Arkansas Health Care Independence Program ('Private Option

ACHI is a nonpartisan, independent, health policy center that serves as a catalyst to improve the health of Arkansans.

1401 W Capitol Avenue, Suite 300 ● Little Rock, Arkansas 72201 ● (501) 526-2244 ● www.achi.net

Arkansas Health Care Independence Program

(‘Private Option’)

Section 1115 Demonstration Waiver

Final Report

June 30, 2018

Arkansas Health Care Independence Program (‘Private Option’) Final Report

About the Evaluators

The Arkansas Center for Health Improvement (ACHI) is a nonpartisan, independent health policy center whose

vision is to be a trusted health policy leader committed to innovations that improve the health of Arkansans. This

vision is carried out through ACHI’s mission to be a catalyst for improving the health of Arkansans through

evidence-based research, public issue advocacy, and collaborative program development. In undertaking this

mission, ACHI has worked hard to develop and adhere to an important set of values that guide the Center.

ACHI’s core values are commitment, initiative, trust, and innovation. These are the fundamental principles that

guide the Center’s collective and individual decisions, strategies, and actions to advance the health of Arkansans.

These core values define what ACHI — the organization and its people — stands for throughout time, regardless

of changes in ACHI’s internal structure and leadership, or in response to external factors and environmental

conditions.

ACHI Team

Joseph W. Thompson, MD, MPH, Director

Anthony Goudie, PhD, Director of Research and Evaluation

Nichole Sanders, PhD, Assistant Director of Analytics

Kanna Lewis, PhD, Health Policy Data Architect

Vanessa White, BA, Project Coordinator

Stephen Lein, MS, Senior Data Analyst

Judy Bennett, MS, Senior Research Analyst

Pader Moua, MPH, Policy Analyst

Tim Holder, BA, Technical Editor

John Lyon, BA, Strategic Communications Manager

Vaishali Thombre, MBA, MS, Senior Research Analyst

Jeral Self, MPH, Senior Data Analyst

Kenley Money, MA, MFA, Director of Information Systems Architecture

University of Arkansas for Medical Sciences Evaluators

 Anthony Goudie, PhD

Director of Research and Evaluation, Arkansas Center for Health Improvement

Assistant Professor, Center for Applied Research and Evaluation, Department of Pediatrics, College of Medicine,

University of Arkansas for Medical Sciences

Assistant Professor, Department of Health Policy and Management, Fay W. Boozman College of Public Health,

University of Arkansas for Medical Sciences

Arkansas Health Care Independence Program (‘Private Option’) Final Report

 John “Mick” Tilford, PhD

Professor and Department Chair, Department of Health Policy and Management, Fay W. Boozman College

of Public Health, University of Arkansas for Medical Sciences

Professor, Department of Pharmaceutical Evaluation and Policy, College of Pharmacy, University of

Arkansas for Medical Sciences

 Bradley Martin, PharmD, PhD

Professor, Division Head, Department of Pharmaceutical Evaluation and Policy, College of Pharmacy,

University of Arkansas for Medical Sciences

Comparative Effectiveness Research Program Co-Director, Translational Research Institute, University of

Arkansas for Medical Sciences

 Teresa Hudson, PharmD, PhD

Associate Professor of Psychiatry and Director, Division of Health Services Research, College of Medicine,

University of Arkansas for Medical Sciences

Director, National Rural Evaluation Center, U.S. Department of Veterans Affairs

Associate Director, Center for Mental Healthcare and Outcomes Research, U.S. Department of Veterans Affairs

 Jeffrey Pyne, MD

Professor, Department of Psychiatry and Behavioral Sciences, College of Medicine, University of Arkansas

for Medical Sciences

Associate Professor, Department of Pharmacy Practice, College of Pharmacy, University of Arkansas for

Medical Sciences

Associate Professor, Department of Epidemiology, College of Public Health, University of Arkansas for

Medical Sciences

Staff Physician, Central Arkansas Veterans Healthcare System

Associate Director for Research and Site Leader for Community Engagement, South Central (VISN 16)

Mental Illness Research, Education, and Clinical Center

Research Health Scientist and Core Investigator, Center for Mental Healthcare and Outcomes Research,

U.S. Department of Veterans Affairs

 Chenghui Li, PhD

Associate Professor, Division of Pharmaceutical Evaluation and Policy, College of Pharmacy, University of

Arkansas for Medical Sciences

 Nalin Payakachat, PhD

Associate Professor, Division of Pharmaceutical Evaluation and Policy, College of Pharmacy, University of

Arkansas for Medical Sciences

Arkansas Health Care Independence Program (‘Private Option’) Final Report

UAMS Research Support Team

Gary Moore, MS, Research Analyst

Xiaotong Han, MS, Biostatistician

Niranjan Kathe, MS, Graduate Research Assistant

Divyan Chopra, MS, Graduate Research Assistant

Naleen Raj Bhandari, MS, Graduate Research Assistant

Mahip Acharya, B Pharm, Graduate Research Assistant

Acknowledged Contributors

Heather L. Rouse, MSEd, PhD

Brady Rice, BS

Grayson Shelton, BS

Donald Poe, BS

Anuj Shah, PhD

Siqing Li, MS

Arkansas Health Care Independence Program (‘Private Option’) Final Report

Acknowledgements

Private Option Evaluation National Advisory Committee

The purpose of the National Advisory Committee (NAC) is to act as an external expert advisory group for the

Arkansas Section 1115 demonstration waiver evaluation. Members of the committee were asked to serve in the

capacity for the duration of the three-year evaluation period (2014-2017) and were selected based on their content

expertise and methodological experience. The committee is comprised of a diverse range of policy perspectives and

professional backgrounds.

 Eduardo Sanchez, MD, MPH, Chief Medical Officer (CMO) for Prevention, American Heart Association National

Center

 Darrell J. Gaskin, PhD, Deputy Director, Center for Health Disparities Solutions, Bloomberg School of Public

Health, Johns Hopkins University

 Daniel Polsky, PhD, Executive Director, Leonard Davis Institute of Health Economics, University of Pennsylvania

 Timothy S. Carey, MD, MPH, Sarah Graham Kenan Professor of Medicine, Departments of Medicine and Social

Medicine, School of Medicine, University of North Carolina at Chapel Hill

Correspondence concerning this report should be addressed to the Arkansas Center for Health Improvement

(ACHI).

Contributing ACHI authors include Joseph W. Thompson, MD, MPH, and Anthony Goudie, PhD.

Suggested Citation

Arkansas Center for Health Improvement. Arkansas Health Care Independence Program (‘Private Option’)

Section 1115 Demonstration Waiver Final Report. Little Rock, AR: Arkansas Center for Health Improvement, June

2018.

Arkansas Health Care Independence Program (‘Private Option’) Final Report

Abbreviations

ACHI Arkansas Center for Health Improvement

AHA The Arkansas Hospital Association

AHRQ Agency for Healthcare Research and Quality

AID Arkansas Insurance Department

AV Actuarial Value

BNC Budget Neutrality Cap

CAHPS Consumer Assessment of Healthcare Providers and Systems

CAST Center for Advanced Spatial Technologies

CDC Centers for Disease Control and Prevention

CMS Centers for Medicare & Medicaid Services

CPT Current Procedure Terminology

CSR Cost-Sharing Reduction

CY Calendar Year

DCO Department of County Operations

DHS Arkansas Department of Human Services

EHB Essential Health Benefit

EPSD Early and Periodic Screening and Diagnostic

ER Emergency Room

FFM Federally Facilitated Marketplace

FFS Fee-For-Service

FPL Federal Poverty Level

FQHC Federally Qualified Community Health Center

GME Graduate Medical Education

HbA1c Hemoglobin A1c

HCIP Health Care Independence Program

HEDIS Healthcare Effectiveness Data and Information Set

HIA Health Independence Accounts

LATE Local Average Treatment Effect

LDL-c Lipoprotein

LSM Least Squares Mean

Arkansas Health Care Independence Program (‘Private Option’) Final Report

Marketplace Individual Health Insurance Marketplace

MDES Minimum Detectable (standardized) Effect Size

MEPS Medical Expenditure Panel Survey

MLR Medical Loss Ratio

NAC National Advisory Committee

NCQA National Committee for Quality Assurance

NEMT Non-Emergency Medical Transportation

NQF National Quality Forum

NYU New York University

PCCM Primary Care Case Management

PCP Primary Care Provider

PMPM Per Member Per Month

PMPY Per Member Per Year

PPACA Patient Protection and Affordable Care Act

PY Person-Years

PY1 Program Year 1

PY2 Program Year 2

PY3 Program Year 3

QHP Qualified Health Plan

Questionnaire Healthcare Needs Assessment Questionnaire

RCT Randomized Controlled Trial

RD Regression Discontinuity

RE Relative Efficiency

RUCA Rural-Urban Commuting Area Code

SIPTW Stabilized Inverse Probability of Treatment Weighting

SNAP Supplemental Nutrition Assistance Program

SSI Social Security Income

TEFRA Tax Equity and Fiscal Responsibility Act

UAMS University of Arkansas for Medical Sciences

UPL Upper Payment Limit

Arkansas Health Care Independence Program (‘Private Option’) Final Report

Table of Contents

Executive Summary ...................................................................................................................................................... i

Background ............................................................................................................................................................... i

Summary of Findings Based on Evaluation Hypotheses ........................................................................................... i

Conclusion ............................................................................................................................................................... vi

Introduction ................................................................................................................................................................. 1

Background .............................................................................................................................................................. 1

Arkansas Profile ....................................................................................................................................................... 1

Arkansas Structure of Commercial Premium Assistance ......................................................................................... 2

Arkansas Structure of PPACA Eligibility and Enrollment ......................................................................................... 3

Arkansas HCIP Program Experience ......................................................................................................................... 6

Arkansas HCIP Evaluation Strategy .......................................................................................................................... 8

Research Design and Approach ................................................................................................................................... 9

Goals and Objectives................................................................................................................................................ 9

Programmatic Timeline and Reporting Requirements .......................................................................................... 10

Theoretical Approach............................................................................................................................................. 11

Hypotheses ............................................................................................................................................................ 13

Data Sources and Analytic Comparison Groups .................................................................................................... 14

Methodological Approaches .................................................................................................................................. 16

The First Year Experience ....................................................................................................................................... 18

Year 2 and Year 3 Waiver Impact Findings ................................................................................................................ 24

Introduction to Analyses ........................................................................................................................................ 24

Geographic and Realized Access Differences and Provider Reimbursement Comparison

between Medicaid and QHP Programs .................................................................................................................. 27

Differences in Comparison Groups by Comparison Populations ........................................................................... 31

Special Populations and Topic Studies ................................................................................................................... 43

Program Costs ............................................................................................................................................................ 71

Overview ................................................................................................................................................................ 71

Cost-Effectiveness .................................................................................................................................................. 72

Program Impact Simulation ................................................................................................................................... 73

Appendices .............................................................................................................................................................. A-1

Appendix A – Arkansas Evaluation Hypotheses: Proposed and Original Test Indicators .................................... A-1

Appendix B – Arkansas Evaluation Hypotheses .................................................................................................... B-1

Appendix C – Data Processing of Carrier Data ...................................................................................................... C-1

Appendix D – Study Subjects: Analytical Sample Extraction ............................................................................... D-1

Arkansas Health Care Independence Program (‘Private Option’) Final Report

Appendix E – Statistical Modeling to Assess Effect Differences for Those Assigned to Medicaid and QHPs ....... E-1

Appendix F – Supplemental Payments, Claims Loading Process, and Per Member Per Month Logic ................. F-1

Appendix G – Geographic Information System (GIS) Processing ......................................................................... G-1

Appendix H – Weighted Average Price Calculations............................................................................................ H-1

Appendix I – Pregnancy Analytic Profile ................................................................................................................ I-1

Appendix J – Opioid Use Outcomes ...................................................................................................................... J-1

Appendix K – Consumer Assessment of Healthcare Providers and Systems (CAHPS) Sample Size

Calculation, Sampling Strategy, and Data Preparation ......................................................................................... K-1

Table of Figures

Figure 1. HCIP Premium and Cost-Sharing Reduction Breakdown ............................................................................... 3

Figure 2. Enrollment Pathways and Plan Assignment Process ..................................................................................... 4

Figure 3. HCIP Monthly Enrollment, January 2014–December 2016 ........................................................................... 6

Figure 4. Enrollment Age Demographics by Category .................................................................................................. 7

Figure 5. Premium and Cost-Sharing Reduction Breakdown, January 2014–December 2016 .................................... 8

Figure 6. Arkansas Demonstration Waiver Evaluation Logic Model............................................................................. 9

Figure 7. Arkansas Health Care Independence Program Period: System Evaluation ................................................. 10

Figure 8. Proportion of Medicaid and QHP Enrollees with a First Outpatient Care Visit, by Day .............................. 31

Figure 9. Health Care Independence Program Enrollment by Program and Month .................................................. 52

Figure 10. Charlson Comorbidity Index by Health Independence Account Non-Participants

(Made No Payment) vs. Participants (Made Payment) .............................................................................................. 56

Figure 11. Access Improvement Simulation Model Studying Scenario-Based Price Increases

to Medicaid Providers to Reach Commercial PMPM Payments ................................................................................. 75

Table of Tables

Table 1. Comparison Group Description and Analytical Data Populations ................................................................ 15

Table 2. Differences in General Population Geographic Access to Health Care between Medicaid and QHP

Enrollees ..................................................................................................................................................................... 28

Table 3. Differences in Higher Needs Population Geographic Access to Health Care between Medicaid and

QHP Enrollees ............................................................................................................................................................. 29

Table 4. Medicaid and Commercial Payer Price Differences for Outpatient Procedures by Provider Type .............. 30

Table 5. Differences in Perceived Access to Health Care between Medicaid and QHP Enrollees (Propensity

Score Matched Comparison) ...................................................................................................................................... 32

Table 6. Differences in Primary Preventive Health Care between Medicaid and QHP Enrollees (Propensity

Score Matched Comparison) ...................................................................................................................................... 32

Table 7. Differences in Secondary and Tertiary Preventive Health Care between Medicaid and QHP Enrollees

(Propensity Score Matched Comparison) ................................................................................................................... 33

Table 8. Differences in the Percentage of Medicaid and QHP Enrollees Receiving Recommended Disease

Management (Propensity Score Matched Comparison) ............................................................................................ 34

Table 9. Differences in the Use of Healthcare Services between Medicaid and QHP Enrollees (Propensity

Score Matched Comparison) ...................................................................................................................................... 35

Arkansas Health Care Independence Program (‘Private Option’) Final Report

Table 10. Differences in Rates of Preventable Hospitalizations and Readmissions between Medicaid and

QHP Enrollees (Propensity Score Matched Comparison) ........................................................................................... 36

Table 11. Differences in Utilization of Emergency Room Services between Medicaid and QHP Enrollees

(Propensity Score Matched Comparison) ................................................................................................................... 36

Table 12. Differences in Perceived Access to Health Care between Medicaid and QHP Enrollees (Regression

Discontinuity Comparison) ......................................................................................................................................... 37

Table 13. Differences in Primary, Secondary, and Tertiary Preventive Health Care between Medicaid and

QHP Enrollees (Regression Discontinuity Comparison ............................................................................................... 38

Table 14. Differences in Secondary and Tertiary Preventive Health Care between Medicaid and QHP Enrollees

(Regression Discontinuity Comparison) ...................................................................................................................... 39

Table 15. Differences in the Percentage of Medicaid and QHP Enrollees Receiving Recommended Disease

Management (Regression Discontinuity Comparison) ............................................................................................... 40

Table 16. Differences in the Use of Healthcare Services between Medicaid and QHP Enrollees (Regression

Discontinuity Comparison) ......................................................................................................................................... 41

Table 17. Differences in Rates of Preventable Hospitalizations and Readmissions between Medicaid and

QHP Enrollees (Regression Discontinuity Comparison) .............................................................................................. 42

Table 18. Differences in Utilization of Emergency Room Services between Medicaid and QHP Enrollees

(Regression Discontinuity Comparison) ...................................................................................................................... 43

Table 19. Maternal Pregnancy and Birth Outcomes, with Description ...................................................................... 44

Table 20. Maternal Pregnancy Differences between Medicaid and QHP Enrollees .................................................. 46

Table 21. Pregnancy Birth Outcomes between Medicaid and QHP Enrollees ........................................................... 46

Table 22. Differences in Non-Emergency Transportation between General Population Medicaid and QHP

Enrollees ..................................................................................................................................................................... 48

Table 23. Differences in Non-Emergency Transportation between Medicaid and QHP Enrollees (Propensity

Score Matched Comparison) ...................................................................................................................................... 48

Table 24. Differences in Non-Emergency Transportation between Higher Needs Population Medicaid and

QHP Enrollees ............................................................................................................................................................. 48

Table 25. Differences in Non-Emergency Transportation between Medicaid and QHP Enrollees (Regression

Discontinuity Comparison) ......................................................................................................................................... 49

Table 26. Health Insurance Carrier at 19 Years of Age for Medicaid Screened and Diagnosed Sickle Cell

Disease, Cystic Fibrosis, or Hemophilia (n=151) ......................................................................................................... 50

Table 27. Medicaid Enrollment History of QHP and 06 Medicaid Individuals Enrolled in June 2015 ........................ 53

Table 28. The Impact of Redetermination on QHP and 06 Medicaid Enrollees (Enrolled in July 2015) .................... 53

Table 29. HCIP Enrollment History of Individuals Enrolled in HCIP During 2016........................................................ 54

Table 30. Health Independence Account Eligible Population Demographic Profile................................................... 56

Table 31. Health Independence Accounts Cost-Sharing Expenditures....................................................................... 56

Table 32. Characteristics of Unmatched and Propensity Score Matched General Population .................................. 59

Table 33. Opioid Use Measures for Propensity Score Matched General Population Medicaid and QHP by

Assessment Period ...................................................................................................................................................... 60

Table 34. Comparison of Opioid Utilization for General Population Medicaid and QHP Enrollees ........................... 60

Table 35. Comparison of Number of Opioid Prescriptions by General Population Medicaid and QHP

(Reference) Enrollees .................................................................................................................................................. 61

Table 36. Characteristics of Higher Needs Population ............................................................................................... 62

Table 37. Opioid Use Measures for Propensity Score Matched Higher Needs Population Medicaid and QHP by

Assessment Period ...................................................................................................................................................... 62

Table 38. Comparison of Opioid Utilization for Medicaid and QHP Higher Needs Population Enrollees .................. 63

Arkansas Health Care Independence Program (‘Private Option’) Final Report

Table 39. Change in Estimated Rate of Non-Emergent Emergency Room Visits per 100 Person-Years .................... 64

Table 40. Unadjusted Analysis of Urgent Care Facility Utilization and Cost ............................................................... 65

Table 41. Medicaid and QHP Enrollee General Population Propensity Score Matched Characteristics .................... 67

Table 42. Medicaid and QHP Enrollees General Population Propensity Score Matched Characteristics .................. 69

Table 43. Observed Utilization Rates (Per Member Per 100 Person-Years) for Medicaid and QHP Enrollees in

2014, 2015, and 2016 ................................................................................................................................................. 71

Table 44. Observed and Estimated PMPM Cost Scenarios for Medicaid and QHP Enrollees by Service

Category and Year ....................................................................................................................................................... 73

Arkansas Health Care Independence Program (‘Private Option’) Final Report i

Executive Summary

Background

In 2013, like many other states across the country, Arkansas faced a complex series of political challenges

following the U.S Supreme Court decision on the Patient Protection and Affordable Care (PPACA) Act. Unlike any

other state in the South, however, Arkansas was able to successfully navigate these challenges to pursue a novel

approach to Medicaid expansion through the commercial sector. Through a Section 1115 demonstration waiver,

the state utilized premium assistance to secure private health insurance, offered on the newly formed individual

health insurance marketplace (the Marketplace), for individuals between 19 and 64 years of age with incomes at

or below 138 percent of the federal poverty level (FPL).

In 2014, Arkansas successfully established the Health Care Independence Program (HCIP),

commonly referred to

as the “Private Option,” as designed under the terms and conditions of the Section 1115 demonstration waiver.

Through 2015, the estimated target-enrollment population of approximately 250,000 was met. Approximately

25,000 additional individuals eligible under the PPACA — and deemed to have exceptional healthcare needs —

were enrolled in the traditional Medicaid program. Finally, approximately 20,000 previously eligible but newly

enrolled individuals also obtained Medicaid coverage. By the end of 2016, the Private Option population totaled

approximately 280,000.

Arkansas’s healthcare providers have reported significant clinical and financial effects under the HCIP. In 2014,

federally qualified community health centers (FQHCs) reported increased success in attaining needed specialty

referrals for their clients.

iii

The Arkansas Hospital Association (AHA) reported significant annualized reductions in

uninsured outpatient visits (45.7 percent reduction), emergency room (ER) visits (38.8 percent reduction), and

hospital admissions (48.7 percent reduction).

The state’s public teaching hospital reported a reduction in uninsured

admissions, from 16 percent to 3 percent, during the same time period.

These reductions persisted through 2016.

Competitiveness and consumer choice in the Marketplace have increased across the seven market regions in the

state with approximately 80 percent of the covered lives in the Marketplace enrolled through the Private Option.

In 2014, individuals in three out of the seven regions of the state — those marked by extreme poverty — only had

access to Arkansas Blue Cross and Blue Shield and Blue Cross Blue Shield Multi-State plans. By 2016, five carriers

were offering coverage across all seven market regions, with one market region having access to six carriers. A

sixth carrier operated in a single region restricted by Medicaid’s purchasing guidance limiting premium assistance

to those plans within 10 percent of the second-lowest cost silver plan in the market region. Over the three year

period, all plans experienced single digit rate increases.

For 2014, the estimated budget neutrality cap (BNC) was exceeded during the initial enrollment phase of the

program. The enrollment of younger individuals over time (affecting net premiums), the rebate of medical-loss

ratio (MLR) payments by one carrier not meeting the minimum MLR requirements in 2014, and inflationary

expectations brought cumulative program costs within the estimated BNC 2015 limit of $500.08 per member per

month (PMPM) and well under the 2016 limit of $523.58 PMPM.

Summary of Findings Based on Evaluation Hypotheses

The HCIP programmatic goals and objectives included successful enrollment, enhanced access to quality health

care, improved quality of care and outcomes, and enhanced continuity of coverage and care at times of re-

enrollment and during income fluctuations. These goals and objectives were to be achieved within a cost-effective

Arkansas Health Care Independence Program (‘Private Option’) Final Report ii

framework for the Medicaid program, compared with what would have occurred if the state had provided

coverage to the same expansion group in Arkansas’s traditional Medicaid fee-for-service (FFS) delivery system.

The state’s required evaluation design under the terms and conditions of the Private Option waiver was

negotiated within 135 days of approval of the waiver on Sept. 27, 2013. The terms and conditions required the

evaluation to meet prevailing standards of scientific rigor, use the best available data, use controls and

adjustments for reporting limitations of the data, and discuss generalizability of results. Additional requirements

included a robust discussion of cost effectiveness. The state was required to submit a final Interim Evaluation

Report within 180 days of the end of the second year of the demonstration. That report is available

at www.achi.net.

The evaluation was conducted independently and with oversight of a National Advisory Committee consisting of

established leaders in major academic and medical centers around the country. The evaluation employed the

most current and well-established research design techniques to optimize confidence in observed findings. There

were two principle comparisons — the experiences of those with “Higher Needs” and that of the “General

Population.”

The Higher Needs population examined the approximately one-half of the newly eligible expansion population

who took a medical frailty screener to detect prior conditions and utilization. This information was used to

identify and compare similar individuals with higher needs that were placed in the traditional Medicaid program

versus those placed in QHPs through a quasi-experimental regression discontinuity approach.

The General Population consisted of the one-half of newly eligible that did not take the screener and were placed

in QHPs. They were paired with the approximately 40,000 newly enrolled adult Medicaid beneficiaries (woodwork

effect) and were matched using advanced propensity score techniques with appropriate statistical tests applied.

This report reflects the experience and findings from the three-year waiver for the Private Option, and major

findings are summarized below, grouped by questions of interest.

1. What were differences across access, quality, and outcomes between those enrolled in Medicaid and

those enrolled in commercial Qualified Health Plans (QHPs)?

A major assumption grounded in Arkansas’s use of premium assistance through the Marketplace was that

by utilizing the delivery system available to the privately enrolled individuals in the Marketplace, the

availability and accessibility of both primary care providers (PCPs) and specialists would be greater than

what would have been expected if Arkansas had utilized a traditional Medicaid expansion strategy. A

three-year enrollment comparison of Medicaid and commercial QHP beneficiaries in both the cohort with

Higher Needs and that in the General Population revealed:

 The geographic proximity of available primary and specialty providers were similar for those

served by Medicaid and the QHP networks, and both met network adequacy requirements of the

Arkansas Insurance Department.

 Initiation of care occurred more rapidly for enrollees in QHPs than for those in the Medicaid

program following enrollment.

 In 2014, differences in the accessibility of both primary care and specialty providers were

reported, with QHP enrollees experiencing increased ability to get needed “care, tests, and

treatment” and receiving “an appointment for a check-up or routine care as soon as needed,”

compared to their Medicaid counterparts.

Arkansas Health Care Independence Program (‘Private Option’) Final Report iii

 Perceived access differences improved after 18 months in the program for the General

Population. However, for individuals in the Higher Needs Population, Medicaid enrollees

continued to report more difficulty “receiving care when they needed it right away” and did not

always “find it easy to get the care, tests, and treatments they need,” compared to QHP enrollees

(range of differences 31-36 percent).

 For Emergency Room (ER) use, differences were only observed within the General Population, in

which Medicaid enrollees experienced more ER visits in total, for both emergent and non-

emergent reasons, compared to QHP enrollees over the three years enrolled.

 With the exception of QHP enrollees experiencing longer hospital stays compared to Medicaid

enrollees, there are no consistent differences across hospitalization measures.

 For clinical services assessed for both populations — and for most measures studied — differences

in care and clinical service delivery were observed.

 QHP enrollees were significantly more likely to receive individual clinical preventive services and

were more likely to receive all recommended screenings (a range of 24-94 percent relative

differences for the General Population).

 QHP enrollees were significantly more likely to receive appropriate disease management services

and more likely to adhere to appropriate medication management than Medicaid enrollees (a

range of 31-55 percent relative differences for the General Population).

 For pregnancy related care, no clinically significant differences were observed in the initiation of

prenatal services, complications of maternity care, or birth outcomes between QHP and Medicaid

enrollees.

 With respect to non-emergency medical transportation, no differences were observed for the

General Population. However, for the Higher Needs Population, those in QHPs were 15 percent

less likely to miss a visit due to transportation issues.

 Opioid use, while similar in the first year, diverged significantly in subsequent years, with

increasing numbers of prescriptions, high dose utilization, and concomitant benzodiazepine use in

the QHPs, compared to Medicaid enrollees.

 With respect to Early Periodic, Screening, Diagnosis, and Treatment (EPSDT) services, we found no

indication that needed services were not available to individuals in premium assistance.

 With respect to continuity of enrollment, both Medicaid and QHP enrollees experienced few

disruptions in coverage, with the exception being a mass eligibility redetermination undertaken in

the summer of 2015.

 There were no statistically significant differences in mortality within the first three years of the

program.

2. What were the differences in costs between Medicaid and premium assistance?

The cost of providing coverage for Medicaid beneficiaries through premium assistance in QHPs was

expected to be greater than providing coverage through the traditional Medicaid FFS system. Exploration

and characterization of the contrasts between the two programs provided a better understanding of the

observed variations in access, utilization, and clinical impacts described above. In addition, dramatic

Arkansas Health Care Independence Program (‘Private Option’) Final Report iv

differences in payment rates were observed, with QHP rates consistently exceeding those in the Medicaid

program:

 Physician payment rates across outpatient services were approximately 95 percent higher in each

of the three years under study for enrollees in a QHP compared to their Medicaid counterparts

(e.g., In 2016, the weighted average per PCP visit was $94.03 for a QHP compared to $47.69 for

Medicaid);

 For inpatient hospital stays, average QHP payments averaged $12,270 per discharge compared to

Medicaid payments of $7,778 (a 53 percent difference); and

 In 2016, administrative costs were estimated to be $91.65 PMPM for QHPs and $64.33 PMPM for

Medicaid (a 29.8 percent difference).

Utilization differences were also observed, but not at the same magnitude as payment differentials.

Medicaid beneficiaries, under the traditional FFS system, experienced increased ER visits and

hospitalizations. Conversely, QHP beneficiaries received more outpatient visit contacts, and by 2016,

almost twice as many prescriptions.

3. What were the cost-effective aspects of premium assistance?

Cost-effectiveness for the purposes of this evaluation considered any benefits associated with care

delivered through QHPs at increased payment rates. To assess cost-effectiveness, total program costs for

enrolled individuals in QHPs were directly compared to their Medicaid counterparts. Ratios of

improvement in care to associated costs were developed (e.g., access improvements, clinical, and

utilization differentials compared to payment rate differentials).

In 2016, the weighted average payment to QHPs (premium and cost-sharing reductions) was $486 PMPM

or $5,832 per year, compared to Medicaid costs of $317 PMPM or $3,804 per year for each enrollee

(using existing Medicaid payment rates). Using the difference of $167 PMPM, select ratios of

improvement to access reflect the following:

 For colorectal cancer screening in the General Population, the QHP cohort had a 94 percent

higher relative difference in screening rates. Thus, the marginal improvement is suggested to be

an increase of 5.6 percent per observed 10 percent increase in program costs associated with use

of premium assistance.

 For the proportion who received all indicated clinical preventive services in the General

Population, the QHP relative difference of 25 percent greater than Medicaid suggests a 1.4

percent improvement in clinical performance per observed 10 percent increase in program costs.

 For individuals with Higher Needs, QHP enrollees were 26 percent more likely to self-report

“always getting care when needed right away” and 18 percent more likely to find it “easy to get

the care, tests, and treatment needed.” This suggests a 1.1 percent improvement in access, per

observed 10 percent increase in program costs.

 For individuals with Higher Needs, Medicaid enrollees experienced fewer outpatient events and a

concurrent higher rate of ER visits and hospitalizations. For each observed 10 percent increase in

program costs, QHPs were projected to achieve seven more physician office visits and avoid 2.5

ER visits per 100 person years.

Arkansas Health Care Independence Program (‘Private Option’) Final Report v

 There were no clinical indicators in which Medicaid was favored.

 Importantly, enrollees in QHPs received twice as many prescriptions than their Medicaid

counterparts. For each 10 percent increase in program costs, QHPs were projected to cover 78

additional prescriptions per 100 person years.

Over the three-year evaluation, PMPM claims costs for QHPs increased, a trend which was associated

with both increased utilization rates and provider rate increases. The established Medicaid rate schedule

does not incorporate inflationary adjustments, so no comparable increases were observed during the

evaluation. Continued divergence in experience, utilization, and payment rates will likely affect ratios of

improvement in care to associated program costs.

4. What would the Medicaid program have experienced if a traditional Medicaid expansion had been

adopted?

A core component of this demonstration evaluation is an examination of the hypothetical costs of

covering the entire expansion population through Arkansas’s traditional Medicaid program and

identifying the programmatic changes that would be necessary to achieve a similar outcome to that

experienced through premium assistance. In 2013, prior to the PPACA expansion, Arkansas had one of the

lowest Medicaid eligibility thresholds for non-disabled adults in the U.S., covering only 24,955 non-

disabled adults with a full benefits package.

In 2014, following PPACA expansion, an additional 267,482 individuals were covered:

 approximately 17,300 (6.5 percent) previously eligible but newly enrolled;

 approximately 25,000 (9.3 percent) PPACA eligible but with exceptional healthcare needs;

 and 225,000 (84.2 percent) PPACA eligible with premiums purchased on the individual marketplace.

These 267,482 individuals represented 16.0 percent of the total 19- to 64-year-old population in the state.

In 2016, this number increased to 330,943 covered lives:

 approximately 22,375 (6.8 percent) PPACA eligible but with exceptional healthcare needs;

 32,427 (9.8 percent) interim status before enrollment in a QHP;

 and 276,141 (83.4 percent) PPACA eligible with premiums purchased on the individual marketplace.

These 330,943 individuals represent 19.1 percent of the working-aged adults within the state.

Thus, because of the high rates of uninsurance and low Medicaid eligibility prior to the PPACA, Medicaid

has experienced a 13-fold increase in coverage for the non-disabled 19- to 64-year-old population.

Traditional microeconomics principles would suggest that increased demand through the expansion of the

Medicaid program would place increasing price pressure on the rate structure of the existing Medicaid

program. The observed differences in payment rates between QHPs and Medicaid described above could

lead to unsustainable access differentials for Medicaid enrollees. Any potential increase in payment rates

would affect not only the new expansion population, but also enrollees under the same payment rate

schedule across the entire Medicaid program. To model the potential effects, a budgetary impact analysis

was conducted on increasing payment rates across the Medicaid program.

Three increasingly fiscally conservative scenarios were simulated for alternative expansion options

through the existing Medicaid FFS system, the counterfactual, to provide policymakers with

conditions under which necessary increases to achieve equitable access could be considered. They

Arkansas Health Care Independence Program (‘Private Option’) Final Report vi

included: 1) claims potentially associated with wage-sensitive services; 2) restricted claims associated

with major medical services; and 3) restricted to claims associated only with physician billed services.

The budget impact analysis revealed that costs to the Medicaid program would exceed the increased

costs associated with premium assistance:

 if wage-sensitive payment rates had increased by 29 percent;

 if claim payments associated with clinical services had increased by 45 percent; or

 if physician-only claim payments had increased by 64 percent;

Importantly, under the most conservative scenario of increases restricted to physician-only claims, the

physician rate increase at which the Medicaid program costs exceed those of premium assistance remains

36 percent below the commercial payment rates observed. This suggests the likelihood of continued

differential access despite increased payments.

These findings suggest that with the 13-fold increase in enrollment of 19- to 64-year-olds, plausible

required increases in Medicaid payment rates across the entire program would exceed the costs

associated with purchasing commercial coverage through premium assistance.

Conclusion

The three-year evaluation of the Private Option is the first direct comparison of commercial insurance with

Medicaid since Medicaid’s inception in 1965. Through Arkansas’s use of premium assistance, and as a

consequence of the rigorous CMS evaluation requirements of the waiver, previously unfeasible direct

comparisons of system performance have been enabled.

Arkansas Medicaid achieves comparable participation in its network, compared to QHPs, for both primary and

specialty providers. However, likely due to markedly higher provider payment rates and more active enrollee

management, the network adequacy and clinical performance of the QHPs exceeds that of Medicaid. These

differences have an impact on the uptake of clinical preventive services, appropriate disease management, and

utilization of emergency room services. For those with Higher Needs, perceived barriers to getting care when

needed and getting necessary tests and treatments are persistent for Medicaid.

For pregnancy, where Medicaid previously covered a majority of deliveries and had invested significant

performance improvement efforts, no advantage was observed for treatment in the commercial sector at higher

reimbursement rates. In addition, Medicaid achieved reduced opioid consumption compared to the commercial

sector, likely due to more assertive programmatic restrictions.

In the first three years of the program, no differences in mortality were observed between enrollees in Medicaid

and those in the commercial sector.

Systemic effects on the Arkansas healthcare landscape through the use of premium assistance was significant. By

guaranteeing purchase for over 250,000 individuals and establishing purchasing guidelines, at a time when

surrounding states were facing major challenges in the stability of their markets, the Arkansas Health Insurance

Marketplace experienced exceptional stability, marked increase in competition, and single-digit rate increases

across the three years. In addition, Arkansas has experienced no hospital closures — with service expansion in

some underserved areas — further distinguishing its overall system stability from neighboring states and from the

rest of the country.

Arkansas Health Care Independence Program (‘Private Option’) Final Report vii

Finally, while the costs of premium assistance exceed those modeled under a static Medicaid expansion scenario,

it is unlikely that Arkansas Medicaid would have been able to absorb a 13-fold increase in enrollees and meet the

federal equal access requirements, under which the state is subject to judiciary review, without considerable

adjustment to provider rates. Although political discourse has highlighted concerns about the differences in

absolute cost between commercial and Medicaid alternatives, Medicaid expansion scenarios under which similar

clinical experiences would be achieved suggest budgetary outcomes that may mitigate these concerns.

These findings may have limited applicability to other states. Few states had the restrictive Medicaid eligibility

threshold, resulting in dramatic increases in coverage through expansion. In addition, provider rate differentials

between Medicaid and commercial plans are also unlikely to exceed those in Arkansas.

However, these findings do support direct comparisons of both commercial and Medicaid costs and performance

across states. These differential payment rates and associated results raise questions regarding the ability of

Medicaid programs nationwide to meet the federal equal access requirements through delivery system strategies

that pay providers significantly lower rates.

The innovative use of premium assistance and the integrated relationship between the individual Marketplace

and the Arkansas Medicaid program merit continued observation. The demonstration waiver through which

Arkansas implemented the Private Option was extended in January of 2017, enabling the successor premium

assistance program, “Arkansas Works.” With work requirements, workforce training, and/or community service

for a subset of enrollees implemented in June 2018, further evaluation of the use of premium assistance and the

impact on upward social mobility is warranted.

The Centers for Medicare & Medicaid Services [Letter] to Andy Allison, Arkansas Department of Human Services. September

2013. https://www.medicaid.gov/Medicaid-CHIP-Program-Information/By-Topics/Waivers/1115/downloads/ar/Health-Care-

Independence-Program-Private-Option/ar-private-option-app-ltr-09272013.pdf.

Ark. Code § 2-77-2405 (2014).

iii

Susan Ward-Jones, MD. Personal Communication. March 12, 2015.

Arkansas Hospital Association. APO’s Hospital Impact Strong in 2014. The Notebook. 2015;22(22):1.

http://www.arkhospitals.org/archive/notebookpdf/Notebook_07-27-15.pdf.

Dan Rahn, M.D., Testimony before Health Reform Legislative Taskforce. August 20, 2015.

Arkansas Department of Human Services (DHS). (2016, December 31). Arkansas Private Option 1115 Demonstration

Waiver. Quarterly Report October 1, 2016-December 31, 2016 [Report]. Retrieved from

https://www.medicaid.gov/Medicaid-CHIP-Program-Information/By-Topics/Waivers/1115/downloads/ar/Health-Care-

Independence-Program-Private-Option/ar-works-qtrly-rpt-oct-dec-2016.pdf

Arkansas Health Care Independence Program (‘Private Option’) Final Report 1

Introduction

Background

The U.S. Supreme Court’s June 2012 ruling

allowed states to decide whether to extend Medicaid benefits to their

citizens who qualified under the Patient Protection and Affordable Care Act (PPACA) expansion. This amplified the

political polarization about the PPACA at the state level, resulting in varied decisions about expansion. Historically,

states have had the option to implement Medicaid coverage through direct provider reimbursement, Medicaid

managed care contracts, or the purchase of coverage with premium assistance through employer-sponsored

coverage.

The state of Arkansas, navigating the political barriers facing many states, pursued a novel approach to expansion

through the commercial sector. Through a Section 1115 demonstration waiver, the state utilized premium

assistance to secure private individual health insurance offered on the newly formed individual health insurance

marketplace (the Marketplace) to individuals between 19 and 64 years of age with incomes at or below 138

percent of the federal poverty level (FPL).

The Arkansas Health Care Independence Program (HCIP),

commonly

referred to as the “Private Option,” provided coverage to more than 225,000 low-income Arkansans through

2015. By the end of 2016, approximately 300,000 individuals were enrolled.

One component of the waiver’s terms and conditions is a required evaluation of differences in access, quality,

outcomes, and efficiencies achieved through the use of commercial coverage for the low-income expansion

population.

The interim evaluation

examined first-year programmatic differences in both effects and costs

through commercial premium assistance compared to the experience that would have been achieved through a

traditional Medicaid expansion as a principal outcome of interest for the demonstration. The final evaluation, and

content of this report, examines the three-year experience of individuals enrolled in the HCIP program.

Arkansas Profile

Arkansas is a largely rural state with approximately 3 million citizens, many of whom face significant healthcare

challenges. These include high health-risk burdens, low median-family income, high rates of uninsured individuals,

and limited provider capacity, particularly in non-urban areas of the state. The Health Resources and Services

Administration has designated 74 of Arkansas’s 75 counties as medically underserved.

Prior to the PPACA, 25

percent of adult Arkansans between 18 and 64 years of age were without health insurance.

Arkansas’s Medicaid program prior to the HCIP had one of the most stringent eligibility thresholds in the nation

for adults, largely limiting coverage to the aged, disabled, and parents with extremely low incomes and limited

assets. Eligibility for adults between 19 and 64 years of age was restricted to parents/caretakers earning at, or

below, 17 percent FPL. Prior to expansion, non-disabled adults with full benefits comprised 18 percent of

Medicaid beneficiaries.

Expansion of the program under the PPACA more than doubled the number of eligible

19- to 64-year-old beneficiaries.

The Arkansas Medicaid program is a Primary Care Case Management (PCCM) fee-for-service (FFS) based delivery

system. Individuals are assigned to a primary care provider (PCP) and providers may limit the number of Medicaid

beneficiaries assigned.

Medicaid provider reimbursement rates are significantly below their commercial

counterparts. Supplemental payments for select hospitals — critical access hospitals, public and private hospitals,

and state teaching hospitals — have been used to support delivery-system stability. Providers elect to join as a

qualified Medicaid provider, but may limit the number of Medicaid beneficiaries they serve.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 2

The commercial insurance marketplace has historically consisted of two carriers with statewide coverage,

including a dominant carrier with over 65 percent of private coverage penetration and other regional carriers. The

predominant network structure is preferred provider organizations with limited managed care and/or the

presence of restricted networks. This is, in part, due to Arkansas’s “any willing provider”

law, requiring insurers

to allow any provider willing to accept terms for the class of providers into their networks. Under the PPACA, the

state elected to utilize the Federally Facilitated Marketplace (FFM) partnership in which the state conducts plan

management and consumer outreach.

Proactive consumer outreach and advertising was limited to responsive

consumer support based upon state legislative restrictions.

Arkansas Structure of Commercial Premium Assistance

The Arkansas approach utilizing commercial premium assistance has several unique attributes that successfully

meet both Medicaid requirements and protections while enabling commercial sector independence. The state’s

approach was in large part based upon the hypotheses that Arkansas could not meet the equal access provision

requiring state Medicaid provider payments to be “consistent with efficiency, economy, and quality of care and

...sufficient to enlist enough providers so that care and services are available under the plan at least to the extent

that such care and services are available to the general population in the geographic area.”

Successful use of the

commercial plans offered on the Marketplace explicitly meet the equal access provision of Medicaid

requirements. However, several structural elements warrant acknowledgement and are described.

First, Medicaid’s purchase of individual commercial coverage via premium assistance is fundamentally different

from the historic use of Medicaid managed care. In premium assistance, the Medicaid program does not directly

contract with the private carrier but rather purchases plans offered on the individual marketplace. While a

memorandum of understanding was established between the state, Medicaid, the Arkansas Insurance

Department (AID), and each carrier to facilitate payments, plans are governed by AID through existing state law

and certification requirements (e.g., network adequacy). The Medicaid population is then integrated into the

privately insured risk pool, and provider payment rates are established by commercial carriers, not through

independent Medicaid contracts. Medicaid beneficiaries engage providers with commercial insurance cards and

are not separated into a Medicaid-specific program or plan.

Second, plans offered under the PPACA were utilized to meet a majority of the Medicaid cost-sharing protections.

For individuals at or below 100 percent FPL, the state utilized the 100 percent actuarial value (AV) plan required to

be available for Native Americans. For Medicaid beneficiaries between 101 and 138 percent FPL, the state utilized

the high-value silver 94 percent AV plan required to be offered on the Marketplace to individuals between 101

and 150 percent FPL. The remaining Medicaid-required cost-sharing protections were achieved through active

structuring of allowable deductibles and other cost sharing.

Importantly, these plans consist both of premiums

subject to medical-loss ratio (MLR) requirements of 80 percent and of cost-sharing reductions (CSRs) that are to

be fully reconciled (see Figure 1).

AID divided the state into seven geographic market regions, and carriers established age-specific premiums within

market regions (one carrier incorporated allowable tobacco use surcharges). The costs of premium assistance

through the individual marketplace was thus influenced by the premium variation based on age within each

market region, the age distribution of those deemed eligible, CSRs paid, and any subsequent repayments for

failure to meet MLR requirements or reconciliation of CSR.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 3

Figure 1. HCIP Premium and Cost-Sharing Reduction Breakdown

Finally, the impact of Medicaid’s guaranteed purchase in the individual insurance market had the potential to

convey stability to the individual marketplace and improve the actuarial profile of the risk pool. The HCIP Act

further extended potential benefits to the actuarial profile of the individual marketplace by requiring that

individuals who were medically frail or had exceptional healthcare needs requiring supplemental Medicaid

benefits would be retained in the traditional Medicaid program.

