in Colorado may have influenced not only ACC participants but also other Medicaid recipients Before addressing results over time and reg ional differences the next section examines differences in cost of health care between the control and ACC groups in each period Cost of Health Care As discussed in the Research Question, Hypothesis, and Variable s section in the previous chapter of this thesis, cost of health care is operationalized in this research study as (a) the cost of claims paid by Medicaid for medical services, (b) the cost of claims paid by Medicaid for prescription drugs, an d (c) PMPM fee s paid to RCCOs and PCMPs in T 2 and T 3 Each of the dependent variables operationalizing cost of health care is continuous, and independence of observations exists in each group. Frequency tables were generated. The independent samples t test was used to d etermine if a difference existed between the mean or average of each continuous dependent variable of interest in the control and ACC groups (Hays, 1994). The mean difference was examined for significance, and 95% confidence intervals were generated to confirm the reasonableness of the mean differences for each variable in each period (Hays, 1994; Healey, 1999). E quality or homogeneity of variances in the variables for the two groups was tested. Results from Lev homogeneity existed and results from in the absence of homogeneity (Hays, 1994; Institute for Digital Research and Education, 2017). Quantile quantile and box plots were generated to conduct a vis ual inspection for outliers, and statistical n ormality of the distributions was determined using the Kolmogorov Smirnov test

PAGE 101

85 (Hays, 1994; Institute for Digital Research and Education 2017 ). In all cases, distributions were not normal due to some above ave rage medical service and prescription drug claim amounts. Outliers in inpatient, outpatient, and professional claims were examined and their presence and amounts were not considered inaccurate or unusual. Additionally, sample sizes of both the number of g roup participants and claims exceeded 100 and were deemed large enough for normality assumptions to be less important (Hays, 1994). Therefore, outliers were not excluded from the analysis. Su pporting tables along with anticipated and actual results for co st of health care represented by medical service and prescription drug claims appear in the next two subsections. The third component in cost of health care is PMPM fees paid to RCCOs and PCMPs in T 2 and T 3 Calculations for PMPM fees appear in the third subsection which is followed by the fourth subsection, Total Cost of Health Care. Medical Services For purposes of analysis in this research study, the cost of medical services is the total cost of claims paid by Medicaid for inpatient, outpatient, and p rofessional services in each period. The ACC anticipated that a coordinated health care system with primary care at its center would improve health outcom es and reduce costs (HCPF, 2012c ; Karabatsos, 2011). A systematic approach to increasing the focus on preventive and primary care, r educing utilization of unnecessary episodic care and duplicative specialty care services, substituting expensive diagnostic tests with less costly alternatives, and decreasing hospital readmissions by managing transitions from inpatient care was viewed as a pathway that would support improved health outcomes while lowering costs (Berwick et al., 200 8 ; Fishe r et al., 2007;

PAGE 102

86 HCPF, 2012c ; Karabatsos, 2011; Wagner et al., 2006). Table 5.7 includes the frequency of medical service cl aims paid by Medicaid for both groups for each of the three periods. Table 5.7: Medical Service Claims -------------T 1 ------------------------T 2 ------------------------T 3 -------------Claims Control Group ACC Group Total Control Group ACC Group Total Contr ol Group ACC Group Total Number 7,424 71,668 79,092 7,754 79,179 86,933 8,307 79,820 88,127 Total Amount in Millions $2.44 $21.81 $24.25 $2.39 $23.76 $26.14 $2.62 $24.07 $26.69 Average Paid Amount $328.88 $304.32 $306.63 $307.73 $300.06 $300.74 $315.39 $301.53 $302.83 If both groups were similar except for ACC participation, no significant difference would be anticipated in the average paid medical service claim amount between the groups in T 1 However, some decrease could be expected between groups in T 2 and T 3 due t o the Table 5 .8 summarizes the results. Table 5.8: ACC Participation and Cost of Medical Service Claims Measure T 1 T 2 T 3 Levene's test for equality of variances a .466 .417 .991 T test for equality of means b .175 .689 .351 Mean difference 24.553 7.674 13.859 95% confidence interval of difference (lower, upper) ( 10.935, 60.042) ( 29.909, 45.258) ( 15.258, 42.977) Kolmogorov Smirnov test of normality c .000 .000 .000 a When p > .05, b When p < .05, significant difference exists between the average paid claim amount for the control and ACC groups c When p < .05, distribution is not normal.

PAGE 103

87 As seen in Table 5.8, no significant difference exists between the average paid cla im amount for medical services between the groups in any of the three periods. Prescription Drugs The cost of claims paid by Medicaid for prescription drugs is also included in the cost of health care for this research study. Collaborative health care delivery systems such as t he ACC anticipate that comprehensive prescription drug reviews and coordinat ed medication management services will increase proper prescription use and adherence and contribute to reduced adverse medication events often resulting from ER visits or hospital readmissions ( Amara, Adamson, Lew, & Slonim, 2014; HCPF, 2012a; Smith, Bates, & Bodenheimer, 2013; Wagner et al., 2006). Consistent use of medications for a chronic condition such as hypertension, for example, can contribute to prevention or delayed onset of more serious complications like cardiovascular illness or ren al disease (HCPF, 2012a) Table 5.9 contains the frequency of prescription drug claims paid by Medicaid for the control and ACC groups during T 1, T 2, and T 3. Table 5.9: Prescription Drug Claims -------------T 1 --------------------------T 2 -------------------------T 3 -------------Claims Control Group ACC Group Total Control Group ACC Group Total Control Group ACC Group Total Number 20,504 131,370 151,874 23,150 146,675 169,825 22,690 148,076 170,766 Total Amount in Millions $1.13 $6.83 $7.96 $1.45 $8.49 $9.94 $1.37 $9.47 $10.83 Average Paid Amount $55.05 $51.99 $52.40 $62.54 $57.87 $58.51 $60.18 $63.93 $63.43 If the control and ACC groups were similar except for ACC participation, no significant difference would be anticipated between the groups in the average paid

PAGE 104

88 prescription drug claim amount in T 1 However, some difference could be expe cted in T 2 and T 3 on participants in the ACC group Table 5.10: ACC Participation and Cost of Prescription Drug Claims Measure T 1 T 2 T 3 Levene's test for equality of variances a .289 .028 .000 T test for equality of means b .030 .000 .021 Mean difference 3.058 4.678 3.747 95% confidence interval of difference (lower, upper) (.301, 5.816) (1.758, 7.599) ( 6.922, .572) Kolmogorov Smirnov test of normality c .000 .000 .000 a riances. When p < .05 equal variances can not be assumed and results u reported. b When p < .05, significant difference exists between the average paid claim amount for the control and ACC groups c When p < .05, distribution is not normal. Results in Table 5.10 indicate a significant difference between the average paid claim amount for prescription drugs between groups in all three periods. average claim amount was lower in T 1 and T 2 but sig nificantly higher than in T 3. However, this change could be attributable to other factors. For example, increased medication compliance for chronic conditions such as asthma, diabetes, or hypertension can lead to initial increases in pr escription utilization (HCPF, 2012a). In such cases, ACC group participants may not have been obtaining or utilizing needed medications increased cost attributable to volume, individual drug price increases occur fr om year to year (CMS, 2015a; MACPAC, 2016).

PAGE 105

89 PMPM As described in the Research Question, Hypothesis, and Variables section in the previous chapter, PMPM c ase management fees paid to RCCOs and PCMPs were calculated for the ACC group as $0 for T 1 respectively, during T 2 ; and $10.50 and $3, respectively, during T 3 (HCPF, 2012a, 2013a). a portion of the PMPM was redirected to an incentive pool that could be earned as additional reimbursement based on regional performance in improvement of health outcomes (HCPF, 2012a). The possibility of earning back the incentive was intended to prompt more effective RCCO and PCMP behavior and to encourage a continued focus on increasing service value rather than volume and reducing the cost of health care (CMS, 2012; HCPF, 2010a, 2012a). No ACC group participants had fewer than 45 out of 48 possible months of Medica id eligibility during the calendar years from 2010 through 2013, but some had varying mont hs of eligibility in each 12 month period. Table 5.11 ill ustrates the calculations and PMPM total s for T 2 and T 3 Table 5.11: ACC Group PMPM Calculations --------------------T 2 ---------------------------------------T 3 -------------------Eligibility in Months Number of Enrollees RCCO PCMP PMPM Number of Enrollees RCCO PCMP PMPM 9 17 $13 $4 $2,601 33 $10.50 $3 $4,010 10 27 $13 $4 $4,590 31 $10.50 $3 $4,185 11 50 $13 $4 $9,350 55 $10.50 $3 $8,168 12 2,589 $13 $4 $528,156 2,564 $10.50 $3 $415,368 Total 2,683 $544,697 2,683 $431,730

PAGE 106

90 Total Cost of Health Care For purposes of this research study, the total cost of health care is the total cost of medical service claims, pr escription drug claims, and PMPM amounts for ACC group participants in T 2 and T 3 The ACC anticipated that its collaborative health care syste m would improve health outcomes for Medicaid recipients and reduc e costs (HCPF, 2012c ; Karabatsos, 2011). Original estimates anticipated the ACC could save Colorado up to $14 million per year by decreasing avoidable, duplicative, or unnecessary use of hea l th care resources and establishing incentives to prompt increased care coordination (Karabatsos, 2011; State Policy Options, 2012). Summarizing the individual components of health care cost reflected earlier in Tables 5.7, 5.9, and 5.11, Table 5.12 illustr ates the total cost of health care for both groups in each period Table 5.12: Total Cost of Health Care ----------T 1 ------------------------T 2 ----------------------T 3 ------------Category Control Group ACC Group Total Control Group ACC Group Total Control Group ACC Group Total Medical service claims a $2.44 $21.81 $24.25 $2.39 $23.76 $26.14 $2.62 $24.07 $26.69 Prescription drug claims a $1.13 $6.83 $7.96 $1.45 $8.49 $9.94 $1.37 $9.47 $10.83 PMPM a $0 $0 $0 $0 $0.54 $0.54 $0 $0.43 $0.43 Total a $3.57 $28.64 $32.21 $3.84 $32.79 $36.62 $3.99 $33.97 $37.95 Unduplicated individuals 380 2,683 3,063 380 2,683 3,063 380 2,683 3,063 Average annual cost per individual $9,395 $10,675 $10,516 $10,105 $12,221 $11,956 $10,500 $12,661 $12,390 a Amounts are expressed in millions.

PAGE 107

91 If the control and ACC groups were similar except for ACC participation, no significant difference in total cost of health care would be anticipated during T 1 but some difference could be expected during T 2 and T 3 Table 5.13 illustrates no significant difference between the groups in any of the three periods Table 5.13: Tot al Cost of Health Care Differences between Groups Measure T 1 T 2 T 3 Levene's test for equality of variances a .869 .503 .229 T test for equality of means b .286 .132 .110 Mean difference 1,196.366 1 760.207 1,974.855 95% confidence interval of difference (lower, upper) ( 3 394.514, 1001.781) ( 4 051.155, 530.742) ( 4,394.563, 444.853) Kolmogorov Smirnov test of normality c .000 .000 .000 a b When p < .05, significant differen ce exists between the average total cost of health care for the control and ACC groups c When p < .05, distribution is not normal. In reviewing cost of health care results between the control and ACC groups in each of the three period ll hypothesis. No significant differences appeared between the groups in medical services claims or total cost of health care. Total cost of health care included PMPM fee s, which applied only to the ACC group However, the additional expense produced no significant effect between the groups in total cost of health care. Prescription drug claim differences were not as expected in T 1 average b eing sig nificantly average Ad average was significantly lower than average in T 2 it was followed by a significantly higher average for the ACC group than for the contr ol group in T 3 As mentioned earlier in this chapter during the discussion of health outcomes, results between groups were not

PAGE 108

92 conclusive enough to be attributed exclusively to participation in the ACC and other influences may be responsible Before discussing regional differences among RCCOs differences in health outcomes and cost of health care are examined between T 1 and T 3 within the control group and within the ACC group. Results o ver Time This research study examined h ealth outcomes and cost of health care differences not only between the control and ACC groups during each of the three periods but also within each group between the first and third periods Although reducing overuse and misuse of services and expenses can drive improvements in he alth outcomes and health care costs, effects typically are not seen immediately (Berwick et al., 2008; Verd ier et al., 2009). Results producing s avings sufficient to offset initial investment s in a collaborative health care delivery system may take two or three years to materialize (Berwick et al., 2008; Verdier et al., 2009; Wagner et al. 2006). The next two subsection s of this thesis discuss differences in health outcomes and health care costs seen in each group between T 1 and T 3 Like findings illustrated earlier in the Health Outcomes and Cost of Health Care sections of this chapter, supporting tables along with anticipated and actual results appear in the following subsections for the control group and the ACC group Additional ly, the same a pproach and assumptions described earlier in the Health Outcomes and Cost of Health Care sections were used to assess dependent variables of interes t within each group over time and between groups in each period. Control Group Results As ill ustrated in Table 4.1 in the fourth chapter of this thesis, control group participa nts differ from those in the ACC group in two ways. They were not AC C

PAGE 109

93 participants because they neither lived in focus community counties whe re ACC implementation began nor had established relationships with PCMPs during the three periods As a result, control group participants would not have experienced direct effects of RCCO and PCMP health c are coordination efforts, which sought to influenc e decreases in acute care and hi gh cost imaging and in creases in preventive and primary care services (Rodin & Silow Carroll, 2013). Table 5.14 illustrates the difference in control group health outcomes between T 1 and T 3 for potentially preventable hospital readmiss ions ER visits, and high cost imaging services Table 5.14: Control Group Health Outcomes Differences between Periods Measure PPHR ER Visits High cost Imaging Number (%) T 1 T 3 Yes 2 (2) 5 (6) 502 (7) 490 (6) 388 (39 ) 201 (1 9 ) No 79 (98) 82 (94) 6,922 (93) 7,817 (94) 619 (61 ) 879 (81 ) Total 81 (100) 87 (100) 7,424 (100) 8,307 (100) 1,007 (100) 1,080 (100) Chi square Value 26.323 .000 7,441.315 42.141 Chi square Significance ab .000 .000 Fisher Significance c .445 Pearson Value 4.944 102.068 Pearson Significance b .026 .000 Phi Value .082 .018 .221 Phi Significance b .288 .026 .000 a When p < .05, observed frequencies in the sample are not the same as expected frequencies in the population; services observed were fewer than expected. b When p < .05, significant difference exists between T 1 and T 3 c When small cell frequencie est are used.

PAGE 110

94 A significant decrease is seen in the control ER visits and high cost imaging services between T 1 and T 3. The changes cannot be attributable to the ACC One factor that can influence changes in cost is service utilization Results in Table 5.14 provide additional context for the results in Tab le 5.15. Cost of health care differences in the control group between T 1 and T 3 are illustrated in Table 5.15, which contains information about changes in medical service and prescription drug claims. Table 5.1 5 : Control Group Cost of Health Care Differences between Periods Measure Medical Service Claims Prescription Drug Claims Number (Average Paid Amount) T 1 T 3 7,424 ($328.88) 8,307 ($301.53) 20,504 ($55.05) 22,690 ($60.18) Levene's test for equality of variances a .250 .000 T test for equality of means b .403 .003 Mean difference 13.491 5.129 95% confidence interval of difference (lower, upper) ( 18.100, 45.083) (8.553, 1.704) Kolmogorov Smirnov test of normality c .000 .000 a When p < .05 equal variances b When p < .05, significant differenc e exists between T 1 and T 3 in the average paid claim amount. c When p < .05, distribution is not normal. Although the difference in the average medical service claim amount was not significant between T 1 and T 3 for control group participants, the difference in the average prescr iption drug claim amoun t was. The average prescription drug claim increase of more than nine percent likely would not have been eroded completely when adjusted for Medicaid inflation and price adjustments dur ing the periods (Catlin & Cowan, 2015; United States Department of Labo r, Bureau of Labor Statistics, 2017). Since control group participants

PAGE 111

95 were not in the ACC, the prescription drug result could be perceived as overutilization not managed by a coordinated approach to care I t could also be attributable to other factors. For example, increased medication compliance for chronic conditions such as asthma, diabetes, or hypertension can lead to increases in prescription utilization and cost (HCPF, 2012a). As noted earlier in this chapter, diff erences in cost may be attributable not only to increasing volume but also to increasing individua l drug prices (MACPAC, 2016). ACC Group Results This subsection examines differences in ACC group health outcomes and health care costs between T 1 and T 3 As Table 4.1 in the fourth chapter of this thesis illustrated, ACC group participants were Medicaid recipients who liv ed in focus community counties where A CC implementation began and had established relationships with PCMPs during the three periods. ACC gro up participants would have experienced RCCO and PCMP heal th care coordination efforts, which focus on preventive and primary care services (Rodin & Silow Carroll, 2013). On the next page, Table 5.16 summ arizes difference s between T 1 and T 3 in ACC group hea lth outcomes for potentially preventable hospital readmiss ions ER visits, and high cost imaging services. ER visits and high cost imaging services between T 1 and T 3. Although participation in the ACC could have influenced positive change in these health outcomes, Table 5.14 illustrated similar changes for the control group.