Arkansas Structure of PPACA Eligibility and Enrollment

As an FFM partnership with the state conducting plan management and consumer assistance combined with

Medicaid’s use of premium assistance, the Arkansas structure for PPACA eligibility and enrollment was complex.

Arkansas used three enrollment pathways for eligibility determination for beneficiaries: the Arkansas Department of

Human Services (DHS) Supplemental Nutrition Assistance Program (SNAP); an Arkansas eligibility web portal

(access.arkansas.gov); and the federal Healthcare.gov portal. Following eligibility determination, individuals were

directed to a separate enrollment portal (insureark.org) to facilitate a healthcare needs assessment and plan selection.

Additional cost-sharing reduction for

HCIP enrollees from 0% to 100% FPL

Deductible buy-down for HCIP enrollees

from 101% to 138% FPL

Cost-sharing reduction for enrollees

from 139% to 150% FPL

Premium

3.5%

2.5%

24.0%

70.0%

100%

PMPM cap including

wrap-around costs:

2014: $477.63

2015: $500.08

2016: $523.58

Audit Processes:

Subject to reconciliation

Subject to medical loss ratio

* Actual PMPM costs were $485.84 (186,950 enrollees) in 2014, $486.86 (200,703 enrollees) in 2015, and $478.61

(276,141 enrollees) in December 2016.

Source: “Arkansas Health Care Independence Program Annual Cap.” Arkansas Department of Human Services

and Arkansas Department of Human Services Final Private Option Report, 2016.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 4

The SNAP-facilitated eligibility determination strategy was a time-limited effort to reach out to and engage with

potentially eligible beneficiaries. Through prior income determination for SNAP benefits, DHS identified

individuals and notified them of their eligibility. Redetermination of income or family composition was not

conducted. Individuals who affirmed their desire for coverage in response to the notice were directed to the

enrollment website. The Arkansas eligibility portal was utilized by DHS county offices, outreach workers,

community and faith-based organizations, and insurance agents across the state. Individuals thought to be

Medicaid-eligible were directed to the portal, where eligibility applications and FPL determinations were

processed. The enrollment pathways and plan assignment process for 2014 enrollment is depicted in Figure 2.

Figure 2. Enrollment Pathways and Plan Assignment Process

As in other states, individuals and the state both experienced challenges in the first year of the federal

Healthcare.gov portal.

Individuals identified through the federal portal and deemed to be Medicaid eligible were

transferred to the state for determination. Frequently, unsuccessful transfer of information resulted in

incomplete enrollments, with similar experiences noted by commercial carriers for individuals above 138 percent

FPL. Over time, the volume of enrollees and accuracy of eligibility information improved.

Two categories of eligible individuals were observed. One category was comprised of individuals who had

previously been eligible for traditional Medicaid benefits but had not enrolled and then subsequently applied and

were determined to be eligible. These individuals were placed into the Medicaid program. The second category

included individuals newly eligible under the PPACA — parents/caretakers from 18 to 138 percent FPL and

childless adults from 0 to 138 percent FPL. These individuals were eligible for commercial premium assistance

under the demonstration waiver. Prior to commercial enrollment, however, individuals were asked to complete a

Enrollment Portals

SNAP

Eligible

PPACA

Newly

Eligible

≤ 138% FPL

Previously

Eligible

Newly Enrolled

Plan

Selected

> 138%

FPL

Auto

Assigned

Federally

Facilitated

Health Insurance

Marketplace

(healthcare.gov)

Arkansas

Eligibility &

Enrollment

System

Arkansas Health Care Independence Program (‘Private Option’) Final Report 5

Healthcare Needs Assessment Questionnaire (the Questionnaire) to retain those with exceptional healthcare

needs in the traditional Medicaid program.

As required in the HCIP Act, these retained individuals were to include those “…determined to be more effectively

covered through the standard Medicaid program, such as an individual who is medically frail

or other individuals

with exceptional medical needs for whom coverage through the Health Insurance Marketplace is determined to

be impractical, overly complex, or would undermine continuity or effectiveness of care.” No previously developed

and validated tool for this purpose was known to exist.

The Arkansas Center for Health Improvement (ACHI) and the Arkansas DHS Division of Medical Services

collaborated with experts at the University of Michigan and the Agency for Health Care Research and Quality

(AHRQ) to develop a screener to identify newly eligible Medicaid applicants who had exceptional healthcare

needs. A pooled subsample from the Household Component of the Medical Expenditure Panel Survey (MEPS),

2005-2010,

was used to develop the Questionnaire and scoring thresholds. Questionnaire responses grouped

individuals into one of three categories:

1) Exceptional healthcare needs (Frail): those who reported exceptional healthcare needs as represented by

deficits in their “activities of daily living,” having severe mental illness, and/or being dependent or homeless;

2) Exceptional healthcare needs (Threshold): those who reported high healthcare use in the prior six months

through hospitalization, emergency room, and/or outpatient visits who met the predetermined threshold; or

3) No exceptional healthcare needs: those completing the Questionnaire who did not meet any criteria

outlined in (1) or exceed the predetermined threshold.

Individuals completing the screener and deemed as not having exceptional healthcare needs proceeded to select

plans offered on the individual marketplace through the state’s enrollment portal. The predetermined threshold

target was selected to achieve 10 percent retention for PPACA-newly eligible within the Medicaid program as

operationalized from the HCIP contract.

Importantly, approximately 50 percent of those determined to be PPACA eligible did not proceed to the

enrollment portal and complete the Questionnaire. After a maximum of 45 days, these individuals entered an

auto-assignment process. Individuals were auto-assigned to carriers based upon previously determined ratios tied

to the number of carriers in each of the seven insurance market regions. Auto-assigned individuals had a time-

limited opportunity to take the Questionnaire and choose to change carriers. Individuals could either choose to

stay with their assigned plan or choose another plan during subsequent open-enrollment periods each year or for

qualifying family events.

Following selection or auto-assignment, DHS executed monthly premium payments to the carriers on behalf of

the individuals. Individuals in the commercial plans received a letter with their Medicaid Identification Number (to

receive services prior to commercial plan coverage start date) and, subsequently, a commercial insurance card

from their carrier. Medicaid-retained individuals received a Medicaid benefit card.

In the second half of 2015 and all of 2016, DHS opted not to use the threshold method to identify individuals who

had exceptional healthcare needs. However, DHS continued use of the question about deficits in activities of daily

living to identify individuals with exceptional healthcare needs. New individuals enrolling in these years who did

not take the Questionnaire were still assigned to a QHP.

Note that “Medically Frail” as defined in the HCIP Act and operationalized in the HCIP preceded the current definition of

“Medically Frail” as found in 42 CFR 440.315(f).

Arkansas Health Care Independence Program (‘Private Option’) Final Report 6

Arkansas HCIP Program Experience

Figure 3. HCIP Monthly Enrollment, January 2014–December 2016

Enrollment in the HCIP and for other newly eligible individuals, both within Medicaid and for those above 138

percent FPL in the individual marketplace, resulted in a reduction in the uninsured rate for adults from 22.5

percent to 9.6 percent in 2014, the largest reduction observed nationwide.

More than 250,000 Arkansans

enrolled, with approximately 45,000 occurring through the SNAP-facilitated eligibility. In 2014, approximately half

completed the Questionnaire. Of the total new enrollees, 10 percent were deemed to have exceptional

healthcare needs and were maintained in traditional Medicaid. In 2014, an additional 22,000 adults previously

eligible (but not enrolled) for traditional Medicaid became newly enrolled in the program. In the three years of

the program, Medicaid has purchased individual plans covering at least one month for 399,330 individuals and

276,081 were enrolled in December 2016 (see Figure 3) through the premium assistance mechanism.

Through 2016, HCIP enrollees represent approximately 80 percent of the covered lives on the individual

marketplace. They are younger than their counterparts above 138 percent FPL participating in the Marketplace

(see Figure 4 for Year 1 age demographic comparison).

50,000

100,000

150,000

200,000

250,000

300,000

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

HCIP Commercial Insurance Coverage

2014

2015

2016

Arkansas Health Care Independence Program (‘Private Option’) Final Report 7

Carrier participation, and thus

beneficiary choice, has increased in

each market region. In 2014, of the

seven market regions, only three

had more than two carrier options.

Statewide, in 2016, every region

had at least five participating

carriers, with one region having six.

Premiums for the benchmark silver

plan in the largest market region

dropped 2.3 percent in 2015 and

experienced an increase of 3.7

percent in 2016.

Because

insurance premiums external to the

Marketplace are tied to those on

the Marketplace, similar rate effects were seen in the non-Marketplace PPACA-compliant market.

Healthcare providers have reported both significant clinical and financial effects. Federally qualified community

health centers (FQHCs) have reported increased success in attaining needed specialty referrals for their clients.

The Arkansas Hospital Association (AHA) compared service use and uninsured volumes between 2013 and 2014.

They found a 5 percent increase in ER visits overall, and a reduction in outpatient visits (45.7 percent), ER visits

(38.8 percent), and hospital admissions for the uninsured population.

The University of Arkansas for Medical

Sciences (UAMS) has reported a reduction in uninsured admissions, from 16 percent to 3 percent, during a similar

time period.

The Section 1115 demonstration waiver required an estimated budget neutrality cap (BNC) represented as a per-

member per month (PMPM) cost for program expenditures.

These expenditures include premiums, CSRs, and

required Medicaid benefits (wrap-around costs associated with required benefits, such as non-emergency

transportation) not covered through premium assistance. Average program expenditures over the program years

have been within the estimated budget neutrality caps established within the conditions of the Section 1115

demonstration waiver with cumulative program expenditures at the beginning of Program Year 3 (PY3) (January

2016) equal to $489.01 PMPM, 7.1 percent below the PY3 — Calendar Year (CY) 2016 — federal cap of $526.58.

During Program Year 1 (PY1) — CY 2014 — cumulative program expenditures exceeded the BNC, but were under

the estimated cap by Program Year 2 (PY2) — CY 2015. The first months of 2016 saw the PMPM under the BNC

but eventually exceeded it by the end of the year. Early in 2017 there was a substantial differential between

PMPM and the 9 percent increase in BNC from 2016.

This includes the effects due to the enrollment of younger

individuals over time affecting net premiums, the rebate of MLR premiums by one carrier who did not meet the

MLR requirements, and inflationary expectations built into the BNC estimates. Importantly, this evaluation will

replace BNC estimates with realized experience.

41%

25%

23%

19%

22%

23%

15% 33%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

With Premium Assistance Without Premium Assistance

Percentage Enrolled Through the

Health Insurance Marketplace

Age 34 and Younger Age 35-44 Age 45-54 Age 55-64

Figure 4. Enrollment Age Demographics by Category

Arkansas Health Care Independence Program (‘Private Option’) Final Report 8

Figure 5. Premium and Cost-Sharing Reduction Breakdown, January 2014–December 2016

$477.63

$500.08

$523.58

$440

$450

$460

$470

$480

$490

$500

$510

$520

$530

50k

100k

150k

200k

250k

300k

J F M A M J J A S O N D J F M A M J J A S O N D J F M A M J J A S O N D

2014 2015 2016

Per Member Per Month

Enrollees

HCIP Monthly Enrollment and Total Costs

Enrollees PMPM Cost 2014 Budget Cap 2015 Budget Cap 2016 Budget Cap

(4.7% increase)

Over the three years of the HCIP program, management has moved from a start-up to a steady-state phase. The

first six months of 2014 contained significant enrollment growth, followed by 12 months of relative steady-state

program performance. In the fall of 2014, income and eligibility redeterminations for Medicaid were delayed due

to information technology limitations on the eligibility and enrollment system. In July 2015, restarting these

determinations resulted in termination of coverage for approximately 10 percent of the covered lives.

Finally, in

2016, DHS implemented purchasing strategies through which premium assistance would only be available for

plans that were priced within 10 percent of the second lowest silver plan within the market region.

Arkansas HCIP Evaluation Strategy

The terms and conditions of the Section 1115 demonstration waiver state the requirements for submission of an

interim report following completion of the second programmatic year. Through a subsequent modification of the

waiver terms and conditions, a final summative report profiling the complete three-year HCIP experience using

fully adjudicated claims data was due by June 30, 2018. This document serves as the final summative report and

represents outcomes from the full three programmatic years of the HCIP. As expected, the Interim Report

reflected the start-up period in the first year and had certain limitations — quality metrics requiring enrollment

periods of 12 months or greater were problematic; continuity of care and coverage was not yet observable; and

steady state comparisons of system performance were premature. However, variations in experienced access,

utilization by types and location of services, and healthcare engagement opportunities to address unmet needs

were the focus of the Interim Report. Costs for the primary comparison of premium assistance to what would

have been experienced under a traditional Medicaid expansion were simulated in Year 1. We update these

simulations based on cost comparisons at the end of 2016. The CSR reconciliation for any of the three years of

Arkansas Health Care Independence Program (‘Private Option’) Final Report 9

HCIP have not been finalized and, hence, are not included in the financial description of the program in this

report. However, this final report presents quality metrics for up to three years of coverage in the program and

also includes a broader set of indicators.

In this report, we profile cost estimates covering the three years of the HCIP program; assess quality of care for

the Medicaid and QHP plans during a steady state period (2016); present an evaluation of continuity of care and

coverage following the redetermination period (July 2015–December 2015); conduct a comparison of Medicaid

and QHP pregnancy quality indicators; include a discussion of the EPSDT population that was eligible for QHP

enrollment; include a special section on opioid utilization; include a section on mortality difference between

Medicaid and QHP enrollees; and re-examine changes in the effectiveness and costs associated with the use of

premium assistance for Medicaid beneficiaries.

Research Design and Approach

Goals and Objectives

The HCIP programmatic goals and objectives included successful enrollment, enhanced access to quality health

care, improved quality of care and outcomes, and enhanced continuity of coverage and care at times of re-

enrollment and income fluctuation. These goals and objectives were to be achieved within a cost-effective

framework for the Medicaid program compared with what would have occurred if the state had provided

coverage for the same expansion group in Arkansas Medicaid’s traditional delivery system.

Figure 6. Arkansas Demonstration Waiver Evaluation Logic Model

Following the evaluation logic model (see Figure 6), this final report presents results from analyses that used

geographic travel time between enrollees and providers, enrollment information, and retrospective claims data,

and sampled survey responses. Using two different comparison groups and approaches, the evaluation team

Improved

Prevention and

Outcomes

Improved

Access to

Providers and

Healthcare

Services

Successful

Enrollment

Through

Marketplace,

State

Enrollment

Portal, and/or

Targeted

Outreach

Continuity of Enrollment and Reduced Churn

Cost-Effectiveness for Medicaid via

System Reform

Arkansas Health Care Independence Program (‘Private Option’) Final Report 10

empirically assessed whether QHP enrollees obtained better access to providers and healthcare services by using

commercial carrier networks and payment rates than comparable groups enrolled in traditional Medicaid. We

also assessed whether QHP enrollees received more appropriate care — including prevention, chronic disease

management, and therapeutic interventions potentially leading to better outcomes than their Medicaid

counterparts. This report also contains a profile of the effects of disruptions in continuous insurance coverage

(“churn”), highlighted by experiences after a universal eligibility redetermination in July 2015. Differences in

program costs between premium assistance and traditional Medicaid were determined and evaluated with

respect to differences in access, utilization, quality, and outcomes. In addition, the alternative of expansion solely

within the traditional Medicaid program was assessed by examining program impact and simulating alternative

payment scenarios with the goal of achieving similar outcomes.

Programmatic Timeline and Reporting Requirements

Under the terms of the Section 1115 demonstration waiver, two reports were required to be submitted that

characterize the experiences of individuals enrolled in Qualified Health Plans (QHPs) using premium assistance — the

Interim Report reflecting the first 12 months of programmatic experience, and the Final Report reflecting the three

years of the originally approved waiver. A key component of both required reports is a comparison of QHP enrollee

experiences to similar cohorts of new enrollees in traditional Medicaid. Another key component is a counterfactual

analysis through budget impact simulations of what it would have cost Medicaid through a traditional expansion to

achieve similar access, quality, and outcomes of the QHPs.

Efforts to optimize comparisons of care between that which is provided through commercial premium assistance

and that which is provided through the Medicaid program have been incorporated into this demonstration

evaluation. Program experience has included a significant uptake period during the initial six months of the

program followed by 12 months of steady state program enrollment due to delays in eligibility redeterminations

by DHS. Redeterminations in the summer of 2015 resulted in a number of previously enrolled individuals being

terminated (10 percent) from Medicaid and premium assistance programs, leading to a stimulus for discontinuity

of coverage and care. Finally, the state extended the HCIP premium assistance program (now called Arkansas

Works) through 2021,

which has continued the three-year HCIP demonstration into Year 4. As depicted in Figure 7

we have modified the original timing of evaluation components in response to these programmatic experiences.

Figure 7. Arkansas Health Care Independence Program Period: System Evaluation

This final report includes assessments of care using enrollment and claims data for the three program years of

2014 through 2016 and survey data from two enhanced Consumer Assessment Health Plan Surveys. Assessments

Arkansas Health Care Independence Program (‘Private Option’) Final Report 11

of quality and utilization data will reflect the steady state periods depicted, and continuity assessments will

evaluate the impact of redeterminations in 2015. In addition, cost-effectiveness and counterfactual simulations

will be reported by year, with 2016 being the first full year of steady state.

Theoretical Approach

The approach and content of this evaluation focuses on a comparison between access, quality, and outcomes

experienced by Arkansans enrolled in traditional Medicaid versus those enrolled in a QHP through premium

assistance. We also evaluate the cost-effectiveness of offering private health insurance by covering costs through

premium assistance, as opposed to expanding the Medicaid program, and assess the counterfactual experience

that would have been expected through the expansion of the traditional Medicaid program. Questions addressed

in this final report include (note that first year experience findings are summarized in this section, under

subsection g):

1. What were the differences associated with access, quality, and outcomes between those enrolled in

Medicaid and those enrolled in QHPs?

A major assumption grounded in Arkansas’s use of premium assistance through the Marketplace was that by

utilizing the delivery system available to the privately enrolled individuals in the Marketplace, the availability

and accessibility of both primary care providers and specialists would have exceeded that of a more

traditional Arkansas Medicaid expansion. By purchasing health insurance offered on the Marketplace and

utilizing private sector provider networks and their established payment rates, traditional barriers to

equitable health care, including limited specialist participation and provider availability, would be minimized.

In fact, through the use of commercial plans offered on the Marketplace, providers were not able to

differentiate between privately insured individuals supported by Medicaid premium assistance (e.g., those

earning less than 138 percent FPL), those supported by tax credits (139 to 400 percent FPL), and those earning

above 400 percent FPL who purchased from carriers offering plans in the Marketplace.

The PPACA required, through federal regulation, that QHPs “…maintain a network that is sufficient in

number and types of providers, including providers that specialize in mental health and substance abuse

services, to assure that all services will be accessible without unreasonable delay.”

AID has developed

network adequacy targets and data submission requirements to ensure adequacy of provider networks in

QHPs offered in the Marketplace. AID established network adequacy requirements to be reported by

participating carriers on an annual basis. This reporting requires data to be submitted, demonstrating a 30-

mile or 30-minute coverage radius from each general/family practitioner or internal medicine provider, and

each family practitioner/pediatrician. In addition, data/maps must be submitted demonstrating a 60-mile or

60-minute coverage radius from each category of specialist including, but not limited to, cardiologists,

endocrinologists, obstetricians, oncologists, ophthalmologists, psychiatric and state licensed clinical

psychologists, and pulmonologists. In addition to geographic access, perceived access indicators from

Consumer Assessment of Healthcare Providers and Systems (CAHPS) surveys and realized access from first

contact with new providers upon becoming insured will identify differences to providers of traditional

Medicaid enrollees to those in QHPs.

To assess quality and outcomes, measures were selected based on National Quality Forum (NQF) guidelines, and

peer reviewed studies that rate and compare health plans and providers. This included reviews of national

healthcare quality guidance, including AHRQ National Healthcare Quality Reports (Agency for Healthcare

Research and Quality, 2012 and 2013 National Healthcare Quality Reports), the U.S. Department of Health and

Arkansas Health Care Independence Program (‘Private Option’) Final Report 12

Human Services Annual Progress Report to Congress (“National Strategy for Quality Improvement: 2013 Annual

Progress Report to Congress,” July 2013) the Centers for Medicare and Medicaid Services (“Medicaid Core Set:

Core Set of Health Care Quality Measures for Adults Enrolled in Medicaid: Technical Specifications and Resource

Manual for Federal Fiscal Year 2014 Reporting,” May 2014, CMS), CAHPS Reporting Guidance (CAHPS Reporting:

Reporting Measures for CAHPS® Health Plan Survey 4.0.), and National Committee for Quality Assurance (NCQA)

Healthcare Effectiveness Data and Information Set (HEDIS).

For this final report that spans individual healthcare coverage experience for up to three years, we used

validated quality, utilization, and outcome measures studied from the sources listed above to meet two

criteria. We chose measures that could be assessed efficiently by patient self-report or by administrative

claims data (measures requiring electronic medical/hospital records or provider reporting sources were

excluded), and that have been previously used to assess care for Medicaid and commercially insured patients.

Final measures for empirical testing were a priori, selected based on oversight from the National Advisory

Committee for this evaluation. The final listing of measures empirically tested to determine differences in

access, quality, and outcomes between Medicaid and QHP enrollees are contained in Appendix A.

2. What were the differences in costs between Medicaid and premium assistance?

Costs to the program were taken from the perspective of the Medicaid program as the payer. For commercial

costs, these included the PMPM premium payments made to commercial carriers on behalf of enrolled

individuals. These PMPM premiums were specific to market region, age, and for one carrier, tobacco status. In

addition to premiums, CSR payments to commercial carriers to achieve Medicaid out-of-pocket cost-sharing

limits were included. Medicaid expenditures for non-covered benefits in the commercial plans — e.g., non-

emergency medical transportation (NEMT) — were also included. Finally, while all MLR rebates were

included, CSR payments, as mentioned previously, have yet to be fully reconciled and are not reflected in cost

estimates presented in this report.

For Medicaid costs, the calculated PMPM expenditure was more complex. Medicaid claims were isolated for

those enrollees in the Medicaid comparison groups. Because Arkansas Medicaid incorporates several

supplemental payment strategies to select providers, allocation of these additional payments was necessary

to achieve a true claim-related expenditure. Hospitals in four categories — teaching, public, private, and

critical access — were eligible for supplemental payments up to the Medicare upper payment limit (UPL) for

both inpatient and outpatient services. In addition, critical access hospitals and UAMS were eligible for cost-

based reimbursement for select services. Total non-claims based payments to providers by Medicaid were

obtained annually for 2014 through 2016 and allocated proportionately to providers by service utilization to

achieve a loaded claims PMPM. Administrative costs were identified from DHS expenditure reports. Non-

changing costs (e.g., Disproportionate Share, Graduate Medical Education, facility costs) were not included in

PMPM estimates, as these would not have changed under alternative expansion approaches. For enrollee-

specific costs (e.g., enrollment, case-management, etc.) a per-enrollee estimate was generated from the

existing Medicaid program and applied to the loaded claims.

3. What were the cost-effective aspects of premium assistance?

Under the premise of the waiver authority, the cost of purchasing healthcare insurance through QHPs using

premium assistance was expected to be greater than purchasing care through the traditional Medicaid system

due to the compressed Medicaid rates publicly available (e.g., $850 per diem for hospitalizations). Concurrently,

the commercial care management strategies and differences in provider payments were hypothesized to

contribute to better access, more appropriate healthcare utilization, and better quality and outcomes.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 13

Effectiveness was assessed through access, utilization, quality, and outcome metrics described above. Costs

were assessed through two approaches. First, by direct comparison of experienced costs for newly enrolled

individuals between those in QHPs and those in the Medicaid program provided absolute differences in

program costs. Modeling efforts to estimate the experience of QHP enrollees and project a PMPM had they

been in traditional Medicaid is employed. In addition, PMPMs from newly eligible and enrolled 19- to 64-year-

old Medicaid beneficiaries are calculated.

Second, where plausible, ratios of improvement in care to associated costs are developed (e.g., access

improvements compared to rate differentials). While no single cost-effectiveness ratio is attainable, observed

effect differences can be interpreted with respect to differential costs between the two programs. If

differential effects on access, utilization, quality, and outcomes are observed, program effects will be

expected to lead to measureable health improvements over time. We will test this expectation by presenting

cost and outcome findings at Baseline (Year 1), Year 2, and Year 3 of the program. This will allow policymakers

to evaluate the potential trade-off of increased costs relative to important indicators focusing on access,

quality of care, and adverse event reduction consistent with health improvement.

4. What would the Medicaid program have experienced if a traditional Medicaid expansion had been adopted?

Examination of the hypothetical costs of covering the entire expansion population in Arkansas’s traditional

Medicaid program and the programmatic changes necessary to achieve a similar outcome to that experienced

through premium assistance is a core component of the demonstration evaluation. Consideration must be

given to the existing Medicaid program, its level of network participation, and the impact of existing payment

rates given differences identified through this evaluation. In addition, the price elasticity of the supply of

medical providers and their ability and/or willingness to provide for the healthcare needs of the expansion

population through the existing Medicaid program must be considered. Finally, if payment rate changes were

required to achieve access and quality outcomes, what would be the financial impact of those modifications

across the entire Medicaid program (e.g., rate changes would apply to all Medicaid rates, not only those

associated with PPACA newly eligible adults)? We provide a basis for this rationale and simulate results under

various pricing scenarios within this report.

Hypotheses

To address the theoretical questions on the previous pages, we tested hypotheses that aligned with the original

12 hypotheses outlined in the Section 1115 Demonstration Waiver Terms and Conditions, STC 70, No. 1 (see

Appendix B). Broadly, the original hypotheses fell into four categories:

1. HCIP beneficiaries will have equal or better access to health care compared with what they would have

otherwise had in the Medicaid system over time. Access will be evaluated using measures for geographic,

perceived, and realized access; use of ER services; potentially preventable ER and hospital admissions; and

non-emergency transportation services.

2. HCIP beneficiaries will have equal or better care and outcomes compared with what they would have

otherwise had in the Medicaid system over time. Quality and outcomes will be evaluated using measures of

preventive (primary, secondary, and tertiary) services and healthcare services, preventable medical events,

and health services utilization.

3. HCIP beneficiaries will have better continuity of care compared with what they would have otherwise had in

the Medicaid system over time. A thorough profile of the experience of churn following the universal

eligibility redetermination in July 2015 will be presented.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 14

4. Services provided to HCIP beneficiaries will prove to be cost-effective. Cost-effectiveness will be evaluated

using findings from testing the aforementioned hypotheses in combination with the following cost

determinations from the three programmatic years spanning 2014–2016:

a. For HCIP beneficiaries, fewer gaps in enrollment, improved continuity of care, and the resultant lower

Medicaid administrative costs would be experienced through premium assistance. Anticipated metrics

include cross-year carrier and Medicaid enrollment, cross-year continuity of primary care provider

engagement, and impact of administrative cost allocation to carriers compared to alternative

Medicaid administration.

b. Through HCIP use of premium assistance in the individual commercial marketplace, performance

characteristics of the Marketplace will be enhanced through increased carrier competition,

stabilization of the actuarial risk pool, and limited premium increases over time.

In this final report, we summarize first year program findings and study trends on premium effects over time.

We include enrollment information, carrier participation, market competition, and premium increases over

the three years of the HCIP. We also present a summary of an actuarial profile of the HCIP population in

comparison to other Arkansas enrollees in the individual marketplace.

c. Use of premium assistance in the individual commercial marketplace will prove to be cost-effective for

the program compared with what it would have cost to cover the same population in the Arkansas

Medicaid system.

Simulation of the counterfactual experience (if all of the PPACA expansion had occurred through the Medicaid

program), including the impact on non-PPACA Medicaid programmatic costs, will reflect one of the primary

outcomes of interest for this hypothesis.

Data Sources and Analytic Comparison Groups

For this final report, data were obtained from three primary data sources: Arkansas Medicaid enrollment files

from the DHS Division of County Operations, administrative claims data (Medicaid and QHPs), and two member

enrollment surveys (CAHPS). In order to construct variables of interest, additional data were obtained from the

Arkansas Department of Health (birth and death certificates from Vital Records), the Arkansas Health Data

Initiative, and the Medicaid-administered exceptional healthcare needs Questionnaire. In addition, enrollee and

network provider addresses were geocoded, allowing distances and drive times between enrollees and providers

to be calculated. Details on data processing can be found in Appendix C.

External claims data were assessed for consistency and integrity before being processed by the analytic data

team. Our team established a step-by-step logic flow to execute data exclusions and to create the final analytical

dataset. See Appendix D for a listing of exclusions and a flowchart of the process used to establish a final

analytical dataset which contains four non-overlapping subpopulations where Medicaid and QHP enrollees were

compared within two groups.

To assess the differences between the programmatic effects of commercial premium assistance to Medicaid, we

utilized two available comparison strategies: 1) a matched comparison group based upon the demographics of

Medicaid and QHP enrollees who did not complete the Questionnaire (the “General Population”); and 2) a

comparison between those who took the exceptional healthcare needs Questionnaire and reported higher

healthcare needs in the six months prior to enrollment (the “Screened Population”) and were subsequently

assigned to either Medicaid or QHP enrollment. For the General Population, we use Year 1 as a Baseline year to

Arkansas Health Care Independence Program (‘Private Option’) Final Report 15

obtain chronic-conditions history from claims data and follow quality and utilization differences across groups in

the second year (those enrolled in the program beginning in 2014 or 2015) and third year (those enrolled in the

program since 2014) of enrollment in the program.

Table 1 summarizes our comparison-group enrollee numbers by year and data source, followed by a description

of the comparison groups. Table 1 in Appendix D (Study Subjects) provides demographic profiles for Baseline/Year

1 for each comparison group by payer.

Table 1. Comparison Group Description and Analytical Data Populations

General Population,

Comparison Group 1

Screened Population,

Comparison Group 2

Year

Data Source

Traditional

Medicaid

(did not complete

the Questionnaire)

QHP

(did not complete the

Questionnaire)

Medicaid

(completed the

Questionnaire and met

the threshold)

QHP

(completed the

Questionnaire but did not

meet the threshold)

Baseline

(Year 1)

Claims

Matching variables from first year including

clinical need obtained

N = 9,037

N = 44,285

Year 2

Claims

N = 27,905

N = 92,940

N = 9,037

N = 44,285

CAHPS II

N = 241

N = 792

N = 985

N = 1,091

Year 3

Claims

N = 13,933

N = 56,874

N = 7,951

N = 37,752

Note: The second CAHPS survey was fielded approximately 18 months after the majority of respondents had enrolled in Medicaid or a

QHP. This was during the second year of insurance coverage in their respective programs. Enrollees in Comparison Group 2 were

oversampled for CAHPS, attributing to the higher number of responses.

Assessing Exceptional Healthcare Needs in Population Completing the Questionnaire (Screener)

A total of 108,690 HCIP enrollees completed a healthcare needs assessment screener. Select individuals who

reported deficits in activities of daily living, severe mental illness, or homelessness, were automatically assigned to

Medicaid and are not included in this evaluation. For the remainder, composite scores were compiled so that

everyone screened had a score on a continuum that ranged between 0.02 (no utilization) and 0.61 (extremely

high utilization). A “threshold” was established at 0.18, based on the assumption that 10 percent of those who

took the Questionnaire would demonstrate a higher need. Individuals with a composite score lower than 0.18 had

the option to choose a QHP or were auto-assigned to a QHP, while those with a composite score of 0.18 or higher

were deemed to have exceptional healthcare needs and were assigned to a Medicaid plan. Because those near

the threshold reported similar experiences in terms of higher utilization, programmatic assignment enables

application of quasi-experimental methods to test for differential program effects.

New Traditional Medicaid Enrollees/New Premium Assistance Commercial Enrollees Balanced on Key

Demographics

The original design to test low-income parents on the traditional Medicaid program (at or below 17 percent FPL)

to low-income parents in the QHPs (more than 17 percent FPL) and replicate quasi-experimental methods similar

to those described previously was not achievable. Fidelity of the income variable and parental status in the DHS

enrollment data violated the assumptions required for this methodological approach. Because of income

discrepancies identified related to eligibility determination, combined with the fact that income is determined at

the point of eligibility determination and varies substantially, the original approach was deemed infeasible.

However, because individuals previously eligible for Medicaid, but newly enrolled, were not screened for exceptional

healthcare needs, and because more than 116,000 of the 2014 newly eligible HCIP enrollees did not take the

screener, comparison of newly enrolled non-screened individuals within the two programs was possible. These

Arkansas Health Care Independence Program (‘Private Option’) Final Report 16

individuals represent the General Population in each program. They were balanced across programs on demographic

variables and, for those enrolled for at least two years, the first year clinical status was also used. This balancing

provided the ability to compare programmatic effects for the General (i.e., non-screened) Population.

Methodological Approaches

Regression Discontinuity

In evaluations such as this, when random assignment to treatment and control groups is not feasible, comparisons

can be performed by examining subgroups of individuals based on scores just below or above a cut-point value of

a predetermined variable. For the Screened population, this approach offered the opportunity to examine

individuals who took the Questionnaire but were assigned to Medicaid or QHP enrollment based upon their

responses and a predetermined composite-score threshold. The assumption is that individuals with very similar

scores on either side of the cut-point (threshold) should not differ significantly in terms of need, even though the

cut-point assigns the individuals into different groups. Regression discontinuity is a quasi-experimental design that

is increasingly being used in evaluation analyses to test differences attributable to group assignment.

In our evaluation, roughly one-half of the new enrollees to Medicaid and premium assistance plans completed a

healthcare needs assessment screener (described on the previous page). Those with a composite score of less

than 0.18 were assigned to a QHP, while those with a score of 0.18 or higher were assigned to a Medicaid plan. If

the regression line examining the association between composite score and an outcome variable of interest (e.g.,

number of ER visits to treat an emergent condition) passes continuously through the cut-point, we would not

expect to see a program effect. If we were to observe a sharp jump at the composite score cut-point where the

program assignment was made, we would have a strong indication that the jump was due to the program effect

and not attributable to individual demographics or traits. Further details on this methodological approach can be

found in Appendix E.

Propensity Score Matching

Enrollees were not randomly assigned to the traditional Medicaid plan or a QHP. Therefore, propensity scores can

be described as the probability of being assigned to a treatment group (here, our QHP group) given a set of

underlying characteristics (or observed covariates).

In our evaluation, for those newly eligible individuals in

either traditional Medicaid or the QHP who were not screened, we calculated the probability of being assigned to

a QHP treatment group (as opposed to the traditional Medicaid control group). This propensity score of

assignment to treatment is calculated based on incorporating age, gender, race/ethnicity, Charlson Comorbidity

Index, insurance region, and census block median income. We utilized first program year individual experience at

Baseline to develop the propensity scores with subsequent observations when individuals were in Years 2 and 3 of

the program. The goal was to balance the groups assigned to traditional Medicaid or a QHP by the underlying

characteristics included in the propensity score models. Using propensity scores in our empirical assessment of

group differences in access, quality, or healthcare outcomes has the potential to reduce biases associated with

imbalanced underlying characteristics across groups. We used a technique called greedy matching to closely

match individuals in Medicaid and QHPs for comparison. Further details on this methodological approach can be

found in Appendix E.

Direct Comparison of Programmatic Costs and Cost-Effectiveness Assessment

As described, while PMPM costs for premium assistance are reflected in the cumulative premiums paid to carriers

when combined with Medicaid payments for wrap-around services, including NEMT, no similar source of PMPM

Arkansas Health Care Independence Program (‘Private Option’) Final Report 17

costs for Medicaid existed. PMPM costs from both programs were constructed to enable program cost

comparisons. Allocation of non-claims related Medicaid payments (supplemental payments) were allocated and

Medicaid administrative costs associated with new enrollees (i.e., non-fixed costs) were identified and

incorporated as described in Appendix F.

We calculated PMPMs for observed Medicaid and QHP costs. To estimate a “what if QHP enrollees had been

enrolled in FFS Medicaid” scenario, we estimated the QHP experience using Medicaid costs with and without

across-program utilization adjustments. We provide direct program cost comparisons based upon actual

cumulative premiums paid; modeled PMPMs calculated from distribution of observed utilization experience

within the QHPs (inpatient, outpatient, ER, prescriptions filled, etc.); modeled projected PMPMs based upon

observed QHP utilization experience and estimated administrative costs within the QHPs adjusted for Medicaid

payment rates; modeled projected PMPMs based upon modeled QHP enrollee utilization in Medicaid with

Medicaid payment rates; and finally, actual Medicaid PMPM rates for newly eligible Medicaid enrollees.

To support policymakers’ decisions surrounding cost-effectiveness determination, we determined trade-offs for

incremental increases in access associated with identifiable payment increases. Specific provider payment

differentials (e.g., primary care outpatient rates) were determined, and observed differences in associated effects

(e.g., primary care accessibility, inappropriate ER use) facilitated the development of incremental cost-

effectiveness ratios for select indicators.

Counterfactual Medicaid Impact Simulation and Sensitivity Analyses

Examination of the hypothetical costs of covering the entire expansion population in Arkansas’s traditional

Medicaid program and the programmatic changes necessary to achieve a similar outcome to that experienced

through premium assistance is a core component of the demonstration evaluation. The price elasticity of the

supply of medical providers and their ability and/or willingness to provide for healthcare needs of the expansion

population through the existing Medicaid program was the central component of the simulation model.

To model the potential counterfactual Medicaid program impact, we examined the potential programmatic

impact on costs if increases in payment rates had been required to maintain provider access. Individuals whose

care would unlikely be exposed to the impact of rate adjustments were excluded (e.g., individuals 65 and older for

whom Medicare would be the primary payer on most medical services, children less than 1 year old covered on a

different payment rate schedule).

Differences observed for total Medicaid expenditures were calculated at incremental increases and expressed as a

total program cost effect. These were allocated onto enrollment month Medicaid PMPMs calculated previously

for the demonstration population. Summation of the experienced Medicaid PMPMs in our comparison groups

with the additional PMPM load caused by potential rate increases resulted in the generation of counterfactual

PMPMs, based upon underlying alternative rate increase scenarios. Importantly, for this simulation model,

increases in utilization due to increasing rates were not included.

Special Studies: Experience with Health Independence Accounts (2015–2016)

To assess the participation and impact of the Arkansas Health Independence Accounts, we examined individuals

required to participate (e.g., those between 101 and 138 percent FPL) and their respective experiences. Through

identified eligibility files, we categorized individuals into two groups: 1) those who received notification but did

not participate and 2) those who participated with at least one monthly contribution.

Comparisons were made between demographics and claims experiences for those who did and did not

participate. In addition, the frequency and duration of monthly contributions, the total contributions, and the

Arkansas Health Care Independence Program (‘Private Option’) Final Report 18

avoided cost-sharing received due to participation were characterized. Site of service for protected cost-sharing

was examined. Finally, net and individual cost-avoidance was compared to net and individual contributions and

assessment of exit payments for individuals from Arkansas DHS was also completed.

The First Year Experience

Overview

Arkansas successfully established the Health Care Independence Program (HCIP), commonly referred to as the

“Private Option,” in 2014, as designed under the Terms and Conditions of the Section 1115 demonstration waiver.

Through the end of 2015, the estimated target enrollment population of approximately 250,000 had been met.

Additionally, approximately 25,000 individuals eligible under the PPACA and deemed to have exceptional

healthcare needs were enrolled in the traditional Medicaid program, while approximately 22,000 previously

eligible, but newly enrolled individuals in 2014, had obtained Medicaid coverage.

After the first year, healthcare providers reported both significant clinical and financial effects. Federally qualified

community health centers (FQHCs) reported increased success in attaining needed specialty referrals for their

clients.

The Arkansas Hospital Association (AHA) reported annualized reductions in uninsured outpatient visits,

emergency department visits, and admissions — by 45.7 percent, 38.8 percent, and 48.7 percent, respectively.

The state’s teaching hospital reported a reduction in uninsured admissions from 16 percent to 3 percent,

attributable to HCIP.

The influence on the risk profile and competitiveness of the individual marketplace was substantive. Representing

84 percent of the covered lives within the individual health insurance marketplace, Medicaid’s premium

assistance lowered the average age of the risk pool(s) by approximately 10 years. The resulting more-favorable

risk within the Marketplace enabled stable premium prices not only in the first year, but across the first three

years of the HCIP. Competitiveness and consumer choice in the Marketplace increased across the seven market

regions in the state. In 2014, three regions had access to only Arkansas Blue Cross and Blue Shield and Blue Cross

Blue Shield Multi-State plans. By 2016, five carriers were offering coverage across all seven market regions, with

one market region having six carriers (the sixth restricted to a single market by Medicaid’s purchasing guidance

limiting premium assistance to those plans within 10 percent of the second lowest-cost silver plan within the

market region).

For 2014, the estimated budget neutrality cap (BNC) was exceeded during the initial enrollment phase of the

program. Enrollment of younger individuals over time affecting net premiums, the rebate of medical-loss ratio

(MLR) payments by one carrier not meeting the MLR requirements in 2014, and inflationary expectations built

into the BNC estimates brought cumulative program costs within the 2015 limit of $500.08 per member per

month (PMPM) and well under the 2016 limit of $523.58 PMPM. Importantly, the first year evaluation allowed

examination of BNC estimates compared to real experience.

Effect Comparison — Access

A key component in this evaluation is the comparison of QHP beneficiary experiences to similar cohorts of new

beneficiaries who were enrolled in traditional Medicaid. We employed two strategies: 1) a comparison of

individuals from a general population of new QHP enrollees or Medicaid (the General Population); and 2) a quasi-

experimental approach to individuals who had reported higher previous healthcare utilization and were assigned

to either QHPs or Medicaid (the Screened population). In general, findings observed in the two comparison

Arkansas Health Care Independence Program (‘Private Option’) Final Report 19

populations were consistent in direction, with the Screened population producing larger differences between

Medicaid and QHP enrollee experiences.

A key component of the demonstration was the degree to which Medicaid and QHP enrollees had access to

providers within their respective networks. Access was framed from three perspectives: 1) the geographic

presence of providers available to enrollees, 2) the experience of enrollees in attaining access at times of need,

and 3) variations in utilization observed between programs.

The location of providers in both the Medicaid and QHP network participation revealed high degrees of

geographic access and minimal variation between programs. Both Medicaid and commercial enrollees’ networks

contained providers who met network adequacy requirements (e.g., 30 minutes from a PCP, 60 minutes from a

specialist). More than 98 percent of enrollees in both Medicaid and QHPs had access to a PCP within a 30-minute

drive time. For specialists within the General Population comparison group, both programs achieved high levels of

geographic access, represented by more than 95 percent of enrollees having no more than a 60-minute drive time

from the enrollee’s home.

Two unexplained statistically significant access differences were observed with QHP enrollees having slightly

higher orthopedic access (98.3 percent of QHP versus 93.7 percent Medicaid enrollees within 60 minutes) and

Medicaid enrollees having slightly more oncological access (99.1 percent of Medicaid versus 95.0 percent of QHP

enrollees within 60 minutes). No meaningful differences were assigned to these statistical findings.

Geographic access assessments represent the Medicaid participating providers compared to commercial

participating providers for all carriers. Because of Arkansas’s “any willing provider” law,

requiring insurers to

allow any provider willing to accept terms for the class of providers into their networks, assessment across all

carriers was deemed appropriate.

By contrast, from the perspective of the beneficiary at times of need, significant differences were observed in

being able to access providers within the networks. Consistently across both the General Population and Screened

Population, enrollees reported improved access within QHPs. Responding to whether it was “always easy to get

care, tests, and treatment needed,” 64.5 percent of General Population QHP enrollees responded affirmatively

compared to 45.9 percent enrolled in Medicaid (a 40.5 percent relative difference). For individuals who

completed the screener, 57.9 percent of QHP enrollees compared to 48.4 percent of Medicaid enrollees

responded affirmatively (a 19.6 percent relative difference). With respect to getting “an appointment for a check-

up or routine care as soon as needed,” enrollees in QHPs reported more accessibility with a 12.1 percent relative

difference in the General Population. Improved accessibility was suggested for ease of appointment availability

for the General Population, with no differences observed for the Screened population.

Because this was the initiation year of the program and many of the newly enrolled lacked prior insurance

coverage, we examined the time to first outpatient visit in the General Population for Medicaid and QHP enrollees

and found significant differences. Within 30 days of enrollment, 21.2 percent of QHP enrollees had accessed an

outpatient visit compared to 8.2 percent of traditional Medicaid enrollees. By 90 days of enrollment, 41.8 percent

of QHP enrollees had accessed an outpatient visit compared to 29.6 percent of Medicaid enrollees. These

differences are dramatic and consistent with the perceived accessibility reported above from enrollees.