PAGE 112

96 Table 5.16 : ACC Group Health Outcomes Differences between Periods Measure PPHR ER Visits H igh cost Imaging Number (%) T 1 T 3 Yes 151 (12) 122 (9) 5,924 ( 8) 5,888 (7) 4,060 (32 ) 2,545 (20 ) No 1,152 (88) 1,178 (91) 65,744 (92) 73,932 (93) 8,438 (68 ) 9,940 (80 ) Total 1,303 (100) 1,300 (100) 79,092 (100) 79,820 (100) 12,498 (100) 12,485 (100) Chi square Value 147.20 65,456.246 805.037 Chi square Significance ab 000 .000 .000 Pearson Value b 3.367 41.540 470.247 Pearson Significance b .067 .000 .000 Phi Value .036 .017 .137 Phi Significance b .067 .000 .000 a When p < .05, observed frequencies in the sample are not the same as expected frequencies in the population; services observed were fewer than expected. b When p < .05, significant difference exists between T 1 and T 3 Noting again that one factor that can influence changes in cost is service utilization, results in Table 5.16 provide additional context for the results in Table 5.17. Like the approach taken in Table 5.15 for the control group cost of health care diffe rences in the ACC medical service and prescription drug claims between T 1 and T 3 are illustrated in T able 5.17 on the next page

PAGE 113

97 Table 5.17 : ACC Group Cost of Health Care Differences between Periods Measure Medical Service Claims Prescription Drug Claims Number (Average Paid Amount) T 1 T 3 71,668 ($304.32) 79,820 ($301.53) 131,370 ($51.99) 148,076 ($63.93) Levene's test for equality of variances a .521 .000 T test for equality of means b .701 .000 Mean difference 2.798 11.934 95% confidence interval of difference (lower, upper) ( 11.508, 17.104) ( 14.000, 9.868) Kolmogorov Smirnov test of normality c .000 .000 a claims, p < .05; equal variances could not be assumed, and subsequent values in the table for prescription drug claims were reported accordingly. b When p < .05, significant difference exists between T 1 and T 3 in the average paid claim amount. c When p < .05 distribution is not normal. Results summarized in Table 5.17 indicate no significant difference between T 1 and T 3 for medical service claims but a significant increase in prescription drug claims. A simil ar result was seen in control gr oup results in Table 5.15 in the previ ous subsection. However, th e participants in Table 5.17 were in the ACC, and it would be expected that PCMPs were managing utilization of prescription drugs. As mentioned earlier, the difference could be attributable t o other factors such as increased medication compliance for chronic conditions like asthma, diabetes, or hypertension I ncreases in prescription utilization and cost also may occur (HCPF, 2012a). Additionally, d ifferences in cost may be partially attributa ble to annual increases in individual drug prices (CMS, 2015a; MACPAC, 2016). 1 and T 3 indicate significant differences in several areas. Both groups showed evidence of improved health outcomes with significant

PAGE 114

98 decreases in both ER visits and high cost imaging. In addition, the ACC group displayed a significant decrease in potentially preventable hospital readmissions. Significant increases in prescription drug costs were also seen in both groups. As discussed ea rlier in this chapter, comprehensive prescription drug reviews and coordinated medication management services supporting proper prescription use and adherence could lea d to increased costs in the ACC group ( Amara et al. 2014; HCPF, 2012a; Smith et al. 20 13; Wagner et al., 2006). For example, this would be expected in the treatment of chronic conditions such as asthma, diabetes, or hypertension in a coordinated health care delivery system like the ACC (HCPF, 2012a). However, control group participants woul d not have received the benefit of medicatio n management from PCMPs. Control group increases in prescription drug costs could be ascribed to possible factors such as higher volume due to overutilization in a fee for service environment as well as annual dr ug price increases (MACPAC, 2016). Because changes occurred in both groups over time, differences cannot be solely attributable to participation in the ACC. Regional Differences This research study examined if participation in the ACC would impact indivi dual health outcomes and cost of health care, and this thesis has focused on examining finding s between the control and ACC groups within each period and within each group between the first and third periods. Because the ACC is a regional health care deliv ery system, data exploration also was conducted to determine if any differences exist among RCCOs. As Institutional Design subsection, the ACC encourage s each RCCO to develop health care delivery approaches suitable for and sustainable by its population and local communities (Karabatsos, 2011; State Policy Options, 2012). As seen in

PAGE 115

99 Appendix A through Appendix D diff er from other RCCOs. Differences among these regional areas of accountability may also influence service delivery and outcomes (HCPF, 2009b; Rodin & Silow Carroll, 2013; State Policy Options, 2012). Health Outcomes among RCCOs B ecause dependent variables operationalizing health o utcomes are categorical and RCCOs represent seven independent group s chi square tests were run to examine the difference in proportions (Institute for Digital Research and Education, 2017). The phi coefficient was used to asses s the strength of associa tions (Healey, 1999). In addition, comparisons of proportions using z tests were generated to identify any significant difference in RCCO pairs and a Bonferroni correction was applied to reduce the probability of false positive results that can occur when running multiple comparisons of proportions on one data set (Institute for Digital Research and Education, 2017). Negligible differences in potentially preventable hospital readmissions within 30 days were observed among RCCOs in both the contr ol and ACC groups in each period; those results are not reported here. S upporting tables and results for ER visits and high cost imaging follow. As seen in Table 5.18, significant differences in ER visits appear among RCC Os in both groups in each period. The phi value in dicates the relationships are n ot strong, but the y are significant.

PAGE 116

100 Table 5.18 : ER Visits among RCCOs Measure ----------T 1 -----------------T 2 ----------------T 3 ---------Control Group ACC Group Control Group ACC Group Control Group ACC Group Pearson Value 44.397 81.271 27.001 182.460 80.184 161.421 Pearson Significance a .000 .000 .000 .000 .000 .000 Phi Value .077 .034 .059 .048 .098 .045 Phi Significance a .000 .000 .000 .000 .000 .000 a When p < .05, significant difference exists among RCCOs. Additional tables in Appendix H provide detailed information indicating paired RCCO relationships where differences exist. In the control group in each period, significant differences exist primarily between RCCO 1 and RCCOs 2, 4, and 7. In the ACC group in T 1 significant differences are seen in RCCO 1 and in RCCO 4 with several other RCCOs. The se differences continu e in T 2 and T 3 with significant differences emerging in additional RCCO pairs. Differences among RCCOs would not necessarily be expected in either group participants. Differences emerging in T 2 and T 3 in the ACC group among various pairs of RCCOs could indicate an ex collaborative approach to coordinated service delivery. R esults similar to those among RCCOs for ER visits were also seen in the category of high cost imaging. Table 5.19 illustrates sig nificant differences in high cost imaging services among RCC Os in both groups in each period. Again, the relationships are not strong as indicated by phi values, but they are significant.

PAGE 117

101 Table 5.19 : High cost Imaging among RCCOs Measure ----------T 1 -----------------T 2 ----------------T 3 ---------Control Group ACC Group Control Group ACC Group Control Group ACC Group Pearson Value 66.555 53.817 75.271 16.285 71.687 94.228 Pearson Significance a .000 .000 .000 .012 .000 .000 Phi Value .060 .017 .065 .009 .062 .022 Phi Significance a .000 .000 .000 .012 .000 .000 a When p < .05, significant difference exists among RCCOs. Details about specific RCCO pairs where significance difference exists are in Appendix I. The main significant differences in the control group appear again in RCCO 1 with RCCOs 2 and 4 in T 1 and in RCCO 1 with almost all other RCCOs in T 2 and T 3 ACC group results indicate differences between RCCO 1 and other RCCOs in all periods, and a few di fferences between RCCO 3 and RCCOs 2 and 7 in T 1 and between RCCO 7 and RCCOs 2, 3, 5, and 6 in T 3 Overall, variation in significant differences among RCCOs in high cost imaging services is not as distinct as in ER visits. Aside from repeated RCCO 1 differences, most variation in high cost imaging occurs in t he ACC group in T 3 Returning to earlier discussion about RCCOs developing approaches suited to their service areas and sus tainable by their local communit ies, results in T 3 are not unexpected (Karabatsos, 2011; State Policy Options, 2012). The time factor is also relevant since improvements in health outcomes typically do not appear immediately; it often takes two or three ye ars for results to begin to materialize (Berwick et al., 2008; Verdier et al., 2009; Wagner et al. 2006). Cost of Health Care among RCCOs D ependent variables operationalizing cost of health care are continuous, and seven independent groups are represented by the RCCOs. As explained in the Cost of Health Care

PAGE 118

102 section earlier in this chapter the presence of above average medical service and prescription drug claim amounts affect the normality of cost distributions (Hays, 1994). O utliers were examined, were not determined inaccurate or unreasonable, and were not excluded from analysis. As a result, the nonparametric Kruskal Wallis H test was used to determine if differences exist among RCCOs in the cost of medical services and pres cription drugs (Institute for Digital Research and Education, 2017 ; Vargha & Delaney, 1998 ). Box plots were generated for visual inspection of d istribution s and the shape was si milar in each RCCO. D ifferences in RCCO median service and prescription costs were assessed (Institute for Digital Research and Education, 2017; Vargha & Delaney, 1998). To ide ntify RCCOs with significant differences, subsequent comparisons between RCCO pairs were generated and a Bonferroni correction was applied (Institute for Dig ital Research and Education, 2017) Results and supporting tables for the comparison of median medical service and prescription drug claims in both the control and ACC groups in each period follow. While Table 5.20 illustrates a significant difference amon g RCCOs in the median cost of medical service claims in the control group in each period Table 5.21 summarizes similar information for the ACC group. Appendix J contains detailed information about significant pairwise comparisons for both study gr oups in each of the three periods.

PAGE 119

103 Table 5.20: Control Group Medical Service Claims among RCCOs ----------T 1 -----------------T 2 ----------------T 3 --------RCCO Number Median Number Median Number Median 1 2,732 $87.60 3,094 $90.12 3,295 $83.28 2 602 $97.71 510 $97.16 492 $90.24 3 408 $58.81 418 $61.12 401 $59.28 4 3,324 $95.14 3,439 $93.29 3,775 $92.59 6 146 $77.99 165 $77.10 217 $71.99 7 212 $72.36 128 $75.77 127 $69.45 Total 7,424 $ 93.29 7,754 $ 92.59 8,307 $ 90.12 Test statistic 35.07 a 24.55 a 42.46 a Degrees of freedom 5 b 5 b 5 b Asymptotic 2 sided test .000 c .000 c .000 c a The test statistic is adjusted for ties, which occur when the same value is in more than one RCCO. b The control group contained no participants in RCCO 5, which represents the Denver metropolitan area. c When p < .05, significant difference in paid medical service claim amounts exists among RCCOs. Table 5.21: ACC Group Medical Service Claims among RCCOs ----------T 1 ----------------T 2 -----------------T 3 --------RCCO Number Median Number Median Number Median 1 4,432 $85.96 5,023 $84.44 5,547 $78.92 2 6,069 $80.30 6,769 $79.49 6,488 $77.10 3 22,884 $78.30 25,491 $77.59 26,897 $77.59 4 10,238 $88.90 10,853 $87.20 10,622 $91.97 5 12,338 $78.17 14,083 $77.10 13,519 $77.59 6 9,133 $77.68 9,960 $77.10 9,563 $77.10 7 6,574 $92.33 7,000 $81.58 7,184 $78.25 Total 71,668 $80.08 79,179 $77.68 79,820 $77.59 Test statistic 60.065 a 66.882 a 142.032 a Degrees of freedom 6 6 6 Asymptotic 2 sided test .000 b .000 b .000 b a The test statistic is adjusted for ties, which occur when the same value is in more than one RCCO. b When p < .05, signifi cant difference in paid medical service claim amounts exists among RCCOs Appendix J indicates few significant differences among RCCOs in the control group in each period compared to those seen in the ACC group. Primary differences in the ACC group in T 1 exist between RCCO 6 and RCCOs 1, 3, 4, and 7 Differences between RCCOs 5 and 6 with RCCOs 1, 2, and 4 appear in T 2 and significant pairwise comparisons exist

PAGE 120

104 between RCCO 4 and RCCOs 1, 2, 3, 5, and 7 and between RCCO 6 and RCCOs 1, 3, 4, 5, and 7 i n T 3 The median cost of medical service claims in RCCO 6 cluster toward the lo w end among all RCCOs in each period whereas median cost in RCCO 4 is consistently at the high end. Tables 5.22 and 5.23 follow and show significant difference s among RCCOs in the median cost of prescription drug claims in each period for both the control group and the ACC group, respectively. D etaile d information indicating RCCO pairs where significant difference exists in both groups in each period is included in Appendix K. For the cost of prescription drug claims, a larger number of significant pairings and greater diversity among RCCO pairs exist in both the control and ACC groups in every period than is observed in medical service claims. Table 5.22 : Control Group Prescr iption Drug Claims among RCCOs ----------T 1 -----------------T 2 -----------------T 3 --------RCCO Number Median Number Median Number Median 1 7,168 $8.99 8,214 $9.44 7,720 $13.33 2 1,684 $11.80 1,727 $12.04 1,594 $12.68 3 1 ,254 $8.04 1,207 $8.31 1,040 $9.81 4 9,411 $10.50 11,003 $9.62 11,211 $11.71 6 295 $8.40 326 $8.74 579 $7.49 7 6 92 $8.18 673 $9.00 546 $11.48 Total 20,504 $9.60 23,150 $9.60 22,690 $12.06 Test statistic 151.139 a 53.089 a 141.720 a Degrees of freedom 5 b 5 b 5 b Asymptotic 2 sided test .000 c .000 c .000 c a The test statistic is adjusted for ties, which occur when the same value is in more than one RCCO. b The control group contained no participants in RCCO 5, which represents the Denver metropolitan area. c When p < .05, significant difference in paid prescription drug claim amounts exist s among RCCOs.

PAGE 121

105 Table 5.23: ACC Group Prescription Drug Claims among RCCOs ---------T 1 ------------------T 2 -----------------T 3 --------RCCO Number Median Number Median Number Median 1 10,506 $8.30 12,091 $8.41 12,598 $10.77 2 11,209 $8.54 12,383 $8.94 12,953 $10.14 3 40,502 $7.77 44,812 $7.78 44,422 $9.19 4 20,227 $10.22 21,439 $10.46 22,627 $11.75 5 20,062 $7.50 22,686 $7.00 22,426 $8.94 6 16,095 $7.79 19,158 $7.44 18,731 $8.94 7 12,769 $8.74 14,106 $8.39 14,319 $10.33 Total 131,370 $8.19 146,675 $8.12 148,076 $9.93 Test statistic 535.848 a 756.898 a 683.090 a Degrees of freedom 6 6 6 Asymptotic 2 sided test .000 b .000 b .000 b a The test statistic is adjusted for ties, which occur when the same value is in more than one RCCO. b When p < .05, significant difference in paid prescription drug claim amounts exists among RCCOs. As seen in Table 5.23, the median cost of prescription drug claims in RCCO 5 is at the low end of median costs among RCCOs in each period. Like results for median cost of is at the high end. Initial examin ation of health outcome and cost of health care differences among RCCOs provides preliminary evidence that significant differences exist in both the control group and the ACC group in each period. In health outcomes for both ER visits and hi gh cost imaging significant RCCO 1 pairings occurred most frequently in the control group in each period. In the ACC group, greater variations in significant RCCO pairings begin to appear in T 2 and increase in T 3 The number of significant pairings between RCCOs was gre ater for ER visits than for high cost imaging. In cost of health care for both medical service and prescription drug costs, fewer variations appeared in the control group in each period. In the ACC group, significant differences in medical service median c osts were mainly observed in RCCO 4 and RCCO 6 pairings with other RCCOs. Greater variation in significant

PAGE 122

106 differences among RCCOs occurred in median prescription drug costs than in median medical service costs. Appendix K reflects numerous RCCO pairings w here significant differences exist in median prescription drug costs. Significant differences among RCCOs in health outcomes and cost of health care indicate the possibility that variations in regional areas of accountability may impact service delivery an d outcomes (HCPF, 2009b; Rodin & Silow Carroll, 2013; State Policy Options, 2012). This chapter concludes with a brief discussion of high level observations before examining key implications of this research study and recommendations in the next chapter. O bservations discussion of findings related to health o utcomes and cost of health care, some differences were noted. As Table 5.24 illustrates at a high level some significant differences between groups and within groups between T 1 and T 3 occurred in health outcomes. Decreases in utilization indicate a favorable outcome and improvement. Significant differences also appeared between groups and within groups between T 1 and T 3 in cost of health care. Increases in costs for prescriptio n drug claims reflect an unfavorable outcome and lack of improvement. With the exception of potentially preventable hospital readmissions within 30 days, significant differences among RCCOs exist in health outcomes and cost of health care Analysis in this research study in both the control group a nd the ACC group, in each period, over time between the first and third periods, and among RCCOs show s greater negative impact and variation in prescription drug costs than any other dependent variable.

PAGE 123

107 Table 5 .24 : Summary of Differences Between Groups a T 1 T 2 T 3 Health outcomes PPHR Yes (decrease) No No ER visits Yes (decrease) Yes (decrease) Yes (decrease) High cost imaging Yes (decrease) Yes (decrease) No Cost of health care Medical service claims No No No Prescription drug claims Yes (increase) Yes (increase) Yes (increase) Total cost of care No No No Within Groups between T 1 and T 3 a Control Group ACC Group Health outcomes PPHR No No ER visits Yes (decrease) Yes (decrease) High cost imaging Yes (decrease) Yes (decrease) Cost of health care Medical service claims No No Prescription drug claims Yes (increase) Yes (increase) Among RCCOs b Control Group (T 1 T 2 and T 3 ) ACC Group (T 1 T 2 and T 3 ) Health outcomes PPHR No No ER visits Yes Yes High cost imaging Yes Yes Cost of health care Medical service claims Yes Yes Prescription drug claims Yes Yes a Significant differences are indicated as Yes Overall impact on the health outcome or cost category is indicated in parentheses. b Significant differences in RCCO pairwise comparisons are indicated as Yes Individual RCCO impacts on the health outcome or cost category are not included. Because significant differences were reported in both the control group and the ACC Although ACC participation cannot be credited as the sole influence for changes, this r findings cannot be dismissed as unimportant. Findings illustrated and discussed in this chapter guide the discussion of implications and recommendations for literature and theory, policy and practice, and future research in the next chapter of this thesis.

PAGE 124

108 CHAPTER VI IMPLICATIONS AND RECOMMENDATIONS Earlier chapters of this thesis focused on the importance of find ing different approaches to climbing costs, conside red colla borative governance a viable foundation to support changes to Medicaid service delivery, collaborative governance effort, specified a research design to examine differences in health care outcomes and cost of health care associated with ACC participation, and reported findings related to variations between the control and ACC groups in each period, within each group from the first to the third period and among RCCOs in the control and ACC groups in each period This chapter concludes the thesis by exploring implications and recommendations for literature and theory, policy and practice and future research. Literature and Theory This research study contributes to collaborative governance literature a nd theory in two ways. First, it introduces the ACC as a n example that satisfie s the six defining criteria, f our common elements and related components, and 10 implications for successful collaboration produced by the Ansell and Gash (2008) research Secon d, this rese arch study provides quantitative evidence of the ACC early impact on the health outcomes and cost of health care for a sample of during its implementation phase. In so doing, this thesis adds to the study of collaborative governance by providing insight into two aspects of ACC performance: process performance, results of the collaborative process, and productivity performance, result s of collaborative actions (Emerson & Nabatchi, 2015).

PAGE 125

109 ACC Process Perform ance The third chapter of this thesis compares the ACC to the Ansell and Gash (2008) collaborative governance criteria, elements and implications for success inception is traced from less than favorable starting conditions to its creation of an institutional design facilitated by state wide and regional leadership and reinforced by b road stakeholder engagement. This thesis describes the laborative process through monthly and quarterly meetings that support face to face dialogue; mediator s and local leaders who build trust among stakeholders; shared ownership of process and outcomes that solidifies commitment; and incremental progress with a strategy for the future that support s intermediate outcomes. In addition, this thesis examines t he role of the RCCOs, seven ACC entities accountable for developing approaches to Medicaid service delivery suited to and sustained by their regional populations and local communities, and illustrates v ariations in RCCO service areas, organizations, and commu nity relationships. From different perspectives in the Ansell and Gash (2008 2012 ) literatu re, this thesis demonstrates that the collaborative process that establish ed develop ed and sustain s the ACC appears designed for success. Ansell and Gash (2008) a lso emphasized interdependence and time as two topics of particular interest to practitioners. Additionally, several scholars subsequently voiced perspec tives about local influences or time in collaborative governance efforts (Andres & Chapain, 2013; Berar d o et al., 2014; de Loe et al., 2015; Gerlak & Heikkila, 2011; Gibson, 2011; Gollagher & Hartz Karp, 2013; Heikkila & Gerlak, 2016; Herranz, 2008; Koebe le, 2015 ; Siddiki et al., 2015 ). This thesis illustrates these two points in the Model Considerations se ction in the second chapter of this thesis Both st rengthen the addition to collaborative governance literat ure and theory As described

PAGE 126

110 earlier in this thesis, t sev en RCCOs lends variation to the statewide model that supplements a key area of interest in col laborative governance A central challenge in collaborative governance is examining how organizations manage to adapt, remain stable, and perform in a new collabo rative environment (Emerson & Nabatchi, 2015). Additionally, this thesis contributes to the time perspective in collaborative governance re implementation period and its first two years of operation. Although t ime will be addressed again in the Future Research section of this chapter, the period examined in this t hesis is the fundamental phase in the life cycle of a collaboration b ecause it establishes tone for phases that follow and tenor for failure or success. Based on current conditions and future uncertainty about increasing enrollment and accelerating cost s described in this thesis, Medicaid service delivery qualifies as a situation that warrants continuing coopera tion, and this thesis broadens the base of collaborative governance literature and theory ACC Productivity Perfor mance This research study also contributes to collaborative governance literature and theory by providing quantitative evidence about the ACC results and outcomes. The previous subsection indicates that examining how and how well the collaborative process functions is important to the study and advancement of collaborative governance As Varda Such collaboration has the potential to improve outcomes by leveraging resources, lowering costs, and identifying solutions that are unachievable by any

PAGE 127

111 Equally important to literature and theory is understanding whether or not the ACC realizes that potential as a collaborative governance effort implementation phase for a sample of Medicaid recipients who were between the ages of 21 and 60 in 2010; who had fee for service Medicaid as their only source of health insur ance coverage; and whose health care claims included physical health diagnoses and behavioral health diagnoses for anxiety or depression Significant changes were observed in health outcomes and costs in the ACC group be tween T 1 and T 3 and in the increase of significant differences between RCCOs fro m T 2 to T 3 Some of the findings discussed in the previous chapter of this thesis were not completely as expected or did not demonstrate that the ACC was the exclusive factor in reported changes Most regional he alth care delivery systems only report data for groups at an aggregate level because smaller numbers of patients can sometimes cause estimates to appear less precise or more variable over time (Wagner et al., 2006). Different approaches and methods complem ent rather than compete in their efforts to increase knowledge (Ostrom, 2007). Through its defined sample of control group and early ACC participants, t his research study provides useful information from a smaller scale pre implementation and post implemen tation perspective. As noted in the Results over Time section in the previous chapter of this thesis, it is not uncommon for a collaborative health care system to need several years to produce improvements in health outcomes and savings in health care cost s sufficient to offset its initial investments (Berwick et al., 2008; Verdier et al., 2009; Wagner et al. 2006). Additionally, any solution may not operate in the same way or produce the same results over time (Ostrom, 2007).

PAGE 128

112 Whether in case studies or t hrough quantitative research of large samples, examining the impacts of collaboration advances theory (McGuire, 2006). In identify ing and presenting the ACC as an example and submitting findings from analysis of its early results, this thesis offers evidence to augment existing collaborative governance literature and theory which can inform future studies. Policy and Practice In addition to contributing to collaborative governance literature a nd theory, this thesis supports policy and practice related to the ACC Four topics surfaced during this research study that warrant further discussion in the following subsections : approach to analysis, data sources, prescription drugs, and contingencies fo r the future. Approach to Analysis As stated in the third chapter of this thesis, the ACC seeks to achieve its goals by improv ing health outcomes, control ling costs and expanding access to comprehensive primary care for population (CMS, 2012; HCPF, 2010a, 2010b). This research study examined health outcomes and cost of health care for a sample of Medicaid operation. Health outcomes and cost of h ealth care were determined from Medicaid claims in the limited data set and PMPM calculations based on published amounts (HCPF, 2012a, 2013a; Kaiser Family Foundation, 2013). Because significant differences appeared in both the control and ACC groups, r es ults could not be attributed ex clusively to ACC participation. As noted in the Research Study Group Characteristics subsection in the fourth chapter of this thesis, it is possible that variations in gender, generation, and RCCO distribution in the control and ACC groups

PAGE 129

113 affected health outcomes and cost of health care examined in this research study. It is also possible that other factors such as race and ethnicity, which could not be determined from the limited data set, may have influenced results. The a pproach taken in this research study differs from the approach the ACC took in assessing health outcomes and health care costs in its first two annual reports (HCPF, 2012a, 2013a). Medicaid recipients represented in the sample for this research study were the same in each period; control and ACC group participants did not change from one period to another. The ACC calculated h eal th outcomes for all Medicaid recipients who were enroll ed in the ACC for three months in any rolling 12 month period and who could be attributed to a PCMP based on claims history (HCPF, 2012a, 2013a; M. Ly and H. Schum, personal communication, November 16, 2015; Treo Solutions, 2012) c alculated total cost of care using counterfactual estimation, which computed ACC impact compared to a hypothetical scenario in which the ACC was not implemented, and difference in differ ence methodology, which compared the cost of ACC enrollees to a simila r group of eligible clients who were not enrolled (HCPF, 2012a). The ACC estimated cost savings or cost avoidance at $20 million for the first year of operation (HCPF, 2012a). As mentioned in the fourth chapter of this thesis, evaluation and analysis are n eeded to inform practice since collaboration has the potential to improve health outcomes and reduce cost of health care (Varda & Retrum, 2012). Estimated cost savings fueled the policy decision to expand ACC enrollment beyond the initial focus community c ounties and had a positive impact on practice because it allowed the collaborative effort to continue (Colorado Access, 2013; Karabatsos, 2011).

PAGE 130

114 invest time, energy and resources in collabor ations that appear to be producing tangible and The report calculated total cost of care using only the counterfactual estimation technique (HCPF, 2013a). The ACC estimated $44 million in cost savings or cost avoidance in the second year of operation (HCPF, 2013a). These estimates afforded additional policy and practice opportunities for the ACC to expa nd the population served, invest in additional preventive benefits adjust incentives for value based care, and improv e the functionality and efficiency of data collection and reporting systems ( CMS, 2012; HCPF, 2012c 2014 a ). The ACC is now moving into its second iteration with plans to connect physical and behavioral health service delivery in one accountable entity, strengthen service coordination through advancements in team based care and health neighborhoods, and pay providers for i ncreased value they deliver (HCPF, 2017 b ) Like the ACC in its first two annual reports, health care delivery syst ems typically report data at the aggregate level because individual outcomes and costs require the use of trusted measures, the presence of data completeness and integrity, and accuracy in reporting (Wagner et al., 2006). As the ACC mature s and include s ad ditional subpopulations of Medicaid recipients, the ability to continue study ing individual characteristics, conditions, RCCO and PCMP relationshi ps, and outcomes over time provides enhanced perspective to policy discussions and approaches to practice (Finney, Humphreys, Kivlahan, & Harris, 2011 ; Friedberg, Hussey, & Schneider, 2010; Singleton & Straits, 2010; Van de Ven, 2007; Verdier et al 2009).

PAGE 131

115 Data Sources Data completeness, accuracy, and timeliness impact policy and practice. As described earli er in this thesis, the SDAC serves as the ACC data repository and provides information to RCCOs, PCMPs, and the state agency that administers the ACC (HCPF, 2010a, 2010b; Kaiser Family Foundation, 2013; Karabatsos, 2011, p. 11). The SDAC began operations i n 2011 (HCPF, 2010b). It operates and maintains data warehousing capability and capacity that Information System (MMIS) and the APCD (Public Knowledge LLC, 2012). Another vend or purchased t he original SDAC contractor in 2014 and Colorado upgraded its requirements T hrough a competitive reprocurement process in 2015 the contract was awarded to a third vendor (3M Health Information Systems, 2014 ; Truven Health Analytics, 2015 ). MMIS is a claims processing and information retrieval system designed to assist states in managing Medicaid program and administrative costs; recording and paying services to recipients and providers; and satisfying reporting needs for management planning and control (CMS, 2017e ). MMIS was a legacy system lacking functionality and capacity for capturing and tracking all data elements essential to ACC operations (Public Knowledge LLC, 2012; Hewlett Pac kard, 2014). Colorado expanded its MMIS requirements and awarded the contract to a different vendor in 2014 after a competitive reprocurement (Hewlett Packard, 2014). As the entity Colorado authorized to provide data sets for research, the APCD began re leasing approved custom data requests in 2013 (CIVHC, 2017a). As mentioned in the Data Source and Considerations subsection in the fourth chapter of this thesis, the APCD was the source of the limited data set used in this research study. Also, as explaine d in the Data

PAGE 132

116 Sample and Research Study Groups subsection in the fourth chapter of this thesis, the limited data set the APCD provided for this research did not contain a designation for those Medicaid providers who were ACC PCMPs but later was able to provide an additional file that allowed ACC PCMPs to be identified and matched to Medicaid providers appearing in the original limited data set. In July 2017, the SDAC transitioned to new data warehouse and analytics partners, Human Services Research Institute and National Opinion Research Center (CIVHC, 2017a). The previous paragraphs provide a snapshot of three interrelated data sources that support the ACC As the ACC continues to evolve mature and stabilize over time data sources and systems related to and supporting it have undergone and continue to undergo changes and improvements as well. T he ability to identify and associate the appropriate Medicaid recipients with their corresponding eligibility date ranges, medical and prescription drugs claims, RCCOs and PCMPs is fundamental to accurate and timely reporting of health outcomes and cost of health care information. Delayed, incomplete, or questionable information affects decisions that state and federal policy makers and regional and local Medicaid program and the ACC. Reported h ealth outcomes and cost of health care could be at risk of overstatement or understatement. Data issues have equal potential to affect research findings. Ongoing improvements and processes that ensure the integrity o f ACC data sources and systems are Prescription Drugs As indicated by findings in this research study, a policy opportunity not only for ACC participants but also for other Medicaid recipients in Colorado lies in prescription drug

PAGE 133

117 utilization and costs. Efforts to increase transparency in drug pricing have stalled nationally and in Colorado (CHI, 2016 b ). Older and lower income Medicaid recipients in Colorado are more li kely than other Medicaid recipients to avoid refilling prescriptions due to their cost (CHI, 2016 b ; MACPAC, 2016 ). N ationally, n o more than half of individuals with chronic conditions adhere to prescribed drug therapies, which exacerbates worsening of thei r health status and contributes to avoidable hospitalizations (Amara et al., 2014). Including pharmacists as members of collaborative health care coordination teams and in initiatives to link electronic health record data among provider s would be a policy decision support ing improved medication compliance, reduced preventable adverse events, and decreased overall health care expenditures (Amara et al., 2014 ; CMS, 2015a ). Another ACC policy consideration would be the addition of a key performance indicator r elated to prescription drugs or medication management services (Amara et al., 2014; Feng & Maini, 2016). If established as a key performance indicator, RCCOs would have an added incentive to collaborate with prescribers and pharmacists to develop innovativ e regional strategies to ensure medication adherence, reduce overutilization, and decrease costs. Current and historical data for Medicaid prescription drug claims could be explored to reveal opportunities by subpopulation, diagnosis, specific drug, and co st per drug Contingencies for the Future Another pressing policy and practice topic relates to contingencies for the ACC if the Affordable Care Act (2010) is revised or repealed Such external shocks can aid in its capacity to continue when faced with challenges and to transform itself over time as needed (Carboni & Milward, 2012). As an ongoing statewide and regionally configured coll aborative governance effort, the ACC may

PAGE 134

118 be positioned to provide an existing platform that is both stable and flexible (Ansell & Gash, 2017). Colorado expanded its Medicaid program under Affordable Care Act (2010) provisions in 2014 and added more than 28 9 thousand recipients as a result (CHI, 2016a). By 2015, the expansion population accounted for more than one fifth of total Medicaid recipients in Colorado (CHI, 2016a). One alternative to Affordable Care Act (2010) provisions for Medicaid that has been discus sed at the national level involve s the federal government limiting its contribution to states to a fixed annual per person amount, which could only increase based on the annual rate of medical inflation (Antonisse, Garfield, Rudowitz, & Artiga, 2017; Rosenbaum, Rothenberg, Schmucker, Gunsalus, & Beckerman, 2017). Theoretically, fixed per person limits encourage innovation by supporting strategies for less costly but equally efficacious care, decreases in overutilization of services that have little im pact on health outcome or cost improvements, or reductions in provider payment amounts (Antonisse et al., 2017; Rosenbaum et al., 2017). The ACC already has made some of the changes mentioned in the previous paragraph. If the Aff ordable Care Act (2010) wer e revised or repealed the ACC could potentially face a reduction in the total number of Medicaid recipients Colorado could serve a reduction in the s cope of services currently covered, an increase in the number of uninsured, or a decrease in investments in infrastructure, technology, and support that stimulate and sustain system transformation (Antonisse et al., 2017; Rosenbaum et al., 2017). Continuing to explore additional opportunities in less costly service categories such as preventive and primary care and medication management is one policy decision that would benefit the ACC regardless of changes to existing legislation. Integrating physica l and behavioral health service delivery to eliminate inefficiencies by more compr ehensively

PAGE 135

119 address ing Medicaid recipient hea lth needs and by reducing servic e duplication or gaps is another policy direction that would directly benefit participants and support the ACC whether or not the Affordable Care Act (2010) is revised or repealed ( Antonisse et al., 2017; CHI, 2016a; HCPF, 2015a, 2017 b ). Future Research After considering implications for literature and theory as well as policy and practice, the path this research study paves for future collaborative governance research is a dynamic one Among areas for further invest igation are individual and multiple RCCOs an expanded time frame the Medicare Medicaid subpopulation, and multivariate analysis. The following subsections consider each of these topics. Individual and Multiple RC COs that merits further research. An ethnographic research might be the most successful strategy for developing g reater insight research into topics such as trust building, development of shared understanding, and commitment formation could be useful in gaining a firmer grasp of the underpinnings of collaborative governance efforts (Ansell & Gash, 2008). This thesis illustrated a number of differences among RCCOs that emerged as a result of this research study. The table of Colorado counties in Appendix A and the state map in Appendix B indicate statewide distribution in seven geographic areas Each RCCO is responsible for cultivating localized resources and developing approaches that fit its population and communities (Karabatsos, 2011; State Policy Options, 2012). V ariation i n

PAGE 136

120 apparent in Appendix C and Appendix D. The Regional Differences section in the previous chapter of this thesis along with information contained in Appendix H through Appendix K show s ignificant differences among RCCOs. Learning about the RCCOs by learning from them is one connection between the ( Higginbottom, Pillay, & Boadu, 2013; LeCompte & Schensul, 2010; Savage, 2000). A focused ethnographic approach to investigating two or more RCCOs could incorporate a range of methods and utilize both qualitative and quantitative data ( Higginbottom et al., 2013; LeCompte & Schensul, 2010; Savage, 2000) Concentrating on one or more discrete com munities or organizations, focused ethnography could provide insight into differences that exist among RCCOs and what contributes to t hose differences, which could be useful in collaborative governance research and in health care service delivery improveme nts ( Higginbottom et al., 2013). In addition, i nterviews with RCCO leadership staff PCMPs, and Medicaid recipients coupled with further documentary analysis could provide an enhanced perspective to additional quantitative dat a analysis (Fetterman, 2010 ; Higginbottom et al., 2013 ) For example, e thnography could illuminate how patient preferences and behaviors influence therapeutic interventions and how provider assumptions can facilitate or impede health promotion (Savage, 2000). While it is likely that s ome RCCO practices would only be successful in a particular region or community, it is also possible that some practices could be effective if generalized across RCCOs. U pcoming transition to the second phase of ACC development also involves transformati on of the RCCOs into Regional Accountable Entities (RAEs) (HCPF, 2015b,

PAGE 137

121 2017 b 2017g ). The RAE in each of the seven ACC regions will be responsible for the duties currently performed by RCCOs and Behavioral Health Organizations (BHOs) (HCPF, 2015b, 2017g). Since the mid 1990s, the Medicaid benefit in Colorado for community behavioral health services has been administered by five managed care entities known as BHOs (HCPF, 2017 b ). The second phase of ACC d evelopment begins in July 2018 (HCPF, 2017g). S ervice areas for the BHOs will be aligned with RCCO service areas, and the RAE will become the single administrative entity in each region responsible for physical and behavioral health outcomes and health care costs for Medicaid recipients as (HCPF, 2015 a 2017 b 2017g). authors note collaborative governance is increasingly viewed as a proactive policy instrument whose strategy can be deployed on a larger scale and extended from one local context to another 17, p. 1) This ACC transition creates an opportunity not only for a more integrated view of physical and behavioral health outcomes and health care costs but also regional differences and the effects of those differences on Expanded Timeframe with enrollment of the first Medicaid recipients as of May 1, 2011 (HCPF, 2012a) This research study examined ACC design and development activities and analyzed health outcomes and cost of health care for a sample of Medicaid recipients from May 1, 2010 through April 30, 2013. Now entering the second half of its sixth year of operations, the ACC as a collaborative governance example invites study through a wide angle lens. Ansell and Gash (2008) acknowledged collaborative governance efforts as time consuming, deliberative arrangements and recognized that early investment in time could facilitate later implementatio n and operations. Additionally, integrated systems of

PAGE 138

1 22 health care delivery typically produce results after two or three years rather than immediately (Berwick et al., 2008; Verdier et al., 2009; Wagner et al., 2006). S cholars also have noted that studying performance with in collaboratives over time represents an important and insufficiently addressed area in collaborative governance research (Gerlak & Heikkila, 2011; Heikkila & Gerlak, 2016). Because the ACC has continued operations more can be learned a bout it from the perspectives of both process performance and productivity performance. In addition to considering all years of operation in further RCCO research or analysis of health ou tcomes and cost of health care, the study of other ACC elements over time could be beneficial. For example, another area of study c ould be to meet Policy and Practice section. Medicare Medicaid Subpopulation Colorado continued to expand ACC enrollment beyond initial focus community counties and expanded ACC eligibility over time to include other subpopulations of s without dependent children be ginning in April 2012 (HCPF, 2012b). In May 2012, Colorado submitted a proposal to CMS to enroll Medicaid recipients who were also fee for service Medicare recipients into the ACC for an initial three year demonstration period (CMS, 2012 2017c ; HCPF, 2014 b). Similar to the managed fee for service approach the ACC took with Medicaid recipients, it intended to coordinate services across the Medicare and Medicaid programs to align services, alleviate fragmentation, enhance quality of care, and reduce costs fo r Medicare Medicaid recipients (CMS, 2012 2017c ).