Finally, comparing utilization patterns for ER use and hospitalizations, the impact of access differences reported

previously are consistently observed for both the General Population and Screened Population. Examining the

rate of total ER visits per 12 months of enrollment, Medicaid enrollees experienced a 13.2 percent higher ER-visit

rate in the General Population and a 50.8 percent higher rate in the Screened Population.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 20

Differentiating between emergent and non-emergent ER visits by a modified NYU algorithm,

programmatic

differences were observed. QHP enrollees were much more likely to utilize ER services for emergent care — 122.1

percent for the General Population and 51.9 percent for the Screened Population. Conversely, non-emergent ER

services were much more likely to be utilized by the Medicaid enrollees — 58.1 percent higher for the General

Population and 63.6 percent higher for the Screened Population. These findings were highly statistically significant

across all ER comparisons. With respect to hospitalizations, the Screened Population demonstrated a 43.2 percent

higher hospitalization rate in the Medicaid program than in the QHP program, with no differences observed in the

General Population.

Considering network participation, self-reported perceived access, and patterns of utilization, a profile of the

differences between Medicaid and QHP program performance emerged. Enrollees were geographically located

near providers who have enrolled as a provider in Medicaid or contracted with a QHP to provide services.

Geographic access, however, does not equate to beneficiary accessibility. Published studies inclusive of Arkansas

providers during the time period reflected in this evaluation have found significant differences across multiple

states, and specifically Arkansas providers, in their acceptance of new patients privately covered by commercial

insurance compared to those with Medicaid coverage. In a 2013 study, Rhodes and colleagues estimated that

rates for new patient availability of appointments for commercial and Medicaid insurance scenarios in Arkansas

were 88.1 percent and 48.7 percent, respectively.

Survey information from this study indicated that fewer

practices were accepting patients with Medicaid coverage when compared with private payer sources. This

finding, combined with responses from practices indicating that their Medicaid patient population comprises less

than 10 percent of all patients, suggests that practices were likely limiting the number of patients with Medicaid

as the primary payer, particularly given that the Medicaid population comprises 26 percent of the state’s

population.

Combined with the time to first outpatient visit and non-emergent ER use rates, all of our findings

suggest individuals in the Medicaid program experience more difficulty accessing care when needed and

subsequently seek care in settings that are less likely to either address unmet needs or successfully establish

clinician-patient relationships to manage chronic conditions.

Effect Comparison — Care and Outcomes

Examination of quality indicators was undertaken in the first year to assess variations in healthcare quality or

outcomes. With a high proportion of the evaluation study population likely not having prior health insurance and

due to the time frame (first 12 months of coverage), assessments focused on the proportion of enrollees who

received appropriate clinical preventive screenings, the proportion of enrollees who received prophylaxis to

prevent influenza, and the proportion of individuals with diabetes who received appropriate management

screenings with an HbA1c and/or LDL-c screening test. Evaluation of a more robust set of quality indicators is

anticipated as person-time accumulates, which enables longer observation periods.

For receipt of preventive screenings, metrics were operationalized, both to compare between enrollee groups if

any clinical preventive screening was obtained and to compare if all recommended clinical preventive screenings

were obtained. In the first year, eligible screenings included the following: breast cancer, cervical cancer,

colorectal cancer, and cholesterol screenings. Across both comparison populations, enrollees in QHPs achieved

higher screening rates than their Medicaid counterparts. For any recommended screening event, the difference

was 29.6 percent to 25.8 percent in the General Population (a relative difference of 14.7 percent) and 66.0

percent to 41.2 percent (a relative difference of 60.2 percent) in the Screened Population. Although still

statistically significant, for those receiving all recommended screening events, the variation was less pronounced

with 16.8 percent versus 15.8 percent in the General Population and 23.2 percent versus 20.3 percent in the

Screened Population for QHP and Medicaid enrollees, respectively.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 21

With respect to receiving infection prevention through influenza prophylaxis (flu shot or spray), observed

differences in the Screened Population again favored enrollees in QHPs over those in Medicaid. A flu shot or nasal

spray was reported by 45.9 percent of enrollees in a QHP versus 38.5 percent in Medicaid (a 19.2 percent relative

difference). While directionally consistent, no statistically significant finding was present in the General

Population comparison.

Of importance, receiving a tertiary screening for HbA1c was more common among QHP enrollees than those

enrolled in Medicaid. In the General Population, QHP enrollees with diabetes received HbA1c assessments 79.1

percent of the time compared to 73.0 percent of the time for Medicaid enrollees with diabetes (an 8.4 percent

relative difference). Equally, in the Screened Population, QHP enrollees with diabetes received HbA1c

assessments 84.8 percent of the time compared to 80.4 percent of the time for Medicaid enrollees (a 5.5 percent

relative difference). No differences were observed for LDL-c screenings.

Ancillary findings reflecting the first year experience of individuals in both comparison groups included

assessments of experience with transportation needs and examination for longer-term health outcomes.

Significant findings were observed from reported transportation needs for individuals in the Screened Population.

QHP enrollees reported no transportation barriers to a personal doctor visit 89.4 percent of the time compared

with 80.4 percent of the time for Medicaid enrollees (an 11.2 percent relative difference). Transportation barriers

and access to specialty visits were not significantly different in the General Population. Preventable

hospitalizations, readmissions, and age-adjusted mortality showed no variation, likely due to this being the first

year of enrollee experiences.

In summarizing and interpreting care and outcomes, it appears that through more accessible and potentially

earlier engagement, the QHP enrollees experienced improved primary prevention (flu prophylaxis) and secondary

prevention (clinical screenings) compared with their Medicaid-enrollee counterparts.

First Year Program Observations

Differences between the costs of QHP enrollees and those managed through the Medicaid system were expected.

Exploration and characterization of cost differences were required to better understand their association with

effect differences in access, utilization, quality, and the outcomes described. Differences in payment rates and

utilization were also anticipated between Medicaid and the QHP carriers. Variations between QHP carriers are

also expected, but for the purpose of this evaluation, weighted averages of their experiences are utilized for the

commercial comparison.

From effect differences that were observed for comparable groups, overall program differences were suggested

and indeed observed. Examination of utilization rates for Medicaid and QHP enrollees with a minimum of six

months of coverage reinforced these findings.

Medicaid enrollees experienced fewer outpatient events and a concurrent higher rate of ER visits and

hospitalizations. Importantly, enrollees within QHPs received twice as many prescriptions than their Medicaid

counterparts. Because Medicaid utilizes different payment mechanisms and provider codes for select services

compared to their QHP counterparts, direct comparison of all services was not feasible. Volume and type of

service utilization have the potential to impact program costs and will be monitored over time to assess

convergence or divergence in experience.

While volume and type of service utilization is important, variation in payment rates and their potential impact on

access, care, and outcomes was a central component of the demonstration waiver justification. We examined

direct comparisons of payment differentials between that paid by Medicaid and by QHPs.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 22

Dramatic differences in payment rates were observed with commercial rates consistently exceeding those in the

Medicaid program. Physician rates for outpatient services were 90 percent higher when the enrollee was in a

QHP, as compared to their Medicaid counterparts. Primary care physicians had 90 percent higher payment rates

under commercial contracts than with the Medicaid payment schedule. Specialist payment differentials ranged

from 90 percent for obstetricians/gynecologists to 165 percent for ophthalmologists.

Higher payment rates for hospital services, as well as ER events, were also observed. For inpatient hospital stays,

average commercial payments were $11,894 per discharge compared to Medicaid payments (with supplemental

additions) of $7,778 — a 52.9 percent difference. For ER non-hospitalized visits, average commercial payments

were $598 per visit compared to Medicaid payments of $196 — a 205.1 percent difference.

Total program cost differences between that of the HCIP enrollees in QHPs and those managed through the

Medicaid system were expected. The cumulative weighted average premium that was paid during 2014 for

commercial premium assistance was $485.05 PMPM. For Medicaid expenditures inclusive of supplemental

payments and beneficiary-related administrative expenses, the observed Medicaid expenditures were $272.01

PMPM. This $213.04 PMPM difference represents a 78.3 percent difference between the commercial and

Medicaid PMPM. This difference largely reflects the variation in provider payments described, modified by

secondary variations in utilization. This is reflected in the modeling of Medicaid estimates based upon payment

rate differences alone ($244.96 PMPM) and in combination with utilization differences ($252.73 PMPM), resulting

in estimates within 9.9 percent and 7.1 percent of the observed, respectively.

The Counterfactual — Simulation of Traditional Medicaid Expansion in Arkansas

Examination of the hypothetical costs of covering the entire expansion population in Arkansas’s traditional

Medicaid program and the necessary programmatic changes necessary to achieve a similar effect outcome to that

experienced through premium assistance is a core component of the demonstration evaluation. Consideration

was given to the existing Medicaid program, its level of network participation, and impact of existing payment

rates given effect changes identified through this evaluation. In addition, the price elasticity of the supply of

medical providers and their ability and/or willingness to provide for healthcare needs of the expansion population

through the existing Medicaid program must be considered. Finally, if payment rate changes were required to

achieve access and quality outcomes, what would be the financial impact of those modifications across the entire

Medicaid program (e.g., rate changes would apply to all Medicaid rates and not only those associated with the

PPACA newly eligible)?

As previously described, Arkansas had one of the lowest Medicaid eligibility thresholds for non-disabled adults in the

United States (below 17 percent FPL for parent/caretakers only). This meant that the majority of the covered lives in

Medicaid were children, low-income Medicare beneficiaries receiving long-term services (not medical), Social

Security Income (SSI) disabled adults, and pregnant women and those with family-planning services with limited

benefit coverage. In 2013, prior to the PPACA expansion, Arkansas Medicaid covered 24,955 non-disabled adults with a

full benefit package. In 2014, following PPACA expansion, an additional 267,482 individuals were covered —

approximately 17,300 (6.5 percent) previously eligible but newly enrolled; approximately 25,000 (9.3 percent) PPACA

eligible but with exceptional healthcare needs; and 225,000 (84.2 percent) PPACA eligible with premiums purchased

on the individual marketplace. Thus, in 2014, Arkansas Medicaid expanded their non-disabled 19- to 64-year-old

population tenfold with 84 percent managed externally in the commercial marketplace. Effect comparisons

represented above draw on the experience of those newly eligible in either the Medicaid FFS or commercial

premium assistance programs during 2014.

Infusion of an additional 297,000 non-disabled 19- to 64-year-olds into the Medicaid system would likely have

resulted in widespread effects to the system. Traditional microeconomics suggests that increased demand

Arkansas Health Care Independence Program (‘Private Option’) Final Report 23

through the Medicaid program would place increasing price pressure on the rate structure of the existing

Medicaid program.

A recent study of appointment availability for Medicaid beneficiaries, inclusive of those in Arkansas, suggests that

increased Medicaid payments result in improved appointment availability.

In this 10-state study, an increase in

availability of primary care appointments of 1.3 percent was observed for each 10 percent increase in Medicaid

reimbursements. These findings are consistent with our first year findings internal to this evaluation, both for

ease of access for receiving needed care and for access with differential ease of receiving needed care, as well as

ease of appointment for routine care. Between Medicaid and QHP enrollees in the General Population, we

observed a 40.5 percent relative difference and for the Screened Population a 19.6 percent relative difference for

ease of access. Similarly, for ease in ability to get an appointment within the general population a 1.4 percent

difference was associated with a 10 percent increase in Medicaid reimbursements. Thus, theoretical, peer-

reviewed, and internal findings suggest upward price pressure on existing Medicaid payment rates in the

counterfactual that would be required to achieve comparable access and potential comparable outcomes to those

experienced in the commercial sector.

The observed differences in payment rates between Medicaid and QHPs described previously would plausibly

lead to increased access differences for Medicaid beneficiaries. As required by federal rule, it would be unlikely

that Arkansas could meet the equal access provision requiring state Medicaid provider payments to be

“consistent with efficiency, economy, and quality of care and ... sufficient to enlist enough providers so that care

and services are available under the plan at least to the extent that such care and services are available to the

general population in the geographic area.”

Importantly, any potential increase in Medicaid payment rates

would necessarily affect not only services for the new expansion population, but also services for beneficiaries

under the same payment rate schedule across the entire Medicaid program.

The simulated incremental effects of inflationary increases and the associated cost impacts were plausible. The

three increasingly conservative scenarios provide policymakers with conditions under which necessary increases

to achieve equitable access can be considered. They include: 1) claims associated with potentially wage-sensitive

services; 2) claims restricted only to those associated with major medical services; and 3) claims restricted to only

those associated with physician-billed services.

The base scenario utilized 2014 actual premiums paid for commercial coverage and observed Medicaid costs for

Medicaid coverage. Under the wage-sensitive scenario, the Medicaid program would achieve budget neutrality if

the Medicaid program experienced a 14.5 percent increase in costs. Under the major medical scenario, the

Medicaid program would achieve budget neutrality at a 24.7 percent increase in clinical-claims cost. Lastly,

restricting to a physician-only scenario, budget neutrality would be achieved at a 34.9 percent increase in

physician-claims costs. In actuality, the market would likely require payment modifications much more complex

than these scenarios. However, these scenarios provide policymakers with a comparison of budget neutrality

estimates based upon actual expenditures.

First year results should be viewed with caution for several reasons. First, cost-sharing reduction reconciliation

with carriers for 2014 had not been executed, and have still not been concluded. In addition, 2014 represented

the initiation phase of the program with significant transitions as reflected in enrollment growth. Future

assessments during steady state periods may provide more accurate reflections of both programmatic effects and

associated costs.

The subsequent sections in this report update findings over the second and third years of the HCIP that include a

period of programmatic steady state in 2016.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 24

Year 2 and Year 3 Waiver Impact Findings

Introduction to Analyses

Overview

While all individuals described in this section of the report are Medicaid-eligible individuals, the following groups

of individuals are categorized for the purposes of our findings:

i. Medicaid enrollees: Patient Protection and Affordable Care Act (PPACA) Medicaid expansion eligible

individuals with exceptional healthcare needs who are enrolled in the state’s fee-for-service Medicaid

program and are not a part of the Health Care Independence Program (HCIP) Section 1115 demonstration

waiver; and

ii. Qualified Health Plan (QHP) enrollees: PPACA Medicaid expansion eligible individuals who are enrolled in

QHPs through premium assistance and are a part of the HCIP Section 1115 demonstration waiver.

Provider Networks

Only providers who practiced within calendar year 2014 were included in these analyses. Medicaid and

commercial QHP enrollees and provider data (i.e., Blue Cross Blue Shield, Ambetter, and QualChoice) were used

to assess network adequacy and perform geographical information system (GIS) mapping. For the analysis,

“access” was defined as the distance (30 or 60 miles) or time of travel (30 or 60 minutes by passenger

automobile) from any enrollee’s location to the nearest provider, within an “access ring” offering any of the

eight categories of healthcare provider types (primary care, orthopedics, ophthalmology, OB/GYN, oncology,

surgical, psychology, and cardiology) under study. These “access rings” were created by computing a total of

four distinct travel distances and time-of-travel attributes from each unique provider location (latitude and

longitude coordinates) for each provider type. From each provider location a set of four miles traveled along

the existing (Environmental Systems Research Institute [ESRI], 2013) street centerline networks were computed

as “ringed-polygons” at the following distance rings: 0-15, 15-30, 30-45, and 45-60 miles. The percentage of

enrollees within a 30-minute travel time to an in-network primary care physician and a 60-minute travel time to

a specialist are compiled. A detailed description of the geospatial analysis is contained in Appendix G. Medicaid

in-network geographic access is compared to QHP in-network access where the three individual commercial

carriers have been aggregated.

Comparison Populations and Comparison Groups

Two comparison populations are available within the analytic framework. The first population compares traditional

Medicaid enrollees — who did not complete the exceptional healthcare needs Questionnaire — with QHP enrollees

who chose not to complete, or simply bypassed the Questionnaire upon applying to Medicaid. As previously stated,

enrollees in these insurance program groups are reflective of the General Population. For this comparison

population, we used enrollee first year program experience as a Baseline to ascertain a clinical needs background

(using the Charlson Comorbidity Index) and created individual propensity scores depicting the probability of being

assigned to a QHP given a set of demographic, clinical needs, and geographic covariates. Medicaid and QHP

enrollees with the same (or very similar) propensity scores were matched and differences in access, quality of care,

utilization, and other health outcomes across these matched pairs were empirically tested. For this population, the

first year program experience and comparison is contained in the Interim Report,

and this final report contains

Medicaid and QHP comparison group differences for coverage in Year 2 and Year 3 of enrollment.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 25

The second analytic population includes individuals who completed the Questionnaire. Those who were deemed

to have exceptional healthcare needs by virtue of attaining a composite score threshold cut-point were assigned

to Medicaid. All others completing the Questionnaire who did not attain the threshold were enrolled in a QHP.

Using a regression discontinuity approach, conclusions about differences between Medicaid and QHP enrollees in

this population were made within a model-derived optimal bandwidth around the threshold cut-point. Enrollees

with composite scores close together but on opposite sides of the cut-point, and thus assigned to different

coverage programs, are the focus of this comparison. Therefore, the comparison is reflective of a population who

completed the Questionnaire, and had at least a measure (close to and around the cut-point) of exceptional

healthcare needs. We called this a Higher Needs Population. For this population, the first year program

experience is a Baseline year and we present Medicaid and QHP comparison group differences for coverage in the

Baseline, and Years 2 and 3 of enrollment. The local average treatment effect (LATE) presented in the regression

discontinuity result tables represents the treatment effect at the discontinuity cut-point that was used to assign

enrollees to Medicaid or QHP programs.

In these analyses, there are two populations containing two comparison groups. Each of the four groups are

mutually exclusive and, in total, represent our complete analytical population (see Appendix D for inclusion and

exclusion criteria for each population over the three program years, 2014-2016).

For a more comprehensive description of the statistical designs used to determine the differences across

comparison groups in the two populations, as well as the technical details, please see the Methodological

Approaches section and Appendix E.

Analytic Approach to Access

For both comparison populations, access differences between the Medicaid and QHP programs were evaluated

using measures for geographic, perceived, and realized access. Other measures influenced by lack of primary care

physician access, including non-urgent emergency room utilization and potentially preventable hospital

admissions, are analyzed. Contributory factors to obtaining primary care physician appointments, and being able

to get to them when scheduled, are provider reimbursement rates and non-emergency transportation.

Differences in Medicaid and QHP reimbursement rates for office or other outpatient services by various provider

types and self-reported differences in the difficulty obtaining non-emergency transportation are highlighted in

this report.

Provider and enrollee populations did not incur any systematic changes in the second and third year of HCIP and,

thus, we present geographic access analyses compiled and presented in the Interim Report.

For the General

Population, this analysis was performed using a stabilized inverse probability of treatment weighting approach

and, for the Higher Needs Population, a regression discontinuity design was applied (see the description in

Appendix E).

For perceived access, there are new findings to present from responses to the second CAHPS survey (CAHPS II)

that was fielded in fall 2016. This survey elicited a perceived access perspective from individuals who were

continuously enrolled in Medicaid or a QHP for a period of approximately 18 months. In the Interim Report,

perceived access was framed based on an interpretation of access at program enrollment or shortly thereafter.

The interpretation in this report will shed light on whether perceived access differences persist after being in a

health insurance program for at least a year — when realized access of first contact with a provider should have

taken place and a familiar provider-patient relationship should have been established.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 26

Analytic Approach to Clinical Preventive Measures Outcomes

In the Interim Report, a series of clinical outcome differences was presented based on primary, secondary, and

tertiary screenings, as well as disease management outcomes that could be effectively measured within one year

of healthcare coverage. For this three-year synopsis of the HCIP program, we again report on these annual

screenings and disease management indicators. The three-year coverage period also allows us to compile

indicators that require more time to measure, such as cancer-screening indicators (i.e., breast, cervical, and

colorectal). A full list, descriptions, and operationalization of indicators included in these analyses can be found in

Appendix H.

Analytic Approach to Healthcare Services and Utilization Outcomes

We tested for differences in healthcare services outcomes (procedures) provided to Medicaid and QHP enrollees.

Not having a needed procedure performed could be an indication of poor quality of care, but at the same time, an

over-supply of unnecessary procedures could also be indicative of poor quality of care.

Based on age- and gender-appropriate inclusion, we present a series of healthcare services outcome rates across

Medicaid and QHP enrollee populations for abdominal hysterectomy, vaginal hysterectomy, cardiac

catheterization, percutaneous coronary intervention (PCI), coronary artery bypass graft, back surgery, open

cholecystectomy, laparoscopic cholecystectomy, prostatectomy, total hip replacement, total knee replacement,

carotid endarterectomy, lumpectomy, unilateral mastectomy, and bilateral mastectomy.

Quality utilization outcomes include not having preventable hospitalizations and non-emergent emergency room

visits. For those who do have hospitalizations, not having to be readmitted following discharge is another

indicator of quality healthcare delivery. We use an algorithm compiled by the Agency for Healthcare Quality and

Research to identify preventable hospitalizations and the modified New York University algorithm to identify non-

emergent emergency room visits.

Special Population and Topic Studies

In this section, we complement the access, clinical preventive measures outcomes, and healthcare services and

utilization program comparisons by studying areas of healthcare delivery where programs may differ in the

provision of quality health care.

The first special population under study was made up of pregnant women and looked at the care delivered in

each program. Women are not enrolled in a QHP if they are pregnant upon application — they are covered by the

traditional fee-for-service Medicaid. If they are diagnosed or become pregnant while enrolled in a QHP they

receive their perinatal care covered by the QHP. From 2014 onward, we have a reasonable population in which to

compare pregnancy and birth outcomes for Medicaid- and QHP-covered women.

In a low-income population, non-emergency transportation is an important factor contributing to gaining access

to the healthcare system. Using responses from the second CAHPS survey we present findings from a comparison

analyses of non-emergency transportation access.

Children who had been identified with chronic and persistent conditions through early and periodic screening,

diagnostic, and treatment benefits while in Medicaid were eligible for HCIP enrollment and a more

comprehensive set of healthcare benefits. In this study, we identify how many adolescents would have been

eligible to take advantage of HCIP enrollment and what level of care they received if they were enrolled in a QHP.

The availability and continuity of healthcare coverage is an important cornerstone to optimal health. Churn — or

gaps in healthcare coverage — has been demonstrated to have negative consequences on individual health,

Arkansas Health Care Independence Program (‘Private Option’) Final Report 27

especially for those with chronic conditions. We studied patterns of continuous enrollment, attrition, and churn

among those who enrolled in HCIP across various periods of the program.

Health Independence Accounts were introduced to the HCIP population, who were earning 101 to 138 percent of

the federal poverty level (FPL) in January 2015. A type of health savings account, the goal was to encourage

enrollees to pre-pay a small premium for next-month coinsurance protection. This study summarizes the buy-in,

impact, and overall effectiveness of the Healthcare Independence Account experience.

Utilization of urgent care clinics continues to increase in Arkansas. In this section, we test to see if emergency

room utilization has been substituted by urgent clinic care utilization to help explain changes in number and rates

of visits.

Given the high rates of opioid use observed in Arkansas statewide and the well documented risks associated with

high dose and long-term opioid use, we compared the rates of opioid analgesic use, high risk opioid use (high dose

opioid use or opioid use with a benzodiazepine), initial opioid use with >7 and >3 days supplied, and the extent of

naloxone use for those prescribed opioids between newly enrolled persons obtaining coverage in traditional

Medicaid and QHPs from 2014-16.

Finally, we present mortality comparison between Medicaid and QHP enrollee groups in both the General and

Higher Needs Populations.

Cost Comparison and Budget Impact Analyses

To determine what cost differentials were present to treat similar patients in Medicaid and QHPs, we identified

per member per month costs associated with the care provided to Medicaid and QHP enrollees. Using a budget

impact analysis simulation model, we used these estimates to present cost-enhancing scenarios to the

Medicaid reimbursement structure to identify how much Medicaid could have increased costs to stay within

HCIP budget neutrality. The goal was to increase Medicaid reimbursement to induce provider engagement to

accept more patients, and in a timelier manner, in order to improve access to care under a counterfactual

scenario that asked, in essence, “What would have happened if adult Medicaid expansion had proceeded in a

standard manner in Arkansas?”

Geographic and Realized Access Differences and Provider Reimbursement

Comparison between Medicaid and QHP Programs

Network Participation (Geographic Access)

Tables 2 and 3 contain findings from a geospatial analysis identifying the proportion of 2014 Medicaid and QHP

General and Higher Needs Population enrollees, respectively, who resided within a 30-minute drive to an in-

network primary care physician (PCP) or within a 60-minute drive to each of seven in-network specialists under

the study.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 28

Table 2. Differences in General Population Geographic Access to Health Care between Medicaid and

QHP Enrollees

Geographic Access Indicators

Comparison

Medicaid

QHP

Relative

Difference

(percent

)

Statistical

Difference

(p-value

)

Proportion of enrollees within 30 minutes

of a primary care physician

(source: GIS)

Crude (n, proportion)

9,604 (0.993)

61,918

(0.986)

Adjusted (LSM, StdErr)

0.994 (0.001)

0.986 (0.003)

- 0.8%

0.494

Proportion of enrollees within 60 minutes

of a cardiologist

(source: GIS)

Crude (n, proportion)

9,604 (0.994)

61,918

(0.977)

Adjusted (LSM, StdErr)

0.995 (0.001)

0.977 (0.001)

- 1.8%

0.116

Proportion of enrollees within 60 minutes

of an obstetrician/gynecologist

(source: GIS)

Crude (n, proportion)

6,541 (0.993)

34,852

(0.990)

Adjusted (LSM, StdErr)

0.994 (0.001)

0.990 (0.001)

- 0.4%

0.743

Proportion of enrollees within 60 minutes

of a psychiatrist

(source: GIS)

Crude (n, proportion)

9,604 (0.994)

61,918

(0.995)

Adjusted (LSM, StdErr)

0.995 (0.001)

0.995 (0.000)

0.0%

0.968

Proportion of enrollees within 60 minutes

of an orthopedist

(source: GIS)

Crude (n, proportion)

9,604 (0.937)

61,918

(0.983)

Adjusted (LSM, StdErr)

0.937 (0.003)

0.983 (0.001)

4.9%

<0.001

Proportion of enrollees within 60 minutes

of an ophthalmologist

(source: GIS)

Crude (n, proportion)

9,604 (0.993)

61,918

(0.979)

Adjusted (LSM, StdErr)

0.995 (0.001)

0.978 (0.001)

- 1.7%

0.140

Proportion of enrollees within 60 minutes

of an oncologist

(source: GIS)

Crude (n, proportion)

9,604 (0.990)

61,918

(0.950)

Adjusted (LSM, StdErr)

0.991 (0.001)

0.950 (0.009)

- 4.1%

<0.001

Proportion of enrollees within 60 minutes

of a general surgeon

(source: GIS)

Crude (n, proportion)

9,604 (0.994)

61,918

(0.995)

Adjusted (LSM, StdErr)

0.995 (0.001)

0.995 (0.000)

0.0%

0.984

Notes: Analysis derived from geospatial database containing travel time between patients and physicians. Relative percentage

calculated as (QHP – Medicaid)/Medicaid x 100. Due to lack of geographic access being a rare event, p-values for differences were

obtained using a Poisson regression.

Abbreviations: n=number of persons; StdErr=standard error of estimated mean.

Results contained in Table 2 assessing network adequacy through geographic proximity of providers reveal

minimal differences between providers accepting Medicaid enrollees and those participating in the

commercial QHP networks. There was no difference in the proportion of Medicaid and QHP enrollees in the

General Population within 30 minutes of a primary care physician and the proportion of enrollees within 60

minutes of most specialists. Minor, but statistically significant differences, were observed for orthopedists

and for oncologists. A higher proportion of QHP enrollees have access to orthopedists when compared to

Medicaid enrollees, with a relative difference of 4.9 percent. Conversely, a higher proportion of Medicaid

enrollees have access to oncologists when compared to QHP enrollees, with a difference of 4.1 percent.

Except for these minor differences, both the commercial and Medicaid networks met the geographic access

standards of AID.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 29

Table 3. Differences in Higher Needs Population Geographic Access to Health Care between Medicaid

and QHP Enrollees

Geographic Access Indicators

Comparison

Medicaid

QHP

Relative

Difference

(percent)

Statistical

Difference

(p-value

)

Proportion of enrollees within 30 minutes

of a primary care physician

(source: GIS)

Crude (n, proportion)

4,487 (0.996)

25,023

(0.976)

Adjusted (LSM, StdErr)

0.998 (0.001)

0.971 (0.007)

- 2.7%

0.578

Proportion of enrollees within 60 minutes

of a cardiologist

(source: GIS)

Crude (n, proportion)

4,489 (0.996)

24,214

(0.944)

Adjusted (LSM, StdErr)

0.997 (0.002)

0.953 (0.009)

- 4.4%

0.100

Proportion of enrollees within 60 minutes

of an obstetrician/gynecologist

(source: GIS)

Crude (n, proportion)

4,491 (0.997)

24,947

(0.973)

Adjusted (LSM, StdErr)

0.998 (0.001)

0.976 (0.003)

- 2.2%

0.466

Proportion of enrollees within 60 minutes

of a psychiatrist

(source: GIS)

Crude (n, proportion)

4,491 (0.997)

25,365

(0.989)

Adjusted (LSM, StdErr)

0.999 (0.001)

0.984 (0.006)

- 1.5%

0.771

Proportion of enrollees within 60 minutes

of an orthopedist

(source: GIS)

Crude (n, proportion)

3,886 (0.863)

24,778

(0.966)

Adjusted (LSM, StdErr)

0.867 (0.007)

0.965 (0.002)

11.3%

0.001

Proportion of enrollees within 60 minutes

of an ophthalmologist

(source: GIS)

Crude (n, proportion)

4,487 (0.996)

24,449

(0.953)

Adjusted (LSM, StdErr)

0.997 (0.001)

0.954 (0.004)

- 4.3%

0.147

Proportion of enrollees within 60 minutes

of an oncologist

(source: GIS)

Crude (n, proportion)

4,453 (0.989)

22,999

(0.897)

Adjusted (LSM, StdErr)

0.989 (0.002)

0.888 (0.006)

- 10.2%

0.001

Proportion of enrollees within 60 minutes

of a general surgeon

(source: GIS)

Crude (n, proportion)

4,491 (0.997)

25,415

(0.991)

Adjusted (LSM, StdErr)

0.998 (0.001)

0.993 (0.001)

- 0.5%

0.872

Notes: Analysis derived from geospatial database containing travel time between patients and physicians. Relative percentage

calculated as (QHP – Medicaid)/Medicaid x 100. Due to lack of geographic access being a rare event, p-values for differences were

obtained using a Poisson regression.

Abbreviations: n=number of persons; StdErr=standard error of estimated mean.

Geographic access differences for the Higher Needs Population are the same as what were identified within

the General Population. Notably, a higher proportion of QHP enrollees have access to orthopedists compared

to Medicaid enrollees, with a relative difference of 11.3 percent. Conversely, a higher proportion of Medicaid

enrollees have access to oncologists compared to commercial QHP enrollees, with a relative difference of

10.2 percent.

Provider Payment Differentials

Table 4 presents a summary of weighted average prices paid to different provider types in 2014, 2015, and

2016. The formal description of the aggregation and estimation of weighted average price is contained in

Appendix H.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 30

Table 4. Medicaid and Commercial Payer Price Differences for Outpatient Procedures by Provider Type

2014

2015

2016

Weighted

Average Price

Weighted

Average Price

Weighted

Average Price

Provider Type

MCD

QHP

Absolute

(Relative)

Diff

MCD

QHP

Absolute

(Relative)

Diff

MCD

QHP

Absolute

(Relative)

Diff

Primary Care

Physician

$48.13

$96.23

$48.10

(99.9%)

$47.61

$92.40

$44.79

(94.1%)

$47.69

$94.03

$46.33

(97.1%)

Advanced Practice

Nurses (APN)

$33.95

$63.45

$29.50

(86.9%)

$33.17

$65.41

$32.24

(97.2%)

$34.62

$65.62

$31.00

(89.5%)

Cardiologists

$44.68

$75.76

$31.08

(69.6%)

$66.61

$117.03

$50.42

(75.7%)

$64.68

$115.41

$50.73

(78.4%)

General Surgery

$58.74

$103.30

$44.56

(75.9%)

$57.83

$104.68

$46.85

(81.0%)

$58.65

$105.81

$47.16

(80.4%)

Obstetrician /

Gynecologist

$33.78

$87.59

$53.81

(159.3%)

$37.78

$90.92

$53.15

(140.7%)

$31.46

$100.94

$69.49

(220.9%)

Oncologist

$64.02

$108.32

($44.30)

(69.2%)

$62.56

$110.60

($48.04)

(76.8%)

$62.94

$109.14

$46.20

(73.4%)

Ophthalmologists

$84.04

$104.65

$20.61

(24.5%)

$57.12

$106.08

$48.96

(85.7%)

$57.37

$107.56

($50.19)

(87.5%)

Orthopedists

$53.50

$97.04

$43.54

(81.4%)

$51.21

$96.13

$44.92

(87.7%)

$51.51

$96.17

$44.66

(86.7%)

Psychologists /

Psychiatrists

$49.94

$83.91

$33.97

(68.0%)

$54.50

$88.95

$34.45

(63.2%)

$56.09

$92.46

$36.38

(64.9%)

Notes: Weighted Medicaid and QHP Average Prices were based on the most common CPT procedures billed for outpatient (non-ER)

services. Only CPT procedures that were represented both in Medicaid and QHP claims are included in the weighted averages. Relative

difference percentage calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: MCD=Medicaid; QHP=Qualified Health Plans; Diff=Difference

For all provider types, QHPs reimburse identical procedures at higher dollar rates than Medicaid. In 2014, average

higher QHP remuneration relative differences ranged from 24.5 percent for ophthalmologists to 159.3 percent for

obstetrician/gynecologist providers. In 2015, average higher QHP remuneration relative differences ranged from

63.2 percent for psychologists/psychiatrists to 140.7 percent for obstetrician/gynecologist providers. In 2016,

average higher QHP remuneration relative differences ranged from 64.9 percent for psychologists/psychiatrists to

220.9 percent for obstetrician/gynecologist providers.

Time to First Visit at Program Initiation

In 2014, realized access differences in the time it took for those enrolled in Medicaid and QHPs to have their first

contact with the healthcare delivery system were identified from a Kaplan-Meier survival analysis (see Figure 8).

This is likely still the best representation of programmatic realized access, as the denominator included those who

were continuously enrolled from program start-up in January 2014 through the end of December 2014. Those

who enrolled after January 2014 may have had initial contact with the healthcare system that was confounded by

seasonal demand for health care.

In 2014, it was identified that more than twice as many QHP enrollees (21.2 percent) had a healthcare visit in the

first 30 days compared to new Medicaid enrollees (8.2 percent). By 90 days, 41.8 percent of those enrolled in a

QHP had at least one outpatient care visit, compared to 29.6 percent of those enrolled in traditional Medicaid.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 31

Figure 8. Proportion of Medicaid and QHP Enrollees with a First Outpatient Care Visit, by Day

Differences in Comparison Groups by Comparison Populations

General Population

Perceived Access

Table 5 presents a comparison of perceived access indicator differences between Medicaid and QHP enrollees in

the General Population. Responses are based on those who completed the CAHPS II survey and with a one-to-one

propensity score matched population of Medicaid and QHP enrollees.

Cumulative Probability of Outpatient Visit

Time to Event Duration: First Outpatient Visit (in Days)

Medicaid

Commercial QHP

% of Enrollees with Visit by

30 Days

Medicaid 8.2%

Commercial QHP 21.2%

% of Enrollees with Visit by

90 Days

Medicaid 29.6%

Commercial QHP 41.8%

Arkansas Health Care Independence Program (‘Private Option’) Final Report 32

Table 5. Differences in Perceived Access to Health Care between Medicaid and QHP Enrollees

(Propensity Score Matched Comparison)

Perceived Access Indicators

Matched N

Medicaid

Mean (StdErr)

QHP

Mean

(StdErr)

Relative

Difference

(percent)

Statistical

Difference

(p-value)

Percentage of enrollees who always received care when

it was needed right away

52.6 (5.1)

66.3 (4.8)

26.0

0.057

Percentage of enrollees who always got an appointment

for a check-up or routine care as soon as needed

149

57.7 (4.1)

59.7 (4.0)

3.5

0.702

Percentage of enrollees who always got an appointment

to a specialist as soon as needed

54.6 (5.1)

48.5 (5.5)

-9.4

0.474

Percentage of enrollees who always found it easy to get

the care, tests, and treatment needed

171

50.3 (3.8)

57.3 (3.8)

14.0

0.195

Notes: Data were obtained from CAHPS II responses. Adjusted analysis was performed using a propensity score matched sample.

Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: Matched N=number of Medicaid and QHP matched enrollee respondents; StdErr=standard error of estimated mean.

When Medicaid and QHP enrollees were matched by propensity score, no differences were observed across any

of the four perceived access indicators.

Primary Preventive Screenings

Receipt of a flu shot or spray is an indicator of quality primary preventive care. Table 6 shows results from

comparing all General Population Medicaid and QHP enrollees responding to CAHPS II and then restricting

analysis to a Medicaid and QHP propensity score matched population.

Table 6. Differences in Primary Preventive Health Care between Medicaid and QHP Enrollees

(Propensity Score Matched Comparison)

Primary

Preventive Healthcare

Indicators

Period

Medicaid

Mean (StdErr)

QHP

Mean (StdErr)

Relative

Difference

(percent)

Statistical

Difference

(p-value)

Percentage of enrollees

receiving flu shot or spray

(primary)

All

Respondents

241 (Medicaid)

792 (QHP)

33.7 (3.1)

36.1 (1.7)

6.9

0.240

Propensity

Score Matched

216

34.3 (3.2)

43.1 (3.4)

25.7

0.063

Notes: Receipt of flu shot or spray was obtained from CAHPS II responses. Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: StdErr=standard error of estimated mean.

There were no differences in the percentages of Medicaid and QHP enrollees who received a flu shot or spray in

the year prior to completing the CAHPS II.

Secondary and Tertiary Preventive Screenings

Secondary and tertiary screenings for medical conditions are also a strong sign of quality preventive care.

Table 7 presents a series of screening comparisons between propensity score matched Medicaid and QHP

enrollees.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 33

Table 7. Differences in Secondary and Tertiary Preventive Health Care between Medicaid and QHP

Enrollees (Propensity Score Matched Comparison)

Secondary and Tertiary

Preventive Healthcare

Indicators

Time Period

Matched

Medicaid

Mean (StdErr)

QHP

Mean (StdErr)

Relative

Difference

(percent)

Statistical

Difference

(p-value)

Yearly Assessments

Percentage of chlamydia

screening (secondary)

Year 2

1,524

36.9 (1.2)

37.5 (1.2)

1.8

0.707

Year 3

610

44.4 (2.0)

45.2 (2.0)

1.8

0.773

Percentage of cholesterol

screening (secondary)

Year 2

9,463

18.8 (0.4)

28.3 (0.5)

50.6

<0.001

Year 3

4,741

17.1 (0.5)

30.8 (0.7)

80.0

<0.001

Percentage of enrollees with

diabetes with evidence of

HbA1C assessment (tertiary)

Year 2

1,156

62.7 (1.4)

82.5 (1.1)

31.6

<0.001

Year 3

1,118

32.9 (1.4)

51.2 (1.5)

55.4

<0.001

Periodic Assessments (if age and gender appropriate, eligible for at least one screening over 3 years)

Percentage of cervical cancer

screening (secondary)

3 years

6,771

27.9 (0.5)

34.9 (0.6)

25.1

<0.001

Percentage of breast cancer

screening (secondary)

3 years

984

16.6 (1.8)

24.2 (1.4)

46.6

<0.001

Percentage of colorectal

screening (secondary)

3 years

647

11.6 (1.3)

22.6 (1.6)

94.7

<0.001

Composite Assessments (if eligible for at least one screening)

Percentage of enrollees who

received at least one eligible

screening (secondary)

3 years

16,902

30.2 (0.4)

39.8 (0.4)

31.8

<0.001

Percentage of enrollees who

received at least half of eligible

screening (secondary)

3 years

16,902

19.9 (0.3)

26.1 (0.3)

30.9

<0.001

Percentage of enrollees who

received all eligible screenings

(secondary)

3 years

16,902

16.9 (0.3)

21.1 (0.3)

24.9

<0.001

Notes: All quality and screening indicators were derived from claims data. Adjusted analysis was performed using a propensity score

matched sample. Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: Matched N=number of Medicaid and QHP matched enrollees; StdErr=standard error of estimated mean.

For every secondary and tertiary screening indicator presented in Table 7, with the exception of chlamydia

screening, QHP enrollees received higher rates of screening than Medicaid enrollees.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 34

Disease Management

Table 8 presents a propensity score matched Medicaid and QHP General Population comparison of treatments and

condition management consistent with quality of care delivery. Diabetic HbA1C assessment is included again in this

section because it is both a condition that is screened for, but also a quality indicator of disease management.

Table 8. Differences in the Percentage of Medicaid and QHP Enrollees Receiving Recommended

Disease Management (Propensity Score Matched Comparison)

Preventable Utilization

Indicators

Time Period

Matched

Medicaid

Mean (StdErr)

QHP

Mean (StdErr)

Relative

Difference

(percent)

Statistical

Difference

(p-value)

Percentage of enrollees with

diabetes with evidence of

HbA1C assessment (tertiary)

Year 2

1,156

62.7 (1.4)

82.5 (1.1)

31.6

<0.001

Year 3

1,118

32.9 (1.4)

51.2 (1.5)

55.4

<0.001

Anti-depressant management

3 years

866

43.2 (1.7)

58.2 (1.7)

34.8

<0.001

Rate of mental/behavioral

disorder inpatient

hospitalizations per 10,000

person-years

Year 2

27,822

193.4 (10.0)

202.4 (10.4)

4.7

0.461

Year 3

13,577

172.0 (15.0)

260.2 (17.4)

51.2

<0.001

Proportion of enrollees with any

re-engagement with outpatient

care per 100 discharges

Year 2

1,693

60.4 (1.6)

65.3 (1.6)

8.0

0.044

Year 3

625

58.6 (2.5)

65.8 (3.0)

12.4

0.054

Notes: All hospitalization indicators derived from claims data. Adjusted analysis was performed using a propensity score matched

sample. Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: Matched N=number of Medicaid and QHP matched enrollees; StdErr=standard error of estimated mean.

QHP enrollees with diabetes had a higher relative percentage in Year 2 (31.6 percent) and in Year 3 (55.4 percent),

of having an HbA1c assessment compared to Medicaid enrollees in these coverage years.

Compared to Medicaid enrollees, 34.8 percent more QHP enrollees (58.2 percent compared to 43.2 percent)

received anti-depressant management over the course of the three years under study.

In the third year of health insurance coverage, compared to those covered by Medicaid, those covered by a

QHP had a higher rate of mental/behavioral disorder hospitalizations (260.2 compared to 172.0 per 10,000

person-years).

Compared to Medicaid enrollees at 60.4 percent in Year 2, 65.3 percent of QHP enrollees had re-engagement with

outpatient care following hospital discharge.

Use of Healthcare Services

Table 9 presents propensity score matched Medicaid and QHP healthcare services and procedure differences for

enrollees entered into the second and third years of coverage.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 35

Table 9. Differences in the Use of Healthcare Services between Medicaid and QHP Enrollees

(Propensity Score Matched Comparison)

Use of Healthcare Services

Indicators

Time Period

Medicaid

Mean (StdErr)

QHP

Mean (StdErr)

Relative

Difference

(percent)

Statistical

Difference

(p-value)

Rate of abdominal

hysterectomy per 10,000

person-years

Year 2

17,633

34.7 (4.5)

46.4 (5.2)

33.7

0.097

Year 3

8,706

30.5 (6.3)

48.8 (8.0)

60.1

0.079

Rate of vaginal hysterectomy

per 10,000 person-years

Year 2

17,633

32.4 (4.7)

42.8 (5.2)

32.4

0.120

Year 3

8,706

45.0 (8.4)

29.0 (6.5)

-35.6

0.108

Rate of cardiac catheterization

per 10,000 person-years

Year 2

27,822

69.2 (5.4)

63.3 (5.1)

-8.6

0.362

Year 3

13,577

75.2 (8.7)

82.3 (8.6)

9.4

0.584

Rate of PCI per 10,000 person-

years

Year 2

27,822

31.8 (4.8)

30.1 (4.2)

-5.3

0.678

Year 3

13,577

14.7 (4.4)

21.6 (4.7)

47.1

0.231

Rate of CABG per 10,000

person-years

Year 2

27,822

6.0 (1.5)

5.0 (1.4)

-17.2

0.599

Year 3

13,577

6.1 (2.3)

1.7 (1.2)

-72.5

0.103

Rate of back surgery per 10,000

person-years

Year 2

27,822

53.4 (4.7)

50.5 (4.9)

5.7

0.672

Year 3

13,577

46.7 (6.6)

53.2 (7.1)

14.0

0.501

Rate of open cholecystectomy

per 10,000 person-years

Year 2

27,822

3.7 (1.2)

1.5 (0.8)

-59.2

0.129

Year 3

13,577

1.7 (1.2)

2.5 (1.9)

44.2

0.700

Rate of laparoscopic

cholecystectomy per 10,000

person-years

Year 2

27,822

72.2 (5.4)

97.2 (6.3)

34.6

0.002

Year 3

13,577

66.6 (7.8)

90.6 (8.6)

36.1

0.036

Rate of total hip replacement

per 10,000 person-years

Year 2

27,822

7.9 (1.9)

8.4 (1.9)

6.8

0.844

Year 3

13,577

5.2 (2.4)

3.4 (1.7)

-35.9

0.482

Rate of total knee replacement

per 10,000 person-years

Year 2

27,822

16.1 (2.7)

12.6 (2.2)

-21.8

0.278

Year 3

13,577

10.4 (2.6)

13.1 (3.2)

36.2

0.425

Rate of lumpectomy per 10,000

person-years

Year 2

17,633

14.7 (3.1)

24.4 (4.0)

65.9

0.047

Year 3

8,706

21.2 (5.3)

23.7 (5.9)

11.9

0.743

Rate of unilateral mastectomy

per 10,000 person-years

Year 2

17,633

4.1 (1.8)

4.8 (1.9)

15.6

0.781

Year 3

8,706

5.3 (3.2)

4.0 (2.2)

-25.4

0.697

Notes: All healthcare service indicators derived from claims data. Adjusted analysis was performed using a propensity score matched

sample. Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100. Prostatectomy, carotid endarterectomy, and bilateral

mastectomy indicators also tested with no significant differences between Medicaid and QHP enrollees.