PAGE 139

123 Enrollment began in September 2014 (HCPF, 2014b). CMS gave Colorado the opportunity to extend the demonstration for Medicare Medicaid recipients beyond the three year demonstration period but the state declined (CMS, 2017b ). By the end of May 2017, the ACC began to implement its phase out plan to end the Medicare Me dicaid demonstration (CMS, 2017b ). Although Colorado will permit enrolled Medicare Medicaid recipients to remain in the ACC i n RCCOs with PCMPs, as of January 1, 2018, RCCOs will no longer be required to provide enhanced care coordination services to these i ndividuals (CMS, 2017b ). As service responsibility is reduced, so too are RCCO administrative and reporting requirements ea sed. Medicare Medicaid demonstration which covered the 16 month period from September 1, 2014 through December 31, 2015. (CMS, 2017b ). Calculations in the report only cover e d Medicare P arts A and B expenditures because same period were no t available (CMS, 2017b ). The report indicated savings occurred in the categories of durable medical equipment, outpatient hospital services, home hea lth, a nd professional services, but all other service categories reflected negative savings (CMS, 2017b ). Early demonstration results of increased costs and absence of savings in Medicare services were an unexpected outcome (CMS, 2017b). Although Medicar e and Medicaid share a number of enrollees, each traditionally has been a siloed system of care. Particularly in a fee for service environment, lack of integration between these two primary payers and providers has contributed to decreased coordination of care, decreased quality of care, reduced access to care, and increased cost of care for Medicare Medicaid recipients ( CMS, 2017c; Grabowski, 2007; Kasper, Watts, & Lyons

PAGE 140

124 2010; Medicare Payment Advisory Commission, 2004, 2011; Vogeli, Shield, Lee, Gibson, Marder, Weiss, & Blumenthal, 2007 ). To a greater extent than Medicaid recipients, Medicare Medicaid recipients are among those whose health conditions are most complex, personal resources are most meager, and health care services are most costly (Jacobson, et al., 2012). Challenges exis t for Medicare Medicaid recipients as noted above, and the area continues to be one of the most prominent in health policy and public management (CMS, 2017a). It also presents a prime area of study for collaborative governance scholars in that the provider and stakeholder base is even broader and more diverse (CMS, 2012). A collaborative governance model addressing issues inherent in serving Medicare Medicaid recipients in a fee for additi onal or different attention in certain areas to be successful. For example, considering the Ansell and Gash (2008, 2012) approach and the ACC as described in the second and third chapters of this thesis, the three threads of time, trust and interdependenc e interwoven through the four common elements of starting conditions, institutional design, facilitative leadership, and collaborative process would warrant further examination. Important lessons can be learned by studying a collaborative governance approa ch when its process performance indicates likelihood for success and productivity performance is other than expected. Multivariate Analysis Research become s an enhanced learning process when data analysis uncovers discrepancies betw een expectations and re sults and the approach to analysis is later modified or replaced (Singleton & Straits, 2010). Several issues with data obtained for this research

PAGE 141

125 prevented multivariate analysis from being feasible at this stage of study. S pecific issues that arose include d abnormal distribution, non linear relationships and heteroscedasticity (Osborne & Waters, 2002). As mentioned in the discussion of research findings in the fifth chapter of this thesis, outliers were found in the data, and a conscious decision was made not to remove them for purposes of this study. Although above average medical service and prescription drug claim amounts would affect the normality of cost distributions, such costs would be expected in ts (Hays, 1994). Outliers were examined, were not found to be inaccurate or unreasonable, and were not excluded from this analysis. To produce accurate estimates, regression models used in multivariate analysis generally rely on linear relationships between independent and dependent variables (Osborne & Waters, 2002). Examination of residual plots between control and ACC group independent variables and categorical health outcomes dependent variables detected nonl inear relationships. No corrections were made in this analysis for the existing and expected nonlinear relationships. Because error variance was not the same across levels of the independent variables, heteroscedasticity was present, which also may have be en the natural result of skewed distribution s and outliers in the data (Institute for Digital Research and Education, 2017). Future research can accommodate multivariate analysis after addressing these data issues, for example, by excluding outliers and tr ansforming variables (Institute for Digital Research and Education, 2017; Osborne & Waters, 2002). Additional issues merit consideration before further analysis occurs. Specification errors can occur when one or more relevant variables are excluded from an alysis or when one or more irrelevant variables are included in the model (Institute for Digital Research and Education, 2017). Using additional

PAGE 142

126 indicators that are available in more recent APCD data sets to configure the control and ACC groups may allow i mproved definition of ACC participation, RCCO assignments, and PCMP relationships Variables other than those the ACC selected as key performance indicators may exist in the data that would provide more information and greater clarity about health outcomes for Medicaid recipients. Pursuing quantitative analysis of ACC health outcomes and cost of health care from a different perspective would contribute to future research efforts. Addressing abnormal distribution and nonlinear relationships, using additiona l indicators now available in APCD data sets to refine definition of control and ACC groups, expanding the analysis timeframe, and examining health outcomes and cost of health care in one or more RCCOs would be useful modifications in further studies. Clus service and prescription drug claims by diagnosis, for example, could provide better insight into where policy and pr actice opportunities lie both for more informed decisions abo ut prescription drug policies and for identification of R CCO and PCMP intervention opp ortunities Multivariate analysis would permit introducing additional variables such as inpatient, outpatient, and professional claim type, gender, and generation, which could lea d to policy changes in how the ACC prioritizes service delivery for Medicaid recipients and in how RCCOs and PCMPs approach not only preventive and primary care but also transitions from one care setting to another. Conclusion A troubled adolescence in 1977 Hubert Humphrey restated a concept that had been expressed over the years by public figures in countries in addition to the United States In what would be his final public speech, he

PAGE 143

127 those who are in the dawn of life, the children; those who are in the twilight of life, the aged; and those in the shadows of life, the Now i n the throes of Medicare and United States continues to confront this challenge. strategies. This thesis emphasizes the need for increased and improved collaboration among state and local resources in serving Medicaid recipients It offers implications for collaborative governance literature and theory, policy and practice, and future research. Moreover, it p rovides evidence that both process performance and productivity performance are crucial when st udying collaborative governance examples. From the perspective of any Medicaid recipient served by a col laborative governance effort process is only as good as individual outcomes. This thesis model that warrants further consideration by scholars and practitioners.

PAGE 144

128 REFERENCES Aaron, H. J. (2007). Budget crisis, entitlement crisis, health care financing problem which is it? Health Affairs, 26 (6), 1622 1633. doi: 10.1377/hlthaff.26.6.1622 Access to Affordable Health Care Act, SB 06 208, Col. Rev. Stat., 10 16 131. (2006). Retrieved from http://www.leg.state.co.us/clics2006a/csl.nsf/billcon tainers/2E0A3C9A1FEA5274872571150 05ECA5F Agbodzakey, J. K. (2012). Collaborative governance of HIV health services planning councils in Broward and Palm Beach Counties of South Florida. Public Organization Review, 12 107 126. doi: 10.1007/s11115 011 0162 7 Agency for Healthc are Research and Quality (2015) All cause readmissions by payer and age, 2009 2013 Washington, DC: United States Department of Health and Human Services. Retrieved from https://www.hcup us.ahrq.gov/reports/statbriefs/sb199 Readmissions Payer Age.pdf Agranoff, R., & McGuire, M. (2001). Big questions in public network management research. Journal of Public Administration Research and Theory: J PART, 11 (3), 295 326. doi: 10.1093/oxfordjournals.jpart.a003504 Alexander, J. A., Comfort, M. E., & Weiner, B. J. (1998). Governance in public private community health partnerships: A surve y of the community care network: SM demonstration sites. Nonprofit Management & Leadership, 8 (4), 311 332. doi: 10.1002/nml.8402 All Payer Claims Database Council. (2017). State efforts: Colorado. Durham, NH: National Association of Health Data Organizatio ns, University of New Hampshire. Retrieved from https://www.apcdcouncil.org/state/colorado Amara, S., Adamson, R. T., Lew, I., & Slonim, A. (2014). Accountable care organizations: Impact on pharma cy. Hospital Pharmacy, 49 (3), 253 259. doi: 10.1310/hpj4903 253 American Medical Association. (2017). Current procedural terminology. Retrieved from https://www.ama assn.org/practice management/cpt current procedural terminology American Recovery and Reinvestment Act of 2009, Pub. L. No. 111 5, 123 Stat. 226 (2009). Retrieved from https://www.gpo.gov/fdsys/pkg/BILLS 111hr1enr/pdf/BILLS 111hr1enr.pdf Andres, L., & Chapain, C. (2013). The integration of cultural and creative industries into local and regional development strategies in B irmingham and M arseille: Toward s an inclusive and collaborative governance? Regional Studies, 47 (2), 161 182. doi: 10.1080/00343404.2011.644531

PAGE 145

129 Ansell, C., & Gash, A. (2008). Collaborative governance theory and practice. Journal of Public Administration Research and Theory 18 (4), 543 571. doi: 10.1093/jopart/mum032 Ansell, C., & Gash, A. (2012). Stewards, mediators, and catalysts: Toward a model of collaborative leadership. The Innovation Journal, 17 (1), 2 21 Retrieved from http://www.innovation.cc/scholarly style/christoffer_nsell_alison_editit_v17i1a7.pdf Ansell, C., & Gash, A. (2017). Collaborative platforms as a governance strategy. Journal of Public Administration Resear ch and Theory, 1 17. doi:10.1093/jopart/mux030 Antonisse, L., Garfield, R., Rudowitz, R., & Artiga, S. (2017). The effects of Medicaid expansion under the ACA: Updated findings from a literature review. Washington, DC: Kaiser Family Foundation. Retrieved from http://files.kff.org/attachment/Issue Brief The Effects of Medica id Expansion Under the ACA Updated Findings from a Literature Review Balanced Budget Act of 1997, Pub. L. No. 105 33 111 Stat. 2 5 1. (1997). Retrieved from https://www.gpo.gov/fdsys/pkg/PLAW 105publ33/pdf/PLAW 105publ33.pdf Ball, C. (2009). What is transparency? Public Integrity, 11 (4), 293 308. doi: 10.2753/PIN1099 9922110400 Berardo, R., Heikkila, T., & Gerlak, A. K. (2014). Interorganizational engagement in collaborative environmental management: Evidence from the s outh F lorida ecosystem restoration task force. Journal of Public Administration Research and Theory, 24 (3), 697 719. doi: 10.1093/jopart/muu003 Berwick, D. M., Nolan, T. W., & Whittington, J. (2008). The triple aim: Care, health, and cost. Health Affairs, 27 (3), 759 769. doi: 10.1377/hlthaff.27.3.759 Big ideas in collaborative management Armonk NY: M. E. Sha rpe. Binz Scharf, M. C., Lazer, D., & Mergel, I. (2012). Searching for answers: Networks of practice among public administrators. The American Review of Public Administration, 42 (2), 202 225. doi: 10.1177/0275074011398956 Blue Ribbon Commission for Health Care Reform. (2008a). Final report to the Colorado general assembly. https://www.colorado.gov/pacific/sites/default/files/700 832 Commission%20Final%20Report Executive%20Summary.pdf Blue Ribbon Commission for Health Care Reform. (2008 b ). Recommendations for health reform in Colorado Retrieved from http://montrosechamber.com/DocumentCenter/Home/View/25

PAGE 146

130 Brewer, G.A., Neubauer, B.J., & Geiselhart, K. (2006). Designing and implementing e government systems: Critical implications for public administratio n and democracy. Administration & Society, 38 (4), 472 499. doi: 10.1177/0095399706290638 Bryson, J. M., & Crosby, B. C. (2005). Leadership for the common good: Tackling public problems in a shared power world (2nd ed.). San Francisco, CA: Jossey Bass. Bryson, J. M., Crosby, B. C., & Stone, M. M. (2006). The design and implementation of cross sector collaborations: Propositions from the literature. Public Administration Review 66 ( Special Issue) 44 55. Retrieved from http://www.jstor.org/stable/4096569 Burger, E., & Stapleton, K. (2009). Colorado health care resource book Denver, CO: Colorado Legislative Council. Retrieved from https://www.colorado.gov/pacific/sites/default/files/09HealthCareResourcebook.pdf Carboni, J. L., & Milward, H. B. (2012). Governance, privatization, and systemic risk in the disarticulated state Public Administration Review, 72 (s1), S36 S44. doi: 10.1111/j.1540 6210.2012.02670.x Catlin, A. C., & Cowan, C. A. (2015). History of health s pending in the United States, 1960 2013 Washington, DC: Centers for Medicare & Medicaid Services. Retrieved from https://www.cms.gov/ Research Statistics Data and Systems/Statistics Trends and Reports/NationalHealthExpendData/Downloads/HistoricalNHEPaper.pdf Center for Improving Value in Health Care. (2017 a ). CO APCD administration. Retrieved from http://www.civhc.org/kb/database administration/ Center for Improving Value in Health Care. (2017b). Companion instruction guide: R equest for CO APCD data. Retrieved from http://www.civhc.org/get data/data release/apply now/ Center for Improv ing Value in Health Care. (2017c ). Data release. Retrieved from http://www.civhc.org/get data/data release/ Center on Budget and Policy Priorities. (2016). Where do our federal tax dollars go? Retrieved from https://www.cbpp .org/sites/default/files/atoms/files/4 14 08tax.pdf Centers for Disease Control and Prevention. (2016). Health, United States, 2016 Retrieved from https://www.cdc.gov/nchs/data/hus/hus1 6.pdf#076 Centers for Medicare & Medicaid Services. (2008). Actuarial report on the financial outlook for Medicaid Retrieved from https://www.cms.gov/Research Statistics Data and Systems/Research/ActuarialStudies/downloads/MedicaidReport2008.pdf Centers for Medicare & Medicaid Services. (2012). State of Colorado demonstration proposal to integrate care for dual eligible individuals. Retrieved from https://www.cms.gov/Medicare Medicaid Coordination/Medicare and Medicaid

PAGE 147

131 Coordination/Medicare Medicaid Coordination Office/FinancialAlignmentInitiative/Downloads/COPropos al.pdf Centers for Medicare & Medicaid Services. (2015a). Actuarial report on the financial outlook for Medicaid. Retrieved from https://www.medicaid.gov/medicaid/financing and reimbursement/downloads/medicaid actuarial report 2015.pdf Centers for Medicare & Medicaid Services. (2015b ). Medicare & Medicaid Milestones 1937 2015. Retrieved from https://www.cms.gov/About CMS/Agency Information/History/Downloads/Medicare and Medicaid Milestones 1937 2015.pdf Centers for Medicare & Medicaid Services. (2017a). About the Medicare Medicaid C oordination O ffice Retrieved from ht tps://www.cms.gov/Medicare Medicaid Coordination/Medicare and Medicaid Coordination/Medicare Medicaid Coordination Office/index.html Centers for Medicare & Medicaid Services. (2017b). Colorado financial alignment demonstration Retrieved from https://www.cms.gov/Medicare Medicaid Coordination/Medicare and Medicaid Coordinatio n/Medicare Medicaid Coordination Office/FinancialAlignmentInitiative/Colorado.html Centers for Medicare & Medicaid Services. (2017c). Financial alignment initiative. Retrieved from https://www.cms.gov/Medicare Medicaid Coordination/Medicare and Medicaid Coordination/Medicare Medi caid Coordination Office/FinancialAlignmentInitiative/FinancialModelstoSupportStatesEffortsinCareCoordinati on.html Centers for Medicare & Medicaid Services. (2017 d ). History Retrieved from https://www.cms.gov/About CMS/Agency Information/History/index.html Centers for Medi care & Medicaid Services. (2017e ). Medicaid management information system Retrieved fro m https://www.cms.gov/research statistics data and systems/computer data and systems/mmis/ Centers for Medicare & Medicaid Services. (2017 f ). Nation al health expenditure data Retrieved from https://www.cms.gov/Research Statistics Data and Systems/Statistics Trends and Reports/NationalHealthExpendData/index.html Chrislip, D. D., & Larson, C. E. (1994). Collaborative leadership: How citizens and civic leaders can make a difference San Francisco, CA: Jossey Bass. Clarke, S. E. (2017). Local place based collaborative governance: Comparing state centric and society centered models. Urban Affairs Review, 53 (3), 578 602. doi: 10.1177/1078087416637126