Abbreviations: Matched N=number of Medicaid and QHP matched enrollees; StdErr=standard error of estimated mean.

With 97.2 procedures per 10,000 person-years in Year 2 of coverage and 90.6 procedures per 10,000 person-

years in Year 3 of coverage, second- and third-year enrollees in a QHP received more laparoscopic

cholecystectomies than Medicaid enrollees with 72.2 and 66.6 procedures per 10,000 person-years,

respectively, over the same coverage periods. Female QHP enrollees into Year 2 of coverage also have more

lumpectomies per 10,000 person-years (24.4) than Medicaid enrollees (14.7) covered for an equal period of

time. No other healthcare services rate differences were found between Medicaid and QHP enrollees in their

second or third year of coverage.

Table 10 shows hospital utilization differences between propensity score matched Medicaid and QHP enrollees

who were into Years 2 and 3 of insurance coverage with the programs.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 36

Table 10. Differences in Rates of Preventable Hospitalizations and Readmissions between Medicaid

and QHP Enrollees (Propensity Score Matched Comparison)

Preventable Utilization

Indicators

Time Period

Matched

Medicaid

Mean (StdErr)

QHP

Mean (StdErr)

Relative

Difference

(percent)

Statistical

Difference

(p-value)

Rate of medical/surgery

hospital discharges per 10,000

person-years

Year 2

27,822

757.7 (23.0)

643.0 (22.2)

-15.1

<0.001

Year 3

13,577

663.0 (28.3)

675.0 (34.2)

1.8

0.773

Rate of preventable

hospitalizations per 10,000

person-years

Year 2

27,822

102.5 (8.2)

97.6 (9.5)

-4.8

0.536

Year 3

13,577

91.6 (13.2)

104.7 (18.4)

14.3

0.325

Average length of stay in days

for all hospitalizations

Year 2

2,463

4.7 (0.2)

6.2 (0.2)

32.6

<0.001

Year 3

946

4.4 (0.2)

6.2 (0.4)

39.2

<0.001

Average length of stay in days

for medical/surgery

hospitalizations

Year 2

1,344

5.1 (0.3)

5.7 (0.3)

11.5

<0.001

Year 3

525

4.6 (0.2)

5.7 (0.4)

24.8

<0.001

Proportion of enrollees with any

all-cause 30-day readmission

per 100 hospitalizations

Year 2

1,593

11.8 (1.0)

13.7 (1.1)

15.7

0.085

Year 3

625

14.3 (2.8)

15.3 (2.5)

7.1

0.581

Notes: All hospitalization indicators derived from claims data. Adjusted analysis was performed using a propensity score matched

sample. Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: Matched N=number of Medicaid and QHP matched enrollees; StdErr=standard error of estimated mean.

Compared to QHP enrollees, Medicaid enrollees in Year 2 of insurance coverage had 15.1 percent more

hospitalizations per 10,000 person-years (757.7 compared to 643.0 discharges per 10,000 person-years).

However, QHP enrollees who had hospitalizations spent more time, on average, in the hospital than Medicaid

enrollees. For those in the second year of insurance coverage, the difference was an average of 1.5 extra days, or

a 32.6 percent greater length of stay. For those in the third year of insurance coverage, the difference was an

average of 1.8 extra days, or a 39.2 percent greater length of stay. No differences between Medicaid and QHP

enrollees were observed for the rate of preventable hospitalizations or all-cause 30-day readmission rates.

Table 11 contains propensity score matched Medicaid and QHP enrollee difference for the rate of total, emergent,

and non-emergent emergency room (ER) utilization.

Table 11. Differences in Utilization of Emergency Room Services between Medicaid and QHP Enrollees

(Propensity Score Matched Comparison)

Emergency Room Indicators

Time Period

Matched

Medicaid

Mean (StdErr)

QHP

Mean (StdErr)

Relative

Difference

(percent)

Statistical

Difference

(p-value)

Rate of total emergency room

visits per 100 person-years

Year 2

27,822

82.4 (1.2)

73.5 (1.0)

-10.8

<0.001

Year 3

13,577

90.3 (1.9)

87.7 (1.7)

-2.9

0.041

Rate of emergent emergency

room visits per 100 person-

years

Year 2

27,822

14.5 (0.4)

12.7 (0.3)

-12.4

<0.001

Year 3

13,577

15.8 (0.6)

14.6 (0.4)

-7.5

0.020

Rate of non-emergent

emergency room visits per 100

person-years

Year 2

27,822

41.6 (0.7)

37.1 (0.6)

-10.8

<0.001

Year 3

13,577

45.5 (1.1)

44.2 (1.0)

-2.8

0.176

Notes: All emergency room indicators derived from claims data. Definitions for emergent and non-emergent care obtained from the

updated NYU algorithm (2017). Adjusted analysis was performed using a propensity score matched sample. Relative percentage calculated

as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: Matched N=number of Medicaid and QHP matched enrollees; StdErr=standard error of estimated mean.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 37

Medicaid enrollees in the second year of coverage had 10.8 percent more total ER visits than QHP enrollees. The

rate difference between Medicaid and QHP enrollees drops to 2.9 percent, but still significant, when comparing

those enrolled in the third year of coverage. When restricting analysis only to emergent emergency room visits,

Medicaid enrollees in the second and third years of coverage had 12.4 and 7.5 percent more visits, respectively,

than QHP enrollees. Compared to QHP enrollees in the second year of coverage, Medicaid enrollees experienced

10.8 percent more non-emergent ER visits. There is no Year 3 difference for non-emergent emergency room visits

between Medicaid and QHP propensity score matched enrollees.

Higher Needs Population

For the Medicaid and QHP statistical comparisons in this section, frequencies and rates for the full subgroup

populations are depicted along with the regression discontinuity local average treatment effects (LATE) for those

with an exceptional healthcare needs assessment composite score within a model-derived optimal bandwidth

around the threshold cut-point. The experience for those around the cut-point may differ from the comparison of

subgroup populations overall and hence effects may be reversed. Significant differences are concluded between

Medicaid and QHP subgroup populations for p-values associated with a LATE of 0.05 or less and the LATE

percentage is described in the results. Negative LATE values indicate that QHP enrollees had higher percentages

or rates compared to Medicaid enrollees.

Perceived Access

Table 12 presents a comparison of perceived access differences between Medicaid and QHP enrollees in the

Higher Needs Population. Responses are based on those who completed the CAHPS II survey.

Table 12. Differences in Perceived Access to Health Care between Medicaid and QHP Enrollees

(Regression Discontinuity Comparison)

Medicaid

QHP

Relative

Difference

(percent)

Average

Treatment

Effect

(percent)

Statistical

Difference

(p-value)

Perceived and Realized Access Indicators

Mean

(StdErr)

Mean

(StdErr)

Percentage of enrollees who always received care

when it was needed right away

537

52.9

(2.2)

505

72.3

(2.0)

36.7

-26.6

0.004

Percentage of enrollees who always got an

appointment for a check-up or routine care as

soon as needed

788

55.2

(1.8)

833

60.4

(1.7)

9.4

1.7

0.790

Percentage of enrollees who always got an

appointment with a specialist as soon as needed

564

52.5

(2.1)

581

55.4

(2.1)

5.6

6.2

0.346

Percentage of enrollees who always found it easy

to get the care, tests, and treatment needed

886

47.2

(1.7)

956

62.0

(1.6)

31.5

-18.7

0.003

Notes: Data were obtained from CAHPS II responses. Adjusted analysis was performed using regression discontinuity with a model-

derived optimal bandwidth around an exceptional healthcare needs composite score cut-point. Average treatment effect depicts

model-estimated difference at cut-point. Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: N=number of Medicaid and QHP enrollees; StdErr=standard error of estimated mean.

Compared to Medicaid enrollees, an estimated 26.6 percent more QHP enrollees indicated that they always

received care when it was needed right away, and an estimated 18.7 percent more QHP enrollees indicated that

they always found it easy to get the care, tests, and treatment they needed.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 38

Primary Preventive Screenings

Receipt of a flu shot or spray is an indicator of quality preventive care. Table 13 presents the results from

comparing all Higher Needs Population Medicaid and QHP respondents to CAHPS II and then restricting to a

regression discontinuity model-derived optimal bandwidth around an exceptional healthcare needs composite

score cut-point.

Table 13. Differences in Primary, Secondary, and Tertiary Preventive Health Care between Medicaid

and QHP Enrollees (Regression Discontinuity Comparison

Primary/Secondary/Tertiary

Preventive Healthcare

Indicators

Medicaid

QHP

Relative

Difference

(Percent)

Average

Treatment

Effect

(Percent)

Statistical

Difference

(p-value)

Period

Mean

(StdErr)

Mean (StdErr)

Percentage of enrollees

receiving flu shot or spray

(primary)

CAHPS II

Crude

967

40.7 (1.6)

1,080

42.2 (1.5)

3.5

0.064

CAHPS II

985

40.3 (1.6)

1,091

44.0 (1.5)

9.2

-22.2

0.001

Notes: Receipt of flu shot or spray, which was derived from CAHPS II. Adjusted analysis was performed using regression discontinuity

with a model-derived optimal bandwidth around an exceptional healthcare needs composite score cut-point. Average treatment effect

depicts model-estimated difference at cut-point. Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: N=number of Medicaid and QHP enrollees; RD=Regression discontinuity; StdErr=standard error of estimated mean.

Using the full Higher Needs Population (Crude) who responded to the CAHPS II survey, there was no difference in

the percentages of Medicaid and QHP enrollees who received a flu shot or spray in the year prior to completing

the survey. When restricting the population to a regression discontinuity model-derived optimal bandwidth

around the exceptional healthcare needs composite score cut-point, 22.2 percent more QHP enrollees were

estimated to have received a flu shot or spray compared to Medicaid enrollees.

Secondary and Tertiary Preventive Screenings

Secondary and tertiary screenings for medical conditions are also a strong sign of qualit y preventive care.

Table 14 presents a series of screening comparisons between Medicaid and QHP enrollees in the Higher

Needs Population.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 39

Table 14. Differences in Secondary and Tertiary Preventive Health Care between Medicaid and QHP

Enrollees (Regression Discontinuity Comparison)

Primary/Secondary/Tertiary

Preventive Healthcare

Indicators

Medicaid

QHP

Relative

Difference

(percent)

Average

Treatment

Effect

(percent)

Statistical

Difference

(p-value)

Period

Mean

(StdErr)

Mean

(StdErr)

Yearly Assessments

Percentage of chlamydia

screening (secondary)

Baseline

253

40.7 (3.1)

1,806

34.9 (1.1)

-14.3

-2.1

0.673

Year 2

395

35.7 (2.4)

2,664

34.2 (0.9)

-4.1

-10.0

0.024

Year 3

224

48.7 (3.3)

1,617

47.1 (1.2)

-3.3

14.5

0.004

Percentage of cholesterol

screening (secondary)

Baseline

1,865

39.6 (0.7)

8,134

40.1 (0.3)

1.1

-17.2

<0.001

Year 2

4,947

31.6 (0.7)

21,354

37.9 (0.3)

20.2

-18.1

0.002

Year 3

4,455

33.3 (0.7)

18,958

41.0 (3.6)

23.4

-19.9

<0.001

Percentage of enrollees

with diabetes with evidence

of HbA1C assessment

(tertiary)

Baseline

978

75.1 (1.4)

2,353

89.1 (0.6)

18.7

-10.7

<0.001

Year 2

1,162

62. 2 (1.4)

2,958

83.7 (0.7)

34.5

-22.8

<0.001

Year 3

1,157

60.6 (1.4)

3,230

77.9 (0.7)

28.6

-21.5

<0.001

Periodic Assessments (if age and gender appropriate, eligible for at least one screening over 3 years)

Percentage of cervical

cancer screening

(secondary)

3 Years

1,313

29.5 (1.3)

5,777

45.8 (0.7)

55.5

-15.9

<0.001

Percentage of breast cancer

screening (secondary)

3 Years

885

44.1 (1.7)

3,721

51.7 (0.8)

17.2

-19.4

0.002

Percentage of colorectal

screening (secondary)

3 Years

1,286

27.3 (1.2)

5,950

28.7 (0.6)

5.0

3.0

0.382

Composite Assessments (if eligible for at least one screening)

Percentage of enrollees who

received at least one

eligible screening

(secondary)

3 Years

6,400

61.6 (0.6)

28,383

62.6 (0.3)

1.7

-15.3

0.007

Percentage of enrollees who

received at least half of

eligible screenings

(secondary)

3 Years

6,400

29.6 (0.6)

28,383

35.5 (0.3)

19.8

-16.2

0.002

Percentage of enrollees who

received all eligible

screenings (secondary)

3 Years

6,400

11.7 (0.4)

28,383

16.8 (0.2)

43.8

-6.9

<0.001

Notes: All quality and screening indicators were derived from claims data. Adjusted analysis was performed using regression

discontinuity with a model-derived optimal bandwidth around an exceptional healthcare needs composite score cut-point. Average

treatment effect depicts model-estimated difference at cut-point. Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: N=number of Medicaid and QHP enrollees; StdErr=standard error of estimated mean.

All secondary and tertiary screening indicators presented in Table 14 with the exception of chlamydia screenings

in 2014 and colorectal screenings, show a higher percentage of QHP enrollees receiving screenings compared to

Medicaid enrollees.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 40

Disease Management

Table 15 presents a Medicaid and QHP Higher Needs Population comparison of treatments and condition

management that are consistent with quality of care delivery.

Table 15. Differences in the Percentage of Medicaid and QHP Enrollees Receiving Recommended

Disease Management (Regression Discontinuity Comparison)

Primary/Secondary/Tertiary

Preventive Healthcare

Indicators

Medicaid

QHP

Relative

Difference

(percent)

Average

Treatmen

t Effect

(percent)

Statistical

Differenc

(p-value)

Period

Mean

(StdErr)

Mean

(StdErr)

Percentage of enrollees with

diabetes with evidence of

HbA1C assessment (tertiary)

Baseline

978

75.1 (1.4)

2,353

89.1 (0.6)

18.7

-10.7

<0.001

Year 2

1,162

62. 2 (1.4)

2,958

83.7 (0.7)

34.5

-22.8

<0.001

Year 3

1,157

60.6 (1.4)

3,230

77.9 (0.7)

28.6

-21.5

<0.001

Anti-depressant

management

3 years

592

45.9 (2.1)

2,477

59.3 (1.0)

29.1

-15.8

0.002

Rate of mental/behavioral

disorder inpatient

hospitalizations per 10,000

person-years

Baseline

9,037

342.1 (20.2)

44,285

75.0 (4.3)

-78.1

304.1

0.003

Year 2

9,037

302.6 (20.0)

44,285

110.8 (5.7)

-63.4

157.5

0.179

Year 3

7,951

334.4 (24.7)

37,752

116.8 (6.5)

-65.1

167.8

0.245

Proportion of enrollees with

any re-engagement with

outpatient care per 100

discharges

Baseline

1,062

76.7 (2.0)

1,765

68.8 (1.3)

-10.3

2.9

0.316

Year 2

979

73.6 (2.9)

2,013

70.4 (1.4)

-4.4

-2.2

0.411

Year 3

905

69.2 (2.0)

1,809

70.1 (1.4)

1.2

-11.4

0.030

Notes: Adjusted analysis was performed using regression discontinuity with a model-derived optimal bandwidth around an

exceptional healthcare needs composite score cut-point. Average treatment effect depicts model-estimated difference at cut-

point. Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: N=number of Medicaid and QHP enrollees; StdErr=standard error of estimated mean.

Compared to Medicaid enrollees with diabetes, a higher percentage of QHP enrollees with diabetes received an

HbA1C test during each of the three years under study. The difference was 10.7 percent at Baseline, 22.8 percent

in Year 2, and 21.5 percent in Year 3. QHP enrollees were also more likely to receive anti-depressant management

compared to Medicaid enrollees (15.8 percent).

Compared to QHP enrollees, those covered by Medicaid had a higher rate of mental/behavioral disorder

hospitalizations in each year of coverage. This effect was estimated to be a 304.1 percent, 157.5 percent, and

167.8 percent higher rate of hospitalizations per 10,000 person-years at Baseline, Year 2, and Year 3,

respectively.

In Year 3 (2016), QHP enrollees had a higher rate of re-engagement with outpatient care compared to Medicaid

enrollees (11.4 percent).

Use of Healthcare Services

The difference in medical procedure rates for Higher Needs Population Medicaid and QHP enrollees is presented

in Table 16.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 41

Table 16. Differences in the Use of Healthcare Services between Medicaid and QHP Enrollees

(Regression Discontinuity Comparison)

Use of Healthcare

Services Indicators

Medicaid

QHP

Relative

Difference

(percent)

Average

Treatment

Effect

Statistical

Difference

(p-value)

Period

Mean (StdErr)

Mean

(StdErr)

Rate of abdominal

hysterectomy per

10,000 person-years

Baseline

5,763

46.4 (9.4)

26,469

57.6 (4.9)

24.1

-12.8

0.715

Year 2

5,763

103.3 (13.6)

26,469

51.7 (4.5)

-49.9

24.6

0.368

Year 3

5,168

42.2 (9.2)

23,241

37.5 (4.1)

-11.2

-11.8

0.655

Rate of vaginal

hysterectomy per

10,000 person-years

Baseline

5,763

79.2 (12.9)

26,469

78.6 (5.8)

-0.9

-57.7

0.011

Year 2

5,763

54.4 (10.2)

26,469

53.3 (4.6)

-2.0

-3.1

0.929

Year 3

5,168

34.1 (8.7)

23,241

51.3 (4.8)

50.2

20.3

0.249

Rate of cardiac

catheterization per

10,000 person-years

Baseline

9,037

280.0 (19.5)

44,285

109.1 (5.3)

-61.0

8.42

0.871

Year 2

9,037

207.2 (15.6)

44,285

100.6 (5.1)

-51.4

-45.3

0.448

Year 3

7,951

213.8 (18.0)

37,752

108.0 (5.6)

-49.5

-59.6

0.089

Rate of PCI per 10,000

person-years

Baseline

9,037

126.9 (19.3)

44,285

34.3 (3.3)

-73.0

35.5

0.337

Year 2

9,037

95.4 (15.7)

44,285

41.5 (3.9)

-59.5

-28.6

0.423

Year 3

7,951

63.0 (11.6)

37,752

31.1 (3.5)

-50.7

-4.6

0.871

Rate of CABG per

10,000 person-years

Baseline

9,037

23.7 (5.4)

44,285

10.9 (1.7)

-53.8

-6.3

0.609

Year 2

9,037

8.1 (3.1)

44,285

8.1 (1.4)

-0.5

-1.5

0.875

Year 3

7,951

15.7 (4.5)

37,752

9.1 (1.6)

-42.4

-5.6

0.661

Rate of back surgery

per 10,000 person-

years

Baseline

9,037

158.0 (15.8)

44,285

46.0 (3.5)

-70.9

27.6

0.199

Year 2

9,037

126.9 (13.4)

44,285

59.6 (4.0)

-53.0

-34.2

0.118

Year 3

7,951

95.7 (12.3)

37,752

75.3 (5.0)

-21.3

-49.1

0.202

Rate of open

cholecystectomy per

10,000 person-years

Baseline

9,037

11.2 (3.7)

44,285

4.1 (1.0)

-63.7

0.7

0.922

Year 2

9,037

7.0 (2.8)

44,285

2.6 (0.8)

-62.4

Year 3

7,951

5.2 (2.6)

37,752

3.0 (0.9)

-42.4

13.1

0.058

Rate of laparoscopic

cholecystectomy per

10,000 person-years

Baseline

9,037

215.2 (16.8)

44,285

123.8 (5.7)

-42.5

9.43

0.868

Year 2

9,037

111.7 (11.6)

44,285

106.1 (5.0)

-5.1

-63.1

0.013

Year 3

7,951

120.7 (12.8)

37,752

88.5 (5.0)

-26.6

-2.6

0.928

Rate of total hip

replacement per

10,000 person-years

Baseline

9,037

23.6 (6.5)

44,285

8.9 (1.6)

-62.4

Year 2

9,037

23.3 (5.7)

44,285

7.4 (1.4)

-68.3

4.2

0.530

Year 3

7,951

15.7 (5.2)

37,752

7.4 (1.4)

-52.8

-12.4

0.070

Rate of total knee

replacement per

10,000 person-years

Baseline

9,037

32.3 (7.1)

44,285

13.5 (2.0)

-58.3

-10.1

0.466

Year 2

9,037

40.7 (7.8)

44,285

22.9 (2.5)

-43.8

-23.1

0.214

Year 3

7,951

51.1 (9.0)

37,752

30.0 (3.3)

-41.4

-27.3

0.033

Rate of lumpectomy

per 10,000 person-

years

Baseline

5,763

38.7 (8.6)

26,469

43.7 (4.5)

13.0

4.8

0.773

Year 2

5,763

38.1 (8.7)

26,469

26.1 (3.3)

-31.5

3.0

0.872

Year 3

5,168

22.1 (6.4)

23,241

24.5 (3.4)

11.0

-36.6

<0.001

Rate of unilateral

mastectomy per

10,000 person-years

Baseline

5,763

9.7 (4.3)

26,469

10.1 (2.2)

4.3

-8.26

0.222

Year 2

5,763

10.9 (5.1)

26,469

9.9 (2.2)

-9.2

-25.0

0.214

Year 3

5,168

8.0 (4.9)

23,241

6.7 (1.7)

-16.7

Notes: All healthcare service indicators derived from claims data. Adjusted analysis was performed using regression discontinuity

with a model-derived optimal bandwidth around an exceptional healthcare needs composite score cut -point. Average treatment

effect depicts model-estimated difference at cut-point. Cells with “--” indicate insufficient data in vicinity of the cut-point to

calculate bandwidth. Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100. Prostatectomy, carotid endarterectomy,

and bilateral mastectomy indicators also tested with no significant differences between Medicaid and QHP enrollees.

Abbreviations: N=number of Medicaid and QHP enrollees; StdErr=standard error of estimated mean.

Sporadic differences in healthcare services are evidenced between Medicaid and QHP enrollees. At Baseline,

there were 57.7 fewer vaginal hysterectomies per 10,000 person-years performed in the eligible female Medicaid

enrollee population compared to QHP enrollees. At Year 2 there were 63.1 fewer laparoscopic cholecystectomies

per 10,000 person-years performed in the Medicaid enrollee population compared to QHP enrollees. At Year 3

Arkansas Health Care Independence Program (‘Private Option’) Final Report 42

there were 36.6 fewer lumpectomies per 10,000 person-years performed in the Medicaid enrollee population

compared to QHP enrollees.

Hospital Utilization

Table 17 compares hospital utilization measures across Medicaid and QHP enrollees in the years that they were

enrolled in the respective programs.

Table 17. Differences in Rates of Preventable Hospitalizations and Readmissions between Medicaid

and QHP Enrollees (Regression Discontinuity Comparison)

Medicaid

QHP

Relative

Difference

(percent)

Average

Treatment

Effect

Statistical

Difference

(p-value)

Preventable Utilization

Indicators

Period

Mean (StdErr)

Rate of

medical/surgery

hospital discharges per

10,000 person-years

Baseline

9,037

1,663.3 (55.7)

44,285

528.0 (12.8)

-68.3

473.4

0.312

Year 2

9,037

1,478.0 (58.4)

44,285

557.3 (14.8)

-62.3

379.6

0.166

Year 3

7,951

1,506.9 (61.7)

37,752

602.2 (17.2)

-60.0

-27.98

0.903

Rate of preventable

hospitalizations per

10,000 person-years

Baseline

9,037

68.9 (4.1)

44,285

233.9 (24.8)

-70.5

22.0

0.845

Year 2

9,037

61.7 (4.7)

44,285

217.6 (21.0)

-71.6

57.3

0.508

Year 3

7,951

69.0 (5.1)

37,752

219.0 (20.6)

-68.5

-10.9

0.846

Average length of stay

in days for all

hospitalizations

Baseline

1,208

5.3 (0.2)

2,259

4.6 (0.2)

-12.5

0.001

0.999

Year 2

1,213

5.1 (0.2)

2,850

5.3 (0.2)

3.5

-0.233

0.640

Year 3

1,072

5.7 (0.2)

2,496

5.7 (0.2)

0.5

-3.012

0.006

Average length of stay

in days for

medical/surgery

hospitalizations

Baseline

952

4.9 (0.2)

1,679

4.4 (0.2)

-10.6

-0.186

0.756

Year 2

881

5.1 (0.2)

1,835

4.9 (0.2)

-5.1

-0.413

0.408

Year 3

796

5.6 (0.3)

1,647

5.3 (0.2)

-5.8

-2.315

0.033

Proportion of enrollees

with any all-cause 30-

day readmission per

100 hospitalizations

Baseline

1,062

12.0 (1.3)

1,765

8.3 (0.8)

-30.7

0.044

0.282

Year 2

979

13.0 (1.4)

2,013

9.1 (0.8)

-30.3

0.101

<0.001

Year 3

905

12.1 (1.5)

1,809

11.6 (1.0)

-4.4

-0.063

0.097

Notes: All hospitalization indicators derived from claims data. Adjusted analysis was performed using regression discontinuity with a

model-derived optimal bandwidth around an exceptional healthcare needs composite score cut-point. Average treatment effect depicts

model-estimated difference at cut-point. Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: N=number of Medicaid and QHP enrollees; StdErr=standard error of estimated mean.

In the three years of program enrollment, no differences in hospital discharge or preventable hospitalization rates

were concluded between Medicaid and QHP enrollees. In addition, in Year 2, Medicaid enrollees had a higher rate

of all-cause 30-day readmission (0.101 readmissions per 100 hospitalizations) compared to QHP enrollees. One

strong difference in program effect is observed for average length of stay in Year 3, where compared to QHP

enrollees, Medicaid enrollees had a lower average length of stay of 3.0 days overall, and 2.3 days for those with

just medical/surgical related hospitalizations.

Emergency Room (ER) Utilization

Table 18 compares emergency room (ER) utilization measures across Medicaid and QHP enrollees in the years

that they were enrolled in the respective programs.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 43

Table 18. Differences in Utilization of Emergency Room Services between Medicaid and QHP Enrollees

(Regression Discontinuity Comparison)

Emergency Room

Indicators

Period

Medicaid

QHP

Relative

Difference

(percent)

Average

Treatment

Effect

Statistical

Difference

(p-value)

Mean

(StdErr)

Mean

(StdErr)

Rate of total

emergency room visits

per 100 person-years

Baseline

9,037

141.9 (2.9)

44,285

58.2 (0.7)

-59.0

80.5

0.085

Year 2

9,037

120.7 (2.7)

44,285

53.9 (0.7)

-55.3

40.8

0.219

Year 3

7,951

112.0 (2.6)

37,752

55.0 (0.7)

-50.8

33.9

0.343

Rate of emergent

emergency room visits

per 100 person-years

Baseline

9,037

27.5 (0.9)

44,285

10.1 (0.2)

-63.4

8.1

0.415

Year 2

9,037

24.6 (0.9)

44,285

10.0 (0.2)

-59.2

4.3

0.602

Year 3

7,951

23.2 (0.9)

37,752

10.7 (0.2)

-53.7

5.1

0.112

Rate of non-emergent

emergency room visits

per 100 person-years

Baseline

9,037

74.4 (1.7)

44,285

31.1 (0.4)

-58.2

41.8

0.109

Year 2

9,037

60.4 (1.5)

44,285

27.8 (0.4)

-54.0

29.3

0.042

Year 3

7,951

53.8 (1.5)

37,752

27.1 (0.4)

-49.7

11.2

0.536

Notes: All emergency room indicators derived from claims data. Adjusted analysis was performed using regression discontinuity with a

model-derived optimal bandwidth around an exceptional healthcare needs composite score cut-point. Average treatment effect depicts

model-estimated difference at cut-point. Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: N=number of Medicaid and QHP enrollees; StdErr=standard error of estimated mean.

With the exception of Medicaid enrollees having a higher rate of non-emergent emergency room visits in Year 2

(29.3 visits per 100 person-years) compared to QHP enrollees, enrollees in the two programs do not differ in ER

utilization.

Special Populations and Topic Studies

Medicaid and QHP Maternal Pregnancy and Birth Outcomes

Overview

Historically, Arkansas Medicaid covered uninsured pregnant women with incomes up to 200 percent FPL,

increasing to 214 percent in 2014 with Modified Adjusted Gross Income calculations and income disregards.

Under the terms and conditions of the Health Care Independence Program (HCIP), all newly pregnant women

applying for Medicaid would continue to be covered through the traditional Medicaid program even if their

income levels (< 138 percent FPL) made them otherwise eligible for enrollment in a Qualified Health Plan (QHP)

covered by premium assistance under the HCIP. However, as QHP-enrolled women became pregnant, they would

remain under a premium assistance program and have their pregnancy care covered by the QHP commercial

carrier in which they were enrolled. There were a number of pregnant women who did enroll in HCIP early in 2014

and this was due to a systematic failure to collect pregnancy information from all new enrollees, or the new

enrollees were diagnosed as pregnant after enrolling in a QHP. The earliest normal full gestational delivery date

for a woman who conceived while a beneficiary of a QHP in HCIP is on, or around, Sept. 30, 2014.

The goal of this study is to compare pregnancy outcomes for women who were covered by traditional Arkansas

Medicaid to those who were covered in a QHP through premium assistance in HCIP.

There is a noteworthy factor with an important impact on this study. Perinatal and obstetrical care for both

Medicaid and the private carriers are bundled and paid a global fee to the delivering provider. As such,

encounters for individual physician visits or treatments received are not consistently available within claims. From

Arkansas Health Care Independence Program (‘Private Option’) Final Report 44

inquiry of both Medicaid and private carriers, routine quality assessments are performed, however claims-based

indicators (e.g., HEDIS) on prenatal and/or pregnancy services are neither generated nor, for Medicaid, reported

to CMS. Consequently, this study relied on pregnancy outcomes and prenatal care recorded in the 2014 and 2015

Arkansas birth certificate vital records.

Data Sources and Analytic Population

Birth certificate data for 2014 and 2015 were obtained from the Arkansas Department of Health (2016 data was

not available at the time of analysis). Mother’s personal identifying information contained on the birth certificate

was assigned a unique identifier using a match engine (described in Appendix C). The same identifier is included

on Medicaid and QHP enrollment files.

Birth certificate and enrollment data were matched by mother and produced information on 34,675 births from

mothers who were covered by Medicaid or an HCIP QHP. This population was then restricted to include only live

births in Arkansas that occurred between Sept. 1, 2014, and Sept. 30, 2015 (15,597). A further restriction required

mothers to be enrolled with Medicaid or an HCIP QHP coverage for a period of at least 43 days prior to the date of

birth and 56 days after date of birth. This minimum enrollment time was based on HEDIS methodology and

provides some assurance that the mother had access to coverage for at least a portion of their prenatal and

postnatal period. Pregnant women who switched from Medicaid to a QHP — or vice versa — were also excluded.

In total, 10,453 births were retained for analysis. Appendix J contains a detailed inclusion/exclusion flow chart to

depict those included in the analytic sample.

Maternal and Birth Outcomes

Table 19 contains a listing of the maternal and birth outcomes under study along with a description and the

unit/scale of the measure. All outcomes were derived from responses on the birth certificate, with the exception

of non-delivery related, preterm hospitalizations.

Table 19. Maternal Pregnancy and Birth Outcomes, with Description

Outcome

Description

Measure

Maternal

Time to First Prenatal Visit

Estimated time from conception to first prenatal visit

Days (continuous)

Number of Prenatal Visits

Self-reported number of prenatal visits

Prenatal visits (days)

Preterm Hospitalization

(Preeclampsia)

Hospitalization (from claims data) was due to a diagnosis

of preeclampsia and not labor/childbirth

Yes or No (dichotomous)

Preterm Hospitalization

(Gestational Diabetes)

Hospitalization (from claims data) was due to a diagnosis

of gestational diabetes and not labor/childbirth

Yes or No (dichotomous)

Preterm Hospitalization (Any

Reason)

Hospitalization (from claims data) was due to any

diagnosis not including labor/childbirth

Yes or No (dichotomous)

Cesarean Section

Delivery was performed by cesarean section

Yes or No (dichotomous)

Birth

Preterm Birth

Birth occurred at less than 38 weeks of gestation

Yes or No (dichotomous)

Low Birth Weight

Baby weighed less than 2,500 grams

Yes or No (dichotomous)

Abnormal Outcome

Combined measure of all abnormal indicators, excluding

meconium

Yes or No (dichotomous)

APGAR5 1

A systematic measure for evaluating the physical

condition of the infant at specific intervals following birth

Very abnormal category or

other category

(dichotomous)

APGAR5 2

A systematic measure for evaluating the physical

condition of the infant at specific intervals following birth

Any abnormal category or

other category

(dichotomous)

NICU Stay

Neonatal intensive care unit stay required following birth

Yes or No (dichotomous)

Arkansas Health Care Independence Program (‘Private Option’) Final Report 45

Propensity Score Matching

A propensity scoring approach was implemented to match QHP mothers with traditional Medicaid mothers,

controlling for demographic and clinical characteristics prognostic for the maternal and infant outcomes assessed.

As a first step, a dichotomous variable indicating the assignment of a pregnant woman in a QHP (1) or Medicaid

(0) was used as an outcome variable in a logistic regression. Covariates obtained from the birth certificate data

used in the regression included mother’s age, race/ethnicity, insurance region, WIC (Women, Infants, and

Children) program status, body mass index category, education and marital status, urban/rural indicator, number

of previous live births, number of previous still births, single/multiple birth status, and number of days enrolled

during the perinatal period. Also included in the model as covariates were indicators for previous cesarean

section, pre-pregnancy diabetes and hypertension, preterm birth, and infections — chlamydia, gonorrhea,

syphilis, or hepatitis B/C — at time of birth. A propensity score was created for each observation to identify the

probability of a delivering mother covered by a QHP (1) or Medicaid (0).

Mothers covered by Medicaid and a QHP were matched, one to one, by their propensity score of program

assignment. Similar to the process used in the findings for the General Population, this was performed using a

greedy matching algorithm.

Crude covariate category values used in the matching and the subsequent matched

sample values are summarized in Appendix I, along with calculation of standardized differences between

Medicaid and QHP groups prior to and after matching. The bias reduction associated with matching is also

presented. In total, our analytic population contains 2,369 birth outcomes in each of the matched mother pairs

from Medicaid and QHP programs.

Statistical Approach

Prenatal visits were measured as a count outcome variable. Time to first visit was measured as a continuous

outcome variable. Frequency distributions for all outcomes under study are included in Appendix I.

Dichotomous maternal quality outcome measures of interest (yes/no responses) included pre-birth

hospitalization for preeclampsia, gestational diabetes, or any reason, and delivery by cesarean section.

Dichotomous birth quality outcome measures of interest (yes/no responses, unless otherwise indicated) included

preterm birth, low birth weight, abnormal outcomes (combined measure of all abnormal indicators, including

meconium), APGAR5 (very abnormal vs. other and any abnormal vs. other), and neonatal intensive care unit

(NICU) stay after birth.

Individual generalized linear models were fit regressing a Medicaid/QHP dichotomous primary independent

indicator on the outcomes of interest. A random variable indicating the matched Medicaid-QHP dyad was also

included. The distribution of the outcome determined the type of regression used. For dichotomous outcomes a

generalized logistic regression was utilized and adjusted odds ratios were reported for the differences between

matched Medicaid and QHP maternal and birth outcomes. For count variables, a generalized Poisson regression

model was utilized and incidence rate ratios reported. For continuous variable outcomes, generalized normal

regression models were utilized with marginal effect sizes reported.

Findings

Tables 20 and 21 contain the results of fitting generalized linear regression models on the matched Medicaid/QHP

data described above.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 46

Table 20. Maternal Pregnancy Differences between Medicaid and QHP Enrollees

Maternal Pregnancy Outcomes

Comparison of Programs

Continuous

Outcomes

Measure

Medicaid (N=2,369)

Mean (StdDev)

QHP (N=2,369)

Mean (StdDev)

Marginal Effect

(95 percent CI)

Time to First Prenatal

Visit

Days

90.81 (49.80)

81.89 (43.79)

-8.71 (-11.62 to -5.80)

Count Outcomes

Measure

Medicaid

Mean (StdDev)

QHP

Mean (StdDev)

Incidence Rate Ratio

(95 percent CI)

Prenatal Visits

Visits

10.55 (3.85)

11.00 (3.81)

0.420 (0.222-0.619)

Dichotomous

Outcomes

Category

Medicaid (N=2,369)

n (percent)

QHP (N=2,369)

n (percent)

Adjusted Odds Ratio

(95 percent CI)

Preterm Hospitalization

(preeclampsia)

Yes vs. No

69 (2.91)

77 (3.25)

1.121 (0.805-1.562)

Preterm Hospitalization

(gestational diabetes)

Yes vs. No

50 (2.11)

1.000 (0.667-1.498)

Preterm Hospitalization

(any reason)

Yes vs. No

225 (9.50)

201 (8.48)

0.885 (0.728-1.079)

Cesarean Birth

Yes vs. No

1,008 (35.17)

1,007 (35.14)

0.999 (0.897-1.111)

Notes: Marginal effect < 0 and confidence interval not containing 0 indicate less time to first visit for QHP. Incident rate ratio > 0

and confidence interval not containing 0 indicate higher number of visits for QHP. Adjusted odds ratios > 1 and confidence

interval not containing 1 indicate a higher rate for QHP.

Abbreviations: QHP=Qualified Health Plan; StdDev=Standard Deviation; CI=Confidence Interval

On average, pregnant women with QHP coverage had a prenatal visit 8.7 days sooner than similarly matched

women who were covered with Medicaid insurance. In addition, QHP-covered women had, on average, 0.4 more

prenatal visits than similarly matched women with Medicaid over the course of a pregnancy.

Table 21. Pregnancy Birth Outcomes between Medicaid and QHP Enrollees

Pregnancy Birth Outcomes

Comparison of Programs

Dichotomous

Outcomes

Category

Medicaid

N (percent)

QHP

N (percent)

Adjusted Odds Ratio

(95 percent CI)

Preterm Births

Preterm vs. Normal

183 (7.72)

214 (9.03)

1.189 (0.966-1.463)

Low Birth Weight

Yes vs. No

196 (8.16)

224 (9.33)

1.079 (0.882-1.321)

Abnormal Outcome

Yes vs. No

144 (6.00)

155 (6.46)

1.079 (0.851-1.369)

APGAR5 1

Very abnormal vs. other

33 (1.38)

36 (1.51)

1.096 (0.875-1.780)

APGAR5 2

Any abnormal vs. other

121 (5.04)

117 (4.88)

0.985 (0.756-1.283)

NICU

Yes vs. No

158 (6.58)

159 (6.63)

1.010 (0.799-1.276)

Notes: Adjusted odds ratios > 1 and confidence interval not containing 1 indicate a higher rate for QHP.

Abbreviations: QHP=Qualified Health Plan; CI=Confidence Interval

Based on matched Medicaid and QHP birth outcomes, our results did not detect any statistically significant

differences in rates of preterm births, low birth weight deliveries, abnormal outcomes, APGAR scores at any

abnormal cut-point, or NICU stays.

Discussion

Except for small differences in prenatal care — one additional visit with initiation approximately one week

earlier — no variation in maternal, delivery, or birth outcomes were observed between women in the

traditional Medicaid program and those with premium assistance in the QHPs.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 47

This finding may be reflective of Medicaid being the source of coverage for the majority of births within the

state — over 50 percent of births covered by Medicaid prior to 2014. In addition, the lack of variation in

chlamydia screening between Medicaid and QHPs — at least in the General Population, noted in Table 7, and in

favor of Medicaid in Year 3 for the Higher Needs Population, depicted in Table 14 — supports the influence of the

Medicaid quality improvement programs over time. These improvement projects and programs have included

focused efforts to increase prenatal visits and screening rates, care coordination and telemedicine support for

high-risk pregnancies, and value-based payment strategies shared with the private sector through the Arkansas

Healthcare Payment Improvement Initiative.

There are limitations to this study, including those previously mentioned involving the lack of detailed claims for

bundled provider payments. There is also not a perfect overlap in income eligibility across Medicaid and QHP

commercial programs. Medicaid covers previously uninsured women (or transition from pre-existing Medicaid aid

categories) with incomes up to 200 percent of FPL, while those receiving coverage in a premium-assisted QHP have

maximum incomes of 138 percent of FPL. To address this, we performed a sensitivity test on all maternal and birth

outcomes where the analytic population was restricted to just those who were WIC recipients during pregnancy.

WIC eligibility is capped at women who earn 185 percent of FPL, thus producing a slightly more balanced distribution

on income levels between Medicaid and QHP enrollees. There were no changes in the findings when analyses were

restricted to WIC beneficiaries (data not shown). On many demographic and geographic variables, the Medicaid and

QHP study populations are also already very similar, or balanced (Appendix I).

Non-Emergency Medical Transportation

Overview

As a required covered benefit within Medicaid, but not a component of the Essential Health Benefit (EHB) in QHP

coverage, non-emergency medical transportation (NEMT) was the most frequent wrapped service for QHP

enrollees. As required in the waiver, and subsequently stated in the research design, availability and accessibility

of NEMT services for QHP enrollees was a concern. Hypothesis 1e: HCIP beneficiaries will have equal or better

access to non-emergency transportation compared with what they would have otherwise had in the Medicaid fee-

for-service system over time.

In the Interim Report

— and based on self-reported responses to the first CAHPS survey — we found that within

the Higher Needs population, Medicaid enrollees had more problems accessing NEMT compared to QHP enrollees

during Program Year 1. There were no self-reported differences in access within the General population.

Data Source and Analytic Population

Using the results from a second CAHPS survey (CAHPS II) that elicited self-reported responses from respondents

who had approximately 18 months of continuous coverage, General Population and Higher Needs Population

results from CAHPS II are presented for Medicaid and QHP comparison group NEMT responses. Enrollees were

asked about transportation barriers for visits to both personal doctors and specialists.

Statistical Approach and Findings

Tables 22 and 23 present a statistical chi-square test of association, and a propensity score matched analysis for

Medicaid and QHP enrollees in the General Population for responses to each of the two non-emergency

transportation questions. Tables 24 and 25 present a statistical chi-square test of association, and a regression

discontinuity analysis for Medicaid and QHP enrollees in the Higher Needs Population for responses to each of the

two non-emergency transportation questions.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 48

Table 22. Differences in Non-Emergency Transportation between General Population Medicaid and

QHP Enrollees

Medicaid

QHP

Perceived Access Indicators

Mean

(StdErr)

Mean

(StdErr)

Relative

Difference

(percent)

Statistical

Difference

(p-value)

Percentage of enrollees who did not miss a visit to

a personal doctor due to lack of transportation

194

89.2 (2.2)

641

81.1 (1.5)

-9.1

0.002

Percentage of enrollees who did not miss a visit to

a specialist due to lack of transportation

89.9 (3.2)

340

88.4 (1.7)

-1.7

0.103

Notes: Data were obtained from CAHPS II responses. Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100. Statistical

difference was concluded from a chi-square test of association.

Abbreviations: N=number of Medicaid and QHP enrollees; StdErr=standard error of estimated mean.

In a direct comparison of responses to NEMT questions, a higher percentage of Medicaid enrollees (89.2 percent),

compared to QHP enrollees (81.1 percent), indicated that they did not miss a visit to a personal doctor due to lack

of transportation.