PAGE 148

132 Colorado Access. (2013). Retrieved from http://coaccess.com/regional care collaborative organization Colorado Access. (2017). Regional care collaborative organization Retrieved from http://coaccess.com/regional care collaborative organization Colorado All Payer Claims Database HB 10 1330, Col. Rev. Stat 25.5 1 204. (2010). Retrieved from http://www.leg.state.co.us/clics/clics2010a/csl.nsf/billcontainers/7772EFE1E998E627872576 B700617FA4/$FILE/1330_enr.pdf Co lorado Center on Law & Policy. (2012). in combating poverty and providing access to health Retrieved from ht tp://cclponline.org/wp content/uploads/2014/03/History of Medicaid in Colorado.pdf Colorado Community Health Alliance. (2017). Advisory committee meeting minutes Retrieved from http://www.cchacares.com/about ccha/governing board stakeholder committees/advisory committee meeting minutes Colorado Department of Health Care Policy and Financing. (1995). Colorado Medicaid 1995 annual r eport. Retrieved from http://www2.cde.state.co.us/artemis/hcpserials/hcp4101internet/hcp41011995internet.pdf Colorado Department of Health Care Policy and Financing. (2009 a ). A nnual report. Retrieved from https://www.colorado.gov/ hcpf Colorado Department of Health Care Policy and Financing. (2009 b ). Request for information #HCPFKQ1001RFIACC accountable care collaborative Retrieved from https://www.colorado.gov/hcpf Colorado Departme nt of Health Care Policy and Financing. (2010a). Request for proposals RFP #HCPFKQ1102RCCO regional care collaborative organizations for the accountable care collaborative program. Retrieved from https://www.colorado.gov/pacific/sites/default/files/Original%20ACC%20RFP.pdf Colorado Department of Health Care Policy and Financing. (2010b). Request for propos als RFP #HCPFKQ1103SDAC statewide data and analytics services for the accountable care collaborative program. Retrieved from https://colorado.gov/pacific/sites/default/files/Statewide%20Data%20Analytics%20Contract or%20Request%20for%20Proposals.pdf Colorado Department of Health Care Policy and Financing. (2011). Colorado accountable care collaborative program summary. Retrieved from https://colorado.gov/hcpf/

PAGE 149

133 Colorado Department of Health Care Policy and Financing. (2012a). Accountable care collaborative annual report. Retrieved from https://colorado.gov/pacific/sites/default/files/Accountable%20Care%20Collaborative%20A nnual%20Report%202012.pdf Colorado Depar tment of Health Care Policy and Financing. (2012b). Adults without dependent children expansion Retrieved from https://biacolorado.org/biac/wp content/uploads/2012/04/Aw DC 3 12.pdf Colorado Department of Health Care Policy and Financing. (2012 c ). A year of transformation cr e ating a culture of coverage, 2012 annual report. Retrieved from https://colorado.gov/hcpf/ Colorado Depa rtment of Health Care Policy and Financing. (2013a). Legislative request for information #2 accountable care collaborative. Retrieved from https://www.colorado.gov/pacific/sites/default/files/Accountable%20Care%20Collaborative %20Annual%20Report%202013.pdf Colorado Department of Health Care Policy and Financing. (2013 b ). Medicaid caseload FY 2014 15 budget r equest Retrieved from https://www.colorado.gov/pacific/hcpf/fy 2014 15 b udget request Colorado Department of Health Care Policy and Financing. (2014 a ). Accountable care collaborative annual report. Retrieved from https://www.colorado.gov/pacific/sites/default/files/Accountable%20Care%20Collaborative %202014%20Annual%20Report.pdf Colorado Department of Health Care Policy and Financing. (2014b). Accountable care collaborative (ACC): Medicare Medicaid program. Retrieved from https://www.colorado.gov/pacific/hcpf/accountable care collab orative acc medicare medicaid program Colorado Department of Health Care Policy and Financing. (2015a). Accountable care collaborative annual report. Retrieved from https://colorado.gov/pacific/sites/default/files/Supporting%20a%20Culture%20of%20Covera ge%20Accountable%20Care%20Collaborative%202014 15%20Annual%20Report.pdf Colorado D epartment of Health Care Policy and Financing. (2015b). Accountable care collaborative phase II concept paper. Retrieved from https://www.colorado.g ov/pacific/sites/default/files/ACC%20Phase%20II%20Concept%20P aper.pdf Colorado Department of Health Care Policy and Financing. (2015 c ). Medicaid enrollment update. Retrieved from https://colorado.gov/hcpf/

PAGE 150

134 Colorado Department of Health C are Policy and Financing. (2016a ). Hospital quality incentive payment 30 day all cause readmission. Retriev ed from https://colorado.gov/pacific/sites/default/files/2016%20March%20HQIP%2030 day%20all cause%20readmission%20measure.pdf Colorado Department of Health Care Policy and Financing. (2016 b ). health first, 2015 2016 ann ual report Retrieved from https://www.colorado.gov/hcpf/hcpf 2015 2016 annual report Colorado Department of Health Care Policy and Financing. (2017 a ). About us Retrieved from https://www.colorado.gov/pacific/hcpf/About HCPF Colorado Department of Health Care Policy and Financing. (2017 b ). Accountable care collaborative phase II Retrieved from https://www.colorado.gov/pacific/hcpf/accphase2 Colorado Department of Health Care Policy and Financing. (2017c ). Accountable care collaborative program improvement advisory committee. Retrieved from https://www.colorado.gov/pacific/hcpf/accountable care collaborative program improvement advisory committee Colorado Department of Health C are Policy and Financing (2017d ). Executive team Retrieved from https://www.colorado.gov/hcpf/executive team Colorado Department of Health Care Policy and Financing. (2017 e ). Health first Col orado primary care providers we need you in the ACC. Retrieved from https://colorado.gov/pacific/hcpf/medicaid primary care providers we need you acc Colorado Department of Health C are Policy and Financing. (2017f ). Regional care collaborative organizations Retrieved from https://www.colorado.gov/pacific/hcp f/regional care collaborative organizations Colorado Department of Health Care Policy and Financing. (2017g). Request for proposals 2017000265 regional accountable entity for the accountable care collaborative Retrieved from https://www.colorado.gov/pacific/sites/default/files/Regional%20Accountable%20Entity%20 for% 20the%20Accountable%20Care%20Collaborative%20RFP%20Body.pdf Colorado Department of Health Care Policy and Financing & Colorado Department of Human Services. (2014). Uniform s ervice c oding s tandards d raft m anual. Gettysburg, PA: Open Minds. Retrieved from http://mstservices.com/files/CO%20MST%20Medicaid%20Standard.pdf Colorado Department of Local Affairs. (2017). State demography office data. Retrieved from https://demography.dola.colorado.gov/data/

PAGE 151

135 Colorado Health Institute. (2005). Colorado Medicaid: A primer. Retrieved from http://www.coloradohealthinstitute.org/sites/default/files/file_attachments/MedicaidPrimer.p df Colorado Health Institute. (2012). An examination of emergency department use in Colorado Denver, CO: The Colorado Trust. Retrieved from http://www.coloradotrust.org/sites/default/files/FINALEmergency_Dept_IB_10 15 12 PDF.pdf Colorado Health Institute. (2016 a ). Medicaid expansion in Colorado Retrieved from http://www.coloradohealthinstitute.org/sites/default/files/migrated/downloads/MK_Expansio n_Report.pd f Colorado Health Institute. (2016 b ). Sticker shock: More Coloradans can afford prescriptions, but problems remain. Retrieved from http s://www.coloradohealthinstitute.org/research/sticker shock prescription drug affordability colorado Colorado Health Institute. (2017). Data: Health insurance. Retrieved from https://www.coloradohealthinstitute.org/data Colorado Multiple Institutional Review Board. (2017). Policies and procedures. Retrieved from http://www.ucdenver.edu/researc h/comirb/policies/Pages/default.aspx Community Care Central Colorado. (2017). News Retrieved from http://www.mycommunitycare.org/news/ financing mechanisms. Health Affairs, 23 (2), 245 257. Retrieved from http://content.healthaffairs.o rg/content/23/2/245.full Cylus, J., Hartman, M., Washington, B., Andrews, K., & Catlin, A. (2010). Pronounced gender and age differences are evident in personal health care spending per person. Health Affairs (Project Hope), 30 (1), 153 160. doi: 10.1377/h lthaff.2010.0216 de Loe R. C., Murray, D., & Simpson, H. C. (2015). Farmer perspectives on collaborative approaches to governance for water. Journal of Rural Studies, 42 191 205. doi: 10.1016/j.jrurstud.2015.10.005 Donz, J., Aujesky, D., Williams, D., & Schnipper, J. L. (2013). Potentially avoidable 30 day hospital readmissions in medical patients: Derivation and validation of a prediction model. JAMA Internal Medicine, 173 (8), 632 638. doi: 10.1001/jamainternmed.2013.3023 el Ansari, W. (2003). Educatio nal partnerships for public health: Do stakeholders perceive similar outcomes? Journal of Public Health Management and Practice, 9 (2), 136 156. Retrieved from

PAGE 152

136 http://journals.lww.com/jphmp/Abstract/2003/03000/Educational_Partnerships_for_Public_ Health__Do.6.aspx Emerson, K., & Nabatchi, T. (2015). Evaluating the productivity of collaborative governance regimes: A performance matrix. Publ ic Performance & Management Review, 38 (4), 717 747. doi: 10.1080/15309576.2015.1031016 Emerson, K., Nabatchi, T., & Balogh, S. (2012). An integrative framework for collaborative governance. Journal of Public Administration Research and Theory, 22 (1), 1 29. doi: 10.1093/jopart/mur011 Farazmand, A. (2004). Sound governance: Policy and administrative innovations Westport, CT: Praeger. Fawcett, S. B., Paine Andrews, A., Francisco, V. T., Schulz, J. A., Richter, K. P., Lewis, R. K., Lopez, C. M (1995). Using empowerment theory in collaborative partnership for community health and development. American Journal of Community Psychology, 23 (5), 677 697. Retrieved from https://www.ncbi.nlm. nih.gov/pubmed/8851345 Fee for service. (n.d.). In Merriam Webster online Retrieved from https://www.merriam webster.com/dictionary/fee for service Feng, J., & Maini, L. (2016) Where is the good stuff? Drug quality measures and economic research. Retrieved from http://www.nber.org/aging/valmed/WhitePaper Feng,Maini12.2016.pdf Fetterman, D. M. (2 010). Ethnography: Step by step (3rd ed.). Thousand Oaks, Ca: Sage Publications. Finney, J. W., Humphreys, K., Kivlahan, D. R., & Harris, A. H. S. (2011). Why health care process performance measures can have different relationships to outcomes for patients and hospitals: Understanding the ecological fallacy. American Journal of Public Health, 101 (9), 1635 1642. doi: 10.2105/AJPH.2011.300153 Fisher, E. S., Staiger, D. O., Bynum, J., & Gottlieb, D. J. (2007). Creating accountable care organizations: T he extended hospital medical staff. Health Affairs, 26 (1), w44 w57. doi: 10.1377/hlthaff.26.1.w44 Freeman, J. (1997). Collaborative governance in the administrative state. UCLA Law Review, 45 (1), 1 77 Retrieved from https://papers.ssrn.com/sol3/papers.cfm?abstract_id=11408 Friedberg, M. W., Hussey, P. S., & Schneider, E. C. (2010). Primary care: A critical review of the evidence on quality and costs of health care. Health Affairs (Project Hope), 29 (5), 766 772. doi: 10.1377/hlthaff.2010.0025

PAGE 153

137 Fung, A., & Wright, E. O. (2001). Deepening democracy: Innovations in empowered participatory governance. Politics & Society, 29 (1), 5 41. doi: 10.1177/00323 29201029001002 Futrell, R. (2003). Technical adversarialism and participatory collaboration in the U.S. chemical weapons disposal program. Science, Technology, & Human Values, 28 (4), 451 482. doi: 10.1177/0162243903252762 Gerlak, A. K., & Heikkila, T. (2 011). Building a theory of learning in collaboratives: Evidence from the E verglades restoration program. Journal of Public Administration Research and Theory: J PART, 21 (4), 619 644. doi: 10.1093/jopart/muq089 Gibson, R. (2011). A primer on collaborative m ulti level governance. Harris Centre, Memorial University. Retrieved from http://research.library.mun.ca/310/ Gifford, K., Smith, V. K., Snipes, D., & Paradise, J. (2011). A profile of Medicaid managed care programs in 2010: Findings from a 50 state survey. Washington, DC: Health Management Associates and Kaiser Family Foundation. Retrieved from http://leg.mt.gov/content/Committees/Interim/2011 2012/Children Family/Topics/Medicaid%20Monitoring/kaiser profile of medicaid managed care.pdf Gilliam, A., Davis, D., Barrington, T., Lacson, R., Uhl, G., & Phoenix, U. (2002). The value of engaging stakeholders in planning and implementing evaluations. AIDS Education and Prevention, 14 (3), 5 17. doi: 10.1521/aeap.14.4.5.23878 Goldfield, N. I., McCullough, E. C., Hughes, J. S., Tang, A. M., Eastman, B., Rawlins, L. K., & Averill, R. F. (2008). Identifying potentially preventable readmissions. Health Care Financing Review, 30 (1), 75 91. Retrieved from https://www.ncbi.nl m.nih.gov/pmc/articles/PMC4195042/ Gollagher, M., & Hartz Karp, J. (2013). The role of deliberative collaborative governance in achieving sustainable cities. Sustainability, 5 (6), 2343 2366. doi: 10.3390/su5062343 Governance. (n.d.). In Merriam Webster online Retrieved from https://www.merriam webster.com/thesaurus/governance Grabowski, D. C. (2007). Medicare and Medicaid: Conflicting incentives for long term care. The Milbank Quarte rly, 85 (4), 579 610. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/18070331 Gray, B. (1989). Collaborating: Finding common ground for multiparty problems San Francisco, CA: Jossey Bass. Hansen, M. K. (2013). Medicaid and the safety n et. Denver, CO: National Conference of State Legislatures. Retrieved from http://www.ncsl.org/documents/health/MedicaidNetTK13.pdf

PAGE 154

138 Hay s, W. L. (1994). Statistics (5th ed.). Belmont, CA: Wadsworth Cengage Learning. Healey, J. F. ( 1999). Statistics: A tool for social research (5th ed.). New York, NY: Wadsworth Publishing Company. Health Insurance Portability and Accountability Act, Pub. L. No. 104 191, 110 Stat. 1936 (1996). Retrieved from https://www.gpo.gov/fdsys/pkg/PLAW 104publ191/pdf/PLAW 104publ191.pdf Health Services Advisory Group. (2013a). FY 2012 2013 accountable care collaborative site review aggregate report Retrieved from https://www.colorado.gov/pacific/sites/default/files/Aggr egate%202013%20RCCO%20Site %20Review.pdf Health Services Advisory Group. (2013 b ). FY 2012 2013 site review report for integrated health care partners (region 4) Retrieved from https://www.colorado.gov/pacific/sites/default/files/Region%204%20Integrated%20Commun ity%20Health%20Partners%202013%20RCCO%20Site%20Review.pdf Health Services Advisory Group (2014). FY 2013 2014 site review report for rocky mountain health plans (region 1). Retrieved from https://www.colorado.gov/pacific/sites/default/files/Region%201%20Rocky%20Mountain% 20Health%20Plans%202014%20RCCO%20Site%20Review.pdf Health Services Advisory Group. (2015). FY 2014 2015 site review report executive summary for Colorado community health a lliance (region 6). Retrieved from https://www.colorado.gov/pacific/sites/default/files/Region%206% 20Colorado%20Communi ty%20Health%20Alliance%202015%20RCCO%20Site%20Review.pdf Heikkila, T., & Gerlak, A. K. (2016). Investigating collaborative processes over time: A 10 year study of the south F lorida ecosystem restoration task force. The American Review of Public Administration, 46 (2), 180 200. doi: 10.1177/0275074014544196 Herranz, J. (2008). The multisectoral trilemma of network management. Journal of Public Administration Research and Theory: J PART, 18 (1), 1 31. doi: 10.1093/jopart/mum004 Hewlett Packard. (2014, April 22). Colorado awards HP $116 million Medicaid contract Retrieved from http://www8.hp.com/us/en/hp news/press release.html?id=163 3252#.WfT8sbpFzIU Higginbottom, G. M. A., Pillay, J. J., & Boadu, N. Y. (2013). Guidance on performing focused ethnographies with an emphasis on healthcare research. The Qualitative Report, 18 (17), 1 16. Retrieved from http://nsuworks.nova.edu/tqr/vol18/iss9/1/

PAGE 155

139 Hill, I. (2004). State responses to budget crises in 2004: Colorado Washington, DC: The Urban Institute. Retrieved from http://www.urban.org/research/publication/state responses budget crises 2004 colorado Hill, I., Courtot, B., Bovbjerg, R. R., & Adams, F. (2012). ACA impl ementation monitoring and tracking: Colorado site visit report. Washington, DC: Urban Institute. Retrieved from http://www.urb an.org/UploadedPDF/412537 ACA Implementation Monitoring and Tracking Colorado Site Visit Report.pdf Holahan, J., Zuckerman, S., Evans, A., & Rangarajan, S. (1998). Medicaid managed care in thirteen states. Health Affairs, 17 (3), 43 63. doi: 10.1377/hlthaff.17.3.43 Hudman, J., & Starfield, B. (1999). Children and welfare reform: What are the effects of this landmark policy change? In Wallace, H. M., & Green, G. (Eds.), Health and welfare for families in the 21st century (pp. 115 133). Sudbu ry, MA: Jones and Bartlett Publishers. Humphrey, H. H. (1997). 123 Cong. Rec. 37287. Retrieved from https://www.gpo.gov/fdsys/pkg/GPO CRECB 1977 pt29/pdf/GPO CRECB 1977 pt29 1 2.pdf Hurley, R. E., & Somers, S. A. (2003). Medicaid and managed care: A lasting relationship? Health Affairs 22 (1), 77 88. doi: 10.1377/hlthaff.22.1.77 Huxham, C., & Vangen, S. (2000). Leadership in the shaping and implementation of collaboration agendas: How things happen in a (not quite) joined up world. Academy of Management Journal 43 (6) 1159 1175. Retrieved from http://oro.open.ac.uk/16221/1/Leadership%20in%20the%20shaping%20....pdf Iglehart, J. K. (2009). Doing more with less: A conversation with Kerry Weems. Health Affairs, 28 (4), w688 w696. doi: 10.1377/hltha ff.28.4.w688 Iglehart, J. K., & Sommers, B. D. (2015). Medicaid at 50 from welfare program to nation's largest health insurer. The New England Journal of Medicine, 372 (22), 2152 2159. doi:10.1056/NEJMhpr1500791 Institute for Digital Research and Education. (2017). Los Angeles, CA: University of California, Los Angeles. Retrieved from https://stats.idre.ucla.edu/ Integrated Community Health Partners. (2017). Performanc e advisory committee Retrieved from http://www.ichpcolorado.com/members/performance committee.html Jacobson, G., Neuman, T., & Damico, A (2012 ). eligible beneficiaries Washington, DC: The Kaiser Family Foundation. Retrieved from http://www.kff.org/medicare/upload/8138.pdf

PAGE 156

140 Joffe, M. D. (2015). Long term trends in Medicaid spending by the states Arlington, VA: Mercatus Center, George Mason University. Retrieved from https://www.mercatus.org/system/files/Joffe State Medicaid Spending.pdf Kaiser Family F oundation. (2011a). Medicaid enrollment by age. Retrieved from http://www.kff.org/med icaid/state indicator/medicaid enrollment by age/?currentTimeframe=0&sortModel=%7B%22colId%22:%22Location%22,%22sort%22: %22asc%22%7D Kaiser Family Foundation. (2011b). Medicaid enrollment by gender. Retrieved from http://www.kff.org/medicaid/state indicator/medicaid enrollment by gender/?currentTimeframe=0&sortModel=%7B%22colId%22:%22 Location%22,%22sort%2 2:%22asc%22%7D Kaiser Family Foundation. (2011 c ). Medicaid enrollment by race/ethnicity. Retrieved from http://www.kff.org/medicaid/state indicator/medicaid enrollment by raceethnic ity/?currentTimeframe=0&sortModel=%7B%22colId%22:%22Location%22,%22s ort%22:%22asc%22%7D Kaiser Family Foundation. (2012). Emerging Medicaid accountable care organizations: The role of managed care Retrieved from https://kaiserfamilyfoundation.files.wordpress.com/2013/01/8319.pdf Kaiser Family Foundation. (2013). accountable care collaborative Retrieved from https://kaiserfamilyfoundation.files.wordpress.com/2013/10/8484 data analytics in medicaid1.pdf Kaiser Family Foundation. (2014). Medicaid moving forward. Retrieved from http://kaiserfamilyfoundation.files.wordpress.com/2014/06/7235 07 medicaid moving forward2.pdf Karabat Policy & Practice, 69 (5), 10 13. Retrieved from http://connection .ebscohost.com/c/articles/66871124/colorados answer integrated delivery system reform Kasper, J., Watts, M.O., & Lyons, B. (2010). Chronic disease and comorbidity among dual eligibles: Implications for patterns of Medicaid and Medicare service use and sp ending. Washington, DC: The Kaiser Family Foundation. Retrieved from http://www.kff.org/medicaid/upload/8081.pdf Keckley, P., & Kalkhof, C (2007). Mending holes in the M edicaid safety net: A str ategy for state healthcare reform. Healthcare Financial Management: Journal of the Healthcare Financial Management Association, 61 (12), 56 62 Retrieved from

PAGE 157

141 http://connection.ebscohost.com/c/articles/27796117/mending holes medicaid safety net strategy state healtcare reform Ket tl, D. F. (2000). The transformation of governance: Globalization, devolution, and the role of government. Public Administration Review, 60 (6), 488 497. doi: 10.1111/0033 3352.00112 Kettl, D. F. (2002). The transformation of governance: Public administration for twenty first century America. Baltimore, MD: Johns Hopkins University Press. Klees, B. S., Wolfe, C. J., & Curtis, C. A. (2010). Brief Summaries of Medicare & Medicaid Baltimore, MD: Centers for Medicare & Medicaid Services. Retrieved from http://ww w.cms.gov/Research Statistics Data and Systems/Statistics Trends and Reports/MedicareProgramRatesStats/downloads/MedicareMedicaidSummaries2009.pdf Klemm, J. D. (2000). Medicaid spending: A brief history. Health Care Financing Review (22) 1, 105 112. Retri eved from https://www.cms.gov/Research Statistics Data and Systems/Research/HealthCareFinancingReview/downloads/00fallpg10 5.pdf Koebele, E. A. (2015). Assessing outputs, outcomes, and barriers in collaborative water governance: A case study. Journal of Contemporary Water Research & Education, 155 (1), 63 72. doi: 10.1111/j.1936 704X.2015.03196.x Ku, L. (2000). Limiting upper payment limit Washington, DC: Center on Budget and Policy Priorities. Retrieved from https://www.cbpp.org/archiveSi te/9 27 00health.pdf Lasker, R. D., Weiss, E. S., & Miller, R. (2001). Partnership synergy: A practical framework for studying and strengthening the collaborative advantage. The Milbank Quarterly, 79 (2), 179 205. doi: 10.1111/1468 0009.00203 LeCompte, M. D., & Schensul, J. J. (2010). Designing & conducting ethnographic research: An introduction (2nd ed.). Lanham, MD: AltaMira Press. Evaluation of the accountable ca re collaborative, methodological appendix. Aurora, CO: University of Colorado Anschutz Medical Campus. Retrieved from https://www.colorado.gov/pacific/sites/default/files/ACC%20Evaluation%20Methodological %20Appendix.pdf Lutzky, A. W., Holahan, J., & Wiener, J. M. (2002). Health policy for low income people: Profiles of 13 states. Washington, DC: The Urban Institute. Ret rieved from http://www.urban.org/sites/default/files/publication/60101/310490 Health Policy for Low Income People P rofiles of States.PDF

PAGE 158

142 Lynn, L. E., Jr., Heinrich, C., & Hill, C. (2000). Studying governance and public management: Challenges and prospects. Journal of Public Administration Research and Theory, 10 (2), 233 262. Retrieved from https://academic.oup.com/jpart/article abstract/10/2/233/986113?redirectedFrom=fulltext Managed care. (n.d.). In Merriam Webster online Retrieved from https://www.merriam webster.com/dictionary/managed%20care Mann, C., & Westmoreland, T. (2004). Attending to Medicaid. The Journal of Law, Medicine & Ethics, 32 (3), 416 425. doi: 10.11 11/j.1748 720X.2004.tb00152.x Mayo Clinic. (2017a). CT scan Retrieved from http://www.mayoclinic.org/tests procedures/ct scan/basics/definition/prc 20014610 Mayo Clinic. (2017b). MRI Retrieved from http://www.mayoclinic.org/tests procedures/mri/home/ovc 20235698 Mayo Clinic. (2017c). Ultrasound Retrieved from http://www.mayoclinic.org/tests procedures/ultrasound/basics/definition/prc 20020341 McGinnis, T., & Small, D. M. (2012). Accountable care organizations in Medicaid: E merging practices to guide program design Hamilton, NJ: Center for Health Care Strategies. Retrieved from https://www.chcs .org/resource/accountable care organizations in medicaid emerging practices to guide program design/ McGuire, M. (2006). Collaborative public management: Assessing what we know and how we know it. Public Administration Review, 66 (s1), 33 43. doi: 10.1111/j.1540 6210.2006.00664.x Means tested. (n.d.). In Merriam Webster online. Retrieved from https://www.merriam webster.com/dictionary/means tested Medicaid and CHIP Payment and Access Commission. (2012 a ). Medicaid UPL supplemental payments Retrieved from https://www.macpac.gov/wp content/uploads/2015/01/MACFacts UPL Payments_2012 11.pdf Medicaid and CHIP Payment and Access Commission. (2012b). Report to the Congress on Medicaid and CHIP Retrieved from https://www.macpac.gov/wp content/uploads/2015/01/2012 03 15_MACPAC_Report1.pdf Medicaid and CHIP Payment and Access Co mmission. (2016 ). Medicaid spending for prescription drugs. Retrieved from https://www.macpac.gov/wp content/uploads/2016/01/Medicaid Spending for Prescription Drugs.pdf Medicare. (n.d.). How Medicare is funded Retrieved from https://www.medicare.gov/about us/how medicare is funded/medicare funding.html

PAGE 159

143 Medicare Payment Advisory Commission. (2004). Report to the Congress: New approach es in Medicare. Washington, DC: MedPAC. Retrieved from http://www.worldcat.org/title/report to the congress new approaches in medicare/oclc/558917 83 Medicare Payment Advisory Commission. (2011). Report to the Congress: Medicare and the health care delivery system. Washington, DC: MedPAC. Retrieved from http://medpa c.gov/docs/default source/reports/Jun11_EntireReport.pdf Michel, C. L. (2017). Collaborative governance: Beyond mere participation. Singapore: Third International Public Policy Conference. Retrieved from http://www.ippapublicpolicy.org/file/paper/594b2070b72f6.pdf Mills, W. (1987). Oral history interview II with Wilbur Mills/Interviewer: Michael L. Gillette LBJ Library, Austin, TX. Retrieved from https://www.ssa.gov/history/pdf/mills2.pdf Milward, H. B., & Provan, K. G. (2000). Governing the hollow state. Journal of Public Administration Research and Theory 10 (2), 359 379. Retrieved from https://academic.oup.com/jpart/article abstract/10/2/359/986124 Mitchell, S. M., & Shortell, S. M. (2000). The governance and management of effective community health partnerships: A typology for research, policy, and practice. The Milbank Quarterly, 78 (2), 241 289. doi: 10.1111/1468 0009.00170 Mizrahi, T., & Abramson, J. S. (2000). Collaboration between social workers and physic ians. Social Work in Health Care, 31 (3), 1 24. doi: 10.1300/J010v31n03_01 Moffitt, R., & Wolfe, B. (1992). The effect of the Medicaid program on welfare participation and labor supply. The Review of Economics and Statistics, 74 (4), 615 626. doi: 10.2307/210 9375 Mosley, J. E. (2014). Collaboration, public private intermediary organizations, and the transformation of advocacy in the field of homeless services. The American Review of Public Administration, 44 (3), 291 308. doi: 10.1177/0275074012465889 Nathan, R. P. (2005). Federalism and health policy. Health Affairs, 24 (6), 1458 1466. doi: 10.1377/hlthaff.24.6.1458 Oh, Y., & Bush, C. B. (2016). Exploring the role of dynamic social capital in collaborative governance. Administration & Society, 48 (2), 216 236. doi: 10.1177/0095399714544941 O'Leary, R., Bingham, L. B., & Choi, Y. (2010). Teaching collaborative leadership: Ideas and lessons for the field. Journal of Public Affairs Education, 16 (4), 565 592. Retrieved from http://www.jstor.org/stable/20790766

PAGE 160

144 collaborative public management. Public Administration Review, 66 (Special Issue), 6 9. Retrieved from http://www.jstor.org/stable/4096565 Osborne, J. W., & Waters, E. (2002). Four assumptions of multiple regression that researchers should always test. Practical Research, Research & Assessment, 8 (2), 1 5. Retrieved from http://PAREonline.net/getvn.asp?v=8&n=2 Ostrom, E. (2002). Policy analysis in the future of good societies. The Good Society, 11 (1), 42 48. Retrieved from http://www.jstor.org/stable/20711048 Ostrom, E. (2007). A diagnostic approach for going beyond panaceas. Proceedings of the National Academy of Sciences of the United States of America, 104 (39), 15181 15187. doi: 10.1073/pnas.0702288104 Padgett, S. M., Bekemeier, B., & Berkowitz, B. (2004). Collaborative partnerships at the state level: Promoting systems changes in public health infrastructure. Journal of Public Health Management and Practice 10 (3), 251 257. Retrieved from http://journals.lww.com/jphmp/Abstract/2004/05000/Collaborative_Partnerships_at_the_Stat e_Level_.9.aspx Page, S. (2010). Integrative leadership for collaborative g overnance: Civic engagement in S eattle. The Leadership Quarterly, 21 (2), 246 263. doi: 10.1016/j.leaqua.2010.01.005 Parker, R. M. (1993). Threats to the validity of research. Rehabilitation Counseling Bulletin, 36 (3), 130 138. Retrieved fro m http://psycnet.apa.org/record/1993 28138 001 Patient Protection and Affordable Care Act, Pub. L. No. 111 148, 124 Stat. 119 (2010). Retrieved from https://www.congress.gov/111/plaws/publ148/PLAW 111publ148.pdf Patrick, S. W., & Freed, G. L. (2012). Intergenerational enrollment and expenditure changes in Medicaid: Trends fr om 1991 to 2005. Health Services Research, 12 (327), 1 7. doi: 10.1186/1472 6963 12 327 Pear, R. (1987, September 20). Medicare expansion imperiled by high costs. The New York Times Retrieved from http://www.nytimes.com/1987/09/20/weekinreview/the nation medicare expansion imperiled by high costs.html Personal Responsibility and Work Opportunity Reconciliation Act, Pub. L. No. 104 193 110 Stat. 2105 (1996). Retrieved from https://www.gpo.gov/fdsys/pkg/BILLS 104hr3734enr/pdf/BILLS 104hr3734enr.pdf Privacy Act, Pub. L. No. 93 579, 88 Stat. 1896 (1974). Retrieved from http://www.justice.gov/opcl/privstat.htm

PAGE 161

145 Public Knowledge LLC. (2012). MMIS procurement analysis report Retrieved from https://www.colorado.gov/pacific/sites/default/files/MMIS%20Procurement%20Analysis%2 0Report%20April%2018%202012.pdf Purdy, J M. (2012). A Framework for assessing power in collaborative governance processes. Public Administration Review, 72 (3), 409 417. doi: 10.1111/j.1540 6210. 2011. 02525.x Rector, R., & Sheffield, R. (2016). Setting priorities for welfare reform. Washington, DC: The Heritage Foundation. Retrieved from http://www.heritage.org/welfare/report/setting priorities welfare reform Regents Health Resources (2012). CT and MRI: Regional variations in utilization and reimbursement. Retrieved from http://www.regentshealth.com/media/regents_imagingmarketfile_rbj_febmar2012.pdf Reilly, T. (2001). Collaboration in action: An uncertain process. Administration in Social Work, 25 (1), 53 74. doi: 10.1300/J147v25n01_06 Ritter, B. (2006). Bill Ritter for governor 2006: The Colorado promise Retrieved from http://coyotegulch.files.wordpress.com/2009/03/ritter_policy_book.pdf Rocky Mountain Health Plans. (2017). Previous community advisory meetings. Retrieved from http://live rocky.pantheonsite.io/content/previous community advisory meetings Rodin, D., & Silow Carroll, S. (2013). Medicaid payment and delivery reform in Col orado: ACOs at the regional level. Washington, DC: The Co mmonwealth Fund. Retrieved from http://www.common wealthfund.org/~/media/Files/Publications/Case%20Study/2013/Mar/16 66_Rodin_Medicaid_Colorado_case_study_FINAL_v2.pdf Rosenbaum, S. (2011). Medic aid and access to h eal th c are a proposal for continued i naction? The N ew England Journal of Medicine, 365 (2) 102 104. doi: 10.1056/NEJMp1106046 Rosenbaum, S., Rothenberg, S., Schmucker, S., Gunsalus, R., & Beckerman, J. Z. (2017). Washington, DC: The Commonwealth Fund. Retrieved from http://www.commonwealthfund.org/~/media/files/publications/issue brief/2017/mar/1936_rosenbaum_repealing_aca_affect_medicaid_ib.pdf Roussos, S. T., & Fawcett, S. B. (2000). A review of collaborative partnerships as a strategy for improving community health. Annual Review of Public Health, 21 (1), 369 402. doi: 10.1146/annurev.publhealth .21.1.369 Savage, J. (2000). Ethnography and health care. BMJ, 321 (7273), 1400 1402. doi: 10.1136/bmj.32 1.7273.1400

PAGE 162

146 Shadish, W. R. (2011). The truth about validity. New Directions for Evaluation, 2011 (130), 107 117. doi: 10.1002/ev.369 Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasiexperimental designs for generalized causal in ference. Boston, MA: Houghton Mifflin. Siddiki, S. N., Carboni, J. L., Koski, C., & Sadiq, A. (2015). How policy rules shape the structure and performance of collaborative governance arrangements. Public Administration Review, 75 (4), 536 547. doi: 10.1111/puar.12352 Singleton, R. A., Jr., & Straits, B. C. (2010). Approaches to social research (5th ed.). New York, NY: Oxford University Press. Smith, M., Bates, D. W., & Bodenheimer, T. S. (2013). Pharmacists belong in accountable care organizations and integrated care teams. Health Affairs (Project Hope), 32 (11), 1963 1970. doi: 10.1377/hlthaff.2013.0542 Smith Bindman, R., Miglioretti, D. L., & Larson, E. B. (2008). Rising use of diagnostic medical imaging in a large integrated health system. Health Affairs, 27 (6), 1491 1502. doi: 10.1377/hlthaff.27.6.1491 Social capital. (n.d.) In Collins English dictionary online. Retrieved from https://www.collinsdictionary.com/dic tionary/english/social capital Social Security Act, Title XVIII, 42 USC 1801 1899. (1965). Retrieved from https://www.ssa.gov/OP_Home/ssact/title18/1800.htm#ft2 Social Security Act, Title XIX 42 USC 1900 1946. (1965). Retrieved from https://www.ssa.gov/OP_Home/ssact/title19/1900.htm Social Security Administration. (1996). Major welfare reforms enacted in 1996. Social Security Bulletin, 59 (3), 56 63. Retrieved from https://www.ssa.gov/policy/docs/ssb/v59n3/v59n3p56.pdf Sommers, B. D. & Epstein, A. M. (2011). Why states are so miffed about Medicaid economics, politics, and the "woodwork effect". The New England Journal of Medicine, 365 (2), 100 102. Retrieved from http://www.nejm.org/doi/full/10.1056/NEJMp1104948 Sorensen, E., & Torfing, J. (2011). Enhancing collaborative innovation in the public sector. Administration & Society, 43 (8), 842 868. doi: 10.1177/0095399711418768 Starr, P. (2015). The health care legacy of the Great Society. In Wilson, R. H., Glickman, N. J., & Lynn, L. E. (Eds.), LBJ's neglected legacy: H ow Lyndon Johnson reshaped domestic policy and government (pp. 235 258). Austin, TX: University of Texas Press.