Table 23. Differences in Non-Emergency Transportation between Medicaid and QHP Enrollees

(Propensity Score Matched Comparison)

Non-Emergency Transportation Indicators

Matched

Medicaid

Mean (StdErr)

QHP

Mean

(StdErr)

Relative

Difference

(percent)

Statistical

Difference

(p-value)

Percentage of enrollees who did not miss a visit to a

personal doctor due to lack of transportation

181

90.1 (2.2)

87.8 (2.4)

-2.5

0.501

Percentage of enrollees who did not miss a visit to a

specialist due to lack of transportation

93.6 (2.8)

92.3 (3.0)

-1.4

0.295

Notes: Data were obtained from CAHPS II responses. Adjusted analysis was performed using a propensity score matched sample.

Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: Matched N=number of Medicaid and QHP matched enrollees; StdErr=standard error of estimated mean.

When General Population Medicaid and QHP enrollees are matched for the comparison of differences, the crude

results reported in Table 23 for personal doctors attenuate, and we find no evidence that access to NEMT is

different across enrollee groups.

Table 24. Differences in Non-Emergency Transportation between Higher Needs Population Medicaid

and QHP Enrollees

Medicaid

QHP

Relative

Difference

(percent)

Statistical

Difference

(p-value)

Non-Emergency Transportation Indicators

Mean

(StdErr)

Mean

(StdErr)

Percentage of enrollees who did not miss a visit to

a personal doctor due to lack of transportation

868

82.2 (1.3)

972

91.9 (0.9)

11.7

0.002

Percentage of enrollees who did not miss a visit to

a specialist due to lack of transportation

533

75.5 (1.9)

560

95.2 (0.9)

26.1

0.084

Notes: Data were obtained from CAHPS II responses. Adjusted analysis was performed using a chi-square test of association. Average

treatment effect depicts model-estimated difference at cut-point. Relative percent calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: N=number of Medicaid and QHP enrollees; StdErr=standard error of estimated mean.

In a direct comparison of responses to NEMT indicators, a higher percentage of QHP enrollees (91.9 percent),

compared to Medicaid enrollees (82.2 percent), indicate that they did not miss a visit a personal doctor due to

lack of transportation.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 49

Table 25. Differences in Non-Emergency Transportation between Medicaid and QHP Enrollees

(Regression Discontinuity Comparison)

Medicaid

QHP

Relative

Difference

(percent)

Average

Treatment

Effect

(percent)

Statistical

Difference

(p-value)

Non-Emergency Transportation Indicators

Mean

(StdErr)

Mean

(StdErr)

Percentage of enrollees who did not miss a

visit to a personal doctor due to lack of

transportation

884

83.0 (1.3)

983

87.9 (1.0)

5.9

-15.0

0.005

Percentage of enrollees who did not miss a

visit to a specialist due to lack of

transportation

544

87.1 (1.4)

564

90.4 (1.2)

3.8

-8.2

0.159

Notes: Data were obtained from CAHPS II responses. Statistical analysis was performed using regression discontinuity with a model-

derived optimal bandwidth around an exceptional needs screener cut-point. Average treatment effect depicts model-estimated

difference at cut-point. Relative percentage calculated as (QHP – Medicaid)/Medicaid x 100.

Abbreviations: N=number of Medicaid and QHP enrollees; StdErr=standard error of estimated mean.

Using a more robust evaluation design with regression discontinuity, we find evidence to confirm that 15.0 percent more

QHP enrollees, compared to Medicaid enrollees, did not miss a visit to a personal doctor due to lack of transportation.

Our findings from the second CAHPS survey indicate that for the Higher Needs population, lack of NEMT is still

more problematic for Medicaid enrollees than for QHP enrollees. This is evidenced even more for those around

the exceptional healthcare needs composite score cut-point, where we would expect no reasonable differences to

be caused by Medicaid or QHP program assignment.

Early and Periodic Screening, Diagnostic, and Treatment (Medicaid Benefit)

Overview

Early and Periodic Screening, Diagnostic, and Treatment (EPSDT) is a required benefit that provides comprehensive

and preventive healthcare services for children and adolescents under 21 who are enrolled in Medicaid. With

expansion, through premium assistance under the HCIP, 19- and 20-year-olds were placed in the QHP and received

the Essential Health Benefit (EHB). Examination of the structure of the EHB determined that the Early and Periodic

Screening and Diagnostic (EPSD) tests would be similarly covered. However, concern over the breadth of potential

treatments available under EPSDT and the definition of commercial coverage within the EHB was a component of

the waiver and the stated hypothesis in the research design. Hypothesis 1d: HCIP beneficiaries will have equal or

better access to EPSDT benefit access for young, eligible adults compared with what they would have otherwise had

in the Medicaid fee-for-service system over time (see Appendix B).

To approach this hypothesis, key informants who manage complex Medicaid-enrolled children and young adults

were consulted.

Concurrence on the overlap in EPSD services was provided and no recognized treatment deficits

in QHP coverage over the program’s initial three years was reported by our key informants.

Dr. Charles Field, former Division Director of Community Pediatrics and Professor in the Department of Pediatrics, University

of Arkansas, Arkansas Head Start Board Member, and clinician at Mid-Delta Community Health Center; Dr. Eldon Shultz, Eldon

G. Schulz, M.D., FAAP, Rockefeller Professor and Endowed Chair for Children with Special Needs, Section Chief, Developmental-

Behavioral and Rehabilitative Pediatrics, Medical Director, Arkansas Children’s Hospital Rehabilitative Services, and Medical

Director, Arkansas Developmental Disabilities Services; and Dan Knight, MD, Chairman of the Department of Family and

Preventive Medicine, and Associate Professor at the University of Arkansas for Medical Sciences.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 50

To further investigate the potential for treatment restrictions, we designed a pre- and post-enrollment

assessment comparing children enrolled in Medicaid as 17- and 18-year-olds, who were subsequently potentially

eligible for HCIP assignment to a QHP. To assess program treatment differences, we selected conditions that were

immutable and had the potential for intense medical treatment needs, as well as conditions and treatments that

would be traceable in both Medicaid and QHP claims. With support from our key informants, we selected sickle-

cell disease, cystic fibrosis, and hemophilia.

Data Source and Analytic Population

From Medicaid enrollment files, all 17- and 18-year-old children with Medicaid coverage in 2013 were identified. We

employed Medicaid claims files from 2013 to identify children who received health care in 2013 to treat a condition

of sickle cell disease, cystic fibrosis, or hemophilia. The premise was that these three conditions would require

specialized chronic care and disease management into young adulthood. We identified 151 Medicaid beneficiaries

who met one of these three conditions and were 17 or 18 years old. These individuals were then tracked to assess

coverage status as well as variations in treatment when they turned 19 and 20, in 2014 and 2015, respectively.

Findings

Efforts to track these 151 individuals into their nineteenth and twentieth year were largely successful. Through

Medicaid programmatic aid categories — including Supplemental Security Income (SSI) disability and the Tax Equity

and Fiscal Responsibility Act (TEFRA), otherwise known as “Katie Beckett options” — combined with the frailty screener

that directed expansion population members with higher needs into the traditional Medicaid-managed population, a

large majority (86.7 percent) of these individuals remained on Medicaid and were not assigned to a QHP.

Table 26 presents the number of beneficiaries who were still in Medicaid in the month following their nineteenth

birthday, as well as the associated aid category and a description of the program under which they were covered.

In total, 131 of the 151 adolescents (86.8 percent) were still enrolled in a Medicaid aid category with access to a

Medicaid in-network specialist provider.

Table 26. Health Insurance Carrier at 19 Years of Age for Medicaid Screened and Diagnosed Sickle Cell

Disease, Cystic Fibrosis, or Hemophilia (n=151)

Type of Health Insurance

Enrolled

Medicaid

131

SSI (aid category 43/45)

Adult Expansion (aid category 06 “Frail”)

Other, including AFDC Grant (aid category 20), Non-Medicaid (aid

category 04), Pregnant (aid category 61), TEFRA (aid category 49)

Not found in either Medicaid or QHP enrollment categories

<20

Enrolled in QHP

Due to the scarcity of transition from Medicaid to QHP premium assistance, no comparative analyses were

achievable. Given the success of programmatic retention in existing aid categories for these individuals in

traditional Medicaid and the frailty screener’s sensitivity in identification of higher needs, this concern is

effectively non-existent. The established programmatic pathway for an individual to transition from QHP to

Medicaid management based on a medical assessment of frailty further allays concerns. However,

modification of the frailty screener to a single question at the time of application may warrant reassessment

in the future.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 51

Continuity of Care

Overview

One form of churn is through gaps in continuous health insurance coverage, when individuals go off coverage and

come back on after a short period of time. Churn can impact health from missed care during the time period when

the individual was uninsured. It can also leave individuals with high medical costs during the uncovered period.

As previously noted, under an alternative demonstration of the implementation of the ACA, the Arkansas adult

Medicaid expansion uses premium assistance to purchase commercial insurance in the Marketplace. Program

income eligibility in the Health Care Independence Program (HCIP) is up to 138 percent of the federal poverty

level. Shore-Sheppard

demonstrated that income can be highly variable from month to month in this low-

income range. Depending on eligibility and redetermination (a process through which income eligibility

determination to remain in the program is updated) periods, individuals eligible for coverage under state -

level Medicaid expansion earning at or less than 138 percent FPL in one month may cross the income

eligibility threshold in a subsequent month — strictly due to a temporary fluctuation in the number of hours

worked that month.

For low-income individuals, there are a number of programs through which they can churn eligibility on or off

simply by crossing a threshold. The traditional Medicaid eligibility in Arkansas is for parents earning at or less than

17 percent FPL; the HCIP threshold is up to 138 percent FPL; and individuals earning up to 400 percent FPL can

enter the Marketplace (with a number of gradients determining individual out-of-pocket premiums and tax

credits). Seamless churning between these programs requires careful application coordination. Another adult

Medicaid expansion population in Arkansas consists of those who were determined to have exceptional

healthcare needs (i.e., “Frail”) after completing a screener (06 Medicaid). Individuals can also churn in and out of

this program.

It was anticipated that churning within state-level ACA adult Medicaid expansions would be problematic, with

some estimates predicting that churn would affect 50 percent of enrollees during the first year.

One factor

that potentially limited the churn effect for the premium assistance HCIP expansion population in Arkansas, was

that commercial carriers received guaranteed monthly premiums to cover individuals enrolled in the HCIP. As

such, QHPs had a stake in retaining members through outreach during periods of redetermination or simply due

to upward-income mobility. Enrollees could also be retained with the same QHP coverage, but through a tax

credit-subsidized plan on the Marketplace. The result was a lower rate of churn for the insured ACA expansion

population.

More than 95 percent of those initiating premium assistance coverage in a QHP in 2014 maintained coverage

through December of that year.

In this report we follow the continuity of coverage, attrition, and churn for the

entire three-year period within the HCIP program. While the start-up period in 2014 included a number of system-

related challenges, those who enrolled during that year and the first six months of 2015 were not subject to

redetermination.

Due to current data limitations, we are unable to track the HCIP expansion population systematically into the

subsidized Marketplace if they leave the HCIP coverage. Therefore we cannot capture potential upward-income

mobility and subsequent insurance coverage into the Marketplace or into employer-sponsored insurance

programs. The Arkansas Center for Health Improvement (ACHI) is working on compiling a customized analytic

database from the Arkansas All-Payer Claims Database (APCD) to track individual insurance coverage across

insurance carriers and identify periods without insurance through machine learning.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 52

In this study, we profile those who enrolled and maintained enrollment in an HCIP QHP or were assigned to Aid

Category 06 of Medicaid coverage by virtue of being determined to have exceptional healthcare needs. We focus

on these two groups during three periods of distinct enrollment and retention processing by Arkansas Medicaid.

The first is an 18-month pre-redetermination period where income eligibility updating was not effectuated. This

was followed by a universal redetermination period that began in July 2015, when all QHP enrollees would have

been subjected to this process, likely for the first time. Finally, we study continuity of coverage, attrition, and

churn in 2016, when a regular monthly process of redetermining groups of enrollees based on individual

enrollment month for income eligibility was reinstated.

Data and Methodology

Data for these analyses came from the Arkansas Medicaid premium payment database. The data contained

indicators for the following: each member enrolled, each month a premium assistance premium was paid to a

QHP, and whether or not each individual was a member of the 06 Medicaid category for that month.

Program Enrollment History

Over the three-year course of HCIP, 399,330 individuals secured health insurance coverage in a QHP or through

the 06 Medicaid category for at least one month. Many who entered the program stayed on through December

2016, as evidenced by 247,579 individuals maintaining continuous coverage with a QHP and 18,660 individuals

maintaining continuous 06 Medicaid coverage from the point at which they enrolled in the program.

Figure 9 depicts enrollment in the start-up month of January 2014 and the steady growth in the number of those

insured over 36 months of the HCIP. The continuous growth in enrollment was only interrupted by the universal

redetermination in July 2015.

Figure 9. Health Care Independence Program Enrollment by Program and Month

Continuity of Coverage During Pre-Redetermination Months

As previously stated, approximately 95 percent of individuals who received health insurance through a QHP

retained continuous coverage in the first year of the program. From January 2014–June 2015, no systematic

redetermination of eligibility was conducted on adults enrolled in the HCIP program or in 06 Medicaid. In June

2015, prior to redetermination, 216,870 Arkansas adults were enrolled in a QHP, while an additional 25,733

expansion adults were retained in 06 Medicaid after being identified as having exceptional healthcare needs.

Table 27 profiles the pre-redetermination enrollment history of these two groups.

50,000

100,000

150,000

200,000

250,000

300,000

350,000

Enrolled

Enrollment Month

QHP 06 Medicaid

Arkansas Health Care Independence Program (‘Private Option’) Final Report 53

Table 27. Medicaid Enrollment History of QHP and 06 Medicaid Individuals Enrolled in June 2015

Enrolled in QHP in June 2015

N=216,870

Percent=100

Continuous Enrollment

213,879

98.6

Experienced at Least one QHP-to-QHP Gap Month

2,167

1.0

Experienced an 06-Medicaid-to-QHP Transition (Seamless)

470

0.2

Experienced an 06-Medicaid-to-QHP Transition (with Gap)

354

0.2

Enrolled in 06 Medicaid in June 2015

N=25,733

Percent=100

Continuous Enrollment

24,525

95.3

Experienced at Least one 06-Medicaid-to-06-Medicaid Gap Month

0.1

Experienced a QHP-to-06-Medicaid Transition (Seamless)

1,093

4.3

Experienced a QHP-to-06-Medicaid Transition (with Gap)

0.3

Overwhelmingly, in the absence of redetermination, those who were enrolled in a QHP in June 2015 did not

experience a gap in coverage (98.6 percent) since enrollment (up to 18 continuous months). Only 1.0 percent of

QHP enrollees enrolled in June 2015 had a gap in coverage of one or more enrollment months. A very small

number of enrollees churned between 06 Medicaid and a QHP (0.4 percent), with the majority having a seamless

transition without a gap in health insurance coverage.

For those determined to have exceptional healthcare needs and assigned to 06 Medicaid and were in the program in

June 2015, 95.3 percent experienced continuous coverage. Another 4.3 percent had seamless coverage — albeit

interrupted with a switch from a QHP to 06 Medicaid coverage.

There was a modest amount of attrition from the respective programs. A total of 9,907 individuals had been

enrolled in a QHP, and 2,364 individuals had been enrolled in 06 Medicaid, in one of the 17 months prior to June

2015, but were not enrolled in either program in June 2015. A small number of 82 individuals had experienced a

switch between programs and were not enrolled in either program in June 2015.

Impact of Redetermination

In this analysis, QHP and 06 Medicaid enrollees who received a notice of redetermination in July 2015 are profiled

to identify the impact on continuity of coverage within both programs. We focus on any gap in coverage during

the three months after July 2015. In July 2015, 222,282 Arkansas adults were enrolled in a QHP, while an

additional 25,930 expansion adults were retained in 06 Medicaid. Table 28 tracks the enrollment of these two

groups through the redetermination period.

Table 28. The Impact of Redetermination on QHP and 06 Medicaid Enrollees (Enrolled in July 2015)

Status of QHP Enrollees in November 2015

N=222,282

Percent=100

Continuous Enrollment

179,426

80.7

Gap Period of One to Three Months, Enrolled in QHP in November 2015

5,072

2.3

QHP-to-06-Medicaid Transition (Seamless), Enrolled in QHP in November 2015

<11

0.0

QHP-to-06-Medicaid Transition (with Gap), Enrolled in QHP in November 2015

<11

0.0

Not Enrolled in November 2015

37,765

17.0

Status of 06 Medicaid Enrollees in November 2015

N=25,930

Percent=100

Continuous Enrollment

20,590

79.4

Gap Period of One to Three Months, Enrolled in QHP in November 2015

776

3.0

06-Medicaid-to-QHP Transition (Seamless)

<11

0.0

06-Medicaid-to-QHP Transition (with Gap)

<11

0.0

Not Enrolled in November 2015

4,552

17.6

Redetermination affected 2.3 percent of QHP enrollees with a gap in coverage of one to three months. Very few

individuals were reassigned from a QHP to 06 Medicaid coverage. Of the 37,765 that were not back in a QHP by

Arkansas Health Care Independence Program (‘Private Option’) Final Report 54

November 2015, we were able to identify (data not shown) that 25,999 individuals (68.8 percent) were

subsequently not enrolled in a QHP or in 06 Medicaid during any period in 2016. Of the 11,766 individuals who

were re-enrolled with QHP or 06 Medicaid coverage in 2016, 6,559 individuals (55.7 percent) were back on in the

first six months of the year. The numbers are similar for the 4,552 06 Medicaid enrollees who lost coverage during

a month impacted by redetermination. A total of 3,293 individuals (72.3 percent) were not enrolled in a QHP or

with 06 Medicaid coverage in 2016.

Continuity of Care in the Post-Redetermination Period

Of the 399,330 individuals who were ever enrolled in a QHP or 06 Medicaid program during the first two HCIP

program years, only 45,387 individuals (11.4 percent) were not enrolled for any time period during 2016. This is a

strong indication of continuity in the program. Following the universal July 2015 redetermination, regular monthly

intervals of redetermination based upon month of birth were re-instituted. For this reason, it is expected that

more frequent gaps in coverage could be observed in 2016.

In 2016, a total of 353,943 individuals were covered in a QHP or in 06 Medicaid at some point in time. This

includes those who had been on the program continuously, those who re-enrolled in 2016 after a gap period of

being uninsured, and those who were covered by a non-HCIP program or were uninsured.

Table 29 presents continuity of coverage and transitions for those in the HCIP program in 2016.

Table 29. HCIP Enrollment History of Individuals Enrolled in HCIP During 2016

Enrolled in QHP in December 2016

N=327,739

Percent=100

Continuous Enrollment

269,091

82.1

Gap in 2016 Coverage, Still Enrolled in December 2016

6,457

2.0

06-Medicaid-to-QHP Transition (Seamless), Still Enrolled in December 2016

464

0.1

06-Medicaid-to-QHP Transition (Gap), Still Enrolled in December 2016

0.0

Not Enrolled in December 2016

51,658

15.8

Enrolled in 06 Medicaid in December 2016

N=26,204

Percent=100

Continuous Enrollment

20,781

79.3

Gap in Coverage, Still Enrolled in December 2016

486

1.8

QHP-to-06-Medicaid Transition (Seamless), Still Enrolled in December 2016

191

0.7

QHP-to-06-Medicaid Transition (Gap), Still Enrolled in December 2016

0.0

Not Enrolled in December 2016

4,731

18.1

The continuous monthly redetermination process did not change the percentage of enrollees maintaining

continuous coverage through 2016. We were unable to ascertain at this time if the attrition observed in 2016 was

due to enrollee-initiated action (i.e., obtaining employer-sponsored insurance) or as a result of redetermination.

Future Analyses

Of the 45,387 individuals (11.4 percent) enrolled in the first two years of the HCIP program, but who were not

enrolled in Program Year 3 (2016), it is unknown whether or not these individuals were eligible but failed to re-

enroll, out-migrated from the state, had an upward change in income status (enabling participation in the

Marketplace) or being employed and obtained insurance through their employer. Through the Arkansas All-Payer

Claims Database — with its unique “hashed” identifier — and the state’s hospital and emergency room discharge

databases, future studies tracking individuals’ longitudinal coverage status is possible. In addition, waiver requests

under consideration by CMS will offer quasi-experimental opportunities to test alternative coverage strategies

and their impact on churn.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 55

Health Independence Accounts

Overview

The Arkansas Health Care Independence Act of 2013 (Ark. Code Ann. § 20-77-2405 (j)) — known as the “Private

Option” — established authority for the state’s Section 1115 waiver for the Medicaid expansion population. The

Act contained language requiring the development and implementation of “health savings or independence

accounts” with required participation by non-aged, non-disabled program-eligible participants. Following waiver

approval and initiation of the Private Option in 2014, the state developed and received federal approval to

implement Health Independence Accounts (HIAs), named MyIndyCards, at the initiation of Program Year 2

(January 2015) for individuals between 100 and 138 percent of the federal poverty level (FPL).

Program Design

Design of the HIAs was to serve as a mechanism to provide protection from cost-sharing, promote appropriate

healthcare utilization, and offer a mechanism to enable savings for future potential premium exposure. All

individuals in qualified health plans (QHPs) between 100 and 138 percent FPL were required to participate and

contribute, or generate a debt to the state (i.e., $10 a month for those earning 100 to 117 percent FPL and $15 a

month for those earning 118 to 138 percent FPL).

Contribution to the HIA in one month resulted in state-funded cost-sharing protection for the following month.

Initial activation of an HIA gained two months’ cost-sharing protection before monthly contributions were

required to maintain cost-sharing coverage. The state debited the HIA balance for a failed payment in a given

month. The state matched the individual’s contribution up to $200, if timely payments were made, and balances

were allowed to roll over annually. Finally, funds were available for premium payments in the marketplace upon

exit from the HIA program.

Data Sources and Analytic Population

We profiled the programmatic experience of the Arkansas HIAs, assessed the characteristics of participating

individuals, and evaluated the financial impact of the HIAs on individuals and on the program.

Data from the 2014 enrollment files for Private Option beneficiaries between 100 to 138 percent FPL and claims

data from QHPs from 2014-2016 were used for this analysis. In addition, third-party HIA transactions for both

payment collection and distribution were employed.

In total, 57,079 individuals were eligible for HIA participation. Demographic characteristics, including age, gender,

race/ethnicity, and rural/urban status, were combined with 2014 claims experience for hospitalizations,

emergency room utilization, and the Charlson Comorbidity Index to profile eligible participants.

Statistical Approach

Outcomes assessed included: initiation of payments; contribution frequency and amounts; cost-sharing protection

received; relative contribution compared to cost-sharing protection of participants both in the aggregate and

individually; and participant contributions compared to programmatic administrative costs. Statistical significance

was assessed using chi-square assessments.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 56

Findings

Of eligible participants (N=57,079) during the 16-month program, only 7,072 individuals (14.1 percent)

contributed at least one valid monthly payment after receiving the card and received cost-sharing protection for

the following month. Table 30 presents a demographic profile of the Health Independence Accounts eligible

population by those who made no payments or made at least one payment.

Table 30. Health Independence Account Eligible Population Demographic Profile

Made No Payments

Made at Least One Payment

Category

N (Percent)

Total

Age (Years)

19-34

24,154 (48.3)

1,508 (21.3)

25,664 (45.0)

35-49

14,279 (28.9)

1,954 (27.6)

16,233 (28.4)

50-64

11,572 (23.1)

3,610 (51.1)

15,182 (26.6)

Gender

Male

20,130 (40.3)

2,831 (40.0)

22,961 (40.2)

Female

29,877 (59.8)

4,241 (60.0)

34,118 (59.8)

Race / Ethnicity

White

32,073 (64.1)

4,400 (62.2)

36,473 (63.9)

Black

7,595 (15.2)

1,050 (14.9)

8,645 (15.2)

Hispanic

1,865 (3.7)

324 (4.6)

2,189 (3.8)

Other

8,474 (17.0)

1,298 (18.4)

9,772 (17.1)

Total

50,007 (85.9)

7,072 (14.1)

57,079 (100)

Those who made payments were disproportionately older

(Table 32) and have more comorbidities (Figure 10).

Compared to non-paying individuals, those making payments

were more likely to have prior hospitalizations (7.1 percent vs.

5.1 percent), but less likely to have had an emergency room

visit (25.7 percent vs. 27.7 percent; p0.0001, data not shown).

No differences in participation were observed for racial or

gender strata or for individuals in the 100-to-117 percent FPL

group ($10/month) compared to the 118-to-138 percent FPL

group ($15/month).

For the 7,072 who contributed, their monthly contributions

totaled $426,670 during the 18-month period — median of

$40 or four months’ payments (interquartile range $15, $90)

and mean of $60 (standard deviation $141) or six months of

payments. Cost-sharing protections totaled $476,843, with a

majority of the protection being for pharmaceutical

copayments (see Table 31).

Table 31. Health Independence Accounts Cost-Sharing Expenditures

Type

Transactions

Percent

Amount ($)

Pharmaceuticals

31,805

61.2

$289,522

Physician

9,198

17.7

$79,482

Non-MD Clinician

7,339

14.1

$59,095

Hospitals

2,884

5.6

$41,744

Other

710

1.4

$7,000

Total

51,936

100

$476,843

Figure 10. Charlson Comorbidity Index by Health

Independence Account Non-Participants (Made

No Payment) vs. Participants (Made Payment)

Arkansas Health Care Independence Program (‘Private Option’) Final Report 57

Cost-sharing protections exceeded HIA payments by $50,172. Cost-avoidance — an individual’s copayment

protections exceeding their payments — was realized by 23.4 percent of participants. Administrative costs to

maintain program and accounts were approximately $9 million throughout the program duration of 18 months.

Postlude

The HIAs were terminated by the Arkansas General Assembly in the 2

Extraordinary Session of 2016. (Ark. Code

Ann. § 20-77-2408). There was no evidence of individuals directly utilizing account balances during the program

for private-sector premium payments. Upon termination of the program, individuals with account balances

(N=2,253) received program termination checks.

Medicaid and QHP Opioid Utilization Outcomes

Overview

Over the last decade, the United States has seen a significant rise in opioid overdose deaths and prescriptions — a

fivefold increase in prescription opioid overdose deaths and a fourfold increase in the sale of prescription

opioids.

Arkansas is no stranger to this epidemic. In 2016, Arkansas had the second highest rate of opioid

prescriptions nationally at 114.6 prescriptions per 100 persons.

While Arkansas has not yet seen the volume of

opioid overdose deaths seen in other states, trends indicate that Arkansas could soon approach those levels

absent intervention.

In an effort to address the issue, the Arkansas General Assembly has enacted legislation on multiple fronts.

1. In 2011, the state authorized a prescription drug monitoring program (PDMP) to enhance patient care by

ensuring legitimate use of controlled substances.

2. In 2015, the state established immunity for individuals seeking medical assistance during a drug overdose and

for healthcare professionals for administering, prescribing, or dispensing naloxone, which is used to treat a

narcotic drug overdose.

3. In 2017, Arkansas joined 38 other states that require prescribers to check the PDMP prior to prescribing

certain controlled substances.

In March 2016, the Centers for Disease Control and Prevention (CDC) issued guidelines

for primary care clinicians

for prescribing opioids for chronic pain outside of cancer treatment, palliative care, and end-of-life care. The

guidelines recommended non-opioid therapy as the preferred method for treatment of chronic pain. However,

when opioids are used, the CDC recommended that clinicians should prescribe the lowest effective dosage, assess

risks when considering increasing dosage to 50 morphine milligram equivalents (MMEs) or more per day, and

avoid concurrent opioids and benzodiazepines.

In response to a 2016 memo from the Centers for Medicare & Medicaid Services (CMS)

indicating that

“Medicaid beneficiaries are prescribed painkillers at twice the rate of non-Medicaid patients and are at three-to-

six times the risk of prescription painkillers overdose,” Arkansas Medicaid implemented best practices cited in the

memo as its new policy. Arkansas has also issued guidelines for emergency department opioid use, and, most

recently, the state has launched an effort to educate doctors on pain management in an effort to curb opioid

misuse and abuse.

Given the high rates of opioids prescribed in Arkansas, the well-documented risks associated with high-dose and

long-term opioid use, and the national and local responses to these issues, we compared the rates of opioid

Arkansas Health Care Independence Program (‘Private Option’) Final Report 58

analgesic use, high-risk opioid use (high dose opioid use or opioid use with a benzodiazepine), initial opioid use

with more than seven and more than three days supplied, and the extent of naloxone use for those prescribed

opioids between newly enrolled persons obtaining coverage in traditional Medicaid and QHPs from 2014-16.

Data Sources and Analytic Population

Data sources and study subjects included in the opioid analysis are the General and Higher Needs Populations

described in previous sections.

Opioid Utilization Outcome Measures

All opioid analgesic prescriptions were converted to morphine equivalent doses using standardized conversion

tables. Opioid-related differences between Medicaid and QHPs were assessed using five outcome measures,

including:

1. Any opioid use, which was calculated annually as a binary indicator whether an individual

received at least one opioid analgesic prescription in a calendar year. It should be noted that only

oral and transdermal opioid analgesics were examined, with opioid-containing cough and cold

preparations excluded.

2. High dose opioid use, which was defined as the receipt of at least one day of an opioid regimen

where the sum daily dose exceeded 90 Morphine Milligram Equivalents (MME).

3. Concomitant benzodiazepine and opioid use, which was defined as observing at least one claim

for both an opioid and a benzodiazepine in a calendar year.

4. Initial duration of the first opioid prescription received by individuals utilized a smaller sample of

new opioid users. In order to identify initial opioid use, this measure was assessed only in those

subjects who had six continuous months of enrollment without any opioid use prior to their first

opioid prescription. Based on the CDC guidelines, we used thresholds of three days and seven

days to identify initial opioid use that is more likely to lead to chronic opioid use.

5. We calculated the total number of opioid prescriptions dispensed per 100 person-years. Opioid

use was further stratified by Schedule III-V opioid use, short-acting Schedule II opioid use and any

long-acting Schedule II opioid use. The number of opioid prescriptions were calculated based on

the number of opioid prescriptions received by an individual in a calendar year and for descriptive

reporting this was converted to a rate by dividing total opioid use by the person-years observed in

each sample for each calendar year.

All measures were calculated annually, except for duration of initial opioid prescription measures, for which we

report use over all three study years.

General Population Propensity Score Matching

The General Population contains all newly enrolled individuals 19-64 years of age with at least 180 days of

continuous enrollment with either Medicaid or QHP coverage in two consecutive years between 2014 and 2016, and

who did not complete an exceptional healthcare needs assessment Questionnaire. In total, this analytic population

contains 120,845 individuals — 92,940 assigned to a QHP and 27,905 covered by Medicaid. A propensity score was

generated using variables from the first year in the program: age, sex, race/ethnicity, median household income

based on census block group residence, insurance market region, Charlson Comorbidity Index and duration of

enrollment. Enrollees were matched using a one-to-one greedy matching algorithm without replacement. A

summary of covariates for unmatched and matched samples are reported in Table 32.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 59

Table 32. Characteristics of Unmatched and Propensity Score Matched General Population

Unmatched (N=120,845)

PS Matched (N=55,676)

Medicaid

QHP

Medicaid

QHP

(N=27,905)

(N=92,940)

(N=27,838)

Continuous Variables

Unit

Mean

(StdErr)

Mean

(StdErr)

p-value

Mean

(StdErr)

Mean

(StdErr)

p-value

Age

Years

36.0 (0.07)

39.0 (0.04)

<0.001

36.0 (0.06)

0.795

Household Income

Dollars

40,876 (95)

38,973 (52)

<0.001

40,854 (99)

40,976 (99)

0.383

Enrollment Duration

Days

817 (1.2)

887 (0.7)

<0.001

817 (1.2)

813 (1.2)

0.010

Categorical Variables

Category

N (percent)

p-value

N (percent)

p-value

Sex

Male

1,015 (36.4)

4,375 (47.1)

<0.001

1,016 (36.5)

1,037 (37.3)

0.066

Female

1,775 (63.6)

4,918 (52.9)

1,767 (63.5)

1,746 (62.8)

Race/Ethnicity

Black

4,657 (16.7)

2,592 (27.9)

<0.001

4,663 (16.8)

4,389 (15.8)

<0.001

Hispanic

828 (3.0)

2,312 (2.5)

827 (3.0)

883 (3.1)

Others

7,092 (25.4)

1,316 (14.2)

7,011 (25.2)

6,646 (23.9)

White

1,532 (54.9)

5,154 (55.5)

1,533 (55.1)

1,592 (57.2)

Insurance Market Region

Central

8,517 (30.5)

2,667 (28.7)

<0.001

8,491 (30.5)

0.290

Northeast

5,309 (19.0)

1,886 (20.3)

5,303 (19.1)

5,395 (19.4)

Northwest

4,664 (16.7)

1,323 (14.2)

4,652 (16.7)

4,722 (17.0)

South

Central

1,639 (5.9)

6,207 (6.7)

1,638 (5.9)

1,722 (6.2)

Southeast

2,396 (8.6)

1,154 (12.4)

2,396 (8.6)

2,305 (8.3)

Southwest

2,303 (8.3)

8,378 (9.0)

2,301 (8.3)

2,196 (7.9)

West

Central

3,077 (11.0)

8,038 (8.7)

3,057 (11.0)

3,007 (10.8)

Charlson Comorbidity Index

1,823 (65.3)

5,681 (61.1)

<0.001

1,817 (65.3)

1,819 (65.4)

0.959

1-2

7,330 (26.3)

2,705 (29.1)

7,329 (26.3)

7,269 (26.1)

3-8

2,235 (8.0)

8,743 (9.4)

2,234 (8.0)

2,274 (8.2)

>=9

106 (0.4)

329 (0.4)

105 (0.4)

104 (0.4)

Note: Household income is compiled as the average overall census block group median household incomes.

Abbreviations: N=number of persons; QHP=qualified health plan; PS=propensity score; StdErr=standard error

Prior to propensity score matching (unadjusted), QHP enrollees were older, were more likely to be white, were

more likely to be male, and had fewer comorbidities than their Medicaid counterparts. After propensity score

matching, no differences between groups were observed with the exception of small differences in

race/ethnicity categories.

The opioid utilization measures for Medicaid and QHP propensity score matched populations are presented in

Table 33.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 60

Table 33. Opioid Use Measures for Propensity Score Matched General Population Medicaid and QHP

by Assessment Period

Medicaid

QHP

Measure

Period

Population

Number

Percent

Population

Number

Percent

Any Opioid Use

Baseline

27,838

9,698

34.8

27,838

8,834

31.7

Year 1

27,837

8,996

32.3

27,837

9,560

34.3

Year 2

12,657

4,097

29.4

12,657

4,832

38.2

Received High Opioid

Dose (>90 MME)

Baseline

27,838

213

2.2

27,838

589

6.7

Year 1

27,837

214

2.4

27,837

685

7.2

Year 2

12,657

103

2.5

12,657

361

7.5

Benzodiazepine and

Opioid Use

Baseline

27,838

2,947

30.4

27,838

2,855

32.3

Year 1

27,837

2,621

29.1

27,837

2,949

30.9

Year 2

12,657

1,039

25.4

12,657

1,389

28.8

Duration of First Opioid Prescription >3 days

6,926

5,112

73.8

6,926

4,479

64.7

Duration of First Opioid Prescription >7 days

6,926

1,711

24.7

6,926

1,571

22.7

Measure

Period

Person-

Years

Number

Dispensed

Rate

Person-

Years

Number

Dispensed

Rate

Opioid Prescriptions

Dispensed (per 100

person-years)

Baseline

23,726

32,658

137.7

24,383

38,274

157.0

Year 1

26,753

34,379

128.5

26,253

46,040

175.4

Year 2

11,820

14,605

123.6

11,333

24,371

215.1

Notes: Year 1 and Year 2 indicate follow-up years after baseline measurement. For those with enrollment beginning in 2014, follow-up

Year 1 is 2015 and follow-up Year 2 is 2016. For those with enrollment beginning in 2015, follow-up Year 1 is 2016, and no follow-up Year

2 is available.

Abbreviation: QHP=qualified health plan; MME=morphine milligram equivalent

Generalized linear models were fitted by regressing a Medicaid/QHP dichotomous primary independent indicator

on the outcomes of interest. For binary outcomes, generalized logistic regression models were used and adjusted

odds ratios (AOR) were reported. For count variables, generalized Poisson regression models were used and least-

square (LS) means were reported for each group. For the initial duration of an opioid prescription measure,

individuals having any opioid use with at least six months of prior enrollment without opioid use were matched.

Additional details about the propensity score approach are described in Appendix E.

A comparison of opioid utilization differences between Medicaid and QHP enrollees in the General Population is

presented in Table 34.

Table 34. Comparison of Opioid Utilization for General Population Medicaid and QHP Enrollees

Measure

(QHP is referent group)

Period

Adjusted

Odds Ratio

95% Confidence

Interval

p-value

Any Opioid Use

Baseline

1.151

1.111

1.193

<0.001

Year 1

0.912

0.881

0.945

<0.001

Year 2

0.682

0.648

0.718

<0.001

Received High Opioid Dose (>90 MME)

Baseline

0.356

0.304

0.417

<0.001

Year 1

0.307

0.263

0.358

<0.001

Year 2

0.281

0.226

0.350

<0.001

Benzodiazepine and Opioid Use

Baseline

1.036

0.981

1.094

0.201

Year 1

0.877

0.829

0.927

<0.001

Year 2

0.735

0.676

0.798

<0.001

Duration of First Opioid Prescription >3 Days

3 years

1.540

1.432

1.657

0.037

Duration of First Opioid Prescription >7 Days

3 years

1.119

1.034

1.210

0.005

Notes: Year 1 and Year 2 indicate follow-up years after baseline measurement. For those with enrollment beginning in 2014, Year 1 is

2015 and Year 2 is 2016. For those with enrollment beginning in 2015, Year 1 is 2016, and no Year 2 is available.

Abbreviations: QHP=qualified health plan; MME=morphine milligram equivalent

Arkansas Health Care Independence Program (‘Private Option’) Final Report 61

During the first year of enrollment (baseline) and compared to QHP enrollees, Medicaid enrollees were 1.15 times

as likely to have had at least one opioid prescribed. However, by the first and second follow-up years of

enrollment the effect reversed and Medicaid enrollees were only 0.91 and 0.68 times, respectively, to have any

opioid use compared to QHP enrollees. Medicaid enrollees were 0.36, 0.31, and 0.28 times as likely to receive a

high opioid dose at baseline and in follow-up years 1 and 2, respectively. Medicaid enrollees were 0.88 and 0.74

times as likely to have both benzodiazepine and opioid use in follow-up years 1 and 2, compared to QHP

enrollees. Compared to QHP enrollees, Medicaid enrollees were 1.54 times more likely to have a first opioid

prescription duration longer than three days and 1.12 times as likely to have a first prescription for duration

longer than seven days.

Table 35 presents analyses comparing the number of opioids prescribed to Medicaid and QHP enrollees.

Table 35. Comparison of Number of Opioid Prescriptions by General Population Medicaid and QHP

(Reference) Enrollees

Number of Opioid Prescriptions

Plan Type

Least Squares

Mean

Difference

Standard Error of

Difference

p-value

Baseline

Medicaid

1.120

0.122

0.008

<0.001

QHP

1.242

Year 1

Medicaid

1.080

0.306

0.007

<0.001

QHP

1.385

Year 2

Medicaid

0.982

0.431

0.011

<0.001

QHP

1.413

Notes: Year 1 and Year 2 indicate follow-up years after baseline measurement. For those with enrollment beginning in 2014,

follow-up Year 1 is 2015 and follow-up Year 2 is 2016. For those with enrollment beginning in 2015, follow-up Year 1 is 2016, and

no follow-up Year 2 is available.

Abbreviation: QHP=qualified health plan

For the General Population, Medicaid enrollees had a lower number of total opioid prescriptions, on average,

than QHP enrollees. There was an adjusted absolute difference of 0.122 prescriptions at baseline, 0.306 at follow-

up Year 1 and 0.431 at follow-up Year 2. Over the three measurement period years, the average number of total

opioids prescribed to Medicaid enrollees continuously declined (1.12, 1.08, 0.98) while there was a steady

increase for QHP enrollees (1.24, 1.39, 1.41).

Higher Needs Population Regression Discontinuity Analysis

The Higher Needs Population contains all newly enrolled individuals 19-64 years of age with at least 180 days of

continuous enrollment with either Medicaid or QHP coverage in two consecutive years from 2014 to 2016, and

who completed an exceptional healthcare needs assessment. For this population, enrollees were assigned to

Medicaid and QHP groups on the basis of responses to the assessment resulting in a continuous composite score

threshold cut-point. A descriptive demographic profile of the Higher Needs Population Medicaid and QHP enrollee

groups contained in these analyses is presented in Table 36.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 62

Table 36. Characteristics of Higher Needs Population

Medicaid (N=9,037)

QHP (N=44,285)

Continuous Variables

Measure

Mean (StdErr)

p-value

Age

Years

36.0 (0.07)

38.9 (0.04)

<0.001

Household Income

Dollars

40,704 (172)

41,327 (78)

<0.001

Enrollment Duration

Days

980 (2.0)

971 (0.9)

<0.001

Categorical Variables

Category

N (percent)

p-value

Sex

Male

3,274 (36.2)

1,781 (40.2)

<0.001

Female

5,763 (63.8)

2,646 (59.8)

Race/Ethnicity

Black

1,431 (15.8)

7,969 (18.0)

<0.001

Hispanic

130 (1.4)

853 (1.9)

Others

1,243 (13.8)

7,675 (17.3)

White

6,233 (69.0)

2,778 (62.8)

Insurance Market Region

Central

2,692 (29.8)

1,383 (31.3)

0.037

Northeast

1,800 (19.9

7,784 (17.6)

Northwest

1,471 (16.3)

8,259 (18.7)

South Central

651 (7.2)

3,430 (7.8)

Southeast

869 (9.6)

3,793 (8.6)

Southwest

670 (7.4)

3,094 (7.0)

West Central

884 (9.8)

4,087 (9.2)

Charlson Comorbidity Index

2,666 (29.5)

2,579 (58.2)

<0.001

1-2

4,178 (46.2)

1,468 (33.2)

3-8

2,104 (23.3)

3,699 (8.4)

>=9

89 (1.0)

109 (0.3)

Notes: Household income is compiled as the average overall census block group median household incomes.

Abbreviations: N=number of persons; QHP=qualified health plan; StdErr=standard error

In general, for the Higher Needs Population, Medicaid enrollees were younger, were more likely to be white, and

were more likely to have higher comorbidity burden than the QHP enrollees.

An unadjusted descriptive analysis of opioid utilization for Medicaid and QHP propensity score matched

populations is presented in Table 37.

Table 37. Opioid Use Measures for Propensity Score Matched Higher Needs Population Medicaid and

QHP by Assessment Period

Medicaid

QHP

Measure

Period

Population

Number

Percent

Population

Number

Percent

Any Opioid Use

2014

9,037

5,165

57.1

44,285

15,951

36.0

2015

9,037

4,653

51.5

44,285

16,600

37.5

2016

7,951

3,943

49.6

37,752

14,386

38.1

Received High Opioid Dose (>90

MME)

2014

9,037

188

3.6

44,285

917

5.8

2015

9,037

210

4.5

44,285

1,115

6.7

2016

7,951

195

5.0

37,752

1,065

7.4

Benzodiazepine and Opioid Use

2014

9,037

2,176

42.1

44,285

4,396

27.6

2015

9,037

2,036

43.8

44,285

4,801

28.9

2016

7,951

1,503

38.1

37,752

3,670

25.5

Duration of First Opioid Prescription >3 days

3,310

2,552

77.1

16,202

10,822

66.8

Duration of First Opioid Prescription >7 days

3,310

1,094

33.1

16,202

3,732

23.0

Measure

Period

Person-

Years

Number

Dispensed

Rate

Person-

Years

Number

Dispensed

Rate

Opioid Prescriptions Dispensed

(per 100 person-years)

2014

8,038

22,883

284.7

39,336

62,379

158.6

2015

8,593

23,549

274.1

41,953

74,393

177.3

2016

7,625

20,182

264.7

36,381

68,197

187.5

Abbreviations: QHP=qualified health plan; MME=morphine milligram equivalent

Arkansas Health Care Independence Program (‘Private Option’) Final Report 63

Appendix E contains the technical statistical details for the regression discontinuity approach. Using this approach,

a comparison of opioid utilization differences between Medicaid and QHP enrollees in the Higher Needs

Population is presented in Table 38.