PAGE 163

147 State Policy Options (2012). Colorado builds Medicaid Accountable Care Collaborative to reduce costs and improve quality Washington, DC: National Governors Association. Retrieved from https://www.nga.org/cm s/center Stern, R. G. (2013). "Can we all get along?" Cooperative strategies to reduce imaging overuse. The American Journal of Medicine, 126 (8), 657. Retrieved from http://www. amjmed.com/article/S0002 9343(13)00278 7/fulltext Stevens, R. A. (1996). Health care in the early 1960s. Health Care Financing Review, 18 (2), 11 22. Retrieved from https://www.ssa.gov/history/pdf/HealthCareEarly1960s.pdf Collaborative partnerships in community education. Journal of Education Policy, 18 (1), 37 51. Retrieved from http://www.tandfonline.com/doi/abs/10.1080/0268093032000042191 The State Health Care Spending Project. (2014). State health care s pending on Medicaid Washington, DC: The Pew Charitable Trusts & John D. and Catherine T. MacArthur Foundation. Retrieved from http://www.pewtrusts.org/~/media/Data Visualizations/Interactives/2014/Medicaid/downloadables/State_Health_Care_Spending_on_ Medicaid.pdf Thomson, Rhodes, & Cowie, P.C. (2000). Kaiser sues Medicaid officials in Colorado Managed Care Law Update 3 (9), 1. Retrieved from http://www.trc law.com/wp content/uploads/news bulletins/managed care/Managed Care Law Update 2000 December.pd f 3M Health Information Systems. (2014, April 1). 3M completes acquisition of treo solutions. Retrieved from http://news.3m.com/press release/company/3m comp letes acquisition treo solutions 3M Health Information Systems. (2015). Potentially preventable readmissions grouping software. Retrieved from http://multi media.3m.com/mws/media/849903O/3m ppr grouping software fact sheet.pdf Ticket to Work and Work Incentives Improvement Act of 1999, Pub. L. No. 106 170, 113 Stat. 1860. (1999). Retrieved from https://www.gpo.gov/fdsys/pkg/PLAW 106publ170/pdf/PLAW 106publ170.pdf Treo Solutions. (2012). Key performance indicator calculations for ACC incentive payments. Retrieved from https://colorado.gov/hcpf/ Truven Health Analytics. (2015, June 16). Truven health analytics awarded Colorado Department of Health Care Policy and Financing business intelligence and data management services contract worth $86M. Retrieved from http://truvenhealth.com/media room/press releases/detail/prid/167/truven health awarded colorado hcpf contract worth $86m

PAGE 164

148 Unite d States Census Bureau. (2015). Millennials outnumber baby boomers and are far more diverse Retrieved from https://www.census.gov/newsroom/press releases/2015/cb15 113.html United States Department of Health and Human Services, Health Care Financing Administration. (2000). A profile of Medicaid: Chartbook 2000. Retrieved from https://www.cms.gov/Research Statistics Data and Systems/Statistics Trends and Reports/TheChartSeries/downloads/2Tch artbk.pdf United States Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation. (2005). Using Medicaid to support working age adults with serious mental illnesses in the community: A handbook. Retrieved fro m https://aspe.hhs.gov/system/files/pdf/74111/handbook.pdf United States Department of Labor Bureau of Labor Statistics (2017). Databases, tables and calculators by subject. Retrieved from https://data.bls.gov/timeseries/CUUR0000SAM?output_view=pct_12mths University of Colorado Denver. (2017). IT security program policy. Retrieved from https://www.cu.edu/ope/aps/6005 Van de Ven, A. H. (2007). Engaged scholarship: A guide for organizational and social research. New York, NY: Oxford University Press. Vangen, S., & Huxham, C. (2003a). Enacting leadership for collaborative advantage: Dilemmas of ideology and pragmatism in the activities of partnership managers. British Journal of Management, 14 (s1), S61 S76. doi: 10.1111/j.1467 8551. 2003. 00393.x Vangen, S., & Huxham, C. (2003 b ). Nurturing collaborative relations: Building trust in interorganizational collaboration. The Journal of Applied Behavioral Science, 39 (1), 5 31. doi: 10.1177/0021886303039001001 Varda, D. M. (2011). Data driven management strategies in public health collaboratives. J ournal of Public Health Management Practices, 17 (2), 122 132. doi: 10.1097/PHH.0b013e3181ede995 Varda, D. M., & Retrum, J. H. (2012). An exploratory analysis of network characteristics and quality of interactions among public health collaboratives. Journal of Public Health Research, 1 (2), 170 176. Retrieved from http://www.jphres.org/index.php/jphres/article/view/jphr.2012.e27 Varda, D. M., & Retrum, J. H. (2015). Collaborati ve performance as a function of network members' perceptions of success. Public Performance & Management Review, 38 (4), 632 653. doi: 10.1080/15309576.2015.1031006

PAGE 165

149 Varda, D. M., Shoup, J. A., & Miller, S. (2012). A systematic review of collaboration and network research in the public affairs literature: Implication for public health practice and research. American Journal of Public Health, 102 (3), 564 571. doi: 10.210 5/AJPH.2011.300286 Vargha, A. & Delaney, H. D. (1998). The Kruskal W allis test and stochastic homogeneity. Journal of Educational and Behavioral Statistics, 23 (2), 170 192. doi: 10.2307/1165320 Verdier, J. M., Byrd, V., & Stone, C. (2009). Enhanced prima ry care case management programs in Medicaid: Issues and options for states Hamilton, NJ: Center for Health Care Strategies, Inc. Retrieved from http://www.chcs.org/media/EPCCM_Full_Report.pd f Vogeli, C., Shield, A.E., Lee, T.A., Gibson, T.B., Marder, W.D., Weiss, K.B., & Blumenthal, D. (2007). Multiple chronic conditions: Prevalence, health consequences, and implications for quality, care management, and costs. Journal of General Internal Medicine, 22 http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2150598/pdf/11606_2007_Article_322.pdf Wagn er, E., Austin, B., & Coleman, C. (2006). It takes a region: Creating a framework to improve chronic disease care. Oakland, CA: The California HealthCare Foundation. Retrieved from http://maccollcenter.org/sites/default/files/CreatingAFrameworkToImproveChronicDiseaseC are.pdf Weber, E. P., & Khademian, A. M. (1997). From agitation to collaboration: Clearing the air through negotiation. Public Administration Review, 57 (5), 396 410. Retrieved from http://journals.sagepub.com/doi/abs/10.1177/0275074004269288 Weech Maldonado, R., & Merrill, S. B. (2000). Building partnerships with the community: Lessons from the C amden health improvement learning collaborative. Journal of H ealthcare Management/ American College of Healthcare Executives, 45 (3), 189 205 Retrieved from https://www.ncbi.nlm.nih.gov/pubmed/11066967 Wei l, A., Wiener, J. M., & Holahan, J. (1998). 'Assessing the new federalism' and state health policy. Health Affairs (Project Hope), 17 (6), 162 164 doi: 10.1377/hlthaff.17.6.162 Weissert, C. S. (2002). State Medicaid agencies as key actors in Medicaid managed care. Albany, NY: The Nelson Rockefeller Institute of Government, State of New York. Retrieved from http://www.rockinst.org/pdf/health_care/2002 state_medicaid_agencies_as_key_actors_in_medicaid_managed care.pdf Whitenhill, K., & Shugarman, L. R. (2011). What is a Medicaid waiver? Long Beach, CA: T he SCAN Foundation. Retrieved from http://www.thescanfoundation.org/sites/default/files/ltc_fundamental_8_0.pdf

PAGE 166

150 Williams, B., & Matheny, A. (1995). Democracy, di alogue, and environmental disputes: The contested languages of social regulation New Haven, CT: Yale University Press. Wilmot, F. (2006). A look at race and ethnicity in Colorado (1860 2005): Census definitions and data. Colorado Libraries, 32 (4), 10 18. Retrieved from http://www.coallnet.org/wp content/uploads/2015/04/WilmotArticle.pdf Wood, D. J., & Gray, B. (1991). Toward a comprehensive theory of collaboration. The J ournal of Applied Behavioral Science, 27 (2), 139 162. doi: 10.1177/0021886391272001 Zhang, N. J., Wan, T. T. H., Rossiter, L. F., Murawski, M. M., & Patel, U. B. (2008). Evaluation of chronic disease management on outcomes and cost of care for medicaid ben eficiaries. Health Policy, 86 (2), 345 354. doi: 10.1016/j.healthpol.2007.11.011

PAGE 167

151 APPENDIX A TABLE OF COLORADO ACC RCCO SERVICE REGIONS RCCO REGIONS COUNTIES SERVED 1 Rocky Mountain Health Plans https://www.rmhpcommunity.org/ Archuleta Jackson Ouray Delta La Plata Pitkin Dolores Larimer Rio Blanco Eagle Mesa Routt Garfield Moffat San Juan Grand Montezuma San Miguel Gunnison Montrose Summit Hinsdale 2 Colorado Access 1 http://www.coaccess.com/regional care collaborative organization Cheyenne Morgan Washington Kit Carson Phillips Weld Lincoln Sedgwick Yuma Logan 3 Colorado Access http://www.coaccess.com/regional care collaborative organization Adams Arapahoe Douglas 4 Integrated Community Health Partners http://www.ichpcolorado.com/ Alamosa Custer Mineral Baca Fremont Otero Bent Huerfano Prowers Chaffee Kiowa Pueblo Conejos Lake Rio Grande Costilla Las Animas Saguache Crowley 5 Colorado Access http://www.coaccess.com/regional care collaborative organization Denver 6 Colorado Community Health Alliance http://www.cchacares.com/ Boulder Jefferson Broomfield Clear Creek Gilpin 7 Community Care Central Colorado http://www.mycommunitycare.org/ El Paso Elbert Park Teller 1 The same entity operates RCCOs 2, 3, and 5.

PAGE 168

152 APPENDIX B MAP OF COLORADO ACC RCCO SERVICE REGIONS

PAGE 169

153 APPENDIX C ORGANIZATION OF ACC RCCOS RCCO REGIONS ORGANIZATION DESCRIPTION 1 Rocky Mountain Health Plans https://www.rmhpcommunity.org/ Independent, not for profit health benefits provider serving the health care needs of Coloradans for more than 30 years 2 Colorado Access http://www.coaccess.com/regional care collaborative organization Nonprofit health plan providing access to behavioral and physical health services for Coloradans since 1994 and sponsored by Children's Hospital Colorado, Colorado Community Managed Care Ne twork, and University of Colorado Hospital/University Physicians, Inc. 3 Colorado Access http://www.coaccess.com/regional care collaborative organization Same as Region 2 above 4 Integrated Community Health Partners http://www.ichpcolorado.com/ A partnership between ValueOptions, Inc. 2 and five Colorado based service delivery organizations including Colorado Community Managed Care Network (CCMCN) and its primary care provider organizations, and four community mental health Centers serving ICHP counties 5 Colorado Access http://www.coaccess.com/regional care collaborative organization Same as Region 2 above 6 Colorado Community Health Alliance http://www.cchacares.com/ A partnership between Centura Health, Physician Health Partners (PHP), and Primary Physician Partners (PPP) 7 Community Care Central Colorado http://www.mycommunitycare.org/ A group of local health care providers united to provide better medical services and led by Community Health Partnership, which was established in 1992 to foster a coordinated approach for addressing health care issues 2 Beacon Health Options acquired ValueOptions, Inc. as of December 31, 2014. More information is available on the following websites: https://www.bloomberg.com/res earch/stocks/private/snapshot.asp?privcapId=36136 and https://www.beaconhealthoptions.com/providers/beacon/network/vo colorado part nerships for colorado medicaid/

PAGE 170

154 APPENDIX D RCCO RELATIONSHIPS AND COMMUNITY INVOLVEMENT RCCO ADVISORY BOARD COMPOSITION HOSPITAL RELATIONSHIPS ANCILLARY PROVIDER RELATIONSHIPS COMMUNITY ENGAGEMENT AND OTHER ADVISORY ACTIVITIES 1 Participating providers (e.g., Community Mental Health Center) Community based care coordinators Hospital representatives Convening organizational leadership and RCCO leadership St akeholders including individual s family members, and advocates Local hospi tal participation on community oversight c ommittees Financial or facility support of community based Care Coordination Team Hospital based case managers assisting in arranging care for current members Local hospital representatives Behavioral health providers Home health agencies Partnering with Quality Health Net Hospital, and the Mesa Coun ty independent practice association Participating in the Comprehensive Primary Care Initiative Conducting quarterly open community a dvisory forums Participating in the Adults without Depe ndent Children advisory c ommittee meetings Participatin g in statewide RCCO care c oordination and leadership meetings Reaching out and being responsive to community providers, advocates and other key community players

PAGE 171

155 RCCO ADVISORY BOARD COMPOSITION HOSPITAL RELATIONSHIPS ANCILLARY PROVIDER RELATIONSHIPS COMMUNITY ENGAGEMENT AND OTHER ADVISORY ACTIVITIES 2, 3, 5 Participating providers (e.g., Colorado Behavioral Health Council, Aurora Health Access, Metro Community Provider Network) University and Hospital representatives Health care legal advocates Convening organizational leadership Stakeholders including individuals family members, caregivers, and advocates Hospital Colorado University of Colorado Health Systems, including Poudre Valley Hospital, University of Colorado Hospital, and Memorial Health Systems Represen tation on Advisory Board and RCCO oversight Con tractual relationships with all systems statewide Development of real time emergency room and hospitalization data to support care management Single entry point agency in each region A RCCO seat on Long Term Community Advisory Board Area Ag encies on Aging in each region Denver Regional Council of Governments, Community Care Transition Programs Market ing outreach projects such as the Salute to Seniors event Participating with community based o rganizations focused on health Creating a stakeh older group for providers for Medicaid, Child Health Plan Plus and the uninsured Partnering with the Board of Alliance to develop a community based care management program Working with Aurora Health Access Exchanging ACC member lists with Community Ment al Health Center and Behavioral Health Organization providers Maintaining a relationship with the F ederally Q ualified H ealth C enter s 4 Representation from local single entry p oint agencies Stakeholders including individuals, family members, caregivers, and advocates Hospitals in each po pulation center, including the emergency, admissions, discharge, and case m anagement departments Shared emergency room and primary care provider data Area Agency on Aging Pueblo Advisory Council on Aging Adult Resources for Care and Help Home Instead Senior Care Participating in the Triple Aim Steering Committee Hosting meetings for court appointed special a dvocates Participating on county committees the Health Department, Senior Service Network,

PAGE 172

156 RCCO ADVISORY BOARD COMPOSITION HOSPITAL RELATIONSHIPS ANCILLARY PROVIDER RELATIONSHIPS COMMUNITY ENGAGEMENT AND OTHER ADVISORY ACTIVITIES between the Community Mental Health Center and PCM Ps on shared client rosters using File Connect Data management process and care coordination support at the provider level, including hospitals, and program reporting at the RCCO level Senior Service Network Golden Gate Transportation Bluesky Enterprises Pueblo Step up InnovAge Community Mental Health Centers Community Health Centers Centers for Improving Value in Health Care, Colorado Health Found ation, state level advisory committees Supporting the development of Community Care Coordination and Case Management Teams Developing a community r esource database 6 Health care professionals, including behavioral health Community resource leaders Non profit health advocacy group representative Centura Health representatives Largest hospital network in Colorado Pledged use of specialists with slots op en specifically for RCCO members Emergency r oom and inpatient admission data feeds to R CCO three times per day for all hospitals Arrangements with Exempla and Health One hospitals Arrangement between Salud Family Health Center and Longmont United Hospital for Salud care managers to do hospital rounds to support care transition efforts Single entry point agencies Local continuing care residents Skilled nursing f acilities Multiple non profits serving seniors (e.g., the Senior Resource Center, the Action Center, Dominican Sisters) Multiple home health agencies Hospice care providers Parti cipating in the Adults withou t Dependent Children initiative and the Colorado Medical Home i nitiative Working philanthropically with The Action Center Participating in health care organizations throughout the metro area, such as Colorado Business Group o n Health, Colorado Health Care Strategy and Management, Colorado Health Foundation, and Colorado Health Institute Leading a community resource outreach campaign

PAGE 173

157 RCCO ADVISORY BOARD COMPOSITION HOSPITAL RELATIONSHIPS ANCILLARY PROVIDER RELATIONSHIPS COMMUNITY ENGAGEMENT AND OTHER ADVISORY ACTIVITIES 7 Health care professionals, including behavioral health Hospital representatives Health care partners (e.g., Colorado Springs Health Partners, Peak Vista, and Rocky Mountain Health Care Services) University of Colorado representatives Local hospital systems Partnering with transition of care program and emergency room diversion and referral prog ram Representation on RCCO management committees Medical service providers Single entry p oint agencies Peak Vista Colorado Springs Health Providers All local safety net clinics El Paso County Health Department Aspen Pointe Local hospital systems Pikes Peak Hospice & Palliative Care Colorado Regional Health Information Organization Participating in the Healthy Community Collaborative initiative of the local health department

PAGE 174

158 APPENDIX E HIGH COST IMAGING SERVICES 3 ENHANCED AMBULATORY PAYMENT GROUPS DESCRIPTION CPT4 CODES 00300 Cat Scan Abdomen 74150, 74170, 74160 00299 Cat Scan Brain 70460, 70470, 70450, 0042T 00301 Cat Scan Other 70488, 73201, 70481, 73200, 76070, 72193, 70480, 0150T, 73702, 76071, 0144T, 70492, 70490, 73202, 71260, 70498, 0151T, 75635, 0152T, 73701, 70487, 71270, 70491, 76497, 71250, 0145T, 72194, 72192, 73700, 70482, S8092, 70486 00298 Cat Scan Back 72125, 72128, 72131, 72130, 72126, 72127, 72133, 72129, 72132 00473 Cat Scan Guidance 76370, 77013, 77011, 77014, 76082, 76355, 77012, 76362, 76360, 76013, 76083, 72292 00292 MRI Abdomen 74181, S8037, 74185, 74182, 74183 00294 MRI Back 72142, 72 158, 72159, 72147, 72157, 72141, 72156, 72148, 72146, 72149 00297 MRI Brain 70552, 70553, 70559, 70551, 70558, 70557 00295 MRI Chest 71552, 71551, C8909, 71550 00475 MRI Guidance 77022, 76393, 76394, 77021 00293 MRI Joints 73221, 73722, 73223, 73723, 73222, 70336, 73721 00296 MRI Other 75555, 73718, 77084, 76093, 73719, 75557, C8905, 75552, 73219, 75562, 70543, 76498, 75561, C8908, 76094, 73720, C8903, 75560, 73218, C8906, 75554, 75553, 72195, 76390, 73220, 75558, C8904, 75565, 70542, 75564, 77059, C8901, 75556, S8035, 72197, 77058, C8900 3112F, 76400, 70540, C8907, 75563, C8902, 75559, 72196, 3111F 3 Treo Solutions provided the list of applicable CPT codes for high cost imaging services that are included for purposes of analyzing ACC key performance indicators (2012, p. 7).

PAGE 175

159 APPENDIX F BEHAVIORAL HEALTH DIAGNOSIS CODES 4 DIAGNOSIS CODE DESCRIPTION 300 Neurotic disorders 300.0 Anxiety states 300.00 Anxiety state, unspecified 300.01 Panic disorder 300.02 Generalized anxiety disorder 300.09 Other 300.1 Hysteria 300.10 Hysteria, unspecified 300.11 Conversion disorder 300.12 Psychogenic amnesia 300.13 Psychogenic fugue 300.14 Multiple personality 300.15 Dissociative disorder or reaction, unspecified 300.16 Factitious illness with psychological symptoms 300.19 Other and unspecified factitious illness 300.2 Phobic disorders 300.20 Phobia, unspecified 300.21 Agoraphobia with panic attacks 300.22 Agoraphobia without mention of panic attacks 300.23 Social phobia 300.29 Other isolated or simple phobias 300.3 Obsessive compulsive disorders 300.4 Neurotic depression 300.5 Neurasthenia 300.6 Depersonalization syndrome 300.7 Hypochondriasis 300.8 Other neurotic disorders 300.81 Somatization disorder 300.82 Undifferentiated Somatoform Disorder 300.89 Other 300.9 Unspecified neurotic disorder 309 Adjustment reaction 4 The list includes the behavioral health diag noses of interest related to anxiety and depression that are included in this study (HCPF & Colorado Department of Human Services, 2014).

PAGE 176

160 DIAGNOSIS CODE DESCRIPTION 309.0 Brief depressive reaction 309.1 Prolonged depressive reaction 309.2 With predominant disturbance of other emotions 309.21 Separation anxiety disorder 309.22 Emancipation disorder of adolescence and early adult life 309.23 Specific academic or work inhibition 309.24 Adjustment reaction with anxious mood 309.28 Adjustment reaction with mixed emotional features 309.29 Other 309.3 With predominant disturbance of conduct 309.4 With mixed disturbance of emotions and conduct 309.8 Other specified adjustment reactions 309.81 Prolonged posttraumatic stress disorder 309.82 Adjustment reaction with physical symptoms 309.83 Adjustment reaction with withdrawal 309.89 Other 309.9 Unspecified adjustment reaction 311 Depressive disorder, not elsewhere classified

PAGE 177

161 APPENDIX G CHARACTERISTICS OF CONTROL AND ACC GROUPS

PAGE 178

162

PAGE 179

163 APPENDIX H ER VISIT COMPARISONS AMONG RCCOS Control Group, T 1 ER Visits RCCO Total 1 2 3 4 6 7 No Number a 2 608a 541b 380 a, b 3 073b 132 a, b 188b 6,922 Expected Number 2,547 561 380 3,099 136 198 6,922 % within RCCO 95 90 93 92 90 89 93 Yes Number a 124a 61b 28 a b 251b 14 a, b 24b 502 Expected Number 185 41 28 225 10 14 502 % within RCCO 5 10 7 8 10 11 7 Number 2,732 602 408 3,324 146 212 7,424 Total Expected Number 2,732 602 408 3,324 146 212 7,424 % within RCCO 100 100 100 100 100 100 100 a No statistically significant difference exists between groups with the same letters. Significant differences exist between RCCO 1 and RCCOs 2, 4, and 7. ACC Group, T 1 ER Visits RCCO Total 1 2 3 4 5 6 7 No Number a 3 979a 5 560b 20 968b 9 597c 11 282b 8 335 a, b 6 023b 65,744 Expected Number 4,066 5,567 20,992 9,392 11,318 8,378 6,031 65,744 % within RCCO 90 92 92 94 91 91 92 92 Yes Number a 453a 509b 1 916b 641c 1 056b 798 a, b 551b 5 924 Expected Number 366 502 1,892 846 1,020 755 543 5,924 % within RCCO 10 8 8 6 9 9 8 8 Total Number 4,432 6,069 22,884 10,238 12,338 9,133 6,574 71,668 Expected Number 4,432 6,069 22,884 10,238 12,338 9,133 6,574 71,668 % within RCCO 100 100 100 100 100 100 100 100 a No statistically significant difference exists between groups with the same letters. Significant differences exist between RCCO 1 and RCCOs 2, 3, 4, 5, and 7 and between RCCO 4 and RCCOs 2, 3, 5, 6, and 7.

PAGE 180

164 Control Group, T 2 ER Visits RCCO Total 1 2 3 4 6 7 No Number a 2,945a 470 a,b, c 399 a, c 3,194 b, c 157 a,b, c 113b 7,278 Expected Number 2,904 479 392 3,228 155 120 7,278 % within RCCO 95 92 95 93 95 88 94 Yes Number a 149a 40 a,b, c 19 a, c 245 b, c 8 a,b, c 15b 476 Expected Number 190 31 26 211 10 8 476 % within RCCO 5 8 5 7 5 12 6 Total Number 3,094 510 418 3,439 165 128 7,754 Expecte d Number 3,094 510 418 3,439 165 128 7,754 % within RCCO 100 100 100 100 100 100 100 a No statistically significant difference exists between groups with the same letters. Significant differences exist between RCCO 1 and RCCOs 4 and 7 and between RCCOs 3 and 7. ACC Group, T 2 ER Visits RCCO Total 1 2 3 4 5 6 7 No Number a 4,534a 6,197a,b 23,470b 10,279c 13,006b 9,071a 6,593c 73,150 Expected Number 4,641 6,254 23,550 10,027 13,011 9,202 6,467 73,150 % within RCCO 90 92 92 95 92 91 94 92 Yes Number a 489a 572a, b 2 021b 574c 1 077b 889a 407c 6,029 Expected Number 382 515 1,941 826 1,072 758 533 6,029 % within RCCO 10 8 8 5 8 9 6 8 Total Number 5,023 6,769 25,491 10,853 14,083 9,960 7,000 79,179 Expected Number 5,023 6,769 25,491 10,853 14,083 9,960 7,000 79,179 % within RCCO 100 100 100 100 100 100 100 100 a No statistically significant difference exists between groups with the same letters. Significant differences exist between RCCO 1 and RCCOs 3, 4, 5, and 7; between RCCOs 3 and 7; between RCCO 4 and RCCOs 2, 3, 5, and 6; between RCCO 5 and RCCOs 6 and 7; and between RCCOs 6 and 7.

PAGE 181

165 Control Group, T 3 ER Visits RCCO Total 1 2 3 4 6 7 No Number a 3186a 444b 384 a, c 3484 b, c 204 a,b, c 115 b, c 7,817 Expected Number 3,101 463 377 3,552 204 120 7,817 % within RCCO 97 90 96 92 94 91 94 Yes Number a 109a 48b 17 a, c 291 b, c 13 a,b, c 12 b, c 490 Expected Number 194 29 24 223 13 7 490 % within RCCO 3 10 4 8 6 9 6 Total Number 3,295 492 401 3,775 217 127 8,307 E xpected Number 3,295 492 401 3,775 217 127 8,307 % within RCCO 100 100 100 100 100 100 100 a No statistically significant difference exists between groups with the same letters. S ignificant differences exist between RCCO 1 and RCCOs 2, 4, and 7 and between RCCOs 2 and 3. ACC Group, T 3 ER Visits RCCO Total 1 2 3 4 5 6 7 No Number a 5,026a 5,947 a, b 24,872b 10,101c 12,483b 8,774 a, b 6,729d 73,932 Expected Number 5,138 6,009 24,913 9,838 12,522 8,858 6,654 73,932 % within RCCO 91 92 92 95 92 92 94 93 Yes Number a 521a 541a, b 2025b 521c 1036b 789 a, b 455d 5 888 Expected Number 409 479 1,984 784 997 705 530 5,888 % within RCCO 9 8 8 5 8 8 6 7 Total Number 5,547 6,488 26,897 10,622 13,519 9,563 7,184 79,820 Expected Number 5,547 6,488 26,897 10,622 13,519 9,563 7,184 79,820 % within RCCO 100 100 100 100 100 100 100 100 a No statistically significant difference exists between groups with the same letters. S ignificant differences exist between RCCO 1 and RCCOs 3, 4, 5, and 7; between RCCOs 3 and 7; between RCCO 4 and RCCOs 2, 3, 5, 6, and 7; between RCCOs 5 and 7; and between RCCOs 6 and 7.

PAGE 182

166 APPENDIX I HIGH COST IMAGING COMPARISONS AMONG RCCOS Control Group, T 1 High cost RCCO Total Imaging 1 2 3 4 6 7 No Number a 7,201a 1,103b 953a, b 7,456b 461a, b 629a, b 17,803 Expected Number 7,130 1,121 953 7,511 458 630 17,803 % within RCCO 99 96 98 97 99 98 98 Yes Number a 84a 42b 21a, b 219b 7a, b 15a, b 388 Expected Number 155 24 21 164 10 14 388 % within RCCO 1 4 2 3 1 2 2 Total Number 7,285 1,145 974 7,675 468 644 18,191 Expected Number 7,285 1,145 974 7,675 468 644 18,191 % within RCCO 100 100 100 100 100 100 100 a No statistically significant difference exists between groups with the same letters. Significant differences exist between RCCO 1 and RCCOs 2 and 4. ACC Group, T 1 High cost Imaging RCCO Total 1 2 3 4 5 6 7 No Number a 10,203a 15,758 b c 56,323d 26,101 b,c, d 31,233 b,c, d 21,696 c, d 16,580b 177,894 Expected Number 10,277 15,714 56,413 26,066 31,189 21,720 16,514 177,894 % within RCCO 97 98 98 98 98 98 98 98 Yes Number a 309a 315 b, c 1,377d 560 b,c, d 668 b,c, d 520 c, d 311b 4,060 Expected Number 235 359 1,287 595 712 496 377 4,060 % within RCCO 3 2 2 2 2 2 2 2 Total Number 10,512 16,073 57,700 26,661 31,901 22,216 16,891 181,954 Expected Number 10,512 16,073 57,700 26,661 31,901 22,216 16,891 181,954 % within RCCO 100 100 100 100 100 100 100 100 a No statistically significant difference exists between groups with the same letters. No statistically significant difference exists between groups with the same letters. Significant differences exist between RCCO 1 and RCCOs 2, 3, 4, 5, 6, and 7 and between RCCO 3 and RCCOs 2 and 7.

PAGE 183

167 Control Group, T 2 High cost Imaging RCCO Total 1 2 3 4 6 7 No Number a 8 261a 870b 762b 7 105b 317b 358 a, b 17 673 Expected Number 8 ,195 871 765 7163 320 355 17,673 % within RCCO 99 98 98 97 97 99 98 Yes Number a 53a 14b 15b 162b 8b 3 a, b 255 Expected Number 118 12 11 103 4 5 255 % within RCCO 1 2 2 3 3 1 2 Total Number 8 314 884 777 7 267 325 361 17 928 Expected Number 8,314 884 777 7,267 325 361 17,928 % within RCCO 100 100 100 100 100 100 100 a No statistically significant difference exists between groups with the same letters. Significant differences exist between RCCO 1 and RCCOs 2, 3, 4, and 6. ACC Group, T 2 High cost Imaging RCCO Total 1 2 3 4 5 6 7 No Number a 11 383a 17 785b 62 455 a, b 28 651 a, b 35 431 a, b 2 4436b 16 470 a, b 196 611 Expected Number 11,421 17,760 62,486 28,655 35,415 24,406 16,464 196,611 % within RCCO 98 99 99 99 99 99 99 99 Yes Number a 197a 222b 900 a, b 403 a, b 477 a, b 310b 223 a, b 2 732 Expected Number 158 246 868 398 492 339 228 2,732 % within RCCO 2 1 1 1 1 1 1 1 Total Number 11 580 18 007 63 355 29 054 35 908 24 746 16 693 199 343 Expected Number 11,580 18,007 63,355 29,054 35,908 24,746 16,693 199,343 % within RCCO 100 100 100 100 100 100 100 100 a No statistically significant difference exists between groups with the same letters. Significant differences exist between RCCO 1 and RCCOs 2 and 6.

PAGE 184

168 Control Group, T 3 High cost Imaging RCCO Total 1 2 3 4 6 7 No Number a 7,900a 839b 827b 8,092b 500b 279b 18 437 Expected Number 7,846 845 840 8,120 502 280 18,437 % within RCCO 99 98 97 99 98 98 99 Yes Number a 32a 16b 23b 117b 8b 5b 201 Expected Number 85 9 9 88 5 3 201 % within RCCO 1 2 3 1 2 2 1 Total Number 7 932 855 850 8 209 508 284 18 638 Expected Number 7,932 855 850 8,209 508 284 18,638 % within RCCO 100 100 100 100 100 100 100 a No statistically significant difference exists between groups with the same letters. Significant differences exist between RCCO 1 and RCCOs 2, 3, 4, 6, and 7 ACC Group, T 3 High cost Imaging RCCO Total 1 2 3 4 5 6 7 No Number a 12 766a 16 563b 65 589b 29 974 b,c 34 708b 23248b 16 675 a, c 199,523 Expected Number 12,866 16,536 65,556 29,976 34,664 23,196 16,730 199,523 % within RCCO 98 99 99 99 99 99 98 99 Yes Number 264a 184b 803b 384 b, c 398b 244b 268 a, c 2,545 Expected Number 164 211 836 382 442 296 213 2,545 % within RCCO 2 1 1 1 1 1 2 1 Total Number 13,030 16,747 66,392 30,358 35,106 23,492 16,943 202,068 Expected Number 13,030 16,747 66,392 30,358 35,106 23,492 16,943 202,068 % within RCCO 100 100 100 100 100 100 100 100 a No statistically significant difference exists between groups with the same letters. Significant differences exist between RCCO 1 and RCCOs 2, 3, 4, 5, and 6 and between R CCO 7 and RCCOs 2, 3, 5, and 6.

PAGE 185

169 APPENDIX J MEDICAL SERVICE COMPARISONS AMONG RCCOS Study Period Group RCCO Pairs Test Statistic Std. Error Std. Test Statistic Sig. Adj. Sig. a a When p < .05, significance exists in the difference between pairwise comparisons for the indicated RCCOs. Only significant pairwise comparisons are included in this table.

PAGE 186

170 APPEND IX K PRESCRIPTION DRUG COMPARISONS AMONG RCCOS Study Period Group RCCO Pairs Test Statistic Std. Error Std. Test Statistic Sig. Adj. Sig. a 3 1 620 136 180 320 3.439 001 009 3 4 1 342 660 177 091 7 582 .000 000 3 2 1 969 397 219 730 8 963 000 000 7 4 884 920 232 027 3 814 .000 002 7 2 1 511 657 266 001 5 683 000 000 6 2 1 463 359 371 813 3 936 .000 001 1 4 722 524 92 352 7 824 000 000 1 2 1 349 261 159 527 8.458 .000 000 4 2 626 736 155 869 4 021 000 001 5 3 1 015.200 325.039 3 123 .002 .038 5 1 2 284 904 453.400 5 039 000 000 5 2 2 934.173 443.970 6 609 .000 .000 5 7 3 949 833 426.216 9 267 000 000 5 4 7 316.365 375.141 19 503 .000 .000 6 1 1 533.604 472.214 3 248 .001 .024 6 2 2 182 873 463 167 4 713 000 000 6 7 3 198 533 446 17 8 7 169 .000 000 6 4 6 565.065 397.674 16 509 000 .000 3 1 1 269 704 412.207 3 080 .002 04 3 3 2 1 918.973 401.812 4 776 000 .000 3 7 2 934 633 382 105 7 680 .000 000 3 4 6 301.165 324.151 19.439 000 .000 1 7 1 664.929 495.908 3 357 001 .017 1 4 5 031.461 452 764 11 113 .000 000 2 4 4 382 192 443 321 9 885 .000 000 7 4 3 366.532 425.540 7 911 000 .000 3 1 775 123 204 961 3 782 000 002 3 4 791.453 201 606 3 926 .000 001 3 2 1 744 611 249.450 6 994 000 000 7 2 1 142 163 302 138 3 780 000 002 1 2 969.488 176 013 5 508 .000 000 4 2 953 158 172 094 5 539 000 000 5 3 1 721.782 342.279 5 030 .000 000 5 7 3 479.424 450.409 7 725 000 .000 5 1 3 948.841 472.985 8 349 .000 000 5 2 4 403 132 469 333 9 382 000 .000 5 4 9 675.565 400.104 24 183 .000 .000 6 7 2 341.167 466.038 5 024 .000 000 6 1 2 810 584 487 891 5 761 000 000 6 2 3 264 875 484 352 6 741 000 .000 6 4 8 537.308 417.620 20.443 .000 000 3 7 1 757 642 405 542 4 334 000 .000 3 1 2 227.059 430.478 5 173 .000 000

PAGE 187

171 Study Period Group RCCO Pairs Test Statistic Std. Error Std. Test Statistic Sig. Adj. Sig. a 3 2 2 681 350 426.462 6 287 000 .000 3 4 7 953.783 348.825 22 802 .000 000 7 4 6 196 141 455.403 13 606 000 000 1 4 5 726 724 477.743 11 987 000 .000 2 4 5 272.433 474.128 11 120 .000 .000 6 3 1 107 874 338.433 3 274 001 016 6 4 1 ,7 49 786 278 164 6 290 .000 000 6 7 1 816 728 389.355 4 666 000 000 6 2 2,330 .461 316 .7 03 7 359 .000 000 6 1 2 510.428 281 235 8 926 000 000 3 4 641 912 211 569 3.034 .002 036 3 2 1 222 587 260 167 4 699 .000 000 3 1 1,402 554 215 592 6 506 000 000 4 2 580 675 174 .7 14 3 .3 24 001 .013 4 1 760.642 96 530 7 880 .000 000 5 3 1 430.845 347.497 4 118 .000 .001 5 2 2 751.489 468 160 5 877 .000 000 5 7 3 418.368 453.784 7 533 .000 .000 5 1 4 795.415 472.322 10 153 .000 000 5 4 8 842.028 399.719 22 121 .000 .000 6 2 2 127.516 484.771 4 389 .000 .000 6 7 2 794 395 470 903 5 934 .000 000 6 1 4 171 .443 488 792 8 534 .000 000 6 4 8 218.056 419.052 19 611 .000 .000 3 2 1 320 643 423 603 3 118 .002 0 38 3 7 1 987.522 407.659 4 875 .000 .000 3 1 3 364 570 428 199 7 85 7 .000 000 3 4 7 411.183 346.470 21 391 .000 .000 2 1 2 043.926 530.824 3 850 .000 .00 2 2 4 6 090 539 467 398 13 031 .000 000 7 4 5 423 660 452 998 11 973 .000 000 1 4 4 046.613 471.567 8 581 .000 .000 a When p < .05, significance exists in the difference between pairwise comparisons for the indicated RCCOs. Only significant pairwise comparisons are included in this table.