Table 38. Comparison of Opioid Utilization for Medicaid and QHP Higher Needs Population Enrollees

Measure (QHP is referent group)

Period

Local Average

Treatment Effect

Standard

Error

p-value

Any Opioid Use

Year 1

-151.5

274.6

0.581

Year 2

-1,132.0

245.2

<0.001

Year 3

-929.7

210.3

<0.001

Received High Opioid Dose (>90 MME)

Year 1

-436.7

45.6

<0.001

Year 2

-442.4

50.5

<0.001

Year 3

-431.0

52.1

<0.001

Benzodiazepine and Opioid Use

Year 1

-21.4

115.5

0.853

Year 2

-182.9

93.3

0.049

Year 3

-323.0

48.8

<0.001

Duration of First Opioid Prescription >3 Days

3 years

0.011

0.026

0.687

Duration of First Opioid Prescription >7 Days

3 years

-0.002

0.037

0.962

Number of Opioid Prescriptions

Year 1

-1.569

0.237

<0.001

Year 2

-1.956

0.253

<0.001

Year 3

-1.662

0.184

<0.001

Notes: Local Average Treatment Effect is per 10,000 persons, with the exception of the duration and number of opioid prescription

measures, which are represented at the person level.

Abbreviations: QHP=qualified health plan; MME=morphine milligram equivalent

Compared to QHP enrollees, 1,132 and 930 fewer Medicaid enrollees per 10,000 persons in the Higher Needs

Population had at least one opioid prescribed in 2015 and 2016, respectively. Medicaid enrollees also had lower

rates of high dose opioid use than QHP enrollees by approximately 440 fewer recipients per 10,000 person-year

between in years 2014 through 2016. Medicaid enrollees also had fewer concomitant benzodiazepine users in

2015 and 2016.

No difference in the duration of first opioid prescription was detected between Medicaid and QHP enrollees. The

number of opioids prescribed was found to be significantly lower for Medicaid enrollees in each of the program

years: 1.57 fewer prescriptions per person-year in 2014; 1.96 fewer prescriptions per person-year in 2015; and

1.66 fewer prescriptions per person-year in 2016.

The adjusted finding in the Higher Needs Population is similar to those concluded in the General Population. More

detailed descriptive results, including schedule of opioids, hydrocodone, oxycodone, and tramadol prescriptions,

are presented in Appendix J.

Discussion

In both the General Population and the Higher Needs Population, especially past the first year of enrollment,

enrollees in QHPs had significantly higher opioid utilization. Of those prescribed opioids, QHP enrollees had higher

percentages of prescribed high doses and had more opioid prescriptions filled than those in Medicaid. Program

differences became more exaggerated the longer enrollees were covered.

These differences can largely be explained by higher levels of drug utilization management in Medicaid. Medicaid

policy allows a maximum of 93 units of any solid oral short-acting opioid per 31 days,

which would substantially

reduce the chances of someone exceeding 90 MME. Medicaid also limits short- and long-acting opioids in addition

to limits for benzodiazepine prescriptions for individuals with a prior diagnosis of poisoning from opioids,

narcotics, barbiturates, or benzodiazepine in the previous three months.

In terms of limiting initial opioid

Arkansas Health Care Independence Program (‘Private Option’) Final Report 64

prescribing to three or seven days, we found that the first opioid prescription for approximately two-thirds of

enrollees exceeds a three-day supply, and a fourth of enrollees receive a prescription for more than seven days

supplied. In the propensity score analysis, we also found that enrollees in Medicaid were more likely to have

opioid prescriptions with longer days supplied. The risk of long-term opioid use begins to increase on the third day

supplied,

and efforts to conform to the CDC guidelines

should be undertaken.

It appears that expanding coverage through a QHP strategy in which there are fewer restrictions to opioid

prescribing may be contributing to the state’s high prescribing rate. Some recent state policy efforts should

mitigate opioid misuse, including requirements for prescribers to check the prescription drug monitoring program

as per state legislative mandate, but this effort will likely only affect opioid use where misuse is suspected.

Strategies should be undertaken to reduce opioid prescribing in Medicaid and QHPs, but additional focus is

warranted in QHP plans. Limitations on dispensed opioids for initial prescriptions should be considered for both

Medicaid and QHPs. This is especially true with respect to high dose opioid use given the escalating opioid

utilization rate and the well-documented overdose risk associated with high dose opioid use.

Urgent Care Utilization

Overview

As shown in Table 39, there was a decrease in the rate of non-emergent emergency room visits between each

year of study in the Higher Needs Population. One possible explanation for this decrease in ER utilization for non-

emergent use might be explained by an increase in urgent care visits. To investigate this, we analyzed urgent care

clinic utilization, which would allow us to determine if there was an associated increase in urgent care utilization.

Table 39. Change in Estimated Rate of Non-Emergent Emergency Room Visits per 100 Person-Years

Baseline

Mean (StdErr)

Year 2

Mean (StdErr)

Year 3

Mean (StdErr)

Percent Change

Baseline to Year 2

Year 2 to Year 3

General Population (Propensity Score Matched Comparison)

Medicaid

41.6 (0.7)

45.5 (1.1)

9.4

QHP

37.1 (0.6)

44.2 (1.0)

19.1

Higher Needs Population (Regression Discontinuity Comparison)

Medicaid

74.4 (1.7)

60.4 (1.5)

53.8 (1.5)

-18.8

-10.9

QHP

31.1 (0.4)

27.8 (0.4)

27.1 (0.4)

-10.6

-2.5

Notes: Due to the manner in which we used Baseline clinical comorbidity to propensity score match, no General Population can be

made with Baseline.

Abbreviations: QHP=qualified health plan; StdErr=standard error

Approach

Both the General Population and the Higher Needs Population were included in this analysis. To determine

whether or not an individual utilized urgent care, we extracted claims based on the appropriate place of service

code (CMS Place of Service Code: 20 – Urgent Care Facility). Unfortunately, after the initial claim extraction, we

discovered that there were only two instances in the Medicaid samples of each population in which an urgent

care facility was listed as the place of service (Table 40).

Arkansas Health Care Independence Program (‘Private Option’) Final Report 65

Table 40. Unadjusted Analysis of Urgent Care Facility Utilization and Cost

Visits

Urgent Care Facility Visits per

100 person-years

Total Costs ($)

General Population

Medicaid

Year 1

27,905

Year 2

27,905

Year 3

13,933

0.02

273

QHP

Year 1

92,940

774

0.93

111,489

Year 2

92,940

1,886

2.11

269,346

Year 3

56,874

2,100

3.97

281,317

Higher Needs Population

Medicaid

Baseline

9,037

0.00

Year 2

9,037

0.00

Year 3

7,951

0.03

237

QHP

Baseline

44,285

459

1.17

62,683

Year 2

44,285

1,018

2.43

151,562

Year 3

37,752

2,086

5.73

286,167

Notes: Visit counts were derived from claims data where the place of service was listed as an urgent care facility. Visits were

defined as the same person on the same day at the same facility. Cells with “--” indicate no associated costs.

Abbreviations: N=number of eligible Medicaid and QHP enrollees per group; QHP=qualified health plan

Discussion

While we could not examine whether or not urgent care clinic utilization reflected the decrease in Medicaid

recipients, we did observe an increasing trend in urgent care facility usage under the QHPs. It may be that

providers are not submitting claims under a place of service code for urgent care because of a Medicaid policy.

However, it is also possible that most urgent care clinics do not take Medicaid because of a programmatic effect.

We did attempt to find claims for urgent care clinics using National Provider Identifier; however, this methodology

also proved to be problematic because it was difficult to identify distinct clinics from other types of providers that

may be associated with the clinics. It would be interesting to look deeper into urgent care use compared to non-

emergent ER use in Arkansas in general to determine if there is an association between the two among a broader

population. If there is, there may be cost benefits for Medicaid to encourage urgent care facilities to agree to take

Medicaid patients.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 66

Mortality

Overview

A clear association exists between having health insurance coverage and a reduced risk of death.

The uninsured

lack the coverage necessary to regularly obtain recommended preventive screenings and appropriate disease

management. The Institute of Medicine — using findings from 130 studies — concluded in 2002 that not having

health insurance was a pre-eminent factor in the increase of all-cause mortality risk.

By expanding Medicaid to low-income uninsured adults, the PPACA has likely improved the survival of those who

were formerly uninsured. Previously in this report, we documented the findings that QHP enrollees had

significantly higher screening and disease-management rates.

Following National Advisory Committee (NAC) review of Interim Report

findings, considered indicators were

modified to include 7.1. Crude and Age adjusted mortality rates (see Appendix A). These represent the ultimate

outcome indicator of differences between QHP- and Medicaid-administered healthcare services.

Data Source and Study Population

Mortality data for 2014–2016 were obtained from the mortality vital records at the Arkansas Department of

Health. A unique study ID was placed on the mortality data files using the same methodology and personal

identifier match engine that was used to generate identifiers on all enrollment and claims data files used in this

evaluation. For this study, we compared newly enrolled traditional Medicaid enrollees (in or later than January

2014) in aid category 20 or 25, with those enrolled in a QHP under the Private Option. We restricted our study

population to enrollees who did not complete the exceptional healthcare needs assessment Questionnaire. To

account for early program switching, we only included those with at least 90 consecutive days in a program and

studied enrollees in programs beginning on their 91st day of enrollment. Observation time for death events

started on day 91 of enrollment through date of death, termination of coverage, or Dec. 31, 2016 (the end of the

evaluation period), whichever occurred first. Any time after a post-91st day Medicaid-to-QHP switch, or vice

versa, was censored and not included.

In the General Population, 36,182 traditional Medicaid enrollees and 176,081 QHP enrollees were eligible for the

mortality assessment. Of those, 72 mortalities were observed among the traditional Medicaid enrollees (0.2

percent) through Dec. 31, 2016, and 827 mortalities were observed among QHP enrollees (0.5 percent).

Traditional Medicaid and QHP enrollees were one-to-one matched based on propensity scores derived from

modeling characteristics of enrollees in traditional Medicaid and QHPs. Table 41 presents enrollee characteristics

after propensity score matching. Characteristics with an absolute standardized difference of 0.10 or higher

indicate significant categorical differences between Medicaid and QHP enrollees. A standardized difference of less

than 0.10 indicates that traditional Medicaid and QHP enrollees are balanced on that characteristic. In total,

35,917 enrollees in each program were matched (99.3 percent of Medicaid enrollees) using a greedy matching

algorithm (see Appendix E for additional matching details).

Arkansas Health Care Independence Program (‘Private Option’) Final Report 67

Table 41. Medicaid and QHP Enrollee General Population Propensity Score Matched Characteristics

Characteristic

Category

Medicaid 20/25

N=35,917 (percent)

QHP

N= 35,917 (percent)

Standardized

Difference

Insurance Market Region

Central

10,646 (29.7)

10,756 (30.0)

0.06

Northeast

6,635 (18.5)

6,730 (18.7)

Northwest

6,464 (18.0)

6,509 (18.2)

South Central

2,087 (5.8)

1,997 (5.6)

Southeast

3,111 (8.7)

2,940 (8.2)

Southwest

3,108 (8.7)

3,023 (8.4)

West Central

3,873 (10.8)

3,962 (11.0)

RUCA

Urban

20,847 (58.0)

20,793 ( 57.9)

0.00

Any Rural

15,070 (42.0)

15,124 (42.1)

Median Household Income

Quintile

7,984(22.2)

8,023 (22.3)

0.00

7,579 (21.1)

7,710 (21.5)

7,053 (19.6)

7,075 (19.1)

6,863 (19.1)

6,851 (19.1)

6,438 (17.9)

6,258 (17.4)

Charlson Comorbidity Index

30,277 (84.3)

31,074 (86.5)

-0.06

3,646 (10.2)

3,254 (9.1)

1,365 (3.8)

1,118 (3.1)

3 or higher

629 (1.8)

471 (1.3)

Sex

Female

25,954 (72.3)

25,995 (72.4)

0.00

Male

9,963 (27.7)

9,922 (27.6)

Race/

Ethnicity

Black

7,478 (20.8)

7,057 (20.0)

0.08

Hispanic

1,423 (4.0)

1,217 (3.4)

Other

2,500 (7.0)

2,290 (6.4)

White

24,516 (68.3)

25,353 (70.6)

Enrollment

Quarter

2014 Q1

2,613 (7.3)

2,671 (7.4)

0.00

2014 Q2

6,132 (17.1)

6,183 (17.2)

2014 Q3

3,304 (9.2)

3,267 (9.1)

2014 Q4

3,462 (9.6)

3,489 (9.7)

2015 Q1

2,385 (6.6)

2,409 (6.7)

2015 Q2

1,774 (4.9)

1,741 (4.9)

2015 Q3

2,184 (6.1)

2,161 (6.0)

2015 Q4

1,751 (4.9)

1,660 (4.6)

2016 Q1

3,732 (10.4)

3,756 (10.5)

2016 Q2

2,484 (6.9)

2,469 (6.9)

2016 Q3

3,180 (8.9)

3,214 (9.0)

2016 Q4

2,916 (8.1)

2,896 (8.1)

Age (years)

18-24

5,566 (15.5)

5,552 (15.5)

0.00

25-29

8,268 (23.0)

8,138 (22.7)

30-34

7,618 (21.2)

7,645 (21.3)

35-39

6,272 (17.5)

6,431 (17.9)

40-44

4,047 (11.3)

4,057 (11.3)

45-49

2,239 (6.2)

2,164 (6.0)

50-54

1,196 (3.3)

1,162 (3.2)

55-59

540 (1.5)

575 (1.6)

60-64

171 (0.5)

193 (0.5)

Notes: If any RUCA classification indicated the enrollee was residing in a rural region, a rural designation was assigned. Median

household income quintiles were derived from assigning enrollees a median household income based on the census block group in

which they resided.

Abbreviations: QHP=Qualified health plan; RUCA=Rural-urban commuting area code; Q1-Q4=quarter that individuals enrolled.

After propensity score matching, the traditional Medicaid and QHP enrollee populations are well balanced over all

characteristics used in the creation of the propensity score.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 68

Statistical Approach and Findings

Proportional hazard models tested for differential time to event across variables. After adjusting for variables

(covariates) associated with variability in death (e.g., age, poor health), we hypothesized no difference in the

number, or timing of deaths experienced by both traditional Medicaid and QHP enrollees. Hazard ratios were

constructed as measures to identify the effect of being enrolled in traditional Medicaid or a QHP on mortality.

Fully adjusted Cox proportional hazards time to event models were estimated using enrollment in Medicaid or

QHPs as the exposure variable and enrollee characteristics as covariates, including: patient age (as of the 91st day

of enrollment), sex, rural/urban status of residence, Charlson comorbidity index based on the first 90 days claims

experience (prior to study start date), an indicator for year and quarter of enrollment date, census block group

median income quintile, and insurance market region.

Table 42 presents the results of the fitted, adjusted Cox proportional hazards model. In the model, the outcome is

all-cause mortality. The main variable under study was plan type and Medicaid enrollees were the

“reference/control group,” while QHP enrollees constituted the “treatment group.”

Arkansas Health Care Independence Program (‘Private Option’) Final Report 69

Table 42. Medicaid and QHP Enrollees General Population Propensity Score Matched Characteristics

Characteristic

Category

Hazard Ratio

95% Confidence Interval

p-value

Plan Type (Reference = Medicaid)

QHP

1.256

0.913

1.728

0.162

AGE in Years (Reference = 18-24)

25-29

1.325

0.466

3.764

0.598

30-34

3.119

1.207

8.062

0.019

35-39

2.816

1.072

7.395

0.036

40-44

4.077

1.553

10.703

0.004

45-49

6.469

2.447

17.097

<0.001

50-54

7.149

2.589

19.738

<0.001

55-59

8.235

2.760

24.573

<0.001

60-64

9.404

2.458

35.984

<0.001

Region Code

(Reference = Northwest)

Central

1.433

0.86

2.386

0.167

Northeast

1.26

0.695

2.285

0.446

South Central

1.534

0.74

3.179

0.250

Southeast

2.934

1.567

5.493

<0.001

Southwest

1.589

0.754

3.349

0.223

West Central

1.111

0.571

2.16

0.756

RUCA (Reference = Urban)

Any rural

0.769

0.536

1.104

0.155

Income Quintile

(Reference = 1)

0.907

0.574

1.431

0.674

1.027

0.639

1.649

0.913

0.846

0.506

1.413

0.522

0.497

0.262

0.944

0.033

Charlson Comorbidity Index

(Reference = 0)

2.038

1.299

3.198

0.002

3.74

2.256

6.199

<0.001

9.935

6.219

15.871

<0.001

Gender (Reference = Female)

Male

2.335

1.691

3.224

<0.001

Race/Ethnicity

(Reference = White)

Black

0.522

0.312

0.875

0.014

Hispanic

0.502

0.123

2.046

0.337

Other

0.669

0.31

1.444

0.306

Enrollment Quarter

(Reference = 2014 Q1)

2014 Q2

1.501

0.563

4.007

0.417

2014 Q3

1.091

0.422

2.819

0.857

2014 Q4

1.656

0.629

4.357

0.307

2015 Q1

1.621

0.605

4.346

0.337

2015 Q2

1.642

0.583

4.621

0.348

2015 Q3

1.357

0.429

4.289

0.603

2015 Q4

1.91

0.579

6.301

0.288

2016 Q1

1.329

0.415

4.254

0.632

2016 Q2

1.799

0.470

6.878

0.391

2016 Q3

2.318

0.593

9.064

0.227

2016 Q4

4.916

0.856

28.218

0.074

Note: Hazard ratio describes the mortality odds ratio of a patient belonging to a certain category relative to the reference case.

Abbreviations: QHP=Qualified health plan; RUCA=Rural-urban commuting area code; Q1-Q4=quarter that individuals enrolled.

After adjusting for individual characteristics (covariates), traditional Medicaid 20/25 enrollees were no more likely

to die than QHP enrollees at any given time between 2014 and 2016.

Increasing age, residing in the southeast region of the state, higher number of comorbidities, and being male were

associated with a higher likelihood of mortality. Of note, we observed that black enrollees were approximately

half as likely to die (52.2 percent) at any time over the study period, compared to white enrollees.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 70

Future Analyses

The finding of no increased mortality in Medicaid administered care compared to that in the QHPs is of significant

importance. The primary justification of the Section 1115 Waiver was the Federal Equal Access Requirements (42

U.S.C. §1396a(a)(30)(A)).

Documented perceived and realized access differences between QHP- and Medicaid-administered clinical

services, combined with less appropriate emergency room service utilizations by Medicaid administered

individuals, suggests differences in access may have resulted in differences in individual clinical outcomes — but

not mortality, as measured over the initial three years of enrollment in the program.

Of interest, the higher but not statistically significant mortality rate observed in the QHP General Population

suggests that mere exposure to the healthcare system for the average Arkansan may convey risk. The marked

reduction in mortality rates for African Americans suggest enhanced mortality benefits of coverage and also

warrants further exploration.

These findings require further exploration to account for additional unmeasured variables potentially contributing

to mortality (e.g., if applicable: smoking, obesity, job sector, and others) and a longer study period. This study may

be the first to formally test programmatic results for adults enrolled in traditional Medicaid and commercial

coverage. We will use the Arkansas All-Payer Claims Database to continue analysis of mortality in different insured

populations.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 71

Program Costs

Overview

In Table 4, the actual weighted average remuneration rates by Medicaid and QHPs for an identical basket of

services for different providers was presented. In this section, the focus is on program costs based on healthcare

visit type or services based on claims in the analytic population.

Utilization rates, as a component of differential cost, varied between Medicaid and QHP premium assistance.

While condition-specific comparisons were impractical due to variations in claim type and provider classifications

between the Medicaid and QHP programs, major categories of service utilization were categorized. Table 43

displays the utilization rates for Medicaid and QHP events including: hospitalizations, non-hospitalized ER visits,

physician’s office visits, outpatient visits, prescription medications, and other claims.

Table 43. Observed Utilization Rates (Per Member Per 100 Person-Years) for Medicaid and QHP

Enrollees in 2014, 2015, and 2016

2014

2015

2016

Visit Type

Medicaid Rate

QHP Rate

Medicaid Rate

QHP Rate

Medicaid Rate

QHP Rate

Hospitalization

0.154

0.097

0.135

0.111

0.120

0.117

Emergency Room

141.4

53.7

121.0

68.3

112.5

69.7

Physician’s Office

393.3

425.9

369.4

441.1

361.2

479.3

Outpatient

103.8

97.1

95.4

102.7

96.6

120.4

Prescription Medications

909.3

1,531.3

997.8

1,903.7

983.7

2,290.2

Other

638.3

499.4

608.5

541.8

605.8

613.5

Notes: The total person member-years used for all utilization metrics for the Medicaid population was 23,660 in 2014; 26,691 in 2015;

and 11,737 in 2016. The total person member-years used for all utilization metrics for the QHP population was 109,292 in 2014; 130,994

in 2015; and 88,954 in 2016.

The Medicaid rate includes individuals in traditional Medicaid and excludes 06 Medicaid (“Frail”).

Other

medical claims not classified above (excludes A6000/ARA6000 administrative claims — e.g., PCCM, Transportation — in Medicaid).

Abbreviation: QHP=qualified health plan

Clear and meaningful differences in utilization consistent with observed effects in the previous section are

demonstrated. Medicaid enrollees experienced fewer outpatient events and a concurrent higher rate of ER visits

and hospitalizations. Importantly, enrollees within QHPs received twice as many prescriptions than their Medicaid

counterparts. Because Medicaid utilizes different payment mechanisms and provider codes for select services

compared to their QHP counterparts, direct comparison of all services was not feasible.

The payment rates presented in the previous section and utilization differences depicted in Table 43 provide an

explanatory window into the effect differences observed between our comparison groups for access, quality, and

utilization indicators. Payments and utilization directly contribute to the absolute cost differences of the Medicaid

and commercial QHP programs and provide the quantitative basis to explore and simulate alternative scenarios of

payment and outcomes.

Using the matched payments from the primary care setting described in Table 4, we examined effect differences

in perceived access to determine the effect impact associated with payments.

Examining first year differences in access between individuals enrolled in Medicaid and those in QHPs (findings

contained in the Interim Report

), we noted a 13.2 percent increase in the number of QHP enrollees in the

General Population who reported that they “always get care when needed right away,” compared to Medicaid

Arkansas Health Care Independence Program (‘Private Option’) Final Report 72

enrollees. After approximately 18 months with coverage, the trend in perceived access difference between

Medicaid and QHP enrollees was greater, at 26.0 percent (bordering on, but not statistically significant at

p=0.057). In the Higher Needs Population, the self-reported differences were higher at both measurement

periods, with a difference of 16.9 percent during the first year, and a 36.7 percent difference at the second

measurement period, based on experience over the first 18 to 24 months with health insurance.

For every year of the HCIP program, QHPs reimbursed primary care physicians at nearly twice the rate as

Medicaid (99.9, 94.1, and 97.1 percent higher in 2014, 2015, and 2016, respectively — see Table 4). From the

observed 97.1 percent difference in payment rates to primary care physicians in 2016 and current access

differences, ratios of incremental access increase per payment difference were calculated. For the General

Population, a 2.68 percent improvement in access per 10 percent increase in payment rate could be expected. For

the Higher Needs Population, a 3.78 percent improvement in access per 10 percent increase in payment rate

could be expected.

Cost-Effectiveness

Based upon observed programmatic costs and utilization, PMPM rates for Medicaid and QHPs under alternative

scenarios were developed. Actual QHP premiums paid represented the cumulative PMPM average of premiums

paid to carriers during the first program year. Conversely, actual Medicaid expenditures paid for newly enrolled

previously eligible 19- to 64-year-old adults were calculated and expressed as a PMPM rate (supplemental

payments and estimates for marginal administrative costs were included — see Appendix F for details). Finally,

from observed utilization and payment differences, a model was generated to inform the counterfactual

assessment of what Medicaid would have experienced if a traditional expansion had been employed.

PMPM payments were calculated similarly for both Medicaid and QHP enrollees by summing adjudicated claims

for the different categories of services over 2014, 2015, and 2016 individually. Administrative costs (based on

medical loss ratio) were estimated to be an average of 18 percent for commercial carriers, while administrative

costs for traditional Medicaid were determined by examining spending across all Department of Human Services

categories that were eligible for matching federal funds. We also constructed an estimate to determine what the

QHP-enrolled individuals would have cost in the Medicaid program. The Estimated Medicaid PMPM cost are from

a model that calculates PMPM costs for QHP enrollees under the assumption that payments for services would be

at the prices paid in the Medicaid program. Under this methodology, prices for services were altered to reflect the

experience of the traditional Medicaid population, while holding utilization of services for the QHP enrollees

constant. Actual and estimated program cost numbers are contained in Table 44.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 73

Table 44. Observed and Estimated PMPM Cost Scenarios for Medicaid and QHP Enrollees by Service

Category and Year

2014

2015

2016

MCD

QHP

Estimated

MCD

QHP

Estimated

MCD

QHP

Estimated

MCD

Inpatient

$97.62

$96.90

$60.87

$100.73

$117.35

$81.39

$86.85

$123.99

$82.58

Emergency Room

$14.29

$40.25

$13.07

$13.27

$43.58

$12.54

$13.61

$44.16

$12.45

Physician’s Office

$35.28

$54.68

$43.13

$33.42

$58.67

$41.49

$31.36

$65.68

$37.27

Outpatient

$8.89

$9.89

$12.58

$8.52

$10.81

$12.04

$9.17

$11.63

$13.37

Prescription Medications

$35.53

$62.84

$41.46

$47.00

$94.37

$57.57

$40.27

$113.18

$48.15

Other

$71.20

$115.12

$67.66

$68.53

$129.40

$68.29

$64.44

$146.49

$62.83

Non-Emergency Transport Fees

$3.69

$4.92

$0.01

$3.30

$3.74

$0.01

$3.65

$4.04

$0.01

Total Claims PMPM

$266.50

$384.60

$238.78

$274.77

$457.92

$273.33

$249.35

$509.17

$256.66

Administrative PMPM

3,4

$67.96

$69.23

$68.28

$70.07

$82.43

$70.63

$63.58

$91.65

$63.58

Total PMPM

$334.46

$453.83

$307.06

$344.84

$540.35

$343.96

$312.93

$600.82

$320.24

QHP PMPM Premiums

$486.30

$486.98

$487.14

(Premium--Observed Variance)

6.7%

-9.9%

-18.9%

Notes:

Average PMPM represents loaded (i.e., cost-based reimbursement, supplemental payments, etc.) claims from Traditional Medicaid.

The

average PMPM payment made to QHP carriers includes cost-sharing and wrap services.

Includes PMPM Administrative costs for Observed QHP claims,

which were set at 18 percent of the total claims paid excluding copayments and deductibles.

Average PMPM represents claims costs based on average

Medicaid pricing.

Average PMPM paid represents claims costs based on average Medicaid pricing.

Abbreviations: MCD=Medicaid; QHP=qualified health plan; PMPM=per member per month

In 2014, the model estimate for commercial PMPMs developed from claims experience was $453.84, and 6.7 percent

lower than the actual average PMPM premium paid. In 2015, the model estimate for commercial PMPMs developed

from claims experience was $540.35, 9.9 percent more than the actual average PMPM premium paid. In 2016, the

model estimate for commercial PMPMs developed from claims experience was $600.82, 18.9 percent more than the

actual average PMPM premiums paid.

The estimated PMPM Medicaid cost of QHP enrollees in 2014 was $307.06 compared to $334.46 for actual

traditional Medicaid enrollees. In 2015, the estimated PMPM Medicaid cost of QHP enrollees was $343.96

compared to $344.84 for actual traditional Medicaid enrollees. In 2016, the estimated PMPM Medicaid cost of

QHP enrollees was $320.24 compared to $312.93 for actual traditional Medicaid enrollees. It is important to keep

in mind these modeled PMPMs reflect the estimated costs of care for QHP enrollees had they been managed

through the existing Medicaid system and experienced effects similar to those reported above. These PMPMs do

not account for potential utilization differences in the QHP enrollee population or reflect any modification of the

existing Medicaid program due to rate modifications to achieve necessary access.

Program Impact Simulation

One of the underlying justifications of Arkansas’s Section 1115 Waiver request to utilize premium assistance in the

new individual marketplace was that the network adequacy of the existing Medicaid program was insufficient to

meet anticipated demand. With Medicaid provider reimbursement rates approximately 50 percent of those in the

commercial sector (see Table 4), it was unlikely that Medicaid participating providers would be able or willing to

dramatically expand the volume of services provided without increases in reimbursement rates.

As previously noted, Arkansas had one of the lowest Medicaid eligibility thresholds for non-disabled adults in the

U.S. (below 17 percent FPL for parent/caretakers only). The result was that a majority of the covered lives were

for children, low-income Medicare beneficiaries for long-term services (not medical), pregnant women (limited

benefits for pregnancy and/or family planning services), and disabled adults (Social Security Income).

In 2013, prior to the Patient Protection and Affordable Care Act (PPACA) expansion, Arkansas Medicaid covered 24,955

non-disabled adults with a full benefits package representing 1.5 percent of the total 19- to 64-year-old population in

Arkansas Health Care Independence Program (‘Private Option’) Final Report 74

the state. In 2014, following PPACA expansion, an additional 267,482 individuals were covered — comprising

approximately 17,300 (6.5 percent) previously eligible but newly enrolled; approximately 25,000 (9.3 percent)

PPACA eligible but with exceptional healthcare needs; and 225,000 (84.2 percent) PPACA eligible with premiums

purchased on the individual marketplace. These 267,482 individuals represented 16.0 percent of the total 19- to

64-year-old population in the state.

In 2016, this number increased to 330,943 covered lives — approximately 22,375 (6.8 percent) PPACA eligible but

with exceptional healthcare needs; 32,427 (9.8 percent) interim status before enrollment in a QHP; and 276,141

(83.4 percent) PPACA eligible with premiums purchased on the individual marketplace.

These 330,943

individuals represent 19.1 percent of the working-aged adults within the state. Thus, because of the high rates of

uninsurance and low Medicaid eligibility prior to the PPACA, Medicaid has experienced a 13-fold increase in their

coverage for non-disabled 19- to 64-year-old population.

Infusion of an additional 330,943 non-disabled 19- to 64-year-olds into the Medicaid program is likely to have had

systemic effects. Traditional microeconomics suggests that increased demand through the Medicaid program

would place increasing price pressure on the rate structure of the existing Medicaid program. Medicaid could not

selectively raise provider rates solely for the expansion population. Thus, any potential increase in payment rates

would necessarily affect not only the new expansion population, but also beneficiaries under the same payment

rate schedule across the entire Medicaid program.

Though initial estimates of this potential systematic effects on the Medicaid program in 2014 were previously

calculated,

we have updated the estimates to reflect the more current environment in 2016.

In this program-impact simulation, we identified all claims paid within 2016 for Medicaid beneficiaries in the

system. We then restricted to individuals whose care would be under payment rates that might be affected (e.g.,

eliminating individuals age 65 and over, where Medicare would be the primary major medical payer, and children

younger than 1 because of a different rate structure in effect). Additionally, we eliminated payments by provider

types that were not likely to be subject to direct inflationary pressure (e.g., durable medical equipment providers,

transportation providers, etc.). Fixed costs for Medicaid, including Disproportionate Share payments and provider

supplemental payments to hospitals serving 19- to 64-year-olds were calculated by using the HP Financial

Transaction file provided by Medicaid.

From these expenses a total Medicaid expenditure of $2,490,804,452.76 for covered adults in the traditional (e.g.,

non-expansion) population was generated. The total member months for the traditional Medicaid population was

9,098,587 — resulting in a traditional Medicaid PMPM of $273.76. (Note that this is not the projected Medicaid

expansion PMPMs in Table 14.)

We then simulated, through a budget impact analysis, the incremental effect of potential inflationary increases in

payment rates and the net associated cost impact for the traditional Medicaid population under three alternative

scenarios:

1. Claims associated with potentially wage-sensitive services.

2. Then we restricted to claims associated with major medical services.

3. Then we further restricted to only claims associated with physician-billed services.

Within the budget impact analysis, increasing costs under the scenarios were applied to the total traditional

Medicaid expenditures. Claims were allocated according to each of the three scenarios. Increases for each

scenario were estimated based upon marginal five percentage payment increases and projected total traditional

Arkansas Health Care Independence Program (‘Private Option’) Final Report 75

Medicaid PMPMs were calculated. Results were plotted for each scenario. These Medicaid projections were then

compared to the actual premiums paid for commercial premium assistance to determine if, and at what

threshold, total Medicaid program costs would have exceeded the increased costs associated with using

commercial premium assistance.

Figure 11. Access Improvement Simulation Model Studying Scenario-Based Price Increases to Medicaid

Providers to Reach Commercial PMPM Payments

Note: PMPM expenditures observed for QHPs (premium assistance) and Medicaid with simulated Medicaid costs under incremental

increases under a scenario for all claims associated with wages, all claims restricted to major clinical services, and all claims restricted to

only physician services.

Figure 11 depicts the alternative scenarios when holding constant the actual commercial PMPM ($487.14) and

the actual traditional Medicaid PMPM ($265.75) in 2016. For each scenario, the budget impact analysis

demonstrated inflationary results and the thresholds where increases in the Medicaid program payment rates

would exceed the differential increase of expenditures associated with the use of premium assistance in the

expansion population.

Under the wage-sensitive scenario, if services for the traditional Medicaid beneficiaries experienced a 29.1

percent increase in wage-sensitive costs, Medicaid would have exceeded the costs associated with the use of

commercial premium assistance for the expansion population. Under the major medical scenario, the Medicaid

program would exceed expenditures at a 45.5 percent increase in clinical claims cost. Finally, restricting to the

Arkansas Health Care Independence Program (‘Private Option’) Final Report 76

physician-only scenario, the Medicaid program would exceed expenditures at a 64.2 percent increase in the cost

of physician claims.

While it is unknown to what degree Arkansas Medicaid would have experienced inflationary pressures on

provider payment rates, it is clear from comparison of access, utilization, and clinical performance between

Medicaid and QHP experiences that differences exist and are likely associated with provider payment

differentials. Had the approximately 330,000 newly covered individuals been placed in the traditional Medicaid

program, increases in payment rates would have been required to avoid potential legal challenges to the equal

access requirement and to assure attainment of needed services.

Arkansas Health Care Independence Program (‘Private Option’) Final Report 77

References

National Federation of Independent Businesses v. Sebelius. 567 U.S. ___ (2012).

The Centers for Medicare & Medicaid Services [Letter] to Andy Allison, Arkansas Department of Human Services. September

2013. https://www.medicaid.gov/Medicaid-CHIP-Program-Information/By-Topics/Waivers/1115/downloads/ar/Health-

Care-Independence-Program-Private-Option/ar-private-option-app-ltr-09272013.pdf.

Ark. Code § 2-77-2405 (2014).

The Centers for Medicare & Medicaid Services. Arkansas Health Care Independence Program Special Terms and Conditions.

September 2013, 11-W-00287/6.

Arkansas Center for Health Improvement. Arkansas Health Care Independence Program (“Private Option”)

Section 1115 Demonstration Waiver Interim Report. Little Rock, AR: Arkansas Center for Health Improvement, March 2016.

Health Resources and Service Administration. Data Warehouse Medically Underserved Areas and Populations.

http://datawarehouse.hrsa.gov/tools/analyzers/muafind.aspx. Published 2016.

United States Census Bureau. Small Area Health Insurance Estimates (SAHIE), 2010 SAHIE Data.

https://www.census.gov/programs-surveys/sahie.html.

Arkansas Health Data Initiative. Report generated March 2016 by the Arkansas Center for Health Improvement.

The Centers for Medicare & Medicaid Services. Managed Care in Arkansas. https://www.medicaid.gov/medicaid-chip-

program-information/by-topics/delivery-systems/managed-care/downloads/arkansas-mcp.pdf. Published August 2014.

Ark. Code § 23-99-201.

Ark. Code § 23-61-801.

42 U.S.C. § 1396a(a)(30)(A)

Arkansas Insurance Department. Requirements for qualified health plan certification in the Arkansas federally facilitated

partnership exchange (marketplace) Bulletin No. 3B-2013. 2013

https://insurance.arkansas.gov/uploads/resource/documents/13b-2013.pdf

U.S. Government Accountability Office. CMS Has Taken Steps to Address Problems, but Needs to Further Implement

Systems Development Best Practices. March 2015. http://www.gao.gov/assets/670/668834.pdf.

Agency for Healthcare Research and Quality. The MEPS Data and Methods Manual. Medical Expenditure Panel Survey.

Rockville, MD. 2013. Available at http://meps.ahrq.gov/mepsweb/.

Witters D. Arkansas, Kentucky Set Pace in Reducing Uninsured Rate. Gallup, February 4, 2016.

http://www.gallup.com/poll/189023/arkansas-kentucky-set-pace-reducing-uninsured-rate.aspx.

Arkansas Department of Human Services. Private Option Enrollment and Premium Information. November 2017.

Calculated by the Arkansas Center for Health Improvement on March 22, 2016, based on enrollment and eligibility data

reported by the Arkansas Department of Human Services report run on April 14, 2015.

U.S. Department of Health and Human Services, Office of the Assistant Secretary of Planning and Evaluation. Profile of

Affordable Care Act Coverage Expansion Enrollment for Medicaid/CHIP and the Health Insurance Marketplace, 10-1-2013

to 3-31-2014, Arkansas. https://aspe.hhs.gov/sites/default/files/pdf/93416/ar.pdf. Published May 1, 2014. Accessed March

22, 2016.

Kaiser Family Foundation. Analysis of 2015 Premium Changes in the Affordable Care Act’s Health Insurance Marketplaces.

http://kff.org/health-reform/issue-brief/analysis-of-2015-premium-changes-in-the-affordable-care-acts-health-insurance-

marketplaces/. Published January 2015. Accessed March 21, 2016.

Susan Ward-Jones, MD. Personal Communication. March 12, 2015.

Arkansas Hospital Association. APO’s Hospital Impact Strong in 2014. The Notebook. 2015;22(22):1.

http://www.arkhospitals.org/archive/notebookpdf/Notebook_07-27-15.pdf.

Dan Rahn, M.D., Testimony before Health Reform Legislative Taskforce. August 20, 2015.

Lyon J. Despite elimination backlog, Medicaid re-determination process not quick. Arkansas News.

http://arkansasnews.com/news/arkansas/despite-elimination-backlog-medicaid-re-determination-process-not-quick

Published September 20, 2015.

Plan Management Frequently Asked Questions. https://static.ark.org/eeuploads/hbe/FAQs-3-13-15.pdf. Published March

5, 2015.

Arkansas Department of Health and Human Services. Arkansas 1115 Waiver-FINAL.

https://medicaid.mmis.arkansas.gov/Download/general/comment/ARWorksAppFinal.pdf. Accessed June 26, 2018.

45 CFR § 156.230

Arkansas Health Care Independence Program (‘Private Option’) Final Report 78

National Committee for Quality Assurance. Healthcare Effectiveness Data and Information Set Technical Specifications for

Health Plans, 2015, Vol. 1-3.

Guo S, Fraser MW. Propensity Score Analysis: Statistical Methods and Applications. Los Angeles, CA: Sage; 2015.

Escanciano J. Regression Discontinuity Designs: Theory and Applications. Bingley, UK: Emerald Publishing Limited, 2017.

Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. JASA.

1983;102:75-83.

New York University Center for Health and Public Service Research. NYU ED algorithm.

http://wagner.nyu.edu/faculty/billings/nyued-background. Published 2011. Accessed November 12, 2015.

Ballard DW, Price M, Fung V, et al. Validation of an algorithm for categorizing the severity of hospital emergency

department visits. Med Care. 2010;48(1):58-63. doi: 10.1097/MLR.0b013e3181bd49ad. PubMed PMID: 19952803.

Rhodes KV, Kenny GM, Friedman AB, et al. Primary Care Access for New Patients on the Eve of Health Care Reform. JAMA

Intern Med. 2014;174(6):861-869. doi:10.1001/jamainternmed.2014.20.

Arkansas Center for Health Improvement. Arkansas Health Care Workforce: A Guide for Policy Action. Little Rock, AR:

Arkansas Center for Health Improvement, March 2013.

Polsky D, Richards M, Basseyn S, et al. Appointment Availability after Increases in Medicaid Payments for Primary. N Engl J

Med 2015; 372:537-545. doi: 10.1056/NEJMsa1413299.

Agency for Healthcare Research and Quality, Prevention Quality Indicators.

http://qualityindicators.ahrq.gov/Modules/pqi_resources.aspx. Accessed March 14, 2018.

Kaskie B, Obrizan M, Cook E, et al. Defining emergency department episodes by severity and intensity: a 15-year study of

Medicare beneficiaries. BMC Health Services Research 2010, 10(1). doi:10.1186/1472-6963-10-173

Parsons LS. Reducing bias in a propensity score matched-pair sample using greedy matching techniques; 26th Annual SAS

Users Group International Conference; Long Beach, CA: 2001

Shore-Sheppard LD. Income Dynamics and Affordable Care Act. Health Services Research 2014; 46(6):2041-61.

Sinners BD, Rosenbaum S. Issues in Health Reform: How Changes in Eligibility May Move Millions Back and Forth between

Medicaid and Insurance Exchanges. Health Affairs (Millwood) 2011; 301(2):228-36.

Buettgens M, Nichols A, Dorn S. Churning under the ACA and state policy options for mitigation [Internet]. Urban Institute;

https://www.urban.org/research/publication/churning-under-aca-and-state-policy-options-mitigation

Published June 14, 2012. Accessed August 3, 2016.

Centers for Disease Control and Prevention. Prescription Opioid Overdose Data. 2017.

https://www.cdc.gov/drugoverdose/data/overdose.html. Accessed September 4, 2017.

Dowell D, Haegerich TM, Chou R. CDC Guideline for Prescribing Opioids for Chronic Pain — United States, 2016. MMWR

Recomm Rep. 2016;65(1):1-49.

Ark. Code Ann. § 20-7-601 (2011)

Ark. Code Ann. § 20-13-1701 (2015)

Ark. Code Ann. § 20-13-1801 (2015)

Ark. Code Ann. § 20-7-604 (2017)

Wachino V. Best Practices for Addressing Prescription Opioid Overdoses, Misuse and Addiction. CMCS Informational

Bulletin. 2016.

Arkansas Medicaid. Arkansas Medicaid Prescription Drug Program Prior Authorization Criteria. 2018.

https://arkansas.magellanrx.com/provider/docs/rxinfo/PACriteria.pdf. Accessed May 15, 2018.

Shah A, Hayes CJ, Martin BC. Characteristics of Initial Prescription Episodes and Likelihood of Long-Term Opioid Use —

United States, 2006-2015. MMWR Morb Mortal Wkly Rep. 2017;66(10):265-9.

Woolhandler S, Himmelstein DU. The Relationship of Health Insurance and Mortality: Is Lack of Insurance Deadly? Ann

Intern Medi. 2017; 167(6):424-431

Institute of Medicine; Committee on the Consequences of Uninsurance Care Without Coverage: Too Little, Too Late.

Washington, D.C., National Academies Press, 2002.

Arkansas Department of Human Services (DHS). (2016, December 31). Arkansas Private Option 1115 Demonstration

Waiver. Quarterly Report October 1, 2016-December 31, 2016 [Report]. Retrieved from

https://www.medicaid.gov/Medicaid-CHIP-Program-Information/By-Topics/Waivers/1115/downloads/ar/Health-Care-

Independence-Program-Private-Option/ar-works-qtrly-rpt-oct-dec-2016.pdf

Arkansas Health Care Independence Program (‘Private Option’) Final Report A-1

Appendices

Appendix A – Arkansas Evaluation Hypotheses: Proposed and Original Test

Indicators

Test Metrics for Incorporation into Year 1 HCIP Assessment

Note: This is an abridged list from hundreds of candidate indicators considered. Due to the short coverage period

of 12 months, observations did not meet the HEDIS indicator inclusion criteria. For others, there were too few

observed episodes to be able to detect significant differences across programs. This final listing was approved by

the National Advisory Committee as reasonable for presentation in this report.

1. Geographic Access

1.1. Was network participation in Medicaid equal to that of the qualified health plans (QHPs)?

1.1.1. Proportion of enrollees within 30 miles of at least one primary care provider (PCP)

1.1.2. Proportion of enrollees within 60 miles of the following specialists:

1.1.2.1. Cardiologist

1.1.2.2. Obstetrician/Gynecologist

1.1.2.3. Psychiatrist

1.1.2.4. Orthopedist

1.1.2.5. Internist

1.1.2.6. Ophthalmologist

1.1.2.7. Oncologist

1.1.2.8. General Surgeon

2. Realized Access – Differences in Utilization

2.1. Was time to engagement with a PCP in Medicaid equal to that in the QHP?

2.1.1. Proportion of enrollees with a PCP visit during first 12 months of continuous enrollment

2.1.2. Of those with a PCP visit, what were the number of days from enrollment date to first visit with a

PCP?

2.1.3. Reengagement with primary care following discharge:

2.1.3.1. Proportion of non-maternal discharges seen in outpatient setting within 10 days

2.1.3.2. Proportion of non-maternal discharges seen in outpatient setting within 30 days

2.2. Was utilization of frequent procedures in Medicaid equal to that in the QHP? Will require summary

indicator stratified by age and gender—HEDIS FSP frequency table

2.2.1. Percutaneous coronary intervention

2.2.2. Cardiac catheterization

2.2.3. Coronary artery bypass graft

2.3. Was inpatient utilization in Medicaid equal to that in the QHP? Reported separately by medical, surgical,

and maternity stratified by age and gender—HEDIS IPU frequency table

2.3.1. Discharges

2.3.2. Discharges per 1,000 member months

2.3.3. Days

2.3.4. Days per 1,000 member months

2.3.5. Average length of stay

Arkansas Health Care Independence Program (‘Private Option’) Final Report A-2

3. Perceived Access – The Consumer’s Experience

3.1. Proportion of CAHPS respondents who report obtaining care as soon as needed

3.2. Primary care—Proportion of CAHPS respondents who report getting an appointment for checkup or

routine care as soon as needed

3.3. Specialty care—Proportion of CAHPS respondents who report getting an appointment for specialty care

as soon as needed

3.4. Proportion of CAHPS respondents who report ease of getting care

4. Transportation

4.1. CAHPS respondents who did not visit their personal doctor in the last six months because they could not

arrange acceptable transportation

4.2. CAHPS respondents who did not visit a specialist in the last six months because they could not arrange

acceptable transportation

5. Clinical Quality Process Measures

5.1. Were primary prevention efforts in Medicaid equal to that in the QHP?

5.1.1. CAHPS respondents who had a flu shot or nasal flu spray since July 1, 2014

5.1.2. CAHPS respondents who use tobacco and reported a cessation discussion with provider

5.2. Did secondary prevention efforts in Medicaid achieve equal results to those in the QHPs?

5.2.1. Combined screening indicator for age-appropriate screening tests—cervical cancer, breast cancer,

colorectal, cholesterol (note deviates from HEDIS):

5.2.1.1. Proportion of enrollees receiving one or more indicated screening tests

5.2.1.2. Proportion of enrollees receiving all indicated screening tests

5.3. Were indicators of chronic disease management (tertiary prevention) in Medicaid equal to those in the

QHP?

5.3.1. Diabetes

5.3.1.1. Proportion of enrollees with diabetes with evidence of HbA1c assessment

5.3.1.2. Proportion of enrollees with diabetes with evidence of two or more HbA1c assessments

5.3.1.3. Proportion of enrollees with diabetes with LDL-c screening

6. Clinical Quality Outcome Measures

6.1. Were preventable medical events in Medicaid equal to those in the QHPs?

6.1.1. Preventable emergency department visits

6.1.2. Preventable hospitalizations

6.2. Were avoidable readmissions in Medicaid equal to those in the QHPs?

6.2.1. Plan all-cause readmission (non-mental health)

7. Were Outcome Differences in Medicaid Equal to Those in the QHPs?

7.1. Crude and age-adjusted mortality rates

Arkansas Health Care Independence Program (‘Private Option’) Final Report B-1

Appendix B – Arkansas Evaluation Hypotheses

Table 1 includes a description of each of the original 12 hypotheses outlined in STC 70 that have been reorganized

into the following four categories:

Table 1. Evaluation Hypotheses

Arkansas Evaluation Hypotheses

Arkansas Original Terms and Conditions Hypotheses (Section

8, STC 70, No. 1)

1 – Access

a. Use of PCP/specialist

i. Premium Assistance beneficiaries will have equal or

better access to care, including primary care and

specialty physician networks and services.

b. Non-emergent ER use

iii. Premium Assistance beneficiaries will have lower non-

emergent use of emergency room (ER) services.

c. Preventable ER use

vii. Premium Assistance beneficiaries will have lower rates of

potentially preventable emergency department and

hospital admissions.

d. EPSDT

ix. Premium Assistance beneficiaries who are young adults

eligible for EPSDT benefits will have at least as

satisfactory and appropriate access to these benefits.

e. Non-emergency

transportation

x. Premium Assistance beneficiaries will have appropriate

access to non-emergency transportation.

2 – Care/Outcomes

a. Preventive and healthcare

services

ii. Premium Assistance beneficiaries will have equal or

better access to preventive care services.

b. Experience

c. Non-emergent ER use*

d. Preventable ER use*

viii. Premium Assistance beneficiaries will report equal or

better experience in the care provided.

iii. Premium Assistance beneficiaries will have lower non-

emergent use of emergency room services.

vii. Premium Assistance beneficiaries will have lower rates

of potentially preventable emergency department and

hospital admissions.

Arkansas Health Care Independence Program (‘Private Option’) Final Report B-2

Arkansas Evaluation Hypotheses

Arkansas Original Terms and Conditions Hypotheses (Section

8, STC 70, #1)

3 – Continuity

a. Gaps in coverage

iv. Premium Assistance beneficiaries will have fewer gaps in

insurance coverage.

b. Continuous access to same

health plans

v. Premium Assistance beneficiaries will maintain

continuous access to the same health plans, and will

maintain continuous access to providers.

c. Continuous access to same

providers

4 – Cost Effectiveness

a. Administrative costs

vi. Premium Assistance beneficiaries, including those who

become eligible for Exchange Marketplace coverage, will

have fewer gaps in plan enrollment, improved continuity

of care, and resultant lower administrative costs.

b. Reduce premiums

xi. Premium Assistance will reduce overall premium costs in

the Exchange Marketplace and will increase quality of

care.

c. Comparable costs

xii. The cost for covering Premium Assistance beneficiaries

will be comparable to what the costs would have been for

covering the same expansion group in Arkansas Medicaid

fee-for-service, in accordance with STC 68 on determining

cost effectiveness and other requirements in the

evaluation design as approved by CMS.

* The outcomes of interest and evaluation approaches associated with hypotheses 2c and 2d are shared with 1b

and 1c.

Arkansas Health Care Independence Program (‘Private Option’) Final Report C-1

Appendix C – Data Processing of Carrier Data

Overview

Data were obtained from two primary data sources: administrative claims data from Medicaid and commercial

Qualified Health Plans (QHPs) and a member enrollment survey, Consumer Assessment of Healthcare Providers

and Systems (CAHPS). In order to construct variables of interest, these data were supplemented by files from the

Department of County Operations (DCO), Arkansas Department of Health (Birth/Death Certificates and Hospital

Discharges), and the Exceptional Health Care Needs Assessment Questionnaire. Administrative claims for

Medicaid and QHP enrollees contained enrollment and all reimbursed care provided by Medicaid and the three

commercial health plans from Jan. 1, 2014, through Dec. 31, 2016. The administrative claims data included

medical claims (provider and facility claims), pharmacy claims, member enrollment details, and provider files.

A survey of a sample of traditional Medicaid and QHP enrollees was fielded by the Arkansas Foundation for

Medical Care (AFMC) with measures derived from CAHPS, SF-12, and supplementary items. The survey was

administered between July and September of 2015 and questioned members about their plan experiences over

the prior six months. AFMC sent an initial survey, a reminder post card, a second survey for non-responses, and, if

no response after the second mailing, a follow-up phone survey.

A unique person identifier, called a POID, was generated using characters contained in the enrollee’s Social

Security number, first and last name, and date of birth. All data were linked by this encrypted identifier.

Data Specifications

Program Enrollment and Eligibility Data

The Department of Human Services provided Medicaid enrollment data from Jan. 1, 2014, through March 31,

2017. This data was used to identify enrollees in the traditional Medicaid population and those assigned to

Medicaid with exceptional needs. Medicaid enrollment data was then integrated with enrollment data received

from each of the QHPs including Blue Cross Blue Shield, Ambetter, and QualChoice.

QHP assignment was validated using Department of Human Services fiscal expenditure data. This file included all

subsidized premiums paid by Medicaid for each QHP enrollee from January 2014 through March 2017.

ii.

Medical and Pharmacy Claims

Administrative claims for Medicaid and QHP enrollees were obtained from all reimbursed care provided by

Medicaid and the three QHPs from Jan. 1, 2014, through Dec. 31, 2016. The administrative claims data included

medical claims (provider and facility claims), pharmacy claims, member enrollment details, and provider files.

Medicaid

Department of Human Services Medicaid medical and pharmacy claims files were used to assess utilization and

cost in the traditional Medicaid population and those assigned to Medicaid with exceptional needs. These data

were in Medicaid Management Information Systems standard format and layout. All claims included in the

analyses were adjudicated and non-voided.

Qualified Health Plans (Blue Cross Blue Shield, Ambetter, and QualChoice)

QHP medical and pharmacy claims files for QHP enrollees were used to identify cost and utilization of services.

Each QHP was asked to provide this data in the Arkansas All-Payer Claims Database (APCD) layout. The APCD was

Arkansas Health Care Independence Program (‘Private Option’) Final Report C-2

established pursuant to the Arkansas Healthcare Transparency Initiative of 2015. The Arkansas Insurance

Department (AID) established an APCD Data Submission Guide which defines file requirements for data

submissions (for the specific requirements please see the APCD Data Submission Guide here:

https://www.arkansasapcd.net/Resources/DataSubmissionGuideResources). All claims received were adjudicated

and non-voided.

iii. Arkansas Department of Health

The 2014-2016 Arkansas Inpatient Hospital Discharge data and 2013-2016 Arkansas Vital Statistics information,

including 2013-2016 birth and death certificate data, were made available through the Arkansas Center for Health

Improvement’s Arkansas Health Data Initiative health policy database, with approval from the Arkansas

Department of Health. Birth certificate data were used to assess pregnancy outcomes measures, differences in

low birth weight and very low birth weight, and the number of prenatal visits. Additionally, the birth certificate

data were used to capture pre-delivery covariates such as prior C-sections, mothers’ education, plurality, and

other measures. Mortality files were used to estimate death rates. Additionally, hospital discharge data was used

in calculations for cost effectiveness supplemental payment allocations.

iv. Income Eligibility Determination

Data was received from the Department of County Operations (DCO) containing all available eligibility and

enrollment data for Medicaid and QHPs covering the time period of Jan. 1, 2013, through Dec. 31, 2016. A key

feature of these data was the income determination levels of each enrollee at the time of application that was

used in survey sample strata creation and other analyses.

v. Consumer Assessment of Health Plans Survey (CAHPS) Data

Nationwide experience with CAHPS has led to important new insights into patient experiences regarding care for

both the Medicaid and the commercially insured populations. Various CAHPS surveys are available to gauge

consumer and patient experience with healthcare services and cover important topics including quality of care,

access to care, and experience with care. Surveys are available in the public domain.

The AFMC is the current contractor that collects CAHPS data for the Arkansas Medicaid program every two years.

In our evaluation, we contracted with AFMC to field a modified CAHPS survey in the third quarter of 2015. AFMC

sent an initial survey, a reminder post card, a second survey for non-responses, and, if no response after the

second mailing, a follow-up phone survey.

We included questions to be able to address access to and availability of services, consistency of care providers

and networks, use of primary and specialty care services, and experience with care. In addition, we included

questions from the 12-item Short Form Health Survey (SF12)

to determine physical and mental health

standardized scores.

In order to attain reasonable power to detect differences between Medicaid and QHP enrollees in responses to

CAHPS questions we oversampled sub-populations that formed comparison and distinct groups of interest. In

addition, we sampled a representative portion of the traditional Medicaid enrollees. The cooperation rate for the

survey was 26.4 percent and a representative response rate was received across sub-populations of interest.

Details on the CAHPS sampling strategy can be found in Appendix K.

Arkansas Health Care Independence Program (‘Private Option’) Final Report C-3

vi. GIS Processed Provider-Enrollee Distance Measures Data

To determine geographic access of Medicaid and HCIP enrollees to providers we contracted with the Center for

Advanced Spatial Technologies (CAST) at the University of Arkansas, J. William Fulbright College of Arts and

Sciences. Medicaid and QHP enrollee addresses were obtained as well as in-network provider address lists.

Enrollee and provider addresses were geocoded. Distances between enrollees and in-network providers were

measured in both 15-minute travel time increments and 15-mile distance increments. For enrollees in a QHP,

additional access metrics were calculated to determine if they had access to out-of-network, as well as in-network

providers.

For the purpose of this evaluation, we defined adequate access if an enrollee resided within a 30-minute travel

time for a primary care physician or within 60 minutes of a specialist. Specialty access was calculated for each of

the following: orthopedics, ophthalmology, obstetrician/gynecologist (OB/GYN), oncology, surgical, psychology,

and cardiology. More details on methods used by the CAST team can be found in Appendix H.

Reference

Burdine JN, Felix MR, Abel AL, Wiltraut CJ and Musselman YJ. The SF-12 as a population health measure: an exploratory

examination of potential for application. Health Services Research. 2000 Oct; 35(4): 885–904.

Arkansas Health Care Independence Program (‘Private Option’) Final Report D-1

Appendix D – Study Subjects: Analytical Sample Extraction

This analysis sought to compare the effects of the Health Care Independence Program (HCIP) versus traditional

Medicaid for subjects who had relatively stable coverage in their respective plans for at least two contiguous

calendar years. For the regression discontinuity (RD) sample, subjects had to have at least two years of coverage

starting in 2014, but could be followed until 2016. Additionally, the RD sample was restricted to those with a

health needs assessment (frailty) score recorded before the end of 2014. For the propensity score sample, a

subject’s first year of contiguous enrollment could start in 2014 or 2015, with the first year serving as the baseline

year to capture covariate history, and the subject had to have at least one year of contiguous enrollment the

following year, but could be followed for up to two years after their baseline year. Further, the propensity score

sample was restricted to persons who did NOT have a health needs assessment score.

Arkansas Medicaid enrollment files for calendar years 2013 and 2014 were used to exclude enrollees of non-

interest and identify the study population. Enrollees who did not meet each of the following criteria were

excluded:

 At least 18 years of age on Jan. 1, 2014, and less than 65 years of age on Dec. 31, 2016;

 Continuously enrolled in a Medicaid or HCIP category, which is defined as enrollment for at least

180 days with, at most, one gap of no more than 13 days in each of a subject’s first two calendar

years of enrollment.

o For the regression discontinuity sample, subjects had to be continuously enrolled

throughout 2014 and 2015.

o For the propensity sample, subjects had to have two consecutive years of continuous

enrollment, beginning in either 2014 or 2015.

 Not dually enrolled with a Medicare benefit;

 Not enrolled in a full-benefit Medicaid category between Jan. 1, 2013, and Oct. 15, 2013.

Figure 1 outlines the study population development starting from a base population of enrollees who had at least

one valid enrollment segment from Jan. 1, 2014, through Dec. 31, 2016. Two important exclusions should be

highlighted. First, members who switched between Medicaid and any of the commercial carriers after their first

90 days enrolled were excluded. Persons could switch between any of the commercial carriers after the 90 days

and still be retained, but could not make a transition to a traditional Medicaid enrollment category. Second,

enrollees with coverage discrepancies were also removed from analytical consideration. A majority of these

people were eligible for commercial qualified health plan (QHP) coverage, but never transitioned into a

commercial plan.

Figures 1 and 2 illustrate the application of the inclusion and exclusion criteria for the regression discontinuity and

propensity score samples, respectively. Once the samples were selected, an analyzable dataset was built, which

included, for each subject, rural/urban designation using rural-urban commuting area (RUCA) classifications,

healthcare services utilization variables, and constructed fields to represent access, experience, and outcomes

indicators. Selected demographics are shown in Table 1.

Arkansas Health Care Independence Program (‘Private Option’) Final Report D-2

Figure 1. Analytical Data Preparation Regression Discontinuity Sample — Final Report

Total newly covered 20/25/06 individuals in premium assistance

(regression discontinuity sample)

N = 765,026

N = 215,450

Exclude persons not enrolled for at least 180 days in 2014

and 2015 and those who died before the time frame of

analysis

N = 545,379

N = 207,484

Exclude persons enrolled in Medicaid prior to the

HCIP program (Oct. 15, 2013) other than Family

Planning

N = 7,966

Analyzable Study Population

N = 53,322

Exclude persons who were dually eligible or

institutionalized in 2014

N = 6,403

Exclude those who made a switch between Medicaid

and commercial plans after the first 91 days of

enrollment but before the beginning of the third year

N = 17,660

Exclude persons less than 18 years of age on Jan. 1, 2014

or older than 65 years of age on Dec. 31, 2016

N = 4,197

Exclude those with coverage discrepancies (i.e.

premium assistance members who never transitioned

to commercial coverage or were frail and placed in

commercial coverage)

N = 1254

Exclude persons with one day or more of pregnancy

coverage

N = 20

Exclude persons without a frailty score

N = 128,825

General Population

Traditional Medicaid

Baseline and Year 2

N = 9,037

General Population

Commercial QHP

Baseline and Year 2

N = 44,285

General Population

Commercial QHP

Year 3

N = 37,752

General Population

Traditional Medicaid

Year 3

N = 7,951

Arkansas Health Care Independence Program (‘Private Option’) Final Report D-3

Figure 2. Analytical Data Preparation-Propensity Score Sample — Final Report

Total newly covered 20/25/06 individuals in premium assistance

(propensity score sample)

N = 765,026

N = 300,549

Exclude persons enrolled for less than 180 days in two

consecutive calendar years and those who died before the

time frame of analysis

N = 458,845

N = 255,182

Exclude persons enrolled in Medicaid, other than

Family Planning, prior to the HCIP program,

Oct. 15, 2013

N = 45,367

Analyzable Study Population

N = 120,845

Exclude persons who are severely developmentally disabled,

dually eligible or institutionalized in 2014

N = 5,210

Exclude those who made a switch between Medicaid and

commercial plans after the first 91 days of enrollment but

before the beginning of the third year

N = 54,501

Exclude persons less than 18 years of age on Jan. 1, 2014

or older than 65 years of age on Dec. 31, 2016

N = 5,632

Exclude those with coverage discrepancies

(I.e. overlapping coverage)

N = 1,193

Exclude persons with one day or more of pregnancy

coverage

N = 2,175

Exclude persons with a frailty score

N = 71,258

General Population

Traditional Medicaid

Baseline Year and Year 2

N = 27,905

Expansion Population

Baseline Year and Year 2

N = 92,940

General Population

Traditional Medicaid

Year 3

N = 13,933

Expansion Population

Year 3

N = 56,874

Arkansas Health Care Independence Program (‘Private Option’) Final Report D-4

Table 1. Demographic Profiles of General Population and Higher Needs Population

General Population

(n=120,845)

Higher Needs Population

(n=53,322)

Medicaid

QHP

Medicaid

QHP

Age

<34

14,731

52.8

39,627

42.6

2,910

32.2

18,165

41.0

35-44

7,399

26.5

22,252

23.9

2,261

25.0

9,872

22.3

45-54

4,052

14.5

20,124

21.7

2,408

26.6

9,374

21.2

55-64

1,723

6.2

10,937

11.8

1,458

16.1

6,874

15.5

Gender

Male

10,155

36.4

43,756

47.1

3,274

36.2

17,816

40.2

Female

17,750

63.6

49,184

52.9

5,763

63.8

26,469

59.8

Race

White

19,645

70.4

58,711

63.2

7,026

77.7

31,776

71.8

Black

5,858

21.0

28,258

30.4

1,588

17.6

8,856

20.0

Hispanic

347

1.2

1,009

1.1

0.6

426

1.0

Other

1,288

4.6

1,440

1.5

152

1.7

900

2.0

Missing/Unknown

767

2.7

3,522

3.8

221

2.4

2,327

5.3

Region

Central

8,517

30.5

26,674

28.7

2,692

29.8

13,838

31.2

Northeast

5,309

19.0

18,867

20.3

1,800

19.9

7,784

17.6

Northwest

4,664

16.7

13,232

14.2

1,471

16.3

8,259

18.6

South Central

1,639

5.9

6,207

6.7

651

7.2

3,430

7.7

Southeast

2,396

8.6

11,544

12.4

869

9.6

3,793

8.6

Southwest

2,303

8.3

8,378

9.0

670

7.4

3,094

7.0

West Central

3,077

11.0

8,038

8.6

884

9.8

4,087

9.2

Household Median Income

$30k or less

6,985

25.0

27,215

29.3

2,213

24.5

10,380

23.4

$30k–$40k

7,589

27.2

25,684

27.6

2,558

28.3

12,310

27.8

$40k–$50k

7,059

25.3

22,319

24.0

2,306

25.5

11,496

26.0

$50k–$60k

3,268

11.7

9,580

10.3

1,019

11.3

5,123

11.6

Greater than $60k

3,004

10.8

8,142

8.8

941

10.4

4,976

11.2

Totals

27,905

92,940

9,037

44,285

Reference

United States Department of Agriculture, Economic Research Services. http://www.ers.usda.gov/data-products/rural-urban-

commuting-area-codes.aspx. Accessed 10/12/2017

Arkansas Health Care Independence Program (‘Private Option’) Final Report E-1

Appendix E – Statistical Modeling to Assess Effect Differences for Those

Assigned to Medicaid and QHPs

In our evaluation plan, we proposed using regression discontinuity as the statistical method to assess the access,

quality, and utilization effect differences between beneficiaries assigned to the Medicaid plan and beneficiaries

assigned to the QHPs. We had two candidate variables on which to use regression discontinuity designs. One was

using the continuum of composite scores obtained from the exceptional healthcare needs assessment

Questionnaire. A threshold cut-point was established along the composite scores to identify those with higher risk

of exceptional healthcare needs, who were then assigned to the Medicaid plan.

In addition, the traditional programmatic assignment of beneficiaries into the traditional Medicaid fee-for-service

plan was based on being a parent earning 17 percent or less of the federal poverty level (FPL). We anticipated

using the programmatic income cut-point assignment for a parent subpopulation to be able to fit a series of

regression discontinuity models. Preliminary data analyses confirmed that nearly one in three enrollees (30.3

percent) in the traditional Medicaid fee-for-service plan (aid categories 20/25) were not parents. Also, many

parents who were assigned to a QHP indicated an income of less than 17 percent FPL, with many of these

indicating an income of zero dollars. For this reason, and lacking the ability to conform to accepted standards for

the use of a regression discontinuity design, we decided to pursue a different statistical method to compare

groups of people who did not complete an exceptional healthcare needs assessment Questionnaire but were

assigned into one of the Medicaid plans, or a QHP.

This document provides an outline of the statistical methods used in producing the results in this report.

Regression Discontinuity Design

Comparison Groups

We included beneficiaries completing the exceptional healthcare needs assessment Questionnaire and meeting

other inclusion criteria for claims-based metrics and for CAHPS respondents. This population is labeled as “Higher

Needs” and there was a sharp assignment into either the Medicaid or a QHP plan based on the exceptional

healthcare needs assessment composite score. Those completing the Questionnaire and receiving a composite

score of 0.18 or higher were all assigned to the Medicaid plan and those with a composite score less than 0.18

were assigned to a QHP. In total, our analytic claims file contained 9,037 assigned to the Medicaid plan and

44,285 assigned to a QHP (see Table 1 in the Final Report). For CAHPS II, where we oversampled populations that

were close to the threshold composite score value of 0.18, our analytic file contained 1,013 assigned to the

Medicaid plan and 1,119 assigned to a QHP.

Figures 1 and 2 depict the composite score frequencies in the Claims and CAHPS II analytic populations.

Arkansas Health Care Independence Program (‘Private Option’) Final Report E-2

5,000

10,000

15,000

20,000

25,000

30,000

35,000

0.02

0.07

0.08

0.10

0.13

0.15

0.17

0.18

0.19

0.20

0.21

0.22

0.23

0.24

0.26

0.27

0.28

0.29

0.31

0.32

0.33

0.34

0.35

0.37

0.38

0.40

0.42

0.43

0.44

0.45

0.46

0.49

0.50

0.51

0.56

0.61

Frequency of Enrollees

Exceptional Health Care Needs Composite Score

Figure 1. Claims Data - Exceptional Health Care Needs Composite Score (N=53,332)

100

150

200

250

300

350

0.02

0.07

0.08

0.10

0.13

0.15

0.17

0.18

0.19

0.20

0.21

0.22

0.23

0.24

0.26

0.27

0.28

0.29

0.31

0.32

0.33

0.34

0.35

0.37

0.38

0.40

0.42

0.44

0.45

0.46

0.50

0.51

0.61

Frequency of Enrollees

Exceptional Health Care Needs Composite Score

Figure 2. CAHPS II Data - Exceptional Health Care Needs Composite Score (N=2,132)

Arkansas Health Care Independence Program (‘Private Option’) Final Report E-3

Theoretical Causal Determination of Program Effect

As stated in the report, regression discontinuity is a quasi-experimental design that is increasingly being used in

evaluation analyses to test differences attributable to group assignment. This approach is useful when random

assignment to groups is not feasible. We adhered to standards for regression discontinuity designs

which included:

1) Treatment assignments were based on a forcing variable (composite score); enrollee assignment to

Medicaid and QHP groups on the basis of a threshold cut-point from the composite score met

assumptions to conduct a regression discontinuity design analysis.

2) The forcing variable must be ordinal with a sufficient number of unique values (higher composite

scores indicated higher needs).

3) There must be no factor confounded with the forcing variable. Overall, those assigned to Medicaid with

higher needs were slightly older and tended to be female and white more often than those assigned to a

QHP. Locally around the cut-point, we did not see any differences across these confounders.

Modeling Strategy

Since our outcome variables are discrete — either counts (e.g., number of emergency room visits) or binary (e.g.

any readmission within 30 days of an indexed hospitalization) variables — we used the local linear regression

modeled as below.





  



 



















) + 















), where 



is the outcome for individual  and 



is a dummy

variable for the treatment (1 = QHP, 0 = Medicaid), 



is the composite score, 



is the threshold (cut-off) for

assignment (0.18), 



is the parameter for the effect of association between the outcome and being in a QHP

(treatment group) compared to Medicaid (comparison group). This model also incorporates an interaction

between treatment assignment and difference in composite score from the cut-off.

For the regression discontinuity models presented in the Final Report, we incorporated optimal bandwidths. An

Imbens-Kalyanaraman optimal bandwidth calculation

and heteroskedasticity-consistent (“HC1”) variance were

estimated.

Since the construction of frailty scores were discrete, standard errors were adjusted for clustering within

frailty scores.

Furthermore, triangular kernel weights were applied which gave observations closer to the threshold of

0.18 higher weights than those further away from the threshold within the optimal bandwidth around the threshold

and zero weights to those outside of the estimated optimal bandwidth.

We used the R statistical computing language (R Foundation for Statistical Computing, Vienna, Austria) to run the

regression discontinuity design (RDD) package developed by Drew Dimmery (2016) and fit the regression models

with the functional form above. For each model we reported linear average treatment effects and the p-value

corresponding to the bandwidth.

Models Using Stabilized Inverse Probability of Treatment Weighting (Geographic Access Models)

Comparison Groups

We included individuals not completing the exceptional healthcare needs assessment Questionnaire and meeting

other inclusion criteria detailed in Appendix D for claims-based metrics and Appendix G for CAHPS respondents,

contained in the Interim Report.

Individuals assigned to the traditional Medicaid fee-for-service plan were mostly

parents, and mostly with very low incomes, while those assigned to a QHP were, in general, a more equal mixture of

parents and individuals with no dependents, with many having higher income levels than those assigned to

Medicaid. Since the presumptive assumption of assignment to Medicaid and QHP did not hold to meet standards of

using a regression discontinuity design based on parenthood status and an income forcing variable we chose to

Arkansas Health Care Independence Program (‘Private Option’) Final Report E-4

match the beneficiaries assigned to Medicaid and QHPs by using propensity scores. In total, our analytic claims file

contained 27,905 assigned to the General Population Medicaid plan and 92,940 assigned to a General Population

QHP. For CAHPS, our analytic file contained 275 assigned to the Medicaid plan and 835 assigned to a QHP.

Theoretical Association Determination of Program Effect

To test for the association of plan assignment and outcome, we mitigated differences in assignment that may

have been due to demographic or other factors attributing to the assignment (income was not used). Propensity

scores are the probability of being assigned to a treatment group (i.e., a QHP) given a set of underlying

characteristics.

Stabilized Inverse Probability of Treatment Weighting (SIPTW)

Following nomenclature outlined in Xu and colleagues,

we let z be a binary outcome where 1 indicates that a person

was assigned to a QHP and 0 as assignment to the Medicaid plan. X is a row vector of covariates (characteristics) and π

indicates for the probability of being assigned to a QHP (the propensity score).

For our claims analytic data the total number (N=80,505) was equal to the 11,006 assigned to Medicaid (n

) plus 69,499

assigned to a QHP (n

). The crude calculation of probability of treatment was p = n

/N = 69,499/80,505 = 0.86. The

probability of no treatment, or being assigned to Medicaid was 1-p = 1 – 0.86 = 0.14.

The propensity score associated with an individual (i) was estimated using a logistic regression with the functional

form of:



















, where β is a vector of parameters to be fit in the model and associated with the 



a row vector of

covariates for individual i (age, gender, race/ethnicity, and parent status for claims data and in addition education,

marital, employment, and obesity status for CAHPS data — no data on 2013 health status was available). We used a

class statement to produce dummy covariate categories. Inherent in the model is a parameter (β

) associated with

the intercept for the model. Our logistic regression model had a reasonably high coefficient of determination (c-

statistic) of 0.718 for claims data and 0.717 for CAHPS data.

The resulting estimated propensity scores were used to produce inverse proportional treatment weights (IPTW)

as follows:

















 















  , where 



denotes the IPTW for individual i.

Weights for individuals assigned to a QHP (treatment group) with a low propensity score and those assigned to

Medicaid with a high propensity score for assignment into a QHP will be high and increase the risk of a Type I

error. To reduce this risk we adjusted, or stabilized, the increased pseudo-sample size that was created by using

large weights as follows:

















 















  , where  is the crude probability of treatment without

considering covariates and



denotes the SIPTW for individual i.





effectively keeps sample sizes used in statistical models the same as if those weights had not been included

and had not artificially increased the risk of a Type I error.

Arkansas Health Care Independence Program (‘Private Option’) Final Report E-5

Statistical Models with Treatment Assignment as the Primary Predictor

Statistical models to fit access, quality, and utilization outcomes incorporated the distributional form of the

outcome that produced the best-fitting model. For example, for the number of total emergency room visits over

the course of 12 months, we used a negative binomial distribution after demonstrating that this distribution

provided a better model fit. The only covariate that was used in models testing for a treatment effect (primary

predictor) was urban/rural designation determined by a RUCA classification (see Appendix D).

General linear models were fit using SAS Enterprise Guide V7.1 HF1 (SAS Institute, Cary, NC). For outcomes that

were affected by duration of tenure with coverage (ranging from six to 12 months) we produced estimated rates

based on 12 months of coverage. We accomplished this by using an offset option in the model statement of the

proc genmod SAS procedure. The SAS statement to produce the results for total emergency room visits is

presented below. The stabilized inverse probability of treatment weights were labeled “siptw.”

proc genmod data = ps_data;

class treatment urban;

model num_poid_er_visits_total = treatment urban/dist = negbin link = log

offset = TotDur_POID_per_yr_ln type1 type3;

weight siptw;

lsmeans treatment/ilink cl alpha = 0.05;

run;

Models Using Propensity Score Matching

Comparison Groups

We included individuals not completing the exceptional healthcare needs assessment Questionnaire and meeting

other inclusion criteria detailed in Appendix D. Individuals assigned to the traditional Medicaid fee-for-service plan

were mostly parents, and mostly with very low incomes, while those assigned to a QHP were, in general, a more

equal mixture of parents and individuals with no dependents, with many having higher income levels than those

assigned to Medicaid. Since the presumptive assumption of assignment to Medicaid and QHP did not hold to

meet standards of using a regression discontinuity design based on parenthood status and an income forcing

variable, we chose to match the beneficiaries assigned to Medicaid and QHPs by using propensity scores. In total,

our analytic claims file contained 27,905 assigned to the General Population Medicaid plan and 92,940 assigned to

a General Population QHP. For CAHPS II, matching respondents meeting the new inclusion criteria produced an

analytic file containing 275 assigned to the Medicaid plan and 835 assigned to a QHP.

Theoretical Association Determination of Program Effect

Propensity score analysis is a post-hoc statistical method that estimates a treatment effect when subjects are not

randomly assigned to a treatment group. To test for the association of plan assignment and outcome, we

mitigated differences in assignment that may have been due to demographic or other factors attributing to the

assignment. Propensity scores are the probability of being assigned to a treatment group (i.e., a QHP) given a set

of underlying characteristics.

Propensity Score Matching

For claims and CAHPS II data separately, logistic regression models were fit modeling assignment to a QHP as the

treatment (1) and assignment to Medicaid as the control (0). Covariates used in the models included age, gender,

Arkansas Health Care Independence Program (‘Private Option’) Final Report E-6

race/ethnicity, Charlson comorbidity index (obtained from the complete first year experience in the program),

insurance region, and census block median income.

The propensity score associated with an individual (i) was estimated from the logistic regression with the

functional form of:



















, where β is the vector of parameters to be fit in the model and associated with the vectors of

covariates listed above. The resulting propensity scores were matched using a greedy matching algorithm.

The

algorithm first attempts matches between Medicaid and QHP where the propensity score is identical to eight

decimal places. Unmatched treatment and controls are then attempted to be matched at a less restrictive seven

decimal places and this process continues until a caliper boundary of 0.005 is met. All remaining unmatched

treatment and controls are not entered in the analysis. For full population denominator indicators, we were able

to find matched one-to-one pairs of Medicaid and QHP enrollees in 99.7 percent of claims for Medicaid enrollees

and 94.2 percent of CAHPS II Medicaid enrollees. There were no standardized differences in any of the covariates

that were included in the logistic regression in claims, CAHPS II, and the special population pregnancy study.

The c-statistic of the claims data logistic regression model used to generate the propensity scores was 0.67, indicating a

moderate ability to differentiate between those assigned to a QHP and those with Medicaid coverage. The c-statistic

represents the discriminative power of the logistic regression model. The c-statistic of the CAHPS II data logistic

regression model was 0.63 and for the pregnancy study Medicaid and QHP match, the c-statistic was 0.87.

Matched pairs were accounted for in individual generalized linear models by using a random effect indicator for the

dyads. In reality, the number of matched pairs was too high for fitting a random effects model and we therefore

collapsed matched propensity scores into 100 groups (percentiles) and used this variable as a random effect.

References

Schochet P, Cook T, Deke J, Imbens G, Lockwood JR, Porter J, and Smith J. Standards for Regression Discontinuity Designs.

Retrieved from What Works Clearinghouse. Website http://ies.ed.gov/ncee/wwc/pdf/wwc_rd.pdf Accessed 8/6/2015.

McCrary, Justin. (2008) "Manipulation of the running variable in the regression discontinuity design: A density test," Journal

of Econometrics. 142(2): 698-714.

Guido Imbens & Karthik Kalyanaraman. Optimal Bandwidth Choice for the Regression Discontinuity Estimator. The

Review of Economic Studies. Vol. 79, No. 3 (July 2012), pp. 933-959.

MacKinnon JG, White H (1985). “Some Heteroskedasticity-Consistent Covariance Matrix Estimators with Improved Finite

Sample Properties.” Journal of Econometrics, 29, 305–325. doi:10.1016/0304-4076(85)90158-7.

Lee, D.S. and Card, D. (2008): “Regression discontinuity inference with specification error,” Journal of Econometrics, 142,

655–674.

Arkansas Center for Health Improvement. Arkansas Health Care Independence Program (“Private Option”) Section 1115

Demonstration Waiver Interim Report. Little Rock, AR: Arkansas Center for Health Improvement, March 2016.

Xu S, Ross C, Raebel MA, Shetterly S, Blachetter C, and Smith D. Use of stabilized inverse propensity scores as weights to

directly estimate relative risk and its confidence intervals. Value Health. 2010;13(2):273–277.

Guo S and Fraser MW. Propensity Score Analysis: Statistical Methods and Applications. Second edition. Thousand Oaks, CA:

Sage Publications, Inc; 2015.

Parsons LS. Reducing bias in a propensity score matched-pair sample using greedy matching techniques; 26th Annual SAS Users

Group International Conference; Long Beach, California: 2001.

Arkansas Health Care Independence Program (‘Private Option’) Final Report F- 1

Appendix F – Supplemental Payments, Claims Loading Process, and Per Member

Per Month Logic

The cost analysis focused on a comparison of health services provided to commercial qualified health plan (QHP)

enrollees in the Health Care Independence Program (HCIP) premium assistance program relative to health service

provision under a traditional Medicaid Fee-for-Service (FFS) program. The analysis required an assessment of the

costs of services provided under the two alternatives that can be compared to assessments of access,

effectiveness, and quality outcome measures. Our main hypothesis was that the provision of health services

through QHPs results in higher per member per month (PMPM) costs (payments) relative to traditional Medicaid,

but quality and access would be rated higher by enrollees in a QHP. Policymakers generally believe, and evidence

suggests, that access to healthcare services is improved within state Medicaid programs when payments for

primary care services increase.

If our hypothesis is confirmed, the findings provide information for policymakers to evaluate the trade-offs

between the cost of care, access to health services, and the quality of services provided. In all of our discussions of

cost, we follow the health services literature and define cost as the amount paid on behalf of beneficiaries by

insurance or Medicaid payers, plus the amount paid by the beneficiary.

In order to obtain a per member per month cost (PMPM) we used the claims data provided by Medicaid and QHPs

and incorporated in the Arkansas All-Payer Claims Database. As claims data are limited to claims paid, we had to

adjust claims for PMPM calculations to include important supplemental payments made by the Medicaid program

to providers. In Arkansas, providers receive payments based on upper payment limits (UPL) and cost-based

reimbursement from the state Medicaid program. We define supplemental payments as any payment made for

the utilization of services that were not paid through claims. Supplemental payments totaled $596,142,904.30 in

fiscal year 2016, including disproportionate share hospital payments (DSH). As these payments are made for

services provided, they should be incorporated into the costs of claims paid, rather than treated as an

administrative cost allocated across all beneficiaries. The process for measuring supplemental payments and

allocating them to claims is discussed below.

Both UPL payments and cost-based reimbursements for inpatient and hospital outpatient services were identified

for every hospital receiving them from the Division of Medicaid Services (DMS). As there is inconsistent use of

cost-reports and different resource pools for allocating supplemental payments to generate payment algorithms,

UPL and cost-based reimbursement were averaged across four categories of hospitals: academic (UAMS), private,

critical access, and public hospitals.

Specifics of this allocation are contained in Table 1. Allocation of inpatient supplemental payments was

straightforward as these payments are based on discharges and not days. However, allocation of outpatient

hospital supplemental payments proved difficult to identify and, as such, were also allocated with inpatient

hospitalizations. Sensitivity analyses were performed to test the effect of adding hospital outpatient to

inpatient claims.

Arkansas Health Care Independence Program (‘Private Option’) Final Report F- 2

Table 1. Allocation for 2016 Supplemental Payments by Hospital Category

Hospital Category

Inpatient

UPL/Discharge

Outpatient

UPL/Discharge

Cost-Based

Reimbursement

Total Supplement

Academic (UAMS)

$2,699.00

N/A

$1,148.00

$3,847.00

Private

$2,946.00

$983.00

$20.00

$3,949.00

Critical Access

$4,172.00

$2,419.96

$5,144.00

$11,735.00

Public

$6,621.00

$150.77

$0.00

$6,771.00

Average/Discharge

$3,093.00

$889.00

$309.00

$4,529.00

Total costs were calculated as the sum of claims costs plus administrative costs for both Medicaid and QHPs. For

Medicaid administrative costs as depicted in Table 2, we identified expenditures for services across all agencies

associated with Medicaid that received matching funds from the federal government. We also identified other

payments made by DMS that could not be included in claims such as physician payments for rate adjustments,

case management services, contracts, and payments for quality improvement initiatives.

Table 2. Administrative Costs for 2016

Administrative Cost

Non-Fixed

Fixed

Others

All Costs

64.10 Base and Waivers

$375,059,126.00

21. Children’s Health Insurance Program (CHIP)

$625,391.00

Inpatient Quality Incentive (IQI) Payment

N/A

64.10's Net

$7,774,710.00

Disproportionate Share to Hospitals (DSH)

N/A

Graduate Medical Education (GME)

$10,319,457.00

Physician Rate Adjustment (UAMS)

$35,954,166.00

Physician Services

N/A

Prescription Drug – Clawback

N/A

Case Management

N/A

Payment to Federally Qualified Health Center

(FQHC)

N/A

Total

$411,013,292.00

$18,719,558.00

$0.00

$429,732,850.00

PMPM Admin

$45.17

$2.06

$47.23

Note: We were unable to report $1,978,130 Total Cost of CQ/64.10B, line 2B (FFY16 Q1), Medicaid Management Information (MMIS)

related cost this quarter due to insufficient allotment available. There will be a PPA in FFY16 Q2. It is not included in this total.

Abbreviations: PMPM=total administrative cost/total enrollee months in Medicaid (excluding 06 commercial months).

Calculation of Per Member Per Month Payments

This section describes the calculation of per member per month (PMPM) payments made by the Arkansas

Medicaid program for services provided to people in one of our study groups — newly insured parents with

incomes at or below 17 percent of the federal poverty level (FPL). People in this group received traditional

Medicaid benefits that did not change as a result of Medicaid expansion. Given the benefit structure and

population, we report in this section actual service utilization and payment measures, including PMPM payments

for the 174,167 enrollees in our analytical sample — all of whom had at least six months of coverage. Payments

were categorized as inpatient, outpatient emergency room (ER) visits, outpatient physician visits, prescription

drugs, and other services that fell outside of these categories.

Utilization and payment of services are presented using the following measures: the percentage of enrollees

who use the service category, the amount of spending for the service category conditional on having positive

spending (or the probability of positive spending), the average spending on the category conditional on

having positive spending, and per person spending defined as total spending for the category divided by the

Arkansas Health Care Independence Program (‘Private Option’) Final Report F- 3

number of enrollees in the group. PMPM payment was calculated by dividing average payment pe r person by

average member months. Per member per year (PMPY) payment was calculated by multiplying PMPM

payment by 12 months.

Table 3 provides observed payments for traditional Medicaid and QHP enrollees on a per member per month

basis by service category, administrative costs, and overall. Also, we present estimates of PMPM payments based

on an additional method. PMPM payments were calculated similarly for both traditional Medicaid and QHP

enrollees by summing adjudicated claims for the different categories of services and then summing the total

months of services for each enrollee. PMPM payments are defined as total payments divided by total months of

coverage. Service categories for both traditional Medicaid and QHP enrollees were defined as inpatient,

prescription drugs, outpatient ER visits, outpatient physician, and all other services including outpatient

ambulatory surgeries, labs, imaging, and other services that could not be assigned. Inpatient services include ER

services that resulted in an admission, as well as other labs, imaging, and other services that could be identified as

occurring during the inpatient hospitalization.

Table 3. Observed and Estimated Per Member Per Month Cost Scenarios for Traditional Medicaid and

QHP Enrollees by Service Category (2016)

Per Member Per Month (PMPM) Observed and Estimated Costs (2016)

Service Category

Observed Medicaid

Commercial

Premiums

Observed

Commercial

Estimated

Medicaid

Inpatient

$86.85

$123.99

$82.58

Emergency Room

$13.61

$44.16

$12.45

Physician’s Office

$31.36

$65.68

$37.27

Outpatient

$9.17

$11.63

$13.37

Prescription Medications

$40.27

$113.18

$48.15

Other

$64.44

$146.49

$62.83

Non-Emergency Transport Fees

$3.65

$4.04

$0.01

Total Claims PMPM

$249.35

$509.17

$256.66

Administrative PMPM

3,4

$63.58

$91.65

$63.58

Total PMPM

$312.93

$478.14

$600.82

$320.24

Notes:

Average PMPM represents loaded (i.e., cost-based reimbursement, supplemental payments, etc.) claims from Traditional Medicaid.

The average PMPM payment made to QHP carriers includes cost-sharing and wrap services.

Includes PMPM Administrative costs for

Observed QHP claims, which were set at 18 percent of the total claims paid excluding copayments and deductibles.

Average PMPM

represents claims costs based on average Medicaid pricing.

Column 1 (Observed Medicaid) provides PMPM payments for enrollees in traditional Medicaid that had

incomes below 17 percent of FPL. PMPM payments were much lower than for QHP enrollees because of the

differences in prices paid for services between the two programs. The largest difference in PMPM payments

were found in the “Other” category, suggesting the need for a greater understanding of spending in this

category across the two programs. Administrative costs for traditional Medicaid were based on an exhaustive

examination of spending across all Department of Human Services categories that were eligible for matching

funds from the federal government. Spending on disproportionate share adjustments and Graduate Medical

Education were included as these are paid by the Medicaid program to deliver services. Total administrative

spending was divided by total enrollee member months excluding QHP enrollees to generate PMPM payments

for administrative services ($63.58).

Arkansas Health Care Independence Program (‘Private Option’) Final Report F- 4

Column 3 (Observed Commercial) of Table 3 shows that PMPM payments for enrollees in a QHP incurred $509.17

in payments per month with the majority of payments incurred for Inpatient and Other services. Total payments

included administrative loading costs that were estimated to be 18 percent of total claims paid ($91.65 PMPM)

resulting in total monthly payments of $600.82.

Column 4 (Estimated Medicaid) provides PMPM costs for QHP enrollees under the assumption that payments for

services would reflect the prices paid in the Medicaid program. Under this methodology, prices for services were

altered to reflect the experience of the traditional Medicaid population, while holding utilization of services for

the QHP enrollees constant. Allowing prices to change for the QHP enrollees substantially reduced estimated

PMPM expenditures with the largest change occurring for Inpatient and Other services. Administrative payments

were calculated by excluding payments that would not change irrespective of how the program was financed. DSH

payments and GME payments would not change with a difference in volume following the implementation of

HCIP and were considered fixed costs. In calculating administrative payments for only services associated with the

volume of the program, there was no change in administrative payments. Total payments were estimated to be

$320.24 PMPM.

References

Decker SL. Medicaid Physician Fees and the Quality of Medical Care of Medicaid Patients in the USA. Rev Econ Household

(2007). 5:95-112.

Pauly MV. Lessons to Improve the Efficiency and Equity of Health Reform. Hastings Center Report (2012). 42(5):21-24.

Arkansas Health Care Independence Program (‘Private Option’) Final Report G- 1

Appendix G – Geographic Information System (GIS) Processing

The Center for Advanced Spatial Technologies (CAST) was founded in September 1991 within the University of

Arkansas, J. William Fulbright College of Arts and Sciences by the Board of Trustees and the Arkansas Department

of Higher Education. Research efforts include multi-disciplinary projects from all around the world. CAST is funded

by competitive grants and research proposals from federal and state government offices, universities, nonprofit

groups, and other organizations.

Medicaid and commercial qualified health plan (QHP) enrollees and provider data (i.e., Blue Cross Blue Shield,

Ambetter, and QualChoice) were used to assess network adequacy and perform GIS mapping. For the analysis,

“access” was defined as the distance (30 or 60 miles) or time of travel (30 or 60 minutes by passenger automobile)

from any enrollee’s location to the nearest provider, within an “access ring” offering any of the eight categories of

healthcare provider types (primary care, orthopedics, ophthalmology, OB/GYN, oncology, surgical, psychology, and

cardiology). These “access rings” were created by computing four distinct travel distances and time-of-travel attributes

from each unique provider location (latitude and longitude coordinates) for each provider type. From each provider

location, a set of four miles traveled along the existing (ESRI, 2013) street centerline networks were computed as

“ringed-polygons” at the following distance rings: 0-15, 15-30, 30-45, and 45-60 miles.

Only providers that practiced within calendar year 2014 were included in the analyses. It is important to note that as

both enrollees and providers included out-of-state addresses, ACHI and CAST agreed, for this analysis, to compile travel

distances or travel-time polygons from any provider that was located within 70 miles of the Arkansas state boundary. It

was not uncommon for enrollees living near a state border to seek health care from a provider across the border.

De-identified data were sent to CAST for analysis. CAST returned to ACHI four provider-enrollee data tables

each — for Medicaid and the three QHPs. An additional data file combining all commercial enrollees and

providers was also prepared and returned. QHP enrollees were encouraged to seek care with a provider inside

a QHP, but were not pre-empted from going out-of-network. Within each enrollee dataset, fields containing

minutes and miles to each provider type were included, along with the geocoded locations for the enrollee. A

geocoded provider file listing was also returned to ACHI.

More specifics of the data processing steps used by CAST can be found on the following pages.

CAST was responsible for the development of the following products for the Arkansas Center for Health

Improvement (ACHI) between Oct. 20, 2015, and Feb. 15, 2016:

1. geocoded locations (latitude/longitude; WGS84) for all ENROLLEES and PROVIDERS of each of four

specified insurance programs,

2. “access zones” for each of the eight provider “specialties” for each of the four specified insurance

programs, based upon both drive-time (minutes) and distance (miles) criteria, plus the

3. updated tables for all geocoded ENROLLEES of each of the specified insurance providers,

indicating the time/distance zones that enrollees fall within for each PROVIDER specialty, and

4. delivery of GIS results as they were completed.

All work was performed on a physically and electronically secure computer, disconnected from any networks. The

specific methods for generating each of these products were as follows:

1. Geocoding and GIS data processing for four Arkansas-based healthcare PROVIDERS and their current

ENROLLEE tables obtained by ACHI and provided as .csv files (with schema descriptions) to CAST:

Arkansas Health Care Independence Program (‘Private Option’) Final Report G- 2

a. CAST imported each provider and their enrollee tables (while maintaining ACHI-generated

unique IDs) into an ArcGIS File Geodatabase format.

b. Geo-locations for each providers’ facility address and each of their enrollee home

addresses were assigned using the best available Arkansas GIS locator(s) as the first-pass

for each provider (and their enrollee tables). The GIS locator used for Arkansas (“GeoStor

Composite” address locator, Arkansas GIS Office) leveraged our statewide 911 street

address datasets from 2015, which included street centerline AND address points (any

911 addressed structure), plus the fallback of the nine-digit zip code centroids (e.g.

73722-2356) located within the state of Arkansas. We set the minimum match score of

“65” (0-100) for these geocoded addresses. Any original misspellings or address issues

found within the source data tables were not corrected by CAST.

c. For the second-pass of geo-locating any remaining records that were not geocoded within

the first pass, we also re-ran the resulting “non-located” records against ESRI’s nationwide

street centerline network (updated in 2013), so that we could achieve a street-level

geocode for any remaining records that MAY have been missed within Arkansas OR that

were located outside of the state boundary.

d. The results of these two attempts to locate each and every record within all four provider

and enrollee tables were then MERGED into a common, single database table schema

grouped by healthcare provider. We maintained the original ACHI attributes contained

within each table and ADDED VALUE to each of these tables by providing the

latitude/longitude coordinates (WGS84), along with additional metadata references such

as: Locator Name, Status, Score, Match Type, and the resulting Matching Address from

the GeoLocator used to geocode each record. These additional attributes were appended

to each of the original data tables from ACHI, which also contained the original data table

“unique ID,” which was generated in-house by ACHI to help maintain anonymity for all

providers and their enrollees.

2. Determination of each healthcare enrollee’s “ACCESS” to each of the eight categories of provider-

specific services WITHIN and OUTSIDE OF each provider network:

a. ACHI originally defined “access” as the distance (30 or 60 miles) or time of travel (30 or 60

minutes by passenger automobile) from any enrollee’s location to the nearest provider

offering any of the eight categories of healthcare provider services (primary care,

orthopedics, ophthalmology, OB/GYN, oncology, surgical, psychological, and cardiology).

CAST proposed that ACHI compute a total of four distinct travel distances and time-of-travel

attributes from each unique provider location (latitude/longitude coordinates) for each

provider type. Therefore, from each provider location a set of four MILES-traveled along the

existing (ESRI, 2013) street centerline networks were computed as “ringed-polygons” at the

following distance rings: 0-15, 15-30, 30-45; 45-60 MILES.

b. The same provider locations and nationwide street network dataset (ESRI, 2013) was used

to ALSO compute a set of TIME-traveled (MINUTES) ringed-polygon buffers out from each

provider location — following the same street network dataset. The TIME-traveled was

computed using the ArcGIS Network Analyst extension for each provider location using

“typical” driving conditions for a passenger car traveling the posted speed limits along

each street segment as a hypothetical enrollee travels to or from each provider location

(geocoded latitude/longitude coordinate). These travel-time-in-minutes were created in

Arkansas Health Care Independence Program (‘Private Option’) Final Report G- 3

sets of four from each provider and attributed as: 0-15, 15-30, 30-45, and 45-60 minute

ringed-polygon geometries.

c. It is important to note that ACHI and CAST agreed to ONLY compile travel distances or

travel-TIME polygons from any provider located within 70 miles of the Arkansas state

boundary. It was determined that it is very common for enrollees in a healthcare network

to travel across state lines to access any of the eight provider specialties as long as they

were “in-network” facilities.

d. The resulting set of provider polygon files enabled us to summarize the following

information for each provider.

e. Joining the original PROVIDER services-offered at each Unique Location enabled us to

select each “Services Provided” attribute and then “Merge Polygons” (actually, DISSOLVE)

by travel time or distance so that we would end up with 32 unique GIS data layers

(polygon geometry) for each provider specialty of the four unique providers (Blue Cross

Blue Shield, Medicaid, QualChoice, and Ambetter).

3. Once each of the four unique providers (Blue Cross Blue Shield, Medicaid, QualChoice, and Ambetter)

had their travel DISTANCE buffers (miles) and TIME-OF-TRAVEL (minutes) computed with four distinct

zones for each provider specialty, they could be used to leverage the spatial analysis of our

geographic information systems by using a “spatial selection” of the complete set of ENROLLEE

address locations (latitude/longitude geocodes) against each of their corresponding “in-network”

provider SPECIALITY with their travel distance/time merged buffers. We prepared each of the four

CAST geocoded provider-ENROLLEE tables with the eight provider services specified by ACHI within

this study and we processed all of those enrollee records using this methodology:

a. Add eight attribute columns needed to store each provider specialty (long integer).

b. Update any enrollee addresses that were NOT geocoded to the value “1111.”

c. Reverse the selection and then update all remaining records to the value “9999” within

EACH of the eight ACHI provider specialty columns.

d. We agreed that this value (“9999”) would represent any member located BEYOND our

cut-off of 60 miles or 60 minutes travel time from any of the eight provider specialty

columns. Therefore, we pre-populated those values before processing a series of “spatial

selections” between our ENROLLEE (point features) and the full-set of 32 travel polygons

grouped by distance and provider specialty (polygon feature classes).

e. For each ENROLLEE table we would methodically work through each of the eight provider

specialty TYPES using a SELECT BY LOCATION command and working our way from the

GREATEST distance or time; INWARD towards the nearest/shortest travel time. Each

enrollee record would be OVERWRITTEN or updated with ONE of these FOUR values (60,

45, 30, 15), which would indicate (depending upon the ENROLLEE table — MILES or

MINUTES — the quickest “level of access” each enrollee has to ANY provider within their

healthcare network for each of the eight specialty types.

4. Resulting Data Delivery: The four provider enrollee data tables were delivered as .csv and .xlsx

formats to ACHI using a secure FTP protocol and with AES-256 encryption applied. Within each

provider dataset a MINUTES and MILES-traveled table for each of the original enrollee records

were provided with a supporting CODE BOOK detailing the attribute columns and attribute values

contained within the delivery

Arkansas Health Care Independence Program (‘Private Option’) Final Report H- 1

Appendix H – Weighted Average Price Calculations

To determine the average weighted price for Medicaid and Qualified Health Plans (QHPs), the paid amount listed

on the paid claims were rolled up on procedure codes for office visits for high utilization provider specialties. It

was anticipated that all payer/provider combinations were likely to have claims with office visits coded in a similar

manner regardless of provider specialty. This allows us to compare pricing across procedures and providers. One

limitation of these analyses is that not all payers used the same provider specialty codes to identify the different

provider types studied. In general, payers used provider specialty codes that were consistent with Medicare

coding. The only difference between payers was that the two smaller payers omitted the leading zero character

from the specialty code. Medicaid provider specialty codes are constructed with both typical Medicare provider

specialty codes and Medicaid specific provider specialty codes (See Table 1).

Table 145. Provider Specialty Codes Selected for Comparison

Provider Specialty

Medicaid

Centene/

QualChoice

BCBS

Primary Care Physician

01, 08, 11

1, 8, 11

01, 08, 11

Advanced Practice Nurse (APN)

N0, N6, N7

Cardiologist

General Surgeon

Obstetrician/Gynecologist (OB/GYN)

16, 19

Oncologist

83, 90, 91, 98

Ophthalmologist

Orthopedist

20, 57

Psychologist/Psychiatrist

26, 62

26, 62, 68

In order to calculate the weighted average price, claims were extracted from Medicaid medical and UB92 paid claims

and the medical paid claims from the QHP files if they contained any of the procedure codes listed in Table 2. These

claims were further reduced by extracting only those with the provider specialties listed in Table 1. The total and

average paid amounts, as well as claim counts were determined by procedure code for each provider specialty, for

each payer type (Medicaid and combined QHPs). For each provider specialty the Medicaid average paid amount was

multiplied by the number of commercial claims for procedure codes and summed, which gave a total cost for

Medicaid procedures at QHP reimbursement rates. It is important to note that only procedure codes that were

represented both in Medicaid and QHP claims were included in the weighted averages. This total was then divided

by the total number of QHP claims for the matching procedures to obtain a weighted average price.

Arkansas Health Care Independence Program (‘Private Option’) Final Report H- 2

Table 2. Procedure Codes and Their Descriptions

Code Ranges

Code Descriptions

CPT Codes

99201–99206

Office or other outpatient visit (New patient 10/20/30/45/60 minutes)

99211–99215

Office or other outpatient visit for the evaluation and management of an established

patient that may not require the presence of a physician or other qualified healthcare

professional (10/15/25/40 minutes)

99241–99245

Office consultation for a new or established patient

99343; 99347–99349

Home visit for the evaluation and management of a new/established patient

99384–99387

Office or other outpatient visit — preventive (new patient)

99394–99397

Office or other outpatient visit — preventive (established patient)

99401–99402

Preventive medicine counseling and/or risk factor reduction intervention(s) provided to

an individual (separate procedure); approximately 15/30 minutes

HCPCS Codes

G0438

Annual wellness visit, initial (AWV)

G0463

Hospital outpatient clinic visit for assessment and management of a patient

T1015

Clinic visit/encounter, all-inclusive

Note: Ranges of procedure codes for office visits are listed for simplicity. The difference within the codes in these ranges is

based on the level of care a patient would need for that visit.

Arkansas Health Care Independence Program (‘Private Option’) Final Report I- 1

Appendix I – Pregnancy Analytic Profile

This appendix contains additional technical and analytical information not presented in the body of the report,

including a detailed description and flow chart of the inclusion/exclusion criteria used to derive the analytical

sample, as well as the socioeconomic and health related demographics that were used to develop the propensity

scores.

Steps for inclusion / exclusion criteria in prenatal and post-partum visits:

1. All births in the birth certificate data having a mother participating in a QHP or in Medicaid were

selected (n = 34,675).

2. All births occurring on or between Sept. 1, 2014, and Sept. 30, 2015, were retained (n = 17,605).

3. Mothers and associated births which did not have delivery claims in the medical claims file were

removed (n = 15,734).

4. Infants must have been live-born according to the birth certificate (n = 15,597).

5. Mothers must have had continuous enrollment for 43 days prior to and 56 days after the date of

birth in any one of the following enrollment aid categories: 06, 20, 25, or 6X (except 69) (n =

14,502).

6. Mothers and associated births were excluded if:

a. The mother had enrollment segments other than 69 prior to Oct. 15, 2013 (n = 10,864).

b. The mothers had enrollment segments for dual eligibility, physical disability or mental

retardation (10, 11, 13, 14, 16, 17, 18, 31, 33, 34, 35, 36, 37, 38, 41, 43, 44, 45, 46, 47, 48,

49, 58, 78, 88) at any time during 2014/2015 (n = 10,862).

c. There were two or more claims more than 30 days apart any time during 2014/15 from

an institutionalized facility based on place of service codes (09 , 13, 14, 31, 34, 55, and

56) (n = 10,862).

7. Births with same mother and date of delivery were only counted once while births on different

dates were counted as separate births (n = 10,722).

8. Mother-birth pairs were excluded if the assigned Private Option identifier was associated with

multiple provider IDs (Medicaid or private) (n = 10,711).

9. Mothers with more than one switch between Medicaid coverage and commercial coverage were

removed (n = 10,692).

10. Mothers must have spent at least 50 percent of the prenatal time in the enrollment segment

closest to the delivery (n = 10,467).

11. Mothers who were missing data in the “Previous live births” or “Previous births now deceased”

were excluded (n = 10,453).

Subpopulation studied: Mothers who did not have continuous enrollment prior to delivery going back up to week

24 of gestation (n = 9,943).

To analyze the population, a propensity scored one-to-one matched sample with a caliper of 0.005 was created

and analyzed (matched n = 2,369). Medicaid beneficiaries were considered as the control group and compared

with those in the QHP.

Arkansas Health Care Independence Program (‘Private Option’) Final Report I- 2

Figure 1. Flow Diagram Depicting the Development of the Analyzable Study Population

Total prenatal and post-partum visits starting with all births found

in the birth certificate data having a mother POID

N = 34,674

N = 15,597

Exclude mother POIDs and associated births with no

delivery claims

N = 1,871

N = 10,864

Exclude persons not continually enrolled for 43 days

prior to date of birth and 56 days after date of birth

N = 1,095

Analyzable Study Population

N = 10,453

Exclude persons who are dually eligible, disabled or

institutionalized in 2014

N = 2

Exclude POIDS and births associated with coverage

discrepancies

N = 11

Exclude births prior to September 1, 2014, and after

September 30, 2015

N = 17,069

Exclude births with deceased status on the birth

certificate

N = 137

Exclude those who made more than one switch between

Medicaid and Commercial plans

N = 19

Deduplicated all births with same mother POID and and date

of delivery

N = 140

Exclude persons that had enrollment segments other

than “69" prior to October 15, 2013

N = 3,638

Exclude those who spent less than 50% of the prenatal time

in the enrollment segment closest to delivery

N = 225

Exclude those who had missing “previous live births” or

“previous births now deceased”

N = 14

Arkansas Health Care Independence Program (‘Private Option’) Final Report I- 3

Table 1. Socio-Economic Demographics Selected as Covariates

Raw Sample

Matched Sample

Medicaid

N=7,583

QHP

N=2,870

Medicaid

N=2,369

QHP

N=2,369

Covariates

Category

Percent

Mother’s age

25 years old or less

3,749

49.4

1,067

37.2

986

41.6

999

42.2

25 ≤ 30 years old

2,218

29.3

1,037

36.1

830

37.4

812

34.3

30 ≤ 35 years old

1,149

15.2

569

19.8

397

16.8

411

17.4

35 ≤ 40 years old

397

5.2

163

5.7

134

5.7

121

5.1

40 and above

0.9

1.2

0.9

1.1

Mother’s

education

No high school

diploma

1,435

18.9

375

13.1

327

13.8

343

14.5

High school diploma

or GED

5,259

69.4

2,056

71.6

1,721

72.7

1,704

71.9

Some college

441

5.8

235

8.2

171

7.2

174

7.3

Bachelor’s degree

340

4.5

163

5.7

112

4.7

115

4.9

Graduate degree

0.8

1.0

0.9

Unknown

0.7

0.5

0.7

0.5

WIC

Yes

5,661

74.7

2,096

73.1

1,750

73.9

1,771

74.8

1,803

23.8

724

25.2

577

24.4

563

23.8

Unknown

119

1.6

1.7

1.8

1.7

Race

Hispanic

1,138

15.0

149

5.2

126

5.3

148

6.3

Non-Hispanic Black

1,635

21.6

898

31.3

647

27.3

630

26.6

Non-Hispanic White

4,605

60.7

1,790

62.4

1,570

66.3

1,558

65.8

Other

205

2.7

1.2

1.1

1.4

City Birth

Yes

5,880

77.5

2,157

75.2

1,778

75.1

1,781

75.2

1,530

20.2

656

22.9

678

23.7

656

22.9

Unknown

173

2.3

2.0

Region code

Central

2,383

31.4

922

32.1

747

31.5

758

32.0

Northeast

1,219

16.1

514

17.9

442

18.7

431

18.2

Northwest

1,502

19.8

463

16.1

392

16.6

397

16.8

South Central

471

6.2

190

6.6

152

6.4

147

6.2

Southeast

683

9.0

333

11.6

239

10.1

246

10.4

Southwest

379

5.0

162

5.6

130

5.5

130

5.5

West Central

917

12.1

282

9.8

263

11.1

256

10.8

Unknown

0.4

Marital Status

Currently married

2,709

35.7

1,010

35.2

825

34.8

826

34.9

Never married

1,985

26.2

722

25.2

634

26.8

604

25.5

Previously married

2,889

38.1

1,138

39.7

910

38.4

939

39.6

Notes: Cells with “--” indicate counts that are 10 or less. Percentages are rounded and may not add up to precisely 100 percent.

Abbreviation: QHP=qualified health plan

Arkansas Health Care Independence Program (‘Private Option’) Final Report I- 4

Table 2. Health-Related Demographics Selected as Covariates

Raw Sample

Matched Sample

Medicaid

N=7,583

QHP

N=2,870

Medicaid

N=2,369

QHP

N=2,369

Covariates

Category

Percent

Enrollment duration during pregnancy

(in months)

219

63.4

275

18.2

274

21.1

274

19.9

Body Mass

Index (BMI)

underweight

279

3.7

2.9

3.0

3.4

normal

2,681

35.4

950

33.1

799

33.4

798

33.7

overweight

1,857

24.5

663

23.1

544

23.0

559

23.6

35 > BMI >= 30

1,252

16.5

470

16.4

424

17.9

403

17.0

40 > BMI >= 35

673

8.9

303

10.6

236

10.0

228

9.6

BMI 40 and above

596

7.9

327

11.4

236

10.0

243

10.3

not populated

245

3.2

2.5

Pre-pregnancy

diabetes

Yes

2,222

29.3

935

32.6

968

33.8

931

32.5

cesarean

Yes

1,242

16.4

545

19.0

438

18.5

432

18.2

Pre-pregnancy

hypertension

Yes

114

1.5

2.1

2.0

1.9

Preterm

birth risk

Yes

286

3.8

125

4.4

4.1

102

4.3

Chlamydia

Yes

270

3.6

3.3

3.1

3.3

Gonorrhea

Yes

0.4

0.6

0.5

Hepatitis B/C

Yes

0.4

0.7

0.5

0.6

Syphilis

Yes

Single/multiple

birth

One baby (single)

7,477

98.6

2,826

98.5

2,336

98.6

2,336

98.6

Multiples

106

1.4

1.5

1.4

Number of living children

1.26

1.32

1.35

1.32

1.34

1.31

1.29

Number of deceased children

0.03

0.2

0.04

0.2

0.03

0.2

0.04

0.2

Notes: Cells with “--” indicate counts that are 10 or less. Percentages are rounded and may not add up to precisely 100 percent.

Abbreviation: QHP=qualified health plan

Arkansas Health Care Independence Program (‘Private Option’) Final Report J- 1

Appendix J – Opioid Use Outcomes

The tables contained in this Appendix complement the data and information presented in the special section

comparing the differences in opioid use between Medicaid and QHP enrollees. Type of opioid was stratified by

Schedule III to V opioid use, short-acting Schedule II opioid use, any long-acting Schedule II opioid use, and by

specific drug types. Only oral and transdermal opioid analgesics were examined — opioid-containing cough and

cold preparations were excluded.

Arkansas Health Care Independence Program (‘Private Option’) Final Report J- 2

Table 1: Schedule of Opioids and Specific Types by General Population Enrollees and Prescriptions

Baseline

Follow-Up Year 1

Follow-Up Year 2

QHP

Medicaid

QHP

Medicaid

QHP

Medicaid

Enrollees

Percent

Enrollees

Percent

Enrollees

Percent

Total Enrollees

27,838

100

27,838

100

Total Enrollees

27,838

100

27,838

100

Total Enrollees

12,657

100

13,937

100

No Opioid Use

19,004

68.3

18,140

65.2

No Opioid Use

18,278

65.7

18,842

67.7

No Opioid Use

7,825

61.8

9,840

70.6

Opioid Use

8,834

31.7

9,698

34.8

Opioid Use

9,560

34.3

8,996

32.3

Opioid Use

4,832

38.2

4,097

29.4

Opioid Schedule

III, IV, V Only

1,605

18.2

1,609

16.6

III, IV, V Only

2,141

22.4

1,748

19.4

III, IV, V Only

1,098

22.7

861

21.0

II SA Only

4,640

52.5

5,766

59.5

II SA Only

4,484

46.9

4,914

54.6

II SA Only

2,294

47.5

2,322

56.7

II LA Any

199

2.3

203

2.1

II LA Any

273

2.9

278

3.1

II LA Any

138

2.9

125

3.1

III, IV, V, II SA Only

2,390

27.1

2,120

21.9

III, IV, V, II SA Only

2,662

27.9

2,056

22.9

III, IV, V, II SA Only

1,302

27.0

789

19.3

Total

8,834

100.1

9,698

100.1

Total

9,560

100.1

8,996

100

Total

4,832

100.1

4,097

100.1

Specific Type

Hydrocodone

6,385

72.3

6,973

71.9

Hydrocodone

6,382

66.8

6,052

67.3

Hydrocodone

3,189

66.0

2,692

65.7

Oxycodone

1,945

22.0

2,078

21.4

Oxycodone

2,248

23.5

2,061

22.9

Oxycodone

1,137

23.5

876

21.4

Tramadol

3,092

35.0

2,609

26.9

Tramadol

3,545

37.1

2,598

28.9

Tramadol

1,699

35.2

1,084

26.5

Baseline

Follow-Up Year 1

Follow-Up Year 2

QHP

Medicaid

QHP

Medicaid

QHP

Medicaid

Prescriptions

Percent

Prescriptions

Percent

Prescriptions

Percent

II SA

26,645

69.6

23,852

73.0

II SA

29,240

63.5

23,738

69.1

II SA

15,563

63.9

10,071

69.0

II LA

1,016

2.9

949

2.9

II LA

1,601

3.5

1,418

4.1

II LA

859

3.5

576

3.9

III, IV, V

10,613

27.7

7,857

24.1

III, IV, V

15,199

33.0

9,223

26.8

III, IV, V

7,949

32.6

3,958

27.1

Total

38,274

100.2

32,658

100

Total

46,040

100

34,379

100

Total

24,371

100

14,605

100

Specific Type

Hydrocodone

21,136

55.2

18,948

58.0

Hydrocodone

21,754

47.3

17,921

52.1

Hydrocodone

11,737

48.2

7,630

52.2

Oxycodone

5,131

13.4

4,533

13.9

Oxycodone

7,261

15.8

5,521

16.1

Oxycodone

3,696

15.1

2,406

16.5

Tramadol

7,844

20.5

5,466

16.7

Tramadol

10,305

22.4

5,999

17.5

Tramadol

5,286

21.7

2,519

17.3

Note: III, IV, V only=either Schedule III or Schedule IV or Schedule V prescription only; II SA only=use of Schedule II Short Acting only; II LA only=use of Schedule II Long Acting only; III, IV, V, II SA Only=use of

Schedule III or Schedule IV or Schedule V drugs with Schedule II Short Acting only. Percentages are rounded and totals may not add up to precisely 100 percent.

Abbreviations: N=number of persons; SA=short acting; LA=long acting; QHP=Qualified Health Plans.

Arkansas Health Care Independence Program (‘Private Option’) Final Report J- 3

Table 2: Schedule of Opioids and Specific Type by Higher Needs Population Enrollees and Prescriptions

2014

2015

2016

QHP

Medicaid

QHP

Medicaid

QHP

Medicaid

Enrollees

Percent

Enrollees

Percent

Enrollees

Percent

Total Enrollees

44,285

100

9,037

100

Total Enrollees

44,285

100

9,037

100

Total Enrollees

37,752

100

7,951

100

No Opioid Use

28,334

64.0

3,872

42.9

No Opioid Use

27,685

62.5

4,384

48.5

No Opioid Use

23,366

61.9

4,008

50.4

Opioid Use

15,951

36.0

5,165

57.2

Opioid Use

16,600

37.5

4,653

51.5

Opioid Use

14,386

38.1

3,943

49.6

Opioid Schedule

III, IV, V only

2,786

17.5

683

13.2

III, IV, V only

3,739

22.5

924

19.9

III, IV, V only

3,243

22.5

795

20.2

II SA only

9,132

57.3

3,095

59.9

II SA only

8,039

48.4

2,398

51.5

II SA only

7,124

49.5

2,061

52.3

II LA any

253

1.6

195

3.8

II LA any

358

2.2

225

4.8

II LA any

403

2.8

267

6.8

III, IV, V, II SA only

3,780

23.7

1,192

23.1

III, IV, V, II SA only

4,464

26.9

1,106

23.8

III, IV, V, II SA only

3,616

25.1

820

20.8

Total

15,951

100

5,165

100

Total

16,600

100

4,653

100

Total

14,386

100

3,943

100

Specific Type

Hydrocodone

11,850

74.3

3,976

77.0

Hydrocodone

11,081

66.8

3,053

65.6

Hydrocodone

9,489

66.0

2,563

65.0

Oxycodone

3,027

19.0

1,110

21.5

Oxycodone

3,658

22.0

1,172

25.2

Oxycodone

3,193

22.2

963

24.4

Tramadol

5,259

33.0

1,500

29.0

Tramadol

6,001

36.2

1,466

31.5

Tramadol

4,968

34.5

1,119

28.4

2014

2015

2016

QHP

Medicaid

QHP

Medicaid

QHP

Medicaid

Prescriptions

Percent

Prescriptions

Percent

Prescriptions

Percent

II SA

44,620

71.5

17,824

77.9

II SA

46,973

63.1

16,535

70.2

II SA

43,201

63.4

14,210

70.4

II LA

1,102

1.8

957

4.2

II LA

2,306

3.1

1,401

6.0

II LA

2,624

3.9

1,350

6.7

III, IV, V

16,657

26.7

4,102

17.9

III, IV, V

25,114

33.8

5,613

23.8

III, IV, V

22,372

32.8

4,622

22.9

Total

62,379

100

22,883

100

Total

74,393

100

23,549

100

Total

68,197

100

20,182

100

Specific Type

Hydrocodone

37,190

59.6

14,161

61.9

Hydrocodone

35,895

48.3

11,729

49.8

Hydrocodone

32,643

47.9

10,273

50.9

Oxycodone

6,933

11.1

3,344

14.6

Oxycodone

10,535

14.2

4,467

19.0

Oxycodone

10,134

14.9

3,862

19.1

Tramadol

13,216

21.2

3,223

14.1

Tramadol

17,731

23.8

3,737

15.9

Tramadol

15,976

23.4

2,978

14.8

Schedule III or Schedule IV or Schedule V drugs with Schedule II Short Acting only. Percentages are rounded and totals may not add up to precisely 100 percent.

Abbreviations: N=number of persons; SA=short acting; LA=long acting; QHP=Qualified Health Plans

Arkansas Health Care Independence Program (‘Private Option’) Final Report K- 1

Appendix K – Consumer Assessment of Healthcare Providers and Systems

(CAHPS) Sample Size Calculation, Sampling Strategy, and Data Preparation

Sample Size Calculation for CAHPS I

The sample size calculation is based on the algorithm described in Lee and Munk (2008).

The calculation assumed

a linear regression as follows:





   



 



  



where T

is an indicator for receiving coverage under the Private Option and τ therefore estimates the effect of the

Private Option at the cut-point compared to traditional Medicaid. g(Xi) is a function of the rating variable and can

accommodate the nonlinear terms of X (such as quadratic or cubic of X) and interactions with T. The sample size

calculation is based on detecting a significantly better experience in those enrolled in Private Option plans

compared to Medicaid as expected with the Arkansas Private Option.

  

  

The sample size n can be calculated as follows:

 

  









 











    







where:

n is the overall sample size of the study;







is the proportion of variation in the outcome (Y) predicted by the rating and other covariates included in the

regression discontinuity (RD) model;







is the proportion of variation in treatment status (T) predicted by the rating and other covariates included in

the RD model;

P is the proportion of sample members in the treatment group;

MDES is the minimum detectable (standardized) effect size, calculated as τ/, where  is the standard deviation

of Y;

α is the desired statistical significance level; and

1- β is the desired power level.

Under the regression discontinuity (RD) design, treatment assignment is highly correlated with the rating variable

because treatment is determined based on rating variable crossing a threshold. Therefore, it is less efficient than a

randomized controlled trial (RCT) if everything is the same except the assignment strategy.

The relative efficiency

(RE) of RCTs compared to a RD design is given by:

 



  





Arkansas Health Care Independence Program (‘Private Option’) Final Report K- 2

This ratio varies from 2.75 to 4 depending on the distribution of the rating variable around the cut-point.

For our

sample size calculation, we assumed RE=3, which means the RD design would need three times as large a sample

size as the RCT to be able to detect the same MDES. This implies that 





=2/3. We calculated the sample size at a

significant level of 0.05 (α=0.05), and a power of 80 percent (β=0.2), and assumed the RD model can explain 20

percent of the variation in Y (





 ). To account for the non-response rates and missing patterns, we used

those obtained from a previous 2013 Arkansas Medicaid CAHPS survey.

Sample size calculations were estimated for the global rating of specialist care. The maximum sample size

needed to detect a significant MDES from 0.2 to 0.4 for all items is shown below. We chose to focus on this

item to generate a more conservative sample size estimate because only about 35 percent had visited a

specialist in the previous six months in the 2013 Arkansas Medicaid CAHPS survey. The last column reports the

number of individuals needed to sample after accounting for the potential loss due to non-response. The

results presented assumed a balanced design with equal number of individuals in each group (p=0.5).

Sensitivity analyses were conducted using imbalanced design and the overall sample size needed is generally

larger. For the subsequent sampling, we chose the sample size required to detect a significant MDES of 0.25 as

the target sample size (i.e., n=9,324).

Table 1. Sample Size Requirement for Rating of Specialist Care

MDES

Assumed

Standard

Deviation

Assumed

Mean

Difference

Assumed

Mean of

Medicaid

Expected

Mean of

QHP

Assumed

Missing (no

visit and /

or no

response)

Overall Rate

of Complete

Surveys

Sample

Size

Adjusted for

Non-

Response

0.2

2.20

0.44

8.54

8.98

65%

29%

1,484

14,568

0.25

2.20

0.55

8.54

9.09

65%

29%

950

9,324

0.3

2.20

0.66

8.54

9.20

65%

29%

659

6,475

0.35

2.20

0.77

8.54

9.31

65%

29%

485

4,757

0.4

2.20

0.88

8.54

9.42

65%

29%

371

3,642

Notes: alpha=0.05, power=0.8, R²m=0.20, RT-sq=2/3

Sampling Strategy for CAHPS I

The above sample calculation is based on a parametric approach using all available data. Including all data may

increase bias if observations far away from the cut-point are included and they are significantly different from

those around the cut-point. Although the local linear regression approach is preferred, sample size determination

prior to data acquisition is difficult to assess because the existing methods are largely data driven.

The two procedures recommended for estimating “optimal bandwidth” described in the statistical analysis section

are data driven and cannot be used to determine the optimal bandwidth prior to data acquisition. A “rule of

thumb of bandwidth” has been proposed by Imbens and Kalyanaraman (2012)

that is based on a rectangular

kernel. This calculation requires only an estimate of the variance of the rating variable as follows:





   



 



where X is the rating variable of interest, 



    



  







and N is the total number of observations

in the data set.

For this calculation, we estimated standard deviations using all available data on income and exceptional

healthcare needs assessment Questionnaire scores from the sampling cohort of Medicaid and commercial

Arkansas Health Care Independence Program (‘Private Option’) Final Report K- 3

qualified health plan (QHP) enrollees prepared by ACHI. Within this cohort, there were 36,434 available

exceptional healthcare needs scores from enrollees who satisfied the inclusion and exclusion criteria and had an

exceptional healthcare needs score that was greater than 0.023 and less than 1. To utilize all available

information, for this calculation, we did not exclude individuals with missing gender, race, or market regions if

they had valid scores. Figure 1 shows the distribution of the scores. The estimated standard deviation of the score

was 0.1121. Based on this, we estimated the 



to be 0.0252.

A similar calculation was made for the income criteria comparison. Within the sampling cohort, there were 43,844

individuals who satisfied the inclusion and exclusion criteria and had valid household income and family size

information. Income was reported as monthly household income at eligibility determination, allowing for negative

values. We then extrapolated this information to 12 months to get the yearly household income. Using family size

information, we calculated the household income as the 2014 percentage of the federal poverty level (FPL). Figure

2 shows the distribution of the family income as a percentage of 2014 FPL. The estimated standard deviation of

the household income as a percentage of FPL was 0.3973. Based on this, we estimated the 



to be 0.0862.

10%

15%

20%

25%

30%

0.07

0.08

0.1

0.13

0.15

0.17

0.18

0.19

0.2

0.21

0.22

0.23

0.24

0.26

0.27

0.28

0.29

0.31

0.32

0.33

0.34

0.35

0.37

0.38

0.4

0.42

0.43

0.44

0.45

0.46

0.49

0.5

0.51

0.56

0.61

Percent of Enrollees

Exceptional Healthcare Needs Score

Figure 1. Exceptional Healthcare Needs Scores (N=36,434)

Arkansas Health Care Independence Program (‘Private Option’) Final Report K- 4

It is important to note that 



is proposed as the initial estimate for the calculation of optional bandwidth in

Imbens and Kalyanaraman (2012)

and can underestimate the bandwidth needed.

Based on the rule of thumb bandwidth calculation, we surveyed all who are within 2*



around the cut-point to

allow potential underestimation of the bandwidth. Therefore, our targeted value range for each comparison was

as following:

 For exceptional healthcare needs threshold criteria, we sampled all individuals with a score in this

range: (0.1294, 0.2304).

 For income criteria, we sampled all individuals with family income as percentage of poverty in this

range: (0 percent, 34.24 percent).

Our strategy was to sample enrollees as close to the threshold as possible. We first sampled within the specified

ranges. If there were more enrollees than the targeted sample size within these ranges, we randomly sampled to

obtain the target sample size. However, if there were insufficient numbers of individuals within the specified

ranges, we included all individuals with values within these ranges and further sampled from individuals with

values just outside of these ranges. This process was repeated until we reached our target sample size.

Table 2 summarizes our sampling strategy for the two comparisons.

10%

12%

14%

16%

10%

15%

17%

20%

25%

30%

35%

40%

45%

50%

55%

60%

65%

70%

75%

80%

85%

90%

95%

100%

105%

110%

115%

120%

125%

130%

135%

138%

Percentage of Individuals

Family Income as a Percentage of FPL

Figure 2. Family Income as a Percentage of 2014 Federal Poverty Level (FPL)

(N=43,844)

Arkansas Health Care Independence Program (‘Private Option’) Final Report K- 5

Table 2. Summary of CAHPS Sampling Strategies for Income and Exceptional Healthcare Needs Criteria

Comparisons

Exceptional Healthcare Needs Sample

Required sample size for MDES = 0.25

9,324

Per group (1:1)

4,662

ROT optimal bandwidth

0.1294

0.2304

Below 0.1799

Sampling

Method

0.15-0.17

3,775

All

Random sampling from 0.07-0.13

17,317

1,000

Random

Subtotal

4,775

Above 0.1799

Sampling

Method

0.18-0.23

3,785

All

Random sampling from 0.24-0.29

3,823

1,000

Random

Subtotal

4,785

Total

9,560

Income Sample

Required sample size for MDES = 0.25

9,324

Per group

4,662

ROT optimal bandwidth

0.3424

17% FPL or less

Sampling

Method

5,611

4,500

Random

0-17%

420

All

Subtotal

4,920

18% FPL or more

Sampling

Method

18-35%

2,669

All

35-50%

3,432

2,300

Random

Subtotal

4,969

Total

9,889

Arkansas Health Care Independence Program (‘Private Option’) Final Report K- 6

Data Preparation for CAHPS I

Arkansas Medicaid enrollment files for calendar years 2012, 2013, and 2014, which included HCIP enrollees, were

examined to determine those who were eligible for the survey. In order to be eligible for the survey, enrollees had

to have the following:

 Be continuously enrolled in Medicaid or a QHP plan from July 2014 through November 2014.

Because enrollment data at the time the CAHPS sample was created was only available through

November 2014, the last five months of enrollment were used based on the premise that the vast

majority of those continuously enrolled in the last five months of the year would have maintained

coverage through the end of the year. A minimum of six months of enrollment was required

based on the CAHPS methodology for Medicaid enrollees, which asks beneficiaries about their

experience with healthcare services received over the preceding six months.

 Be at least 19 years old and less than 65 years old on Nov. 15, 2014.

 Have full coverage under traditional Medicaid or a QHP for at least six months in 2014 in Medicaid

Aid categories, 06, 20, and 25.

 Not have had any prior coverage in aid coverage 06, 20, or 25 in 2013.

 Not be determined to be auto-frail (score>=1).

 Not have missing information on gender, race, or market region.

 Received exceptional healthcare needs assessment screening but were not in aid category 06.

 Have a valid mailing address.

The resulting subset of enrollees eligible for mailing was 188,213. A random sample of 5,000 enrollees were

selected from Medicaid and QHP enrollees separately. Further sampling steps were applied to these samples of

enrollees in order to obtain proper income and exceptional healthcare needs groups for comparison.

The final sample used for CAHPS survey distribution was 29,164. Of those surveyed, 6,568 were completed and

valid for analysis. The cooperation rate for the survey was 26.4 percent and a representative response rate was

received across populations of interest. Details of this calculation are contained in Table 3.

Table 3. Calculation of CAHPS Cooperation Rate

Sample

Size

No. of

Pieces

Received

(by Mail)

Complete

and Valid

Surveys

Received

by Mail

(M10)

Complete

and Valid

Surveys

Received

in Phone

Follow-Up

(T10)

Ineligibles

Excluded from

Analysis (Mail

Only) (M21 +

M22 + M24)

Ineligibles

Excluded from

Analysis

(Phone Only)

(T21 + T22 +

T24)

Total

Complete

and Valid

(M10 +

T10)

Forwarding

Addresses

and/or

Invalid

Phone

Numbers

(M23 + T23)

Cooperation

Rate*

29,164

6,438

6,404

164

6,568

4,228

26.4%

*Cooperation rate = (M10 + T10) / (Total Sample - (M21 + T21 + M22 + T22 + M23 + T23 + M24 + T24))

Figure 3 outlines the overall CAHPS study population development. The general comparison sample groups used

in the stabilized inverse probability of treatment weighting methodology are in boxes highlighted in blue. The

Higher Needs Population used in the regression discontinuity models comparing those who were assigned to

Medicaid or a QHP based on the medical needs assessment screener score are in boxes highlighted in pink. The

groups contained with the CAPHS sample will overlap with those contained within our analytical study population

for comparison.

Arkansas Health Care Independence Program (‘Private Option’) Final Report K- 7

Figure 3. CAPHS Analytical Data Preparation

Total Cohort

N = 225,168

Exclude those not continuously enrolled in

the last 6 months of 2014

N = 8,945

Exclude those age 65 or older as of

November 15, 2014, with missing

demographic information, or were frail and

placed in commercial coverage

N = 18,904

Exclude those with prior Medicaid coverage

in Aid Categories 06, 20, or 25

N = 2,124

Exclude those auto-frail (Frailty score >= 1)

N = 6,982

Final Overall Cohort

N = 188,213

Exclude those with invalid addresses

N = 7,007

Final Eligible Cohort for Mailing

N = 181,206

Medicaid Cohort

N = 26,544

HCIP Cohort

N = 154,662

General Population

Traditional Medicaid

N = 648

High Needs

Medicaid

N = 1,569

Commercial Coverage

N = 2,809

High Needs

Commercial QHP

N = 1,914

General Population

Commercial QHP

N = 895

Total Surveys Complete and Valid for Analysis

N = 6,568

Expansion Population

“06”

Sample Size for CAPHS survey distribution

N = 29,164

Arkansas Health Care Independence Program (‘Private Option’) Final Report K- 8

Sampling Strategy for CAHPS II

The sampling strategy for the second CAHPS survey used a pragmatic design that included four distinct

subpopulations, all following on the sampling universe used in CAHPS I. All 5,026 respondents for CAHPS I were

included in the sampling frame for CAHPS II. In addition, 5,000 participants assigned to a QHP and 5,000 assigned

to Medicaid, by virtue of exceeding a composite score threshold on the exceptional healthcare needs assessment

Questionnaire, were randomly selected from the 154,662 HCIP cohort with a valid mailing address. A further

5,000 were randomly selected from the Medicaid 20/25 aid category CAHPS I cohort (26,544). In total, 19,999

participants (excluding duplicates) were sent a CAHPS II survey that was fielded during September through

December 2016.

Data Preparation for CAHPS II

The final sample used for CAHPS survey distribution was 19,999. Of those surveyed, 4,338 were completed and

valid for analysis. The cooperation rate for the survey was 24.7 percent and a representative response rate was

received across all populations of interest. Of note, there was no telephone follow-up pursued in CAHPS II. Details

of this calculation are contained in Table 4.

Table 4. Calculation of CAHPS II Cooperation Rate

Sample Size

No. of Pieces

Received (by

Mail)

Complete and

Valid Surveys

Received by

Mail (M10)

Ineligibles

Excluded from

Analysis (Mail Only)

(M21 + M22 + M24)

Total

Complete and

Valid (M10 +

T10)

No Forwarding

Addresses (M23)

Cooperation Rate*

19,999

4,356

4,338

2,738

25.2%

*Cooperation rate = (M10) / (Total Sample - (M21 + M22 + M23 + M24)

References

Lee M and Munk T. Using Regression Discontinuity Design for Program Evaluation. Journal of the American Statistical

Association Joint Statistical Meetings, Section on Survey Research Methods. Denver, CO. 2008

Imbens GW and Kalyanaraman K. Optimal Bandwidth Choice for the Regression Discontinuity Estimator. Review of

Economics and Statistics. 2012;79(3): 933-959